the gcvT 5¶-UTR of B. subtilis responds to
glycine with characteristics that parallel those
observed when conducting inline probing of
the cooperative VC I-II RNA.
To assess whether glycine binding and in
vitro transcription control correspond to genet-
ic control events in vivo,we generated reporter
operon from B. subtilis to a $-galactosidase
reporter gene and integrated them into the
bacterial genome (17). The reporter fusion
constructcarrying the wild-type IGR expresses
a high amount of $-galactosidase whenglycine
is present in the growth medium, whereas a
low amount of gene expression results when
alanine is present (Fig. 4C). These results
indicate that the gcvT motif is part of a
glycine-responsive riboswitch with a default
state that is off. Glycine binding is required to
activate gene expression, as was also observed
with the in vitro transcription assays (Fig. 4B).
The importance of several conserved
features of the motif were examined by
mutating the P1 and P2 stems of the first
aptamer domain to disrupt (variants M1 and
M3, respectively) and restore (M2 and M4,
respectively) base pairing (Fig. 4A). Result-
ing gene expression levels from constructs
carrying the mutant IGRs are consistent with
base-paired elements predicted from phylo-
genetic analyses (14) (fig. S1). Furthermore,
the introduction of mutations into the con-
served cores of either aptamer I or aptamer II
(variants M5 and M6, respectively) caused a
complete loss of reporter gene activation.
This latter result suggests that glycine
binding to both aptamers is necessary to
trigger gene activation, which is consistent
with a model wherein cooperative glycine
binding is important for riboswitch function.
The glycine-dependent riboswitch is a
remarkable genetic control element for several
reasons.First,glycine riboswitchesform selec-
tive binding pockets for a ligand composed of
only 10 atoms and thus bind the smallest
organic compound among known natural and
engineered RNA aptamers. This observation is
consistent with the hypothesis that RNA has
sufficient structural potential to selectively
bind a wide range of biomolecules.
Second, the 5¶-UTR of the B. subtilis gcvT
operon is a genetic on switch, and thus joins
the adenine riboswitch (23) as a rare type of
RNA that has been proven to harness ligand
binding and activate gene expression. In most
instances, riboswitches cause repression of
their associated genes, which is to be expected
because many of these genes are involved in
biosynthesis or import of the target metabo-
lites. However, the glycine riboswitch from B.
subtilis controls the expression of three genes
required for glycine degradation. A ligand-
activated riboswitch would be required to
determine whether sufficient amino acid sub-
strate is present to warrant production of the
glycine cleavage system, thereby providing a
rationale for why this rare on switch is used.
Third, this is the only known metabolite-
binding riboswitch class that regularly makes
use of a tandem aptamer configuration. In both
V. cholerae and B. subtilis, the juxtaposition of
aptamers enables the cooperative binding of
two glycine molecules. For the B. subtilis ribo-
switch, this characteristic is expected to result
in unusually rapid activation and repression of
genes encoding the glycine cleavage system in
response to rising and falling concentrations of
glycine, respectively. Given the prevalence of
the tandem architecture of glycine ribo-
switches, this more Bdigital[ switch likely
gives the bacterium an important selective
advantage by controlling gene expression in
response to small changes in glycine.
References and Notes
1. W. C. Winkler, R. R. Breaker, ChemBioChem 4, 1024
2. A. G. Vitreschak, D. A. Rodionov, A. A. Mironov, M. S.
Gelfand, Trends Genet. 20, 44 (2004).
3. E. Nudler, A. S. Mironov, Trends Biochem. Sci. 29, 11
4. M. Mandal, B. Boese, J. E. Barrick, W. C. Winkler, R. R.
Breaker, Cell 113, 577 (2003).
5. A. S. Mironov et al., Cell 111, 747 (2002).
6. W. C. Winkler, S. Cohen-Chalamish, R. R. Breaker,
Proc. Natl. Acad. Sci. U.S.A. 99, 15908 (2002).
7. A. Nahvi et al., Chem. Biol. 9, 1043 (2002).
8. W. Winkler, A. Nahvi, R. R. Breaker, Nature 419, 952
9. N. Sudarsan, J. E. Barrick, R. R. Breaker, RNA 9, 644
10. W. C. Winkler, A. Nahvi, A. Roth, J. A. Collins, R. R.
Breaker, Nature 428, 281 (2004).
11. M. Ptashne, A. Gann, Genes & Signals (Cold Spring
Harbor Press, Cold Spring Harbor, NY, 2002).
12. B. I. Kurganov, Allosteric Enzymes (Wiley, New York,
13. A. A. Antson et al., Nature 374, 693 (1995).
14. J. F. Barrick et al., Proc. Natl. Acad. Sci. U.S.A. 101,
15. G. Kikuchi, Mol. Cell. Biochem. 1, 169 (1973).
16. R. Duce, J. Bourguignon, M. Neuburger, F. Re ´beille ´,
Trends Plant Sci. 6, 167 (2001).
17. Materials and methods are available on Science Online.
18. G. A. Soukup, R. R. Breaker, RNA 5, 1308 (1999).
19. M. Mandal et al., unpublished data.
20. A. M. Jose, G. A. Soukup, R. R. Breaker, Nucleic Acids
Res. 29, 1631 (2001).
21. W. C. Winkler, A. Nahvi, N. Sudarsan, J. E. Barrick,
R. R. Breaker, Nature Struct. Biol. 10, 701 (2003).
22. N. Sudarsan, J. K. Wickiser, S. Nakamura, M. S. Ebert,
R. R. Breaker, Genes Dev. 17, 2688 (2003).
23. M. Mandal, R. R. Breaker, Nature Struct. Mol. Biol.
11, 29 (2004).
24. A. Nahvi, J. E. Barrick, R. R. Breaker, Nucleic Acids Res.
32, 143 (2004).
25. A. V. Hill, J. Physiol. 40, iv (1910).
26. M. Weissbluth, in Molecular Biology Biochemistry and
Biophysics, A. Kleinzeller, Ed. (Springer-Verlag, New
York, 1974), vol. 15, pp. 27–41.
27. S. J. Edelstein, Annu. Rev. Biochem. 44, 209 (1975).
28. I. Gusarov, E. Nudler, Mol. Cell 3, 495 (1999).
29. W. S. Yarnell, J. W. Roberts, Science 284, 611 (1999).
30. We thank members of the Breaker laboratory for
helpful discussions and G. Reguera and B. Bassler for
providing genomic DNA for V. cholerae. This work
was supported by grants from the NIH and the NSF.
R.R.B. is also grateful for support from the Yale Liver
Center and the David and Lucile Packard Foundation.
Supporting Online Material
Materials and Methods
Figs. S1 to S5
27 May 2004; accepted 24 August 2004
Human PAD4 Regulates Histone
Arginine Methylation Levels
Yanming Wang,1,2Joanna Wysocka,1,2Joyce Sayegh,3
Young-Ho Lee,4Julie R. Perlin,1Lauriebeth Leonelli,1
Lakshmi S. Sonbuchner,1Charles H. McDonald,5Richard G. Cook,5
Yali Dou,6Robert G. Roeder,6Steven Clarke,3
Michael R. Stallcup,4C. David Allis,2* Scott A. Coonrod1*
Methylation of arginine (Arg) and lysine residues in histones has been cor-
related with epigenetic forms of gene regulation. Although histone methyl-
transferases are known, enzymes that demethylate histones have not been
identified. Here, we demonstrate that human peptidylarginine deiminase 4
(PAD4) regulates histone Arg methylation by converting methyl-Arg to
citrulline and releasing methylamine. PAD4 targets multiple sites in histones
H3 and H4, including those sites methylated by coactivators CARM1 (H3 Arg17)
and PRMT1 (H4 Arg3). A decrease of histone Arg methylation, with a con-
comitant increase of citrullination, requires PAD4 activity in human HL-60
granulocytes. Moreover, PAD4 activity is linked with the transcriptional
regulation of estrogen-responsive genes in MCF-7 cells. These data suggest
that PAD4 mediates gene expression by regulating Arg methylation and
citrullination in histones.
Posttranslational histone modifications, such
as phosphorylation, acetylation, and methyl-
ation, regulate a broad range of DNA and
chromatin-templated nuclear events, including
transcription (1–3). Pairs of opposing en-
zymes, such as acetyltransferases-deacetylases
R E P O R T S
www.sciencemag.orgSCIENCE VOL 3068 OCTOBER 2004
and kinases-phosphatases, regulate the steady-
state balance of histone acetylation and phos-
phorylation, respectively. In contrast, although
Arg- and Lys-specific methyltransferases have
been identified (3–5), enzymes that remove
methyl groups from histones or any other
cellular proteins remain unknown (6).
Arg methylation has been identified on
many nuclear and cytosolic proteins involved
in various cellular processes, including tran-
scription and cell signaling (7–10). The
methylation of histones by PRMT1 and
CARM1 facilitates transcription in associa-
tion with nuclear hormone coactivators and
p53 (11–15). Here, we demonstrate that
peptidylarginine deiminase 4 (PAD4), an en-
zyme previously known to convert Arg to
citrulline (Cit) in histones (16–19), can also
demethyliminate histones in vitro and in
vivo, thus regulating both histone Arg
methylation and gene activity.
Multiple Arg residues in H3 and H4 can
be methylated by CARM1 and PRMT1,
respectively (fig. S1A). Free methyl-Arg
amino acids (monomethyl-Arg and asymmet-
ric dimethyl-Arg) can be converted to Cit by
(DDAH) (20–21). To identify enzymes that
might catalyze a similar reaction on protein
methyl-Arg substrates as that catalyzed by
DDAH, we searched the Homologous Struc-
ture Alignment database for proteins
homologous to DDAH and identified PAD4
(22) (fig. S1). Peptidylarginine deiminases
are a family of enzymes known to convert
protein Arg to Cit in a calcium- and
dithiothreitol (DTT)-dependent reaction
Ereviewed in (16)^. These findings prompted
us to test the hypothesis that PAD4 can
convert histone methyl-Arg to Cit.
Previous studies have correlated PAD4 ex-
pression with histone citrullination (17–18).
We purified a glutathione S-transferase
(GST)–PAD4 (human) full-length fusion pro-
tein from Escherichia coli and tested it on
reversed-phase high performance liquid chro-
matography (RP-HPLC)–purified cellular H3
and H4 as substrates. In the presence of cal-
cium and DTT, GST-PAD4 generated Cit in
H3 and H4 as detected by an antibody against
a chemically modified form of Cit (Fig. 1, A
and B). Cellular H3 and H4 either treated or
untreated was probed with site-specific anti-
bodies against methyl-H3 Arg17and -H4 Arg3
residues (for antibody specificity, see fig. S2).
A dramatic diminishment of H3 Arg17and H4
Arg3methylation was detected after PAD4
treatment (Fig. 1, A and B), suggesting that
PAD4 targets select methyl-Arg sites in H3
and H4. Protein microsequencing showed that
the N-terminal tail of PAD4-treated H3 and
H4 was not being randomly degraded (table
S1). To identify potential PAD4 target site(s)
in the N-terminal tail of H3, we quantified the
amount of Cit detected at cycles of micro-
sequencing. As shown in table S1, PAD4
deiminated multiple Arg residues in H3 (e.g.,
È93.6% of H3 Cit2compared to È98.9% of
H3 Cit8) in vitro. Cellular H4 is N-terminally
acetylated, thus preventing direct microse-
quencing analyses. Therefore, we analyzed
recombinant H4 after PAD4 treatment and
found that its N terminus also remained intact
and that È99.6% of H4 Arg3was citrullinated
sites of H3 and H4 to Cit with low site
preference in vitro. Neutralization of the pos-
itive charge of multiple Lys residues by
acetylation alters the electrophoretic behavior
of histones in SDS–polyacrylamide gel elec-
trophoresis (SDS-PAGE) gels (23). Because
the positive charge of Arg is neutralized by
citrullination, we postulate that the mass shift
of histones observed on SDS-PAGE gels after
PAD4 treatment is caused by deiminating
multiple Arg residues in H3 and H4 and that
the varying degrees of citrullination at differ-
ent Arg residues caused the expansion of the
band width of H3 and H4 (Fig. 1).
Two possible pathways can lead to the
loss of methyl-Arg epitope (Fig. 2A). Either
PAD4 removes the methylimine group from
methyl-Arg, thus producing Cit and releas-
ing methylamine (pathway 1), or the imine
group is removed by PAD4 thereby produc-
ing methyl-Cit and releasing ammonium
(pathway 2). To distinguish between these
two pathways, we radioactively labeled
recombinant H3 and H4 with CARM1 and
PRMT1, respectively, and with E3H^-S-
adenosylmethionine as a methyl donor. The
amount of E3H^-methyl in H3 and H4 was
then detected by fluorography. We found
that the amounts of E3H^-methyl in histones
were dramatically decreased by PAD4 treat-
ment (Fig. 2B). These results suggest that the
methyl-group produced on H3 and H4 by
CARM1 and PRMT1, respectively, is direct-
ly removed by PAD4.
We sought to analyze the biochemical
nature of the released product. If PAD4 acts
via pathway 1 (Fig. 2A), methylamine would
be generated. To detect methylamine, we
first took advantage of the solubility differ-
ence of methylamine in H2O at different pH
values (methylamine pKa0 10.4) (24, 25).
After PAD4 treatment of recombinant H4
radioactively labeled by PRMT1, released
volatile E3H^-methyl radioactivity was de-
tected from samples adjusted to a high pH
(pH 0 12, at which methylamine becomes
volatile) but not from various control sam-
ples (Fig. 2C). The identity of methylamine
as a methyl product was further confirmed
by chromatography using an amino acid
cation-exchange column (26). Radioactivity
released from PAD4-treated E3H^-methyl-H4
co-migrated with both an unlabeled mono-
methylamine standard (absorbance, 570 nm)
and a E14C^-dimethylamine standard, indi-
cating that the volatile E3H^-methylamine
could be released in a monomethyl or
dimethyl form (Fig. 2D) (27). In contrast,
E3H^-methylamine was not detected in the
untreated E3H^-methyl-H4 samples (Fig. 2E).
These results support the hypothesis that
PAD4 can convert methyl-Arg in histones
to Cit and methylamine. Hereafter, we will
1Department of Genetic Medicine, Weill Medical
College of Cornell University, 1300 York Avenue,
New York, NY 10021, USA.2Laboratory of Chromatin
Biology, Rockefeller University, Box 78, 1230 York
Avenue, New York, NY 10021, USA.3Department of
Chemistry and Biochemistry and Molecular Biology
Institute, University of California at Los Angeles, Los
Angeles, CA 90095–1569, USA.
Pathology, University of Southern California, Los
Angeles, CA 90089–9092, USA.5Department of Mi-
crobiology and Immunology, Baylor College of
Medicine, Houston, TX 77030, USA.
Biochemistry and Molecular Biology, Rockefeller
University, New York, NY 10021, USA.
*To whom correspondence should be addressed.
E-mail: email@example.com (C.D.A.); scc2003@
Fig. 1. PAD4 reduces Arg methyla-
tion levels and generates citrulline
(Cit) in H3 (A) and H4 (B). (Left) Cit
was detected in H3 or H4 when
treated with PAD4. (Middle) After
PAD4 treatment, the signal of H3
Arg17or H4 Arg3methylation was
dramatically diminished (see fig. S2
for antibody specificity). (Right)
Silver staining shows H3 and H4,
as well as citrullinated H3 and H4
(H3* and H4*) in SDS-PAGE gels.
Note the increased mobility of H3*
PAD4 - +
- +- +
PAD4 - +
R E P O R T S
8 OCTOBER 2004VOL 306 SCIENCEwww.sciencemag.org
refer to this reaction as demethylimination to
reflect these findings.
We next examined whether PAD4 mod-
ulates histone Arg methylation and citrullin-
ation in vivo. We chose to test this in HL-60
granulocytes where PAD4 expression can be
induced by dimethyl sulfoxide (DMSO) and
PAD4 can be activated by calcium ionophore
(17, 18) (Fig. 3A). When total histones were
probed with site-specific antibodies against H3
methyl-Arg17or H4 methyl-Arg3, the signals
were dramatically reduced after PAD4 activa-
tion (Fig. 3B). In addition, calcium ionophore
treatment did not either increase histone
citrullination (Fig. 3A) or decrease histone
Arg methylation in undifferentiated HL-60
cells (28). These results correlate the activa-
tion of PAD4 with a loss of histone Arg meth-
ylation in a cellular context.
To further analyze the change of Arg
methylation in individual cells, we carried
out immunofluorescence analyses of HL-60
granulocytes. Before treatment, amounts of
H3 Arg17methylation in each cell were rough-
ly comparable (Fig. 3C, top). In contrast, after
15 min of calcium ionophore treatment, H3
Arg17methylation dramatically decreased in
most of the cells (È57.3%, n 0 200) (Fig. 3C,
bottom). In contrast, amounts of H3 Lys4
methylation were unchanged in calcium
ionophore-treated cells (fig. S4), suggesting
that the N terminus of H3 is intact and
that PAD4 does not affect Lys methylation.
To directly demonstrate the conversion of
particular H3 Arg residues to Cit in vivo, we
Fig. 2. PAD4 demeth-
yliminates H3 and H4
and produces methyl-
amine and Cit as reac-
tion products. (A) Two
of PAD4 reaction on
methyl-Arg in a pro-
tein substrate. (B) Re-
combinant H3 or H4
was first radioactively
labeled by CARM1 or
After PAD4 treat-
ment, the [3H]-methyl
radioactivity in H3 and
H4 dramatically de-
creased. (C) A volatility
assay (25) to detect
leased from radio-
actively labeled H4
after PAD4 treatment.
[3H]-activity was found
only from samples at
a high pH (12) after-
PAD4 treatment. Error
means T SD of three
(D) A nonradioactive methylamine standard [10 6mol, detected by
absorbance (A) at 570 nm] was co-eluted with the released radioactive
products generated by PAD4 from radioactively labeled recombinant H4,
suggesting that [3H]-methylamine was produced. (E) [3H]-methylamine
was not detected in radioactively labeled recombinant H4 samples that
were not treated with PAD4.
[3H-methyl]-H4 + PAD4
40 50 607080 90 100
40 50 6070 8090100
1 2 3 4
H3 cycle 8
Relative abundancy (%)
H3 cycle 8
H3 cycle 17H3 cycle 17
Fig. 3. Linking PAD4 activity with the regulation of H3 Arg17methylation. (A) PAD4 protein was
expressed in HL-60 granulocytes upon DMSO treatment (lanes 3 and 4). Citrullinated H3 and H4
(denoted by asterisks) were detected in histones purified from cells treated with both DMSO and
calcium ionophore (lane 4). (B) Amounts H3 Arg17methylation and H4 Arg3methylation
decreased in HL-60 granulocytes after calcium ionophore treatment. (C) Before calcium ionophore
treatment, H3 Arg17methylation signals (red) are present at comparable levels in each HL-60
granulocyte. After 15 min of calcium ionophore treatment, methylation of H3 Arg17strongly
decreased in the majority of cells. (D) Protein microsequencing of H3 and citrullinated H3. Cit was
not detected before calcium ionophore treatment. After PAD4 activation, È27.3% of H3 Arg8is
citrullinated (2.52 pmol of Cit versus 6.72 pmol of Arg), and È6.5% of H3 Arg17is citrullinated
(0.21 pmol of Cit versus 3.02 pmol of Arg).
R E P O R T S
www.sciencemag.orgSCIENCEVOL 306 8 OCTOBER 2004
performed microsequencing with H3 isolated
from HL-60 granulocytes. We found that H3
was only citrullinated after treatment with cal-
cium ionophore and identified major PAD4
target sites at Arg8(È27.3% Cit) and Arg17
(È6.5% Cit) (Fig. 3D). Although the H3 Arg2
site was deiminated by PAD4 in vitro, its
deimination was not detectable in HL-60
granulocytes. In addition, although only
È6.5% of H3 Cit17was detected, the majority
of methyl-Arg17signal was lost (Fig. 3B),
suggesting that methyl-Arg17was selectively
targeted by PAD4. Furthermore, the high
percentage of histone Cit8detected after
calcium activation demonstrates that PAD4
can deiminate Arg in vivo.
To investigate whether PAD4 can citrul-
linate H4 at Arg3, we developed a specific
antibody against H4 Cit3("-Cit3H4). West-
ern blot analyses showed that the Cit3H4
antibody strongly recognized H4 after treat-
ment of HL-60 granulocytes with calcium
ionophore (Fig. 4A). This reactivity was spe-
cifically decreased by the Cit3H4(1-8) pep-
tide (Fig. 4A). These data suggest that PAD4
can target H4 Arg3site for citrullination.
To analyze the temporal changes in H4
Arg3methylation and citrullination, we per-
formed Western blot experiments at different
time points after calcium ionophore treat-
ment. A gradual loss of H4 Arg3methylation
was observed (Fig. 4B), which is directly
correlated with a concomitant increase of H4
Cit3(Fig. 4B). The dynamic and complemen-
tary change of H4 Arg3methylation and
citrullination in HL-60 granulocytes suggests
that PAD4 either preferentially targets methyl-
Arg3in vivo or reacts with both H4 methyl-
Arg3and Arg3equally well.
As is the case of H3 Arg17methylation
(Fig. 3), H4 methyl-Arg3antibody staining
was greatly reduced in the majority of cells
(È55.2%, n 0 200) after 15 min of calcium
ionophore treatment (Fig. 4C). By using an
H2A/H4 phospho-Ser1 antibody (29), we
found that this phosphorylation mark was not
decreased after calcium ionophore treatment
(fig. S4), suggesting that the extreme N termi-
nus of H4 is unaltered. In contrast, although
HL-60 granulocytes were not stained with the
Cit3H4 antibody before calcium ionophore
treatment (merged images in Fig. 4C), the ma-
jority of cells (È63.8%, n 0 1178) were
positively stained with the Cit3H4 antibody
after 15 min of calcium ionophore treatment
To address whether the observed de-
crease of H4 Arg3methylation and increase
of H4 Cit3was dependent on PAD4 activity,
we carried out PAD4 RNA interference
experiments in HL-60 cells. As shown in
Fig. 4D, the amount of PAD4 protein dra-
matically decreased after PAD4 small inter-
fering RNA (siRNA) treatment but was not
affected by a control siRNA (Fig. 4D). As
expected, the ability of HL-60 granulocytes
to decrease H4 Arg3methylation and to in-
crease H4 Cit3was lost when PAD4 ex-
pression was inhibited (Fig. 4D). These data
illustrate that PAD4 is the major, if not the
only, enzyme that directly mediates the dy-
Fig. 5. PAD4 and the
regulation of estrogen-
responsive genes (A)
Luciferase activity of
an EREII-LUC reporter
gene transfected into
MCF-7 cells was dra-
matically increased in
response to estradiol
amounts of plasmids
(0.1 to 0.3 6g) express-
ing wild-type PAD4 ef-
ficiently inhibited the
reporter gene activity
in a dose-dependent
manner. In contrast, a
catalytic inactive form
of PAD4 (C645S) dis-
played significantly re-
duced inhibitory effect.
Error bars indicate the
means T SD of three
and the dynamic
change of methylation
and citrullination of H4
Arg3on the pS2 gene
promoter in MCF-7 cells. (C) As controls, PAD4 was not associated with the promoter of CIITA gene
(specific to immune cells). On the ubiquitously expressed GAPDH promoter, background levels of
polymerase chain reaction signals were detected from PAD4 ChIP.
Cit3H4 (1-8) pep: SGCit3GKGGK
Ponceau S staining
Ca2+ 0 5 10 15 30 60 (min.)
PAD4-siRNA - - + +
Control-siRNA + + - -
Ca2+ - + - +
Fig. 4. PAD4 regulates H4 Arg3methylation and citrullination levels in HL-60 cells. (A) An antibody
generated against an H4 Cit3peptide (amino acids 1 to 8 of H4) detects H4 after calcium
ionophore treatment (left). This signal is specifically competed by the H4 Cit3peptide (middle).
Equal loading of samples is shown by Ponceau S staining. (B) A dynamic decrease of H4 Arg3
methylation mirrored by a concomitant increase in H4 Arg3citrullination after calcium ionophore
treatment. (C). After calcium ionophore treatment, H4 Arg3methylation staining in the majority
of cells was dramatically reduced (top). In contrast, a vast majority of cells became positively
stained with the Cit3H4 antibody after calcium ionophore treatment (bottom). (D). PAD4 siRNA
experiments in HL-60 cells. PAD4 protein amounts were dramatically reduced with PAD4 siRNA
treatment (top). Cells treated with PAD4 siRNA had no obvious decrease in H4 Arg3methylation
and little production of H4 Cit3after calcium ionophore treatment (middle). Equal protein loading
was confirmed by Coomassie Blue staining (bottom).
R E P O R T S
8 OCTOBER 2004VOL 306 SCIENCE www.sciencemag.org
namic change of histone H4 Arg3methylation
and citrullination in HL-60 granulocytes.
Histone Arg methylation at H3 Arg17and
H4 Arg3is known to regulate estrogen-
responsive genes, such as the pS2 gene in
MCF-7cells(11, 30). The observed demethyl-
imination activity of PAD4 suggests it might
regulate histone Arg methylation on specific
promoters, leading to a change of gene
expression. To test this idea, we first analyzed
the effect of PAD4 and an enzymatically
inactive form of PAD4 (PAD4C645S) (fig. S3)
on the activity of an EREII-luciferase reporter
gene, which can be strongly induced by $-
estradiol in MCF-7 cells (Fig. 5A). We found
that the wild-type PAD4 effectively repressed
the activity of the luciferase reporter in a
dose-dependent manner (Fig.5A), whereas the
PAD4C645Smutant displayed weaker inhibito-
ry effects. Intriguingly, the PAD4C645Smutant
displays partial repressive activity when pres-
ent at higher doses. Whether the mutant retains
partial enzymatic activity, recruits additional
cofactors, or heterodimerizes with endogenous
PAD4 in MCF7 cells Eas does wild-type PAD4
(19)^ remains unclear.
The repressive activity of PAD4 on the
EREII-luciferase reporter gene prompted us
to test whether PAD4 plays a role in reg-
ulating the endogenous pS2 gene in MCF-7
cells after estradiol stimulation. We found
both PAD4 expression and low amounts of
H4 Cit3in MCF-7 cells (28). With chromatin
immunoprecipitation (ChIP) analyses, we
showed that PAD4 is associated with the
pS2 gene promoter before the addition of
estradiol and that PAD4 amounts increased
Ètwofold at 40 and 60 min after estradiol
induction (Fig. 5B). We observed a strong
increase of H4 Arg3methylation at 20 min
and a decrease at subsequent time points,
whereas H4 Cit3increased at 40 and 60 min.
Therefore, the decrease of H4 Arg3methyla-
tion correlates with the increase of PAD4
protein and H4 Cit3levels on the pS2 gene
promoter. In addition, PAD4 was not associ-
ated with the control CIITA gene and
(GAPDH) gene promoters before or after
estradiol treatment (Fig. 5C). These data sug-
gest that PAD4 acts specifically at the pS2
promoter and that its recruitment does not
simply result from increased PAD4 expression
upon hormone induction. Thus, our data sup-
port the conclusion that the demethylimination
activity of PAD4 is likely involved in the
subtle balance of the estrogen-inducible pS2
gene expression in MCF-7 cells.
Our finding that PAD4 can both de-
that PAD4 may affect chromatin structure
and function via two related but different
mechanisms (fig. S5). Regarding demethyli-
mination, histone Arg methylation mediated
by secondary co-activators, such as CARM1
and PRMT1, has been correlated with gene
activity (11–15) (fig. S5). Given the para-
digm already established by reversible acetyl-
ation (31–33), it seems reasonable that Arg-
directed methylation events, particularly
those that lead to gene activation, would be
reversible. In the case of estrogen-induced
genes in MCF-7 cells, we favor the view that
PAD4 also functions to remove histone Arg
methylation marks, thereby reversing the
transcriptional activation brought about by
nuclear hormone receptor coactivators and
histone arginine methyltransferases, likely in
concert with other chromatin modifying ac-
tivities (e.g., histone deacetylases) (fig. S5).
It remains a formal possibility, however, that
the repressive effect of PAD4 may be due to
its deimination activity, which, in turn, pre-
vents histone methylation by CARM1 and
PRMT1. Because of the dual enzymatic ac-
tivities of PAD4, deimination versus de-
methylimination, separating any observed
transcriptional or other biological effects
brought about by PAD4 at target Arg residues
will represent a challenge for future studies.
References and Notes
1. T. Jenuwein, C. D. Allis, Science 293, 1074 (2001).
2. B. D. Strahl, C. D. Allis, Nature 403, 41 (2000).
3. Y. Zhang, D. Reinberg, Genes Dev. 15, 2343 (2001).
4. T. Kouzarides, Curr. Opin. Genet. Dev. 12, 198
5. M. Lachner, R. J. O’Sullivan, T. Jenuwein, J. Cell Sci.
116, 2117 (2003).
6. A. J. Bannister, R. Schneider, T. Kouzarides, Cell 109,
7. F. M. Boisvert, J. Cote, M. C. Boulanger, S. Richard,
Mol. Cell. Proteomics 2, 1319 (2003).
8. J. D. Gary, S. Clarke, Prog. Nucleic Acid Res. Mol. Biol.
61, 65 (1998).
9. K. A. Mowen et al., Cell 104, 731 (2001).
10. W. Xu et al., Science 294, 2507 (2001); published
online 8 November 2001 (10.1126/science.1065961).
11. U. M. Bauer, S. Daujat, S. J. Nielsen, K. Nightingale,
T. Kouzarides, EMBO Rep. 3, 39 (2002).
12. D. Chen et al., Science 284, 2174 (1999).
13. B. D. Strahl et al., Curr. Biol. 11, 996 (2001).
14. H. Wang et al., Science 293, 853 (2001); published
online 31 May 2001 (10.1126/science.1060781).
15. W. An, J. Kim, R. Roeder, Cell 117, 735 (2004).
16. E. R. Vossenaar, A. J. Zendman, W. J. van Venrooij,
G. J. Pruijn, Bioessays 25, 1106 (2003).
17. K. Nakashima, T. Hagiwara, M. Yamada, J. Biol.
Chem. 277, 49562 (2002).
18. T. Hagiwara, K. Nakashima, H. Hirano, T. Senshu, M.
Yamada, Biochem. Biophys. Res. Commun. 290, 979
19. K. Arita et al., Nat. Struct. Biol. 11, 777 (2004).
20. T. Ogawa, M. Kimoto, K. Sasaoka, J. Biol. Chem. 264,
21. J. Murray-Rust et al., Nat. Struct. Biol. 8, 679 (2001).
22. See more information on the Fugue program at
23. E. I. Georgieva, R. Sendra, Anal. Biochem. 269, 399
24. H. Xie et al., Methods 1, 276 (1990).
25. Materials and methods are available on Science
26. T. L. Branscombe et al., J. Biol. Chem. 276, 32971 (2001).
27. J. Sayegh, S. Clarke, unpublished data.
28. Y. Wang, S. Coonrod, unpublished observations.
29. C. Barber et al., Chromosoma 112, 360 (2004).
30. R. Metivier et al., Cell 115, 751 (2003).
31. S. Y. Roth, J. M. Denu, C. D. Allis, Annu. Rev. Biochem.
70, 81 (2001).
32. C. Tse, T. Sera, A. P. Wolffe, J. C. Hansen, Mol. Cell.
Biol. 18, 4629 (1998).
33. M. Grunstein, Nature 389, 349 (1997).
34. We are grateful to members of the Allis and Coonrod
laboratories, X. Zhang and X. Cheng (Emory Univer-
sity) for insightful discussions and comments, E.
Smith for critical reading of the paper, M. Myers
(Cold Spring Harbor Laboratory) for help on mass
spectrometry analysis and discussions, T. Senshu and
M. Yamada for PAD4 reagents, and F. Campagne for
bioinformatics expertise. Upstate Biotech, Incorpo-
rated participated in Cit3H4 antibody development.
This work was supported by NIH grants GM R01
26020 (S.C.), DK55274 (M.R.S.), GM R01 50659
(C.D.A.), and HD R01 38353 (S.A.C.). J.W. is a fellow
of the Damon Runyon Cancer Research Fund.
Supporting Online Material
Materials and Methods
Figs. S1 to S5
14 June 2004; accepted 25 August 2004
Published online 2 September 2004;
Include this information when citing this paper.
Prebiotic Formation of Peptides
Luke Leman,1Leslie Orgel,2M. Reza Ghadiri1*
Almost all discussions of prebiotic chemistry assume that amino acids, nu-
cleotides, and possibly other monomers were first formed on the Earth or
brought to it in comets and meteorites, and then condensed nonenzymati-
cally to form oligomeric products. However, attempts to demonstrate plau-
sibly prebiotic polymerization reactions have met with limited success. We
show that carbonyl sulfide (COS), a simple volcanic gas, brings about the
formation of peptides from amino acids under mild conditions in aqueous
solution. Depending on the reaction conditions and additives used, exposure
of !-amino acids to COS generates peptides in yields of up to 80% in minutes
to hours at room temperature.
The first suggestion that COS might be a pre-
biotic condensing agent appears in a footnote
of a paper by Hirschmann and co-workers on
peptide synthesis from 2,5-thiazolidinediones
(1). The authors reported that traces of
dipeptide are formed from phenylalanine
R E P O R T S
www.sciencemag.org SCIENCEVOL 306 8 OCTOBER 2004