Dimerization of a viral SET protein
endows its function
Hua Wei and Ming-Ming Zhou1
Department of Structural and Chemical Biology, Mount Sinai School of Medicine, 1425 Madison Avenue, Box 1677, New York, NY 10029
Edited by Steven Henikoff, Fred Hutchinson Cancer Research Center, Seattle, WA, and approved September 7, 2010 (received for review July 13, 2010)
Histone modifications are regarded as the most indispensible
phenomena in epigenetics. Of these modifications, lysine methyla-
tion is of the greatest complexity and importance as site- and
state-specific lysine methylation exerts a plethora of effects on
chromatin structure and gene transcription. Notably, paramecium
bursaria chlorella viruses encode a conserved SET domain methyl-
transferase, termed vSET, that functions to suppress host transcrip-
tion by methylating histone H3 at lysine 27 (H3K27), a mark for
eukaryotic gene silencing. Unlike mammalian lysine methyltrans-
ferases (KMTs), vSET functions only as a dimer, but the underlying
mechanism has remained elusive. In this study, we demonstrate
that dimeric vSET operates with negative cooperativity between
the two active sites and engages in H3K27 methylation one site
at a time. New atomic structures of vSET in the free form and a
ternary complex with S-adenosyl homocysteine and a histone
H3 peptide and biochemical analyses reveal the molecular origin
for the negative cooperativity and explain the substrate specificity
of H3K27 methyltransferases. Our study suggests a “walking”
mechanism, by which vSET acts all by itself to globally methylate
host H3K27, which is accomplished by the mammalian EZH2 KMT
only in the context of the Polycomb repressive complex.
chromatin biology ∣ crystallography ∣ histone lysine methylation
states to regulate gene expression in the chromatin context
(1). Histone lysine methylation is unique in that one lysine resi-
due can have one, two, or three methyl groups; such different
methylation states have functionally distinctive outcomes (2).
Further, methylation at different lysine sites in histones can
typically result in totally different consequences. For instance,
histone H3 lysine 4 (H3K4) methylation is normally associated
with gene activation (3), whereas H3K9 or H3K27 methylation
leads to chromatin compaction and transcriptional silencing
(4, 5). The core catalytic domain of KMTs shares a conserved
structural fold called the SET domain (6), so named after its
founding members suppressor of variegation [Su(var)3-9], enhan-
cer of zeste [E(z)], and trithorax.
Given the importance of lysine methylation, it was not a sur-
prise but interesting to see that some viruses or bacteria encode
SET domain proteins that target cellular proteins to aid pathogen
replication (7). Specifically, we discovered that a SET domain
protein, vSET, found in a mature viral particle of paramecium
bursaria chlorella virus 1 (PBCV-1) (8), has explicit methyltrans-
ferase activity for host histone H3 lysine 27 (8, 9). Moreover, we
demonstrated that upon PBCV-1 infection, vSET is released
into the host cells and translocates to the cell nucleus where it
catalyzes H3K27 methylation globally, resulting in genome-wide
repression of host gene transcription (10). vSET likely exerts its
activity by mimicking the function of EZH proteins of mamma-
lian Polycomb repressive complex 2 (PRC2) that are the only
KMTs known to methylate H3K27. Despite the indispensible
role of PRC2 in Hox gene silencing during development (11),
in maintaining embryonic stem cell identity (12), and in cancer
pathophysiology (13), thus far there is no structure available
for EZH proteins, nor do we understand the detailed molecular
he flexible tails of histones undergo a variety of posttransla-
tional modifications to generate a diverse set of epigenome
basis of their KMT specificity for H3K27 methylation. vSET
serves as an attractive model to gain previously undescribed in-
sights into the activity and specificity of EZH proteins.
vSETis one of a few known examples of viral encoded enzymes
that directly modify host chromatin to benefit viral transcription
and replication (14). Uniquely, vSET functions as a dimeric
enzyme and is active on its own in cells, which contrasts sharply
to most other KMTs that are monomeric and rely on partners to
achieve optimal activity on their biological substrates (15, 16).
We sought to determine the role of vSET dimerization in its
structure and activity as an independent H3K27-specific KMT,
which has remained unresolved in our previous NMR structural
analysis of vSET due to limited structural resolution (8, 9). In this
study, we determined a high-resolution crystal structure of free
vSET, which reveals a detailed network of interactions at the
dimer interface. Guided by the atomic structure, we generated
numerous monemeric mutant proteins and found them to be
inactive, testifying to the functional importance of vSET dimer
formation. Our isothermal titration calorimetry (ITC) and enzy-
matic activity analyses of the monemeric mutants as well as a het-
erodimer vSETreveal that the enzyme exhibits independence of
the two active sites for the first binding event but strong negative
cooperativity for the second. The single-site binding model was
further strengthened by FRET studies. The atomic structure of
the enzyme/substrate/cofactor ternary complex explains precisely
the binding determinants of the H3K27 specificity by vSET
and suggests the molecular origin of the negative cooperativity.
Finally, we discuss the functional implications of these findings in
a unique viral mechanism for transcriptional gene silencing on
host chromatin that operates on vSET dimerization.
Crystal Structure of vSET. Overall structure. vSETcrystallized in two
different crystal forms. The structures were solved by single-
wavelength anomalous dispersion to 1.85 Å (condition A, see
Methods) and 1.60 Å (condition B), respectively (Fig. 1 and
Fig. S1; see Table S1 for statistics). Condition A crystal contains
one monomer per asymmetric unit that forms a dimer of C2
symmetry with another symmetry-related monomer (Fig. 1A),
whereas condition B crystal contains one dimer per asymmetric
unit (Fig. S1A). The two protomers of condition B structure are
virtually the same with an rms deviation of 0.09 Å for the Cα
atoms. Among the differences between the two structures are
the conformation of the loop containing residues 9–17 (Fig. S1C)
and the C-terminal tail of residues 106–119 that was seen only
Author contributions: H.W. and M.-M.Z. designed research; H.W. performed research;
H.W. contributed new reagents/analytic tools; H.W. and M.-M.Z. analyzed data; and
H.W. and M.-M.Z. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Data deposition: Coordinates and structure factors for free form and ternary complex
vSET have been deposited in Protein Data Bank, www.pdb.org (PDB ID codes 3KMJ,
3KMA, and 3KMT).
1To whom correspondence should be addressed. E-mail: Ming-Ming.Zhou@mssm.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/
www.pnas.org/cgi/doi/10.1073/pnas.1009911107 PNAS ∣ October 26, 2010 ∣ vol. 107 ∣ no. 43 ∣ 18433–18438
under condition A but not B presumably due to high mobility
(Fig. S1A). The tail confirmation observed in condition A is
stabilized by a few hydrogen bonds (Fig. S1D). We describe
the condition A structure for the rest of study.
The overall fold of each protomer of either structure is similar
to SET domain structures solved previously (17, 18, 19, 20, 21),
which consists of three β-sheets positioned around a pseudoknot
structure at the C terminus (9, 22). One major difference in the
crystal structure is a long continuous β-strand around the dimer
interface that connects two β-sheets, which is made up by two
connecting β-strands in the solution (8). The two structures
obtained under different conditions are similar with an rms
deviation of 0.61 Å for the Cα atoms (Fig. S1C).
Dimerinterface.The dimer interface consists of many hydrophobic
residues (Fig. 1 A and B). There are also main chain hydrogen
bonds across the dimer interface at two discrete locations. The
top site contains two symmetry-related hydrogen bonds between
the carbonyl of Asp42 of one protomer and the amide of Thr45′
At the middle site, the carbonyl group of Gly63 interacts with the
amide group of the corresponding Gly63′ in the other protomer,
thus forming an eight-membered ring due to C2symmetry
(Fig. 1D). The combination of hydrophobic and polar interac-
tions creates a stable dimer interface burying over 1;800 Å2of
solvent accessible surface area upon dimer formation.
Dimer Interface Residue Mutants of vSET. Oligomeric state of mutants.
To determine whether dimer formation is critical for the activity
of vSET, we mutated residues at the dimer interface including
Ile36, Met60, and Leu62. We checked the elution profile of
wild-type and mutant proteins at loading concentrations from
3 to 300 μM to assess the oligomeric state as a function of protein
concentration. Wild-type vSET is a dimer at all concentrations
tested (Fig. 2A, Upper Left). Conservative mutations such as
L62V and I36V shift the elution curves slightly to the right
(Fig. 2A, Upper Right), indicating that they are still dimers. Muta-
tion of Met60 or Leu62 to a charged amino acid, such as M60E,
M60D, and L62K, rendered the protein monomeric even at a
loading concentration of 300 μM (Fig. 2A, Lower Left). Alanine
mutations of the interfacial residues I36A, M60A, L62A, as well
as I36K, produced proteins that behave monomeric at low con-
centration and become dimeric at high concentration (Fig. 2A,
Activity of monomeric mutants. The activity of wild-type and mu-
tant vSETwas measured by in vitro methyltransferase assay using
histone H3 peptide substrate (residues 13–33). The reaction mix-
tures were analyzed by mass spectrometry so that at each reaction
time point relative abundances were known of the original unmo-
dified substrate and the products with different numbers of
methyl group added (Fig. 2B, Left). The relative abundance
was plotted against reaction time for each species. Curve fitting
was performed assuming a three-step irreversible consecutive
structure of dimeric vSET, solved under crystallization condition A. The
two subunits are shown in yellow and cyan, respectively. Cβ atoms of hydro-
phobic residues at the dimer interface are shown as spheres. (B) Hydrophobic
interactions at the dimer interface. The ’sign is used to differentiate residues
from different protomers. (C) Intersubunit main chain hydrogen bonds at
the top site. (D) Intersubunit main chain hydrogen bonds at the middle site.
(E) Stereo view of the 2Fo− Fcelectron density map of the condition A struc-
ture contoured at 1σ.
Crystal structure of vSET. (A) Cartoon representation of the overall
Elution volume (mL)
520 530 540 550
me0 me2 me3+1m/Z (Da)
0 100 200 300 400
Reaction time (min)
Relative abundance (%)
0 500 1000 1500 2000
Reaction time (min)
Relative abundance (%)
0 250 500 750 1000 1250
Reaction time (min)
Relative abundance (%)
Elution volume (mL)
8 1012 14 16
Elution volume (mL)
Elution volume (mL)
8 1012 1416
cal size exclusion chromatography elution profiles of vSET and its mutants.
The elution profiles of selected mutants representing different behaviors
are shown. The elution volumes of pure monomer and dimer are shown
based on protein standards. The y axes are not drawn to scale. (B) Mass
spectrometry results of wild-type vSETreaction at different time points. Peaks
are labeled that correspond to unmodified peptide (me0) as well as mono-,
di-, or trimethylated peptide (me1, me2, or me3). The me3 þ 1 peaks repre-
sent low level and nonspecific activity of vSET. The relative abundance of
each species is plotted against reaction time (Right). (C) Reaction time course
for vSET-Strep(II) homodimer. (D) Reaction time course for His-Y105F/
WT-Strep(II) heterodimer. Only k1and k2are obtained.
Oligomeric state and methyltransferase activity of vSET. (A) Analyti-
www.pnas.org/cgi/doi/10.1073/pnas.1009911107 Wei and Zhou
reaction model (Fig. 2B, Right). Three pseudo first-order rate
constants for three methylation steps were obtained through this
analysis, k1, k2, and k3, respectively. The rate constants for
wild-type vSETare listed in Table 1. The activities of monomeric
mutants were found to be much lower than those of the wild type;
only k1was obtained by analyzing samples taken at two time
points, 5 or 29.3 h of the reaction (Table 2). Mutants L62D,
M60D, and M60E did not produce any trace amount of mono-
methylated product even after 29.3 h of reaction, whereas the
activities of I36A, I36K, and L62A were reduced by more than
a thousandfold (Table 2). Even the most active monomeric
mutant L62K has only 0.35% of wild-type activity. Therefore,
we concluded that dimer formation of vSET is indispensible
for its enzymatic activity.
We ruled out a possibility that the extremely low activities of
these mutants were caused by misfolding, as the solubility and
stability of these mutants under normal buffer condition are
similar to those of the wild type. This was further supported
by the observation that 1D proton NMR spectrum of I36K shows
no significant differences from that of wild type (Fig. S2), thus
confirming that the overall fold of vSET is maintained in
I36K. So what could cause the loss of activity after dimer disrup-
tion? We postulated that dimerization maintains the two active
sites in an optimal conformation for catalysis and that vSET
monomeric mutations allosterically alter the optimal active site
conformation for cofactor and substrate binding required for a
productive catalysis. We performed isothermal titration calorime-
try measurements for some of the mutants to test this hypothesis.
Although wild-type vSETshowed nice binding (Fig. 3A), mutant
M60E showed almost no binding to H3K27 peptide substrate in
the presence of cofactor SAH (S-adenosyl homocysteine) at the
protein concentration of 300 μM (Fig. 3B). Similar results were
obtained for L62K (Fig. S3A).
Activity of a Heterodimer. To test whether the two active sites
in vSET dimer are interdependent, we generated a heterodimer
of one wild-type and one dead mutant. Y105F mutation was cho-
sen as it was shown to abolish the enzymatic activity (8). Y105F
mutant behaves as a dimer in solution as supported by size exclu-
sion chromatography, but shows abrogated peptide binding as
confirmed by ITC (Fig. S3B). pETDuet-1 vector from Novagen
was utilized for the coexpression of N-terminal His-tagged Y105F
and C-terminal Strep(II)-tagged wild-type vSET (Fig. S4A).
A two-step purification procedure with Strep-Tactin and Ni-NTA
affinity chromatography columns afforded a purified heterodi-
mer with a nearly equal amount of the wild-type and the Y105F
mutant (Fig. S4B).
Because the C-terminal residues of vSETare involved in both
cofactor and substrate binding (9), we determined as a control the
activity of the C-terminal Strep(II) tagged wild-type vSET homo-
dimer (Fig. 2C) to be k1(0.026 min−1) and k2(0.0099 min−1),
which are 42% and 47% of the rate constants of the C-terminal
untagged vSET, respectively (Table 1). We then measured the
heterodimer activity and obtained k1 and k2 of 0.013 and
0.0054 min−1, respectively (Fig. 2D). Notably, rate constants
for the first two steps are almost exactly half of those of the
WT-Strep(II) homodimer (Table 1), arguing that the activity of
the wild-type active site in the heterodimer was not affected
because the concentration of functional active site was reduced
by half in the heterodimer at a given protein concentration. In
essence, even if one active site in the vSET dimer is knocked
out, the other site is still functional.
Peptide Binding Stoichiometry of vSET Dimer. As described above,
vSET binds to the H3K27 peptide in the presence of saturating
concentration of cofactor SAH with a dissociation constant (Kd)
of 5 μM (Fig. 3A). Astonishingly, we found the binding stoichio-
metry of H3 peptide to vSET dimer to be 1.1, which means that
each dimer can bind only one peptide molecule. This finding
along with the data of the enzymatic activity of the monomeric
mutants and heterodimer vSET strongly suggests that peptide
binding at one site triggers some conformation change that pro-
pagates through the dimer interface to the other active site so that
binding affinity at the other site becomes much lower.
To validate this single-site binding model, we performed a
fluorescence resonance energy transfer study as follows: If a his-
tone H3 peptide substrate could bind both active sites of vSET
dimer simultaneously, we would detect a FRET signal between
two vSET-bound H3 peptides (residues 24–35), one fused to
cyan fluorescence protein (H3-CFP) and the other to yellow
fluorescence protein (H3-YFP) (Fig. 4A). However, we observed
virtually no change of the ratio of emission at 532 nm to that at
481 nm (an indicator of FRETsignal) for the mixture as a func-
tion of vSET concentration from 0.78–100 uM while keeping
the concentrations of H3-CFP and H3-YFP constant (Fig. 4B).
One representative emission spectrum of the mixture shows no
0 20 40 60 80 100
0 0.5 1 1.5 2 2.5
( peptide/vSET dimer)
kcal/mole of injectant
( peptide/vSET dimer)
kcal/mole of injectant
0 20 40 60 80 100
0 0.5 1 1.5 2 2.5
wild-type vSET. The solid line at the lower panel represents the fit of the
calorimetric data to a single binding site model. The vSET dimer and SAH
were kept at 0.05 mM and 1 mM, respectively, whereas the H3 peptide used
in the titration ranged from 0 to 0.12 mM. The curve fitting gave a peptide to
protein dimer ratio of 1.1 for binding. The association constant is 2.0E5 M−1
and the ΔH is −60 kJmol−1. (B) ITC data for vSET monomeric mutant M60E.
The vSET mutant and cofactor SAH concentrations were kept at 0.3 mM and
1 mM, respectively, whereas the histone H3 peptide was added in incremen-
tally from 0 to 0.72 mM, as indicated.
Isothermal titration calorimetry analysis of vSET. (A) ITC data for
Table 1. H3K27 methylation activity of vSET with different
tags and the heterodimer
Table 2. H3K27 methylation activity of vSET and its mutants
after 5 h, %
after 29.3 h, %
Wei and Zhou PNAS
October 26, 2010
prominent peak at 532 nm, which is almost identical to that of the
mixture of H3-CFP and H3-YFP without vSET (Fig. 4C). These
two lines of evidence argue that H3-CFP and H3-YFP could not
bind to vSET at the same time, consistent with the single-site
occupancy model. To generate direct evidence for this model,
we mixed vSET-YFP and H3-CFP and observed a large increase
of E532∕E481from 0.73 to 1.02 as a function of SAH concentration
from 7.8 uM–4 mM (Fig. 4 D and E). A peak at 532 nm, as a
FRET signal, was clearly seen at high concentration of SAH
(Fig. 4F). Notably, as SAH concentration passed 4 mM, a
decrease of FRET signal was observed (Fig. 4E). This could
be explained by ligand binding to the second site at very high con-
centration, which would weaken the binding of peptide at the first
site. The antagonistic effect of the second binding event is also in
line with our hypothesis that only one active site of vSET dimer is
suitable for stable binding at any given moment under physiolo-
gical conditions. To understand the molecular basis for this nega-
tive cooperativity between these two sites and to find out how
substrate specificity is achieved, we solved the crystal structure
of a ternary complex of vSET.
Crystal Structure of vSET/SAH/H3 Peptide Ternary Complex. Overall
structure. The crystal structure of the ternary complex of vSET,
SAH, and H3K27me1 peptide (resides 25–32) was solved to
1.78-Å resolution by molecular replacement (Fig. 5 A and B).
There are three monomers in the asymmetric unit, two of which
form a dimer. The other monomer forms another dimer through
symmetry operation. The rms deviation is 0.29 Å between the Cα
Emission wavelength (nM)
5 50 5005000
E532 / E481
E532 / E481
CFP or YFP
494 534 574
Emission wavelength (nM)
showing hypothetical FREToutcome resulting from one-site binding (Left) vs.
two-site binding (Right). (B) E532∕E481as a function of vSETconcentration for
a vSET/H3-CFP/H3-YFP system. Each sample has 1 μM H3-CFP, 1 μM H3-YFP,
and 0.5 mM SAH in a 20 mM Tris buffer of pH 8, containing 300 mM NaCl
and 0.01% triton X-100. The excitation wavelength is 434 nm. (C) Emission
spectrum of one sample from B with 12.5 μM of vSET. (D) Illustration of
H3-CFP’s single-site binding to vSET-YFP that would lead to a robust FRETsig-
nal. (E) E532∕E481as a function of SAH concentration for a vSET-YFP/H3-CFP
system. Each sample contains 1.6 μM vSET-YFP and 0.8 μM H3-CFP. (F) Emission
spectrum of one representative of E with 1 mM of SAH.
FRET analysis of vSET binding to histone H3. (A) Schematic diagrams
SAM binding site
Peptide binding site
d i t p
toon representation of the ternary structure. One protomer is shown in yel-
low, and the other in cyan. SAH is shown as sticks in green. H3-K27me1
peptide is shown in red. Residues 39–57 of the two subunits that undergo
rotation upon complex formation (see H) are depicted in marine and orange.
(B) Stereo view of the 2Fo− Fcelectron density map around the cofactor and
H3 peptide contoured at 1σ. (C) Electrostatic surface potential representation
of vSET. H3 peptide is shown as sticks. (D and E) Two different views of the
polar interactions between the vSET C-terminal tail and the rest of the pro-
tein in the complex. The tail is shown in red. (F) Hydrogen bonds involved in
cofactor recognition. Side chains of vSET residues, cofactor and peptide are
color-coded by atom type with carbon shown in green, cyan, and light cyan,
respectively. See Fig. S4E for another view. (G) Interactions between the
H3-K27me1 peptide and vSET. Another angle is shown in Fig. S4F. (H) Super-
imposition of the complex (green) and condition A free form (magenta)
structures using one protomer from each structure. (I) Close-up view of
the left protomers in G. vSETresidues 39–57 of the complex and the free form
are highlighted in green and magenta, respectively. (J) A simplified model
that explains the negative cooperativity between the two sites. vSETresidues
39–57 are depicted as squares. S-adenosyl methionine and peptide are shown
as orange and red objects, respectively.
Crystal structure of vSET/SAH/H3-K27me1 ternary complex. (A) Car-
www.pnas.org/cgi/doi/10.1073/pnas.1009911107 Wei and Zhou
atoms of these two dimers so that each is representative of both.
We focus on the dimer in the asymmetric unit for the rest of the
paper. The overall fold of vSET in the ternary complex is similar
to that of free protein with an rms deviation of 0.87 Å (Fig. 5 A
and Fig. S5A). The rms deviation between the two protomers in
the ternary complex dimer is 0.20 Å, so the ternary complex still
maintains the C2symmetry. Peptide and SAH molecules are
found at both active sites in the dimer. This seeming discrepancy
with the ITC data could be explained by the much higher concen-
tration of all components needed for crystallization. Like all
other SET domain complex structures (21), the cofactor and pep-
tide are found to bind two opposite sides of the vSETsurface that
are connected through a narrow pore (Fig. 5C and Fig. S5B).
Ordering of the C-terminal tail upon complex formation. Despite the
overall structural similarity between the complex and free form,
the C-terminal tail undergoes a dramatic conformational change
upon complex formation adopting a one-and-a-half turn α-helix
and a type I β-turn around Pro114 (Fig. 5A vs. Fig. 1A). The tail is
anchored to the rest of the protein through a myriad of polar in-
teractions (Fig. 5 D and E). Arg113 forms a salt bridge with
Asp49, the main chain carbonyl of which forms a hydrogen bond
with Tyr109 (Fig. 5D). Mutation of the residue in Drosophila E(z)
that corresponds to Asp49 of vSETwas shown to abrogate its ac-
tivity (23). Side chains of Arg113 and Tyr109 stack, contributing
to the stability of the α-helix of the tail. Tyr105 is hydrogen-
bonded to carbonyl of Phe68 with a short distance of 2.61 Å.
These hydrogen bonds explain the importance of Tyr105 and
Tyr109, mutation of either of which abrogates the methyltransfer-
ase activity of vSET (8, 9). The Tyr105 hydrogen bond also ex-
plains why even the conservative phenylalanine mutation
eliminates binding as observed in our ITC study. The very last
two residues of vSET are also important for the positioning of
the tail. Gln118 contacts His70 and is also a member of a
water-mediated hydrogen-bonding network involving the C-term-
inal carboxylate of Asn119 and the amide of Ala72 (Fig. 5E). The
side-chain amide of Asn119 in addition interacts with the carbo-
nyl of Leu13. The conformation of the tail is further stabilized by
hydrophobic interactions of Trp110 and its surrounding residues
(Fig. S5C). It is interesting to see that through this combination of
interactions vSETadopts the conformation of its tail suitable for
the catalysis, which is achieved by zinc sulfur cluster in some other
SET KMTs (24) (Fig. S5D). Just as substrate and cofactor binding
is a prerequisite for the tail to adopt this conformation, this con-
formation in turn is also essential for their binding (seen below).
Cofactor binding.The SAH cofactor fits snugly in vSETupon com-
plex formation with its adenine ring inserted deeply into a pocket
(Fig. S5B). One face of the adenine is covered by the aforemen-
tioned Trp110 (Fig. S5C), whereas the other face interacts with
Leu13. Mutation of corresponding latter residue in Drosophila E
(z) leads to its loss of methyltransferase activity (23). All hydro-
gen-bonding capabilities of the amino acid moiety of SAH are
satisfied (Fig. 5F and Fig. S5E). The extracyclic amino group
of adenine interacts withboth the carbonyl of His70and the term-
inal carboxylate of Asn119. The involvement of Tyr109 and
Asn119 in cofactor binding partially explains that the emergence
of the new conformation of the C-terminal tail happens only after
ligand and/or peptide binding. The fact that the C-terminal car-
boxylate of vSET is involved both in the ordering of the tail
(Fig. 5E) and in cofactor binding (Fig. 5F) explains why adding
a C-terminal tag decreases its activity to about half of its original
value (Table 1).
Substrate specificity. The H3K27me1 peptide binds to vSET in an
extended conformation (Fig. 5 A, C, and G). One outstanding
feature is the recognition of H3R26 through exhaustive hydro-
gen-bonding/salt bridge network. H3R26 has a divalent salt
bridge with Glu48 using Nε and Nη2 atoms (Fig. 5G). The same
Nη2 forms water-mediated hydrogen bonds with the carbonyls of
Gly44 and Asn40. Nη1 of H3R26 is contacted by both the carbo-
nyl of Asn40 and the side chain of Ser58. All other polar inter-
actions between the peptide and vSET involve only the main
chain atoms of residues 25 to 30 of the peptide. The amide
and carbonyl of H3K27me1 interact with the carbonyl of
Leu51 and amide of Ser53, respectively. The same groups of
Leu51 and Ser53 have been found to interact with the tail for
the protein alone under crystallization condition A (Fig. S1D).
To investigate whether side chains other than those of H3R26
and H3K27 contribute to vSET specificity for H3K27 methyla-
tion, we looked at the surface complementarity between the
enzyme andthe substrate. H3R26 packs against a flat surface with
one tall wall on one side and a small bump on the other (Fig. 5C).
The methyl group of H3A25 is pointed to a wall. Any residue
with large side chain at this position would induce congestion.
Another critical residue H3A29 is found to point to a shallow
concave surface. Replacement of A29 with any nonglycine amino
acid would cause steric clashes between the substrate and vSET.
This explains vSET’s specificity for H3K27 but not H3K9.
Although the H3K9 region also contains ARKS sequence as
found around H3K27, it is followed by threonine instead of
alanine. This threonine would clash with the protein surface if
a H3K9 peptide were to bind. Indeed, the methyltransferase ac-
tivity of vSETwas found to be 20-fold lower when a H3K9 peptide
was used as a substrate (8). Taken together, the preferred sub-
strate sequence for vSET can be described as #RKXA where
# represents an amino acid with small side chain and X repre-
sents any amino acid.
StructuralBasisfortheNegativeCooperativity.Aligning the complex
and the free structures using just one subunit from each structure
reveals previously unseen conformation differences at the other
subunit. (Fig. 5H vs. Fig. S5A). So substrate and cofactor binding
does seem to affect dimer organization. Moreover, significant
conformational changes are seenaround resides 39to 57(Fig.5A,
residues in marine and orange) for the aligning subunits (Fig. 5I),
which corresponds to some rotational motion from the top view.
Upon complex formation, Tyr50 is pulled away and makes room
for cofactor binding. Arg54 also becomes ordered and the chain
around this residue gets closer to the peptide substrate. These
seemingly minor but large-scale conformational changes may
explain the negative cooperativity between the two active sites
(Fig. 5J). In the free state, the original cofactor pocket is slightly
smaller than needed for tight binding, whereas the peptide pocket
is slightly larger. Because interprotomer main chain hydrogen
bonds involve residues 42 and 45 (Fig. 1C), it is conceivable that
binding of cofactor and peptide could trigger local conforma-
tional adjustments in the region of residues 39 to 57, which could
cause similar rotational movement in the other protomer at the
opposite direction. This ligand binding-induced trans protomer
motion would render the cofactor-binding pocket even smaller
and the peptide pocket even larger, making them unfit for pro-
ductive ligand and substrate binding at the second site simulta-
neously as the first site at their physiological concentrations.
The current study reports the atomic resolution structures of an
H3K27-specific lysine methyltransferase. The ternary complex
structure illustrates how H3K27 recognition is achieved. R26 of
histone H3 is recognized by five ionic/hydrogen-bonding interac-
tions (Fig. 5G). H3A29, on the other hand, is identified through
the existence of a shallow pocket around its side chain on the
protein surface (Fig. 5C). But can we apply the same principles
to the important mammalian EZH KMTs that have the same spe-
cificity? Sequence alignment of vSET with human EZH1 and
Wei and Zhou PNAS
October 26, 2010
EZH2, and Drosophila E(z) shows a sequence identity of 26% Download full-text
and a similarity of 53% (Fig. S6). The residues that are respon-
sible for creating the shallow pocket around peptide H3A29 are
Arg54, Leu80, and Leu84 in vSET, which are replaced by Phe,
Met, and Asp, respectively, in E(z) proteins (Fig. S6, residues
in red). One common feature about these residues (the original
and the replacements) is their bulkiness. So this critical recogni-
tion mode that distinguishes H3K27-specific KMTs from H3K9
KMTs may also exist in E(z) family KMTs. In vSET, Glu48 forms
a salt bridge with H3R26. It is substituted by Asp in E(z) proteins
(Fig. S6, green residues), so bonding possibility remains. The
other H3R26-contacting residue in vSET, Ser58, appears to be
changed to Asn. Collectively, it seems that whereas the precise
conformation of active site residues in the EZH KMTs required
for productive catalysis is dependent on and controlled by their
interactions with othercomponent proteins in the PRC2 complex,
E(z) family SET domain KMTs likely use similar strategies to
recognize R26 and A29 of histone H3.
Our combined structural, biophysical and enzymatic study il-
luminates how vSET functions as a dimer, and uniquely operates
H3K27 methylation catalysis one active site at a time. Our results
lead us to the following in vivo model (Fig. 6). When in its free
form, each subunit of the vSET dimer has equal chance of finding
its substrate and cofactor due to the C2symmetry. However,
ligand occupancy of one active site triggers conformational
change around residues 39 to 57, which propagates across the
dimer interface so that the other active site is not suitable for
simultaneous binding. Binding at the secondsite could occur after
the methylation reaction and release of cofactor and H3 at the
first site; the second binding could also facilitate ligand dissocia-
tion at the first site. The above sequence of events could repeat
numerous times on chromatin only at the expense of methyl
donor cofactor, S-adenosyl methionine, which would result in
the deposition of H3K27 trimethylation marks over large regions
of chromosome. In this proposed mechanism, vSET remains
associated with chromatin all the time while methylating H3K27
by simply “walking” on chromatin. This mechanism would endow
this viral enzyme an efficiency much higher than that by constant
dissociation and reassociation were it monomeric. This would
thus make it possible for vSET to execute genome-wide methyla-
tion of histone H3K27 to suppress global host gene transcription.
Similar walking mechanisms are best known in motor proteins
such as myosins and kinesins (25), which are driven by ATP
hydrolysis. But in vSET, cofactor conversion is used concurrently
for the viral enzyme to catalyzetarget histone H3K27 methylation
and feed itself the host substrate, a unique function that vSET
owes all to its dimerization.
The experimental procedures are briefly described here. A detailed descrip-
tion is provided in SI Methods. vSET proteins were obtained from BL21(DE3)
cells. The protein used for structural studies was the full-length vSETwithout
any tag, whereas the protein used for activity and binding studies contained
an N-terminal His tag by default. Crystal structures were solved first by single-
wavelength anomalous dispersion using selenomethionine-labeled protein
and subsequently by molecular replacement. Similar conditions were used
for methyltransferase assay as described previously (6). Isothermal titration
calorimetric measurements were performed with an Omega instrument
ACKNOWLEDGMENTS. We acknowledge the staff at the X6A and X4C beam-
lines of the National Synchrotron Light Sources at the Brookhaven National
Laboratory for facilitating X-ray data collection. We also thank X. Q. Wang
and A. N. Plotnikov for technical advice and helpful discussion. The work was
in part supported by the grants from the National Institutes of Health
(GM073207 and DA028776) (to M.-M.Z.).
1. Jenuwein T, Allis C (2001) Translating the histone code. Science 293:1074–1080.
2. Barski A, et al. (2007) High-resolution profiling of histone methylations in the human
genome. Cell 129:823–837.
3. Noma K, Grewal SI (2002) Histone H3 lysine 4 methylation is mediated by Set1 and
promotes maintenance of active chromatin states in fission yeast. Proc Natl Acad
Sci USA 99(Suppl 4):16438–16445.
4. Nakayama J, Rice J, Strahl B, Allis C, Grewal S (2001) Role of histone H3 lysine 9
methylation in epigenetic control of heterochromatin assembly. Science 292:110–113.
5. Cao R, et al. (2002) Role of histone H3 lysine 27 methylation in Polycomb-group
silencing. Science 298:1039–1043.
6. AravindL, Iyer L (2003) Provenance of SET-domain histone methyltransferases through
duplication of a simple structural unit. Cell Cycle 2:369–376.
7. Qian C, Zhou M (2006) SET domain protein lysine methyltransferases: Structure,
specificity and catalysis. Cell Mol Life Sci 63:2755–2763.
8. Manzur K, et al. (2003) A dimeric viral SET domain methyltransferase specific to Lys27
of histone H3. Nat Struct Biol 10:187–196.
9. Qian C, et al. (2006) Structural insights of the specificity and catalysis of a viral histone
H3 lysine 27 methyltransferase. J Mol Biol 359:86–96.
10. Mujtaba S, et al. (2008) Epigenetic transcriptional repressionof cellular genes by a viral
SET protein. Nat Cell Biol 10:1114–1122.
11. Wang J, Mager J, Schnedier E, Magnuson T (2002) The mouse PcG gene eed is
required for Hox gene repression and extraembryonic development. Mamm Genome
12. Lee TI, et al. (2006) Control of developmental regulators by Polycomb in human
embryonic stem cells. Cell 125:301–313.
13. Pasini D, Bracken AP, Helin K (2004) Polycomb group proteins in cell cycle progression
and cancer. Cell Cycle 3:396–400.
14. Wei H, Zhou MM (2009) Viral-encoded enzymes that target host chromatin functions.
Biochim Biophys Acta 1799:296–301.
15. Dou Y, et al. (2006) Regulation of MLL1 H3K4 methyltransferase activity by its core
components. Nat Struct Mol Biol 13:713–719.
16. Kuzmichev A, Nishioka K, Erdjument-Bromage H, Tempst P, Reinberg D (2002) Histone
methyltransferase activity associated with a human multiprotein complex containing
the enhancer of Zeste protein. Genes Dev 16:2893–2905.
17. Min J, Zhang X, Cheng X, Grewal SI, Xu RM (2002) Structure of the SET domain histone
lysine methyltransferase Clr4. Nat Struct Biol 9:828–832.
18. Wilson J, et al. (2002) Crystal structure and functional analysis of the histone methyl-
transferase SET7/9. Cell 111:105–115.
19. Trievel R,Beach B, DirkL, HoutzR, HurleyJ (2002) Structure and catalyticmechanism of
a SET domain protein methyltransferase. Cell 111:91–103.
20. Jacobs SA, et al. (2002) The active site of the SET domain is constructed on a knot. Nat
Struct Biol 9:833–838.
21. Xiao B, et al. (2005) Specificity and mechanism of the histone methyltransferase
Pr-Set7. Genes Dev 19:1444–1454.
22. Dillon SC, Zhang X, Trievel RC, Cheng X (2005) The SET-domain protein superfamily:
protein lysine methyltransferases. Genome Biol 6:227.
23. Joshi P, et al. (2008) Dominant alleles identify SET domainresiduesrequired for histone
methyltransferase of Polycomb repressive complex 2. J Biol Chem 283:27757–27766.
24. Southall SM, Wong PS, Odho Z, Roe SM, Wilson JR (2009) Structural basis for the
requirement of additional factors for MLL1 SET domain activity and recognition of
epigenetic marks. Mol Cell 33:181–191.
25. Cross RA (1999) Molecular motors: Walking talking heads. Curr Biol 9:R854–856.
Histone H3 tail
for the initial binding. Once that happens, it triggers some conformational
change around the first binding site that propagates across the dimer inter-
face to the other active site making it unfit for simultaneous binding. After
the reaction takes place at the first site, binding occurs at the second site with
concurrent release of cofactor and peptide at the first site. The aforemen-
tioned events could be repeated many times to allow efficient deposit of
H3K27 methylation marks along long stretches or clustered regions of chro-
matin. The methylated H3K27 is shown as red dots that represent a mixture
of products at different methylation stages. The movement of vSET along
chromatin may be a random walk.
A walking model for vSET. Each site in the dimer has an equal chance
www.pnas.org/cgi/doi/10.1073/pnas.1009911107Wei and Zhou