Molecular Cell, Vol. 18, 83–96, April 1, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.molcel.2005.02.034
PARP-1 Determines Specificity
in a Retinoid Signaling Pathway
via Direct Modulation of Mediator
Rushad Pavri,1Brian Lewis,1Tae-Kyung Kim,1
F. Jeffrey Dilworth,2Hediye Erdjument-Bromage,3
Paul Tempst,3Gilbert de Murcia,4Ronald Evans,5
Pierre Chambon,2and Danny Reinberg1,*
1Department of Biochemistry
Howard Hughes Medical Institute
Division of Nucleic Acids Enzymology
University of Medicine and Dentistry of New Jersey
683 Hoes Lane
Piscataway, New Jersey 08854
2Institut de Genetique et de Biologie Moleculaire
Centre National de la Recherche Scientifique
Institut National de la Santé
et de la Recherche Médicale
67404 Illkirch Strasbourg
3Molecular Biology Program
Memorial Sloan Kettering Cancer Center
New York, New York 10021
4Unité 9003 du Centre National
de la Recherche Scientifique
École Supérieure de Biotechnologie de Strasbourg
Boulevard Sébastien Brant
F-67412 Illkirch Cedex
5Howard Hughes Medical Institute
The Salk Institute
10010 North Torrey Pines Road
La Jolla, California 92037
The nuclear receptor (NR) superfamily consists of a
variety of DNA binding transcription factors that regu-
late diverse and critical physiological processes starting
from early embryogenesis well into adulthood (Kastner et
al., 1995). NRs are generally classified as steroid recep-
tors, such as the estrogen (ER) and glucocorticoid (GR)
receptors, and nonsteroid receptors, such as retinoic
acid (RAR) and thyroid (TR) receptors, and orphan re-
ceptors (Mangelsdorf and Evans, 1995). The steroid re-
ceptors form homodimers, whereas the nonsteroid
class forms heterodimers with the retinoid X receptor
(RXR) and targets promoters by binding specific re-
cognition sequences (Mangelsdorf and Evans, 1995).
The model for receptor-mediated transcription regula-
tion indicates that the receptor in its unliganded state
recruits corepressor complexes (SMRT, NCoR) in asso-
ciation with histone deacetylases (HDACs) to silence
the target gene. Upon binding its cognate ligand, the
receptors undergo a conformational change that dis-
lodges the corepressor complexes, exposing a “dock-
ing” site for coactivators (Glass and Rosenfeld, 2000;
McKenna and O'Malley, 2002).
The essential role of Mediator in NR -dependent tran-
scription (as well as/and in other activators) has been
well documented (Boyer et al., 1999; Fondell et al., 1996;
Gu et al., 1999; Kim et al., 1994; Malik et al., 2000b; Naar
et al., 1999; Rachez et al., 1999; Ryu et al., 1999; Sun et
al., 1998). It is believed that Mediator acts as a “bridge”
between the upstream DNA bound activator and the
basal transcriptional machinery via protein-protein in-
teractions (Hampsey and Reinberg, 1999; Lewis and
Reinberg, 2003; Malik and Roeder, 2000a; Rachez and
Freedman, 2001; Taatjes et al., 2004).
Mediator appears to have a dynamic constitution, ex-
hibiting slightly altered subunit compositions resulting
from the interaction of Mediator with different activa-
tors (Hampsey and Reinberg, 1999; Taatjes et al., 2004).
Consistent with this, structural studies have demon-
strated that Mediator adopts different conformations
upon binding different activators (Taatjes et al., 2002).
Most of the biochemical studies demonstrating Me-
diator requirement for activator-dependent transcrip-
tion involved the use of affinity purifications employing
the activator of interest or affinity pull-down experi-
ments from epitope-tagged cell lines. Scoring for direct
interactions, however, may preclude the identity of
other potential candidates essential for transcriptional
regulation. Moreover, the use of crude fractions in tran-
scription assays limits the ability to comprehensively
identify the contributing factors. Hence, it is currently
not clear whether Mediator (together with TFIID) is suf-
ficient for activator-dependent transcription. We de-
vised an unbiased functional transcription assay con-
sisting of defined purified transcription factors to
identify activities required for RAR/RXR-mediated, li-
gand-dependent transcription. This approach led to the
isolation of a bipartite activity from HeLa cell nuclear
extracts comprised of Mediator and PARP-1.
We show that PARP-1 is indispensable to retinoic acid
receptor (RAR)-mediated transcription from the RAR?2
promoter in a highly purified, reconstituted transcrip-
tion system and that RA-inducible expression of all
RAR? isoforms is abrogated in PARP-1−/−
in vivo. Importantly, PARP-1 activity was independent
of its catalytic domain. PARP-1 directly interacts with
RAR and Mediator. Chromatin immunoprecipitation
experiments confirmed the presence of PARP-1 and
Mediator on RAR-responsive promoters in vivo. Im-
portantly, Mediator was inactive (Cdk8+) under basal
conditions but was activated (Cdk8−) upon induction.
However, in PARP-1−/−cells, Mediator was retained in
its inactive state (Cdk8+) upon induction consistent
with the absence of gene expression. PARP-1 became
dispensable for ligand-dependent transcription in a
chromatin reconstituted transcription assay when
Mediator was devoid of the Cdk8 module (CRSP).
PARP-1 appears to function as a specificity factor
regulating the RA-induced switch of Mediator from
the inactive (Cdk8+) to the active (Cdk8−) state in
In proliferating cells, PARP-1 is an abundant, nonspe-
cific DNA binding protein that catalyzes the transfer of
the ADP-ribose moiety from NAD+to nuclear target
substrates involved in chromatin architecture and in
DNA metabolism when DNA strand breaks are pro-
duced (Ame et al., 1999; D’Amours et al., 1999; Shall
and de Murcia, 2000). This enzyme is highly conserved
and contains an N-terminal zinc finger domain (domain
A) that acts as a molecular sensor for DNA breaks, a
central automodification domain bearing a BRCT motif
(domain D), and a C-terminal catalytic domain involved
in poly-ADP-ribosylation (PAR) (domain F) upon DNA
Although PARP-1 has been best studied for its role
in DNA repair (reviewed in Ame et al., 2004; D’Amours
et al., 1999; Shall and de Murcia, 2000), there have been
several reports demonstrating its role in both activation
and repression of transcription (Ju et al., 2004; Kraus
and Lis, 2003, and references therein). In some cases,
this coactivator function was shown to be independent
of its enzymatic activity (Hassa et al., 2003; Meister-
ernst et al., 1997; Simbulan-Rosenthal et al., 2003; Tulin
et al., 2002).
The studies here demonstrate a functional and physi-
cal interaction between PARP-1 and Mediator that is
essential for RAR-mediated transcriptional activation.
PARP-1 is an essential coregulator for RA-induced
gene expression in vivo and thus a component of
growth and developmental pathways dependent on RA.
Although Mediator is known to exist in inactive and
active states, this study demonstrates that the switch
from inactive to active Mediator is a regulated process
executed by a gene-specific cofactor like PARP-1. This
switch is determinant to the transcriptional status and
constitutes an additional mechanism for gene regu-
transcription factors TFIIA, TFIIB, TFIIE, and TFIIF, and
highly purified TFIID, TFIIH, and RNA polymerase II (Pol
II) from HeLa cells failed to exhibit activation (data not
shown). Therefore, HeLa nuclear extracts were fraction-
ated over several different chromatographic steps (Fig-
ure 1B), and fractions were scored for their ability to
restore ligand-dependent transcriptional activity to the
This procedure resolved two distinct components
necessary to reconstitute ligand-dependent transcrip-
tion. One component eluted from a Superose 6 gel
filtration column with an apparent mass of 1–2 MDa,
whereas the other eluted as an w400 kDa species (Fig-
ure 1B and data not shown). Based on the reported
purification scheme, we surmised that the higher mo-
lecular weight component was likely the Mediator com-
plex. Indeed, Mediator subunits were detected in the
1–2 MDa fractions by Western blot analysis. This Medi-
ator contained Cdk8, Med6, and Trap220, among other
subunits (Figure 1C, data not shown; see below).
To further analyze Mediator requirement, we per-
formed immunodepletion experiments by using anti-
bodies against Cdk8. The Mediator fraction used in the
following and subsequent assays was highly purified
(Figure 1C, silver stain). In one case, the addition of
recombinant Cdk8 (rCdk8) resulted in efficient blocking
of Mediator-antibody interaction such that Mediator
was clearly detectable in the flowthrough fraction (FT/
rCdk8, Figure 1C, Western blot). In the second case,
the addition of equimolar amounts of BSA was ineffec-
tual such that Mediator was undetectable in the
flowthrough fraction (FT/BSA). FT/rCdk8 containing
Mediator was able to confer ligand-dependent tran-
scription, as opposed to the FT/BSA fraction (Figure
1C, bottom). These results confirmed the identity of the
1–2 MDa component as Mediator. Increasing amounts
of the Mediator fraction gave rise to increased tran-
scription in a ligand-dependent manner (Figure 1D).
Interestingly, the addition of Mediator appeared to
suppress the levels of ligand-independent transcription
in this reconstituted system. This was not observed
when crude extracts were used (Figure 1A). However,
crude extracts contain multiple corepressors, such as
NCoR and SMRT, that can effectively suppress ligand-
independent expression in the presence of Mediator.
Reconstitution of Receptor-
and Ligand-Dependent Transcription
Our initial goal was to use chromatinized DNA to mech-
anistically analyze RAR-directed transcription. How-
ever, our attempts at reconstituting ligand-dependent
transcription by using purified transcription factors
were unsuccessful (Figure 1A). Thus, we analyzed the
factors necessary to reconstitute ligand-dependent
transcription from the RARβ2 promoter by using naked
DNA. The p(DR5)5β2G template (Figure 1A, top) con-
sists of the natural promoter of the RA-inducible RARβ2
gene fused to a β-globin sequence (Dilworth et al., 1999,
2000). Five tandem RA response elements (RAREs) are
located upstream of the transcription start site with the
site most proximal positioned similarly as in the natu-
We began by analyzing a crude, extract-based sys-
tem for ligand-dependent transcription. For this and all
future transcription assays, we used recombinant RAR/
RXR heterodimers and all-trans retinoic acid (tRA), the
natural ligand for RAR. HeLa nuclear extracts were able
to efficiently support receptor- and ligand-dependent
transcriptional activation (Figure 1A, bottom). In con-
trast, a reconstituted system consisting of recombinant
PARP-1 Is Necessary
for RA-Dependent Transcription
Fractions derived from the last step of purification of
the 400 kDa factor (Figure 1B) were analyzed for tran-
scription activity (Figure 2A) and by SDS-PAGE fol-
lowed by silver staining (Figure 2B). Selected polypep-
tides (marked by asterisks in Figure 2B) were excised
and analyzed by mass spectrometry. The large, concen-
trated species at 115 kDa was identified as full-length
poly-ADP-ribose polymerase-1 (PARP-1). This was con-
firmed by Western blot analysis (Figure 2C). The other
polypeptides subjected to mass spectrometry were
identified as degradation products of PARP-1.
Recombinant PARP-1 (rPARP-1) efficiently substi-
tuted for native PARP-1 in the reconstituted transcrip-
tion system (Figure 2D), thus confirming that PARP-1
was the active and sole component of the purified ac-
Gene-Specific Modulation of Mediator by PARP-1
Figure 1. Identification of Ligand-Dependent
Activity in Nuclear Extracts
(A) Top: diagram of p(DR5)5β2G template
(described in text). The region of primer hy-
bridization is indicated (arrow), resulting in a
150 nt extension product. Bottom: in vitro
transcription assay with HeLa nuclear ex-
tracts (NE). Recombinant purified RAR/RXR
heterodimers were preincubated with the
template, with or without 1 ?M tRA, prior to
NE addition and primer extension. The arrow
indicates the position of the transcript.
(B) Purification scheme for ligand-dependent
activity (see Supplemental Data).
(C) Immunodepletion of Mediator from 1–2
MDa Superose 6 fraction by using anti-
Cdk8-coupled protein A beads. Beads were
blocked with either rCdk8 or an equimolar
amount of BSA prior to incubation with Su-
perose 6 material. The Western blot shows
complete depletion of Mediator components
from the BSA-blocked flowthrough (FT) as
compared to the rCdk8-blocked FT. Silver
staining revealed that the Superose 6 mater-
ial was a highly purified Mediator fraction. In
the reconstituted transcription assay, the
BSA-blocked FT (lacking Mediator) was un-
able to confer ligand dependence, whereas
the rCdk8-blocked FT (containing Mediator)
efficiently substituted for the 1–2 MDa frac-
tion. Transcription assays were performed as
described (Lewis et al., 2000).
(D) Titration of the Superose 6 Mediator frac-
tion in the reconstituted transcription assay.
The Mediator fraction supported ligand-
dependent transcription in association with
the 400 kDa Superose 6 fraction.
tivity. Importantly, PARP-1 was not functioning here
through nonspecific DNA binding activity. The coactiva-
tor PC4, which also binds to DNA nonspecifically, was
unable to substitute for PARP-1 in the reaction (See
Figure S1 in the Supplemental Data available with this
We next tested for the PARP-1 functional domain(s)
required in the transcription assay. Figure 2E displays
the different domains of PARP-1 and the various con-
structs used in this and subsequentexperiments. Although
the full-length recombinant protein (PARPFL) was able
to restore ligand-dependent transcription, neither the
catalytic domain (PARPEF) nor the ABC fragment con-
taining the DNA binding domain (DBD) (PARPABC) were
able to restore ligand-dependent transcription (Figure
2F). However, a truncated protein with the ABCD do-
mains of PARP-1 (PARPABCD), including both the DBD
and the BRCT motif, was able to efficiently restore li-
gand-dependent transcription (Figure 2G). Therefore,
PARP-1 activity in this system was independent of its
catalytic domain, which resides in the C terminus of the
protein (see Figure 2E).
performed reconstituted transcription assays as above.
On chromatin templates, the addition of PARP-1 and
Mediator was ineffectual with respect to ligand-depen-
dent transcription (Figure 3B, lanes 15 and 16). This ne-
cessitated the further addition of the chromatin remod-
eling factors p300 and Swi/Snf (lanes 5 and 6). Under
these conditions, the requirement for both PARP-1 and
Mediator in ligand-dependent transcription was now
apparent. Thus, the results with both naked and chro-
matinized DNA templates suggested a functional in-
teraction between PARP-1 and Mediator. Importantly,
the chromatin system revealed a coactivator role for
PARP-1 similar to that seen with crude extracts (Figure
1A). As shown below (Figure 4), PARP-1 functions as a
coactivator in vivo, suggesting that the chromatin sys-
tem is truly reflective of the in vivo scenario. The repres-
sive activity of PARP-1 on naked DNA templates may
nevertheless be significant in other contexts.
Similar to the results obtained with naked DNA,
PARPABCD, but not PARPABC, efficiently substituted for
the full-length protein in the chromatin transcription as-
say (Figure 3C), suggesting that even in a physiological
context, the catalytic domain of PARP-1 is not involved
in RAR-dependent transcription. This was not surprising
because our system lacks NAD+, which is essential to
activate PARP-1 catalytic function. Our native PARP-1
PARP-1 Is a Coactivator of RAR-Dependent
Transcription in a Physiological Context
To study the role of PARP-1 in a more physiological
context, we chromatinized the template (Figure 3A) and
Figure 2. Identification of PARP-1 as the Sole Component of the 400 kDa Superose 6 Activity
(A) Reconstituted transcription assay with phenyl sepharose fractions (Figure 1B).
(B) Silver stain of phenyl sepharose fractions. Asterisks (*) indicate bands analyzed by mass spectrometric analysis. The arrow indicates the
position of full-length PARP-1 (115 kDa). Molecular weight markers are shown on the left.
(C) Western blot of phenyl sepharose fractions used in (B) with antibodies to PARP-1. The arrow indicates position of full-length PARP-1 (115
kDa). Molecular weight markers are shown on the left.
(D) Ligand-dependent transcription can be reconstituted with recombinant PARP-1 (rPARP-1). rPARP-1 (10, 20, 50, 100, and 200 ng) efficiently
substituted for native PARP-1 (phenyl sepharose material), confirming that PARP-1 was responsible for ligand dependence in the system.
(E) Diagram of PARP-1. FI and FII indicate zinc fingers I and II, respectively; NLS is the nuclear localization signal, and BRCT is the BRCA1 C
terminus. The location of the active site is shown. Also shown are the various mutants and fragments used in the transcription, transfection,
and IP assays.
(F) PARPABCand PARPEFwere tested for their ability to confer ligand dependence. PARPABCand PARPEF(50, 100, and 200 ng in both cases)
were added to the transcription reaction as shown. In contrast to the full-length protein, neither the ABC nor the catalytic domains were
capable of restoring ligand dependence.
(G) PARPABCand PARPABCDwere tested for ligand-dependence as in (F) above. PARPABCD, but not PARPABC, was able to efficiently restore
ligand dependence, demonstrating that the BRCT domain of PARP-1 was essential for RAR-mediated transcription.
fraction, though catalytically active, was not ADP-ribo-
sylated (Figure S2B and data not shown). In fact, addi-
tion of NAD+at lower concentrations led to a complete
loss of ligand dependence and, at higher concentra-
tions, led to a loss of basal transcription. This is most
likely due to ADP-ribosylation of PARP-1 itself as well
as other known PARP-1 substrates such as TBP, TFIIF,
and/or Pol II (Figure S2A) (Oei et al., 1998). Nonetheless,
our functional experiments confirmed that the catalytic
activity of PARP-1 is not necessary for the cofactor ac-
tivity in RAR-dependent transcription.
PARP-1 Is Essential for RAR-Dependent
Transcription In Vivo
The studies described above uncovered a requirement
for PARP-1 in RAR-mediated transcription in vitro. To
test for PARP-1 requirement in vivo, we scored for RAR-
mediated transcription by using mouse embryonic fi-
broblasts (MEFs) derived from PARP-1 knockout mice
(PARP-1−/−). We used the RARβ2 gene for our in vivo
studies because this promoter was used in all the
in vitro studies mentioned above and because RARβ2
is a well-characterized, immediate-early RA-responsive
Gene-Specific Modulation of Mediator by PARP-1
Figure 3. PARP-1 Functions as a Coactivator
in a Physiological Context
(A) Micrococcal nuclease digestion of chroma-
tin-assembled p(DR5)5β2G. The p(DR5)5β2G
template (Figure 1A) was assembled into
chromatin with S190 Drosophila embryo ex-
tract, and chromatin assembly was analyzed
by micrococcal nuclease digestion.
(B) Transcription assay by using chromatin
templates. S190-assembled chromatin was
used in the reconstituted transcription sys-
tem, and reactions were performed as be-
fore. PARP-1, Mediator, p300, and Swi/Snf
were added where indicated. PARP-1 (native
and recombinant) and Mediator established
ligand dependence on chromatin templates
only when p300 and Swi/Snf were present in
the system (compare lanes 5 and 6 to lanes
15 and 16). This assay thus shows a coacti-
vator role for PARP-1 in RAR-mediated, li-
(C) PARPABCD, but not PARPABCor PARPEF, ef-
ficiently coactivated RAR-mediated ligand-
dependent transcription in a chromatin con-
text, underscoring the requirement of the
BRCT domain for PARP-1 function.
gene (Bouillet et al., 1995). Treatment of PARP+/+MEFs
with RA resulted in the induction of the RARβ2 tran-
script as detected by reverse transcriptase-polymerase
chain reaction (RT-PCR) (Figure 4A, lanes 3 and 4). In
contrast, the PARP-1−/−MEFs showed a drastic de-
crease in RARβ2 expression upon ligand addition (lanes
1 and 2). Because all four RARβ isoforms have been
shown to be RA inducible (Bouillet et al., 1995), we per-
formed RT-PCR with primers amplifying a region com-
mon to all of the isoforms. PARP-1+/+MEFs exhibited
RA-inducible RARβ transcripts, whereas PARP-1−/−
MEFs displayed nearly a complete loss of RA-inducible
expression (Figure 4A). Transfection of constructs en-
coding either the wild-type (wt) or the catalytic mutant
of PARP-1 (PARPFL and PARPcat) rescued ligand-
dependent expression of RARβ (Figure 4B). Such res-
cue was greatly compromised in the case of the DNA
binding mutant of PARP-1 (PARPDBD). As shown below,
Figure 4. RA-Inducible Expression of RARβ Is Impaired in PARP-1−/−MEFs
(A) PARP-1+/+and PARP-1−/−MEFs were treated with tRA for 15 hr, followed by RNA extraction and RT-PCR. The top panel represents an
RARβ2-specific amplicon. Conserved sequences in RARβ, RARα, and RARγ were amplified as shown in subsequent panels. RA-inducible
expression of all RARβ isoforms was abolished in PARP-1−/−MEFs, whereas RARα and RARγ expression profiles were similar in wt and
(B) PARP-1−/−cells were transfected with constructs expressing PARP-1 wt (PARPFL), DNA binding mutant (PARPDBD), and catalytic mutant
(PARPcat) proteins. After transfection for 48 hr, cells were induced with tRA for 15 hr followed by RNA extraction and RT-PCR as before. RA-
inducible expression of RARβ2 was efficiently restored either by wt PARP-1 (PARPFL) or by the catalytically inactive PARP-1 mutant (PARPcat),
but not by the DNA binding mutant (PARPDBD).
(C) PARP-1−/−cells were transfected with constructs expressing PARPABC, PARPABCD, and PARPEF. Experiments were performed as in (B)
above. PARPABCD, but not PARPABCor PARPEF, was able to completely restore RA-dependent expression of RAR.
the DBD was found to bind RAR. Thus the interaction
of RAR and PARP-1 via the DBD is important for
PARP-1 function in RAR-mediated transcription. The
essential role of PARP-1 in ligand-dependent expres-
sion of all RARβ isoforms was independent of its cata-
lytic activity. As shown in Figure 4C, PARPABCD, but not
PARPABCor PARPEF, was able to completely restore
RA-dependent activation in a PARP−/−background.
This finding corroborated our in vitro data on chromatin
templates showing that PARPABCD, but not PARPABC,
was necessary and sufficient for establishing ligand de-
pendence (Figure 2G). This result also confirmed the
essential role of the BRCT domain (domain D) in RAR-
mediated transcription in vivo.
The RAR family consists of three receptor subtypes,
α, β, and γ that exhibit functional redundancy (Kastner
et al., 1995) (see Discussion). We tested if PARP-1 was
required for the expression of RARα and RARγ genes.
RT-PCR analysis, which used primers amplifying con-
served regions, showed that RARα and RARγ were ex-
pressed constitutively in PARP-1+/+and PARP-1−/−
MEFs (Figure 4A). Thus PARP-1 was indispensable for
RAR-dependent transcription of a subset of RAR-regu-
lated genes in vivo.
Trap220 subunit of Mediator (Wada et al., 2004), and, in
agreement, we found that the BRCT domain of PARP-1
(PARPD) efficiently bound to Mediator (Figure 5G). Thus
PARP-1 appeared to be engaged in physical and func-
tional interactions with both RAR and Mediator at
target promoters. This was supported further by the
PARP-1 Localizes to an RAR-Responsive
Promoter In Vivo
We next performed chromatin immunoprecipitation
(ChIP) assays on the mouse RARβ2 gene to test for
PARP-1 recruitment in vivo (Figure 6A, top). We used
the well-studied P19 cells, which are pluripotent embry-
onic carcinomas that can be induced to differentiate
upon RA treatment (Jones-Villeneuve et al., 1982). In
agreement with previous reports (Lefebvre et al., 2002),
RAR was constitutively bound to the promoter (Figure
6A, bottom). Of note, PARP-1 was also constitutively
promoter bound in both uninduced and induced states
(Figure 6A, lane 4). Upon induction, the nuclear receptor
corepressor NCoR was displaced from the promoter
(lane 6), whereas the histone acetyltransferase p300
was recruited (lane 5). Importantly, we did not detect
any PARP-1 signal in the coding region of the gene,
although RNA polymerase II was detected after tRA
treatment (Figure 6B).
Interestingly, Mediator was also detected under unin-
duced and induced states (Figure 6A, lanes 10, 11, and
13). Yet the Mediator complex present in resting cells
contained Cdk8, whereas the complex present upon in-
duction was devoid of Cdk8 (lane 13). Cdk8-containing
Mediator has been shown to be inactive in activator-
dependent transcription (Taatjes et al., 2002; Sun et al.,
1998) and to repress activator-dependent transcription
(Gu et al., 1999; Sun et al., 1998). This repression was
attributed to the Cdk8 module itself (see Discussion). It
is clear from the ChIP data that Cdk8 was lost (or its
epitope masked) upon induction with RA, thus convert-
ing Mediator into its transcriptionally active state.
As reported elsewhere (Lefebvre et al., 2002), RNA
polymerase II (Pol II) was constitutively engaged on the
RARβ2 promoter (Figure 6A, lane 7). We next probed
for the status of the preinitiation complex. In agreement
with the finding that Pol II was associated with pro-
moter sequences, the factors required for its recruit-
ment, i.e., TBP (lane 8) and TFIIB (lane 12), were also
promoter bound. Importantly, TFIIH, which is not nec-
essary for the recruitment of Pol II to promoter se-
quences but is necessary for initiation of transcription
(Kim et al., 2000), was recruited to the promoter upon
RA treatment (lane 9). Thus the preinitiation complex
assembly on this promoter was poised for transcription
dependent upon the recruitment of TFIIH.
To rule out the possibility that PARP-1 was recruited
nonspecifically to inducible promoters, we performed
ChIPs on the promoter of the Brachyury T gene, which
is induced upon treatment of P19 cells with DMSO
(Skerjanc, 1999; Yamaguchi et al., 1999). PARP-1 was
absent from this promoter irrespective of DMSO treat-
ment, and Pol II was recruited to this promoter only
after DMSO induction (Figure 6C). None of the other
non-RAR-regulated promoters analyzed showed the
PARP-1 Physically Interacts with RAR and Mediator
Because our system consists of individually defined
components, we hypothesized that PARP-1 makes di-
rect physical contact with one or more of them. A likely
candidate was RAR. Immunoprecipitation (IP) experi-
ments demonstrated that native purified PARP-1 binds
to FLAG-tagged RAR (f-RAR) (Figure 5A). This was con-
firmed via a reciprocal IP by using anti-PARP-1 anti-
body (Figure 5B). Importantly, the interaction appeared
to be slightly stimulated upon ligand addition. We
mapped the RAR binding domain of PARP-1 to the DBD
(domain A, Figure 5C). Although the AB domain was the
minimal one used, the B domain contains only 30 amino
acids and has been shown to serve as a nuclear local-
ization signal. It is likely that the interaction of PARP-1
with RAR is via the DBD. RAR did not bind to the cata-
lytic or to the BRCT domains of PARP-1 (Figure 5D).
We analyzed whether PARP-1 could interact with
other NRs. As shown in Figure 5E, PARP-1 could bind
to the thyroid receptor (TR), and this interaction was
enhanced in the presence of the TR-specific ligand 3,
3#,5-triiodo-L-thyronine (T3). Thus, PARP-1 interacts di-
rectly with RAR and TR in a ligand-stimulated manner.
Interaction between NRs and Mediator has been pre-
viously documented and shown to be ligand dependent
(Burakov et al., 2000; Yuan et al., 1998). In agreement
with these results we detected such interaction (Figure
5F). However, whereas Mediator-RAR interaction is
strong in the presence of ligand, a small but reproduc-
ible amount of RAR-Mediator interaction was consis-
tently detected in the absence of ligand (see Dis-
As both the DBD and the BRCT domains of PARP-1
were essential for establishing ligand-dependent tran-
scription in vivo and in vitro and RAR was shown to
bind to the PARP-1 DBD, we next investigated the role
of the PARP-1 BRCT domain. Previous studies have
shown that the BRCT domain of BRCA1 binds to the
Gene-Specific Modulation of Mediator by PARP-1
Figure 5. PARP-1 Binds Directly to RAR and TR
(A) Immunoprecipitation (IP) assay for RAR and PARP-1 interaction. Native PARP-1 (phenyl sepharose fraction) and f-RAR were incubated
together in equimolar amounts with anti-FLAG agarose beads in the presence or absence of 1 ?M tRA. The beads were then washed, and
the bound material was eluted and analyzed by SDS-PAGE and Western blot by using antibodies as indicated.
(B) Interaction between PARP-1 and RAR via reciprocal IP. The experiment was performed as in (A) with the exception that anti-FLAG
beads were replaced with anti-PARP-1-coupled protein A beads. The RAR-PARP-1 interaction was independent of, but stimulated upon,
(C) and (D) Mapping of the RAR-PARP-1 interaction via anti-FLAG IPs. Experiments were performed as in (A). RAR bound domain AB (DBD),
ABC, and ABCD, but not domain D or the catalytic domain.
(E) PARP-1 interacts with TR. Equimolar amounts of f-TRα and native PARP-1 were incubated with anti-FLAG beads in the absence or
presence of 0.1 ?M T3, and the IP was performed as in (A). Antibodies to TRα were used for the Western blot. RAR-TR interaction, although
independent of ligand, was markedly stimulated upon ligand addition.
(F) RAR interacts with Mediator. The active Superose 6 Mediator material (1–2 MDa) was incubated with f-RARα in the presence or absence
of 1 ?M tRA, and IP was performed as above.
(G) PARP-1 binds Mediator via the BRCT domain (PARPD). Left: His-PARPDwas mixed with purified Mediator and incubated with Ni-NTA
beads. Bound material was eluted with Imidazole and analyzed as above. Right: reciprocal IP for Mediator-PARPDinteraction by using
antibodies to Cdk8.
presence of PARP-1 (Figure 6D). Thus PARP-1 recruit-
ment to the inducible RARβ2 promoter was specific and
likely mediated by its interactions with RAR and Me-
Given that PARP-1 also interacted with the thyroid
receptor (TR), we analyzed PARP-1 localization at the
promoter of a TR-dependent gene, dio1, in HeLa cells
(Figure 6E). As reported previously, TR was constitu-
tively promoter bound, whereas Mediator was recruited
upon ligand induction (Sharma and Fondell, 2002).
PARP-1 was also localized to the dio1 promoter under
uninduced and induced conditions (Figure 6E), sup-
porting a role for PARP-1 in TR-mediated transcription.
Thus, recruitment of PARP-1 to RAR- and TR-respon-
sive promoters appeared to be ligand independent.
PARP-1 Association at the RAR?2 Promoter Is
Essential for Activation of Mediator
To address the precise role of PARP-1 in RAR-depen-
dent transcription, we compared the occupancy of vari-
ous factors on the RARβ2 promoter in PARP-1+/+versus
PARP-1−/−cells (Figure 6F). The profile of factors occu-
pying the promoter in the PARP-1+/+cells was similar to
that of P19 cells (Figure 6A). Specifically, PARP-1, RAR
and Mediator were constitutively present at the RARβ2
promoter and, upon induction with RA, Mediator was
Figure 6. ChIP Assays
(A) Top: diagram of the endogenous mouse RARβ2 (mRARβ2) promoter. Positions of primers for amplification of promoter and coding regions
are indicated with arrows. TRE is tetradecanoyl phorbol ester-like response element, CRE is cAMP response element, Inr is initiator. Bottom:
P19 cells were treated with tRA for 1 hr followed by ChIP analysis. Input material was incubated with a panel of antibodies as shown. PARP-1 was
constitutively promoter bound as was RAR. The Mediator subunits Med6 and Med10 were also constitutively engaged on the promoter,
whereas the Cdk8 subunit was lost upon ligand addition. The presence of Pol II, TBP, and TFIIB under uninduced and induced states indicated
that the preinitiation complex was assembled at this promoter even in the transcriptionally repressed state. However, TFIIH (ERCC3), which
is required for transcription initiation, was recruited upon RA-induction.
(B) Multiplex PCR by using primers specific to the coding region (exon 3) of the mRARβ2 gene along with the promoter-specific primers used
in (A). The products are of different sizes and can be separated by electrophoresis. PARP-1 was absent in the coding region as was RAR,
whereas Pol II was detected upon RA treatment.
(C) ChIP on the Brachyury T promoter. P19 cells were induced with 1% DMSO for 24 hr followed by ChIP analysis. PARP-1 and RAR were
constitutively absent on this promoter.
(D) ChIP on non-RAR regulated promoters, β-Actin and GAPDH in P19 cells. Neither RAR nor PARP-1 was localized to these promoters,
showing that PARP-1 recruitment to RAR-regulated promoters was specific.
(E) ChIP on the dio1 promoter by using antibodies as indicated. Experiments were performed as above except that the cells were induced
with T3for 1.5 hr. PARP-1 was constitutively localized to this.
(F) ChIP analysis comparing RARβ2 promoter occupancy in PARP-1+/+and PARP-1−/−MEFs. Experiments were performed as in (A) above.
Significantly, chromatin remodeling upon RA induction was likely unaffected in the absence of PARP-1, as evidenced by the similar profiles
of NCoR and p300 occupancy in both PARP-1+/+and PARP-1−/−cells. However, in the absence of PARP-1, Mediator did not attain its active
conformation upon RA induction as evidenced by the retention of Cdk8 after RA treatment. Consequently, TFIIH (ERCC3) was not recruited,
and the gene was silent.
(G) Presence of CRSP bypasses the requirement of PARP-1 for ligand-dependent transcription in a physiological context. Transcriptions were
performed on chromatin templates as described for Figure 3. Highly purified CRSP (see silver stain) was added alone or with PARP-1. CRSP
was able to confer ligand-dependence by itself (lanes 3 and 4), and addition of PARP-1 had no effect on ligand dependence (lanes 5 and 6),
indicating that in the presence of active Mediator (CRSP) PARP-1 was not essential for ligand dependence. In contrast, the inactive Cdk8-
containing Mediator (Med) required PARP-1 for ligand dependence (compare lanes 1 and 2 with lanes 7 and 8).
Gene-Specific Modulation of Mediator by PARP-1
Figure 7. PARP-1 Functions at a Step Prior
to TFIID and Mediator
(A) The order-of-addition experiments were
performed as shown in the scheme. The nor-
mal transcription reaction was slightly modi-
fied, including pre or postincubation steps
(B) and (C) Preincubation of PARP-1 alone
or with TFIID and/or Mediator confers ligand
dependence ([C], compare lanes 3–6 with
lanes 11 and 12). However, preincubation of
TFIID, or Mediator, or both abolished ligand
dependence ([C], compare lanes 7–10 with
lanes 13 and 14). PARP-1 did not confer li-
gand dependence if added after PIC forma-
tion ([B], compare lanes 5 and 6 with lanes
9 and 10). Additionally, the system was TAF
dependent ([B], lanes 9 and 10) and TFIIH
addition after PIC formation did not affect li-
gand dependence ([B], lanes 7 and 8). See
text for details.
converted to the active conformation. This was accom-
panied by the loss of the repressive Cdk8 module (Fig-
ure 6F, top). TFIIH was recruited only upon induction. In
the unstimulated state, PARP-1−/−and PARP-1+/+cells
exhibited similar profiles of promoter occupancy for all
factors tested. However, in PARP-1−/−cells, Mediator
was retained in its inactive state upon RA induction.
TFIIH was not recruited consistent with the lack of tran-
scription activation (Figure 6F, bottom). Of note, there
was no change in the recruitment profiles of NCoR and
p300, indicating that the initial chromatin remodeling
and modification steps required for activation were not
affected in the absence of PARP-1 (Figure 6F, right).
This data indicated that PARP-1 functions at the level
of Mediator by converting its inactive conformation to
an active one, through direct interaction. We specu-
lated that PARP-1 should no longer be necessary for
ligand-dependent transcription in the case of Mediator
lacking the Cdk8 module (CRSP). Indeed, PARP-1 was
now dispensable for CRSP-mediated, ligand-activated
RARβ2 transcription in vitro (Figure 6G).
PARP-1 conferred ligand dependence only when pre-
sent before or during preinitiation complex (PIC) as-
sembly (Figures 7B and 7C). Preincubation of TFIID
(lanes 7 and 8) and/or Mediator (lanes 9 and 10), fol-
lowed by PARP-1 addition, abolished ligand depen-
dence completely. Promoter bound TFIID or Mediator
somehow impinged on the ability of PARP-1 to form a
functionally productive association at the promoter. In
agreement with this, PARP-1 addition post-PIC assem-
bly could not confer ligand dependence (Figure 7B,
lanes 5 and 6). Although PARP-1 may still have associ-
ated with the promoter when TFIID and/or Mediator
were prebound, this association was nonproductive.
These experiments also confirmed the importance of
TAFs in this process because TBP alone could not sub-
stitute for TFIID (Figure 7B, lanes 9 and 10). Consistent
with our previous observations using ChIPs, the addi-
tion of TFIIH after PIC assembly did not affect ligand
dependence (lanes 7 and 8).
By using a highly purified reconstituted transcription
system, we identified PARP-1 as a coregulator of RAR-
mediated gene expression acting in concert with the
Mediator complex. PARP-1 interacts directly with RAR,
and Mediator and ChIP experiments corroborate its
presence at the endogenous promoters of RAR- and
PARP-1 Functions at a Step Prior to the Association
of TFIID and Mediator with Promoter Sequences
Given this role for PARP-1 in RAR gene regulation, we
sought the precise step in the transcription cycle where
PARP-1 functions. For this, we employed order-of-addi-
tion experiments (shown schematically in Figure 7A).
TR-regulated genes. RA-induced expression of all
RARβ isoforms is abolished in PARP−/−cells consistent
with an essential role for PARP-1 in RAR-regulated
gene expression in vivo. Both in vitro and in vivo experi-
ments demonstrate that the coregulatory function of
PARP-1 is independent of its catalytic activity. Finally,
our studies demonstrate that PARP-1 functions as a
gene-specific cofactor by directly modulating Media-
is enhanced upon ligand binding (Figure 5). It is pos-
sible that in the presence of promoter DNA the interac-
tion between Mediator and the receptor is stabilized
even in the absence of ligand. Indeed, recent studies
showed that Mediator was associated with the estro-
gen receptor (ER) in the absence of ligand when pro-
moter DNA was present (Acevedo et al., 2004). Another
study showed that ER can bind Trap220 independently
of the LXXLL motifs and that this interaction was suffi-
cient for coactivation (Wu et al., 2004). This ligand-inde-
pendent association may pertain to Mediator recruit-
ment at silenced promoters, as evidenced by the ChIP
studies (Figure 6).
The ChIP analysis of an RAR-responsive promoter
shows that the repressed promoter is Cdk8 bound
whereas, upon induction, this factor is lost (Figure 6A).
Cdk8, along with cyclin C and TRAP230/240, are be-
lieved to be part of a repressive module within the large
Mediator complex (ARC/NAT) that is displaced upon
activation of transcription (Taatjes et al., 2004). This
was recently proposed to be the case for the promoter
of the C/EBPβ target gene in response to Ras signaling
in vivo (Mo et al., 2004). Cdk8 was also shown to phos-
phorylate the cyclin H subunit of TFIIH (Akoulitchev et
al., 2000). This modification inactivated both the CTD
kinase activity of TFIIH and its ability to initiate tran-
scription. However, some Mediator complexes that
have the Cdk8 module still possess activation func-
tions, such as the TRAP and DRIP complexes (Fondell
et al., 1996; Rachez et al., 1999). It may be that these
Cdk8-containing complexes also lose the Cdk8 module
(or undergo conformational changes) upon activation,
although this has not yet been tested.
PARP-1−/−cells are deficient in RA-inducible gene
expression because Mediator cannot adopt its active
conformation (Figure 6F). It is important to note that the
loss of PARP-1 in these cells does not impinge on the
recruitment of chromatin modifying activities (NCoR,
p300), indicating that chromatin decondensation likely
occurs upon RA induction. PARP-1 apparently func-
tions downstream of chromatin remodeling and modifi-
cation steps. Yet, PARP-1 together with Mediator and
TFIID must be present at the promoter before the re-
cruitment of RNA polymerase II and the general tran-
Because PARP-1 physically interacts with Mediator,
it is likely that this interaction is critical for conversion
of Mediator to its active conformation. Consistent with
this, RA-induced transcription is abrogated in the ab-
sence of this interaction in PARP−/−cells in vivo and by
using PARPABCin vitro on chromatin templates (Figures
3 and 4). Thus, Cdk8-containing Mediator requires an
interaction with PARP-1 to lose the Cdk8 module and
adopt a conformation conducive to transcriptional acti-
vation. Remarkably, PARP-1 is superfluous for ligand
dependence when Mediator lacks the Cdk8 module
(Figure 6G). Thus, PARP-1 serves as a specificity factor
for RAR-dependent transcription by directly regulating
the activity of Mediator.
PARP-1 Is a Classical Coregulator of Transcription
Although PARP-1 has been widely studied for its role in
the DNA damage response and in caspase-indepen-
dent cell death (Ame et al., 2004; Shall and de Murcia,
2000), recent reports have clearly demonstrated a role
for PARP-1 in transcriptional regulation (Kraus and Lis,
2003). Although in some cases PARP-1 enzymatic ac-
tivity is required for the ADP-ribosylation of transcrip-
tion cofactors (Ju et al., 2004; Yu et al., 2004) or his-
tones (Butler and Ordahl, 1999; Tulin and Spradling,
2003), in other cases, such as those involving NFκB
(Hassa et al., 2003), B-MYB (Cervellera and Sala, 2000),
and Tax (Anderson et al., 2000), PARP-1 enzymatic ac-
tivity is not required for regulation of gene expression.
Additionally, in Drosophila, certain PARP-1-regulated
genes require PARP-e, an isoform lacking the catalytic
domain (Tulin et al., 2002). Our work supports that
PARP-1 functions as a classical transcriptional coregu-
lator modulating, in this case, RAR-mediated ligand-
dependent transcription through Mediator. This does
not require the PARP-1 catalytic domain.
The ChIP assays on RAR- and TR-regulated genes
show that PARP-1 is promoter bound constitutively
(Figure 6A). In vitro binding studies reveal that PARP-1
interacts with RAR via the DBD of PARP-1 and with
Mediator via the BRCT domain of PARP-1 (Figure 5).
Consistent with this, we found that both the DBD and
the BRCT domains are essential for PARP-1 function in
vitro and in vivo regardless of the nature of the DNA
(naked or chromatinized). These studies confirm that
the interaction of PARP-1 with RAR and Mediator is
Mediator Is Insufficient in Directing Activator-
The identification of an activity in addition to Mediator
and TAFs (subunits of TFIID) for supporting activator-
mediated transcription on naked and chromatinized
templates is intriguing because it had been hitherto be-
lieved that Mediator with TFIID alone was sufficient
(Baek et al., 2002; Johnson et al., 2002). Instead, our
studies show that in the case of RARβ2 gene regula-
tion, Mediator undergoes a dynamic alteration to opti-
mally engage in gene activation and that this transition
Our results demonstrate the requirement of at least
two different Mediator-like complexes, one associated
with the silent promoter and the other with the acti-
vated promoter. Ligand-dependent interaction between
Mediator and NRs is believed to occur via LXXLL motifs
situated within the Trap220 subunit of Mediator (Ito et
al., 2000; Ren et al., 2000). In the RAR-Mediator IPs,
we consistently observe an interaction of Mediator with
RAR in its unliganded state, although this association
Is PARP-1 Critical to Normal Growth
and Development via Its Essential Role
in Retinoid-Mediated Signaling?
Retinoic acid is essential for normal pre and postnatal
development and, after birth, is indispensable for
Gene-Specific Modulation of Mediator by PARP-1
growth, reproduction, and vision (Kastner et al., 1995).
Extensive mouse knockout studies have confirmed the
importance of the RARs in these processes. Interest-
ingly, knockout of only one particular isoform of a given
RAR subtype in mice did not give rise to any abnormali-
ties, and these mice appeared normal in every respect
(Kastner et al., 1995; Lohnes et al., 1993; Mendelsohn
et al., 1994b). However, RAR double null mutant mice
exhibited in utero or perinatal lethality associated with
a wide array of congenital malformations. This under-
scored the functional redundancy amongst RARs
(Lohnes et al., 1994; Mendelsohn et al., 1994a).
Embryonic cells lacking PARP-1 exhibit a complete
loss of RARβ induction upon RA treatment (Figure 4A).
This directly links PARP-1 to normal growth and devel-
opment via the retinoid signaling pathways. Yet PARP-
1−/−mice do not exhibit developmental abnormalities
(de Murcia et al., 1997; Wang et al., 1995). The func-
tional redundancy amongst RARs provides a cogent
explanation for this. The loss of RARβ in PARP-1−/−ani-
mals would be compensated for by the other RARs as
the above-mentioned mouse studies showed. There-
fore, the phenotype of PARP-1/RARα or PARP-1/RARγ
double null-mutant mice may be more revealing. Our
studies predict that such animals would show the de-
creased viability and congenital developmental defects
seen in the RAR double-knockout mice.
PARP-1 Functions as a Gene-Specific
Coregulator of Mediator
Our model presented in Figure 8 takes the following
findings into consideration. (1) PARP-1 functions in as-
sociation with Mediator. (2) PARP-1 binds Mediator via
the BRCT domain. (3) The BRCT domain is essential for
PARP-1 function. (4) In the absence of this interaction,
i.e., by using PARPABCor in a PARP-1−/−background,
RA-inducible gene expression is abolished. (5) The loss
of this interaction is manifested by the inability of Medi-
ator to switch to its transcriptionally active confor-
The steps shown in the model for RARβ gene activa-
tion are as follows. (1) In the inactive state, the pro-
moter is bound and silenced by corepressor complexes
(NCoR, SMRT, HDACs). RAR/RXR is bound to the RARE
in association with PARP-1. The preinitiation complex
and Pol II are also assembled at the promoter. Mediator
is also present, associated most likely via interactions
with Pol II and/or RAR. (2) Upon ligand binding, RAR
undergoes a conformational change, resulting in the
dissociation of corepressor complexes and the recruit-
ment of coactivator complexes (p300, Swi/Snf) that re-
model and decondense the chromatin at the promoter.
(3) The Trap220 subunit of Mediator now associates
with RAR (via LXXLL motifs), and PARP-1 makes an ad-
ditional contact with Mediator. (4) This association be-
tween RAR, PARP-1, and Mediator triggers the release
of the Cdk8 module such that Mediator can now adopt
its active conformation. (5) Change in Mediator confor-
mation induces a change in PIC/Pol II, resulting in re-
cruitment of TFIIH and initiation of transcription.
The PARP-Mediator-RAR association is the critical
event that determines whether the gene will be tran-
scribed or not. In the absence of PARP-1, the RAR-
Mediator association is unable to switch Mediator to its
Figure 8. Mechanistic Model for the Role of PARP-1 in RAR-Medi-
The model depicts the role of PARP-1 as a gene-specific coactiva-
tor responsible for the ligand-dependent activation of Mediator in
RAR-dependent transcription. The model has been simplified for
clarity. See text for a detailed description.
active form, and gene activation is abrogated. In this
mechanism of gene regulation, modulation of Mediator
by a gene-specific factor is determinant to activation.
It was hitherto unknown just how Mediator conforma-
tion is switched between inactive and active states and
how such a switch is regulated. Our work shows that
gene-specific cofactors (like PARP-1) are key to this
switch, determining the on/off status of inducible
genes. It is likely that other factors regulate gene ex-
pression in a similar manner. Indeed, other genes regu-
lated by PARP-1 independently of ADP-ribosylation
may be subject to a similar mechanism.
Templates and Constructs
The p(DR5)5β2G template has been described elsewhere (Dilworth
et al., 1999). PARP-1 expression vectors have been previously de-
scribed (Molinete et al., 1993; Schreiber et al., 2002).
Cell Culture and Transfections
P19 cells were maintained in MEMα medium (Invitrogen) with 10%
FBS and antibiotics. PARP-1+/+and PARP-1−/−MEFs were main-
tained in Dulbecco’s modified Eagle’s medium with 10% FBS and
antibiotics. Transfections were performed by using Lipofectamine
2000 (Invitrogen) according to the manufacturer’s protocols.
Akoulitchev, S., Chuikov, S., and Reinberg, D. (2000). TFIIH is nega-
tively regulated by cdk8-containing mediator complexes. Nature
Ame, J.C., Spenlehauer, C., and de Murcia, G. (2004). The PARP
superfamily. Bioessays 26, 882–893.
Anderson, M.G., Scoggin, K.E., Simbulan-Rosenthal, C.M., and
Steadman, J.A. (2000). Identification of poly(ADP-ribose) polymer-
ase as a transcriptional coactivator of the human T-cell leukemia
virus type 1 Tax protein. J. Virol. 74, 2169–2177.
Baek, H.J., Malik, S., Qin, J., and Roeder, R.G. (2002). Requirement
of TRAP/mediator for both activator-independent and activator-
dependent transcription in conjunction with TFIID-associated TAF(II)s.
Mol. Cell. Biol. 22, 2842–2852.
Bouillet, P., Oulad-Abdelghani, M., Vicaire, S., Garnier, J.M., Schuh-
baur, B., Dolle, P., and Chambon, P. (1995). Efficient cloning of
cDNAs of retinoic acid-responsive genes in P19 embryonal carci-
noma cells and characterization of a novel mouse gene, Stra1
(mouse LERK-2/Eplg2). Dev. Biol. 170, 420–433.
Boyer, T.G., Martin, M.E., Lees, E., Ricciardi, R.P., and Berk, A.J.
(1999). Mammalian Srb/Mediator complex is targeted by adenovi-
rus E1A protein. Nature 399, 276–279.
Burakov, D., Wong, C.W., Rachez, C., Cheskis, B.J., and Freedman,
L.P. (2000). Functional interactions between the estrogen receptor
and DRIP205, a subunit of the heteromeric DRIP coactivator com-
plex. J. Biol. Chem. 275, 20928–20934.
Butler, A.J., and Ordahl, C.P. (1999). Poly(ADP-ribose) polymerase
binds with transcription enhancer factor 1 to MCAT1 elements to
regulate muscle-specific transcription. Mol. Cell. Biol. 19, 296–306.
Cervellera, M.N., and Sala, A. (2000). Poly(ADP-ribose) polymerase
is a B-MYB coactivator. J. Biol. Chem. 275, 10692–10696.
D’Amours, D., Desnoyers, S., D’Silva, I., and Poirier, G.G. (1999).
Poly(ADP-ribosyl)ation reactions in the regulation of nuclear func-
tions. Biochem. J. 342, 249–268.
de Murcia, J.M., Niedergang, C., Trucco, C., Ricoul, M., Dutrillaux,
B., Mark, M., Oliver, F.J., Masson, M., Dierich, A., LeMeur, M., et
al. (1997). Requirement of poly(ADP-ribose) polymerase in recovery
from DNA damage in mice and in cells. Proc. Natl. Acad. Sci. USA
Dignam, J.D., Lebovitz, R.M., and Roeder, R.G. (1983). Accurate
transcription initiation by RNA polymerase II in a soluble extract
from isolated mammalian nuclei. Nucleic Acids Res. 11, 1475–1489.
Dilworth, F.J., Fromental-Ramain, C., Remboutsika, E., Benecke, A.,
and Chambon, P. (1999). Ligand-dependent activation of transcrip-
tion in vitro by retinoic acid receptor alpha/retinoid X receptor al-
pha heterodimers that mimics transactivation by retinoids in vivo.
Proc. Natl. Acad. Sci. USA 96, 1995–2000.
Dilworth, F.J., Fromental-Ramain, C., Yamamoto, K., and Chambon,
P. (2000). ATP-driven chromatin remodeling activity and histone
acetyltransferases act sequentially during transactivation by RAR/
RXR In vitro. Mol. Cell 6, 1049–1058.
Fondell, J.D., Ge, H., and Roeder, R.G. (1996). Ligand induction of
a transcriptionally active thyroid hormone receptor coactivator
complex. Proc. Natl. Acad. Sci. USA 93, 8329–8333.
Giner, H.A.C., Simonin, F., de Murcia, G., and Menissier-de Murcia,
J. (1992). Overproduction and large-scale purification of the human
poly(ADP-ribose) polymerase using a baculovirus expression sys-
tem. Gene 114, 279–283.
Glass, C.K., and Rosenfeld, M.G. (2000). The coregulator exchange
in transcriptional functions of nuclear receptors. Genes Dev. 14,
Gu, W., Malik, S., Ito, M., Yuan, C.X., Fondell, J.D., Zhang, X., Marti-
nez, E., Qin, J., and Roeder, R.G. (1999). A novel human SRB/MED-
containing cofactor complex, SMCC, involved in transcription regu-
lation. Mol. Cell 3, 97–108.
Hampsey, M., and Reinberg, D. (1999). RNA polymerase II as a con-
trol panel for multiple coactivator complexes. Curr. Opin. Genet.
Dev. 9, 132–139.
Hassa, P.O., Buerki, C., Lombardi, C., Imhof, R., and Hottiger, M.O.
(2003). Transcriptional coactivation of nuclear factor-kappaB-
Antibodies and Primers
All antibodies and primers are listed in the Supplemental Data.
Expression of Recombinant Proteins
RAR/RXR heterodimers were purified as described (Dilworth et al.,
2000). FLAG-tagged proteins (RAR, TR, p300) were purified by anti-
FLAG M2 agarose affinity resin (Sigma), and His-tagged proteins
(RXR, PARPD, Cdk8) were purified via a Ni2+affinity column (Qia-
gen) per the manufacturer’s instructions. Recombinant PARPFL,
PARPAB, PARPABC, and PARPEFproteins were purified as described
(Giner et al., 1992; Molinete et al., 1993). PARPABCDwas expressed
in E. coli.
HeLa nuclear extracts were prepared as described (Dignam et al.,
1983). The purification scheme is shown in Figure 1B. See the Sup-
plemental Data for details.
GTFs and Pol II were purified as described (Maldonado et al.,
1996). Transcription assays and primer extension analyses were
performed as described (Lewis et al., 2000), and transcripts were
visualized by autoradiography or phosphorimager analysis (Molec-
Chromatin was assembled and purified as described (Orphanides
et al., 1998).
IP and ChIP Assays
IPs were performed by standard methods (see Supplemental Data).
ChIPs were performed as described (Vaquero et al., 2004).
RNA Extraction and RT-PCR
See the Supplemental Data.
Supplemental Data including two figures and Materials and Meth-
ods are available online with this article at http://www.molecule.
We thank Dr. Dylan Taatjes for generously providing purified CRSP
complex. We thank Dr. Joe Fondell for f-TRα baculovirus and the
HeLa α2 cell line expressing f-TRα. We thank Drs. Josiane Menis-
sier de Murcia, Lynne Vales, Alejandro Vaquero, and Bing Zhu for
helpful comments and suggestions. This work was supported by a
grant from the National Institutes of Health (GM37120) and the
Howard Hughes Medical Institute to D.R. Work performed at
IGBMC was supported by a grant from the Centre National de la
Recherche Scientifique (CNRS), INSERM, College de France, and
Association pour la Recherche sur le Cancer. G.d.M.’s laboratory is
supported by CNRS, Association pour la Recherche sur le Cancer,
La Ligue contre le Cancer, and Commissariat à l’Energie Atomique.
Received: November 13, 2004
Revised: January 26, 2005
Accepted: February 28, 2005
Published: March 31, 2005
Acevedo, M.L., Lee, K.C., Stender, J.D., Katzenellenbogen, B.S.,
and Kraus, W.L. (2004). Selective recognition of distinct classes of
coactivators by a ligand-inducible activation domain. Mol. Cell 13,
Gene-Specific Modulation of Mediator by PARP-1
dependent gene expression by p300 is regulated by poly(ADP)-
ribose polymerase-1. J. Biol. Chem. 278, 45145–45153.
Ito, M., Yuan, C.X., Okano, H.J., Darnell, R.B., and Roeder, R.G.
(2000). Involvement of the TRAP220 component of the TRAP/
SMCC coactivator complex in embryonic development and thyroid
hormone action. Mol. Cell 5, 683–693.
Johnson, K.M., Wang, J., Smallwood, A., Arayata, C., and Carey, M.
(2002). TFIID and human mediator coactivator complexes assem-
ble cooperatively on promoter DNA. Genes Dev. 16, 1852–1863.
Jones-Villeneuve, E.M., McBurney, M.W., Rogers, K.A., and Kal-
nins, V.I. (1982). Retinoic acid induces embryonal carcinoma cells
to differentiate into neurons and glial cells. J. Cell Biol. 94, 253–262.
Ju, B.G., Solum, D., Song, E.J., Lee, K.J., Rose, D.W., Glass, C.K.,
and Rosenfeld, M.G. (2004). Activating the PARP-1 sensor compo-
nent of the Groucho/ TLE1 corepressor complex mediates a CaM-
Kinase IIdelta-dependent neurogenic gene activation pathway. Cell
Kastner, P., Mark, M., and Chambon, P. (1995). Nonsteroid nuclear
receptors: what are genetic studies telling us about their role in real
life? Cell 83, 859–869.
Kim, T.K., Ebright, R.H., and Reinberg, D. (2000). Mechanism of
ATP-dependent promoter melting by transcription factor IIH. Sci-
ence 288, 1418–1422.
Kim, T.W., Kwon, Y.J., Kim, J.M., Song, Y.H., Kim, S.N., and Kim,
Y.J. (2004). MED16 and MED23 of Mediator are coactivators of lipo-
polysaccharide- and heat-shock-induced transcriptional activa-
tors. Proc. Natl. Acad. Sci. USA 101, 12153–12158.
Kim, Y.J., Bjorklund, S., Li, Y., Sayre, M.H., and Kornberg, R.D.
(1994). A multiprotein mediator of transcriptional activation and its
interaction with the C-terminal repeat domain of RNA polymerase
II. Cell 77, 599–608.
Kraus, W.L., and Lis, J.T. (2003). PARP goes transcription. Cell 113,
Lefebvre, B., Brand, C., Lefebvre, P., and Ozato, K. (2002). Chromo-
somal integration of retinoic acid response elements prevents co-
operative transcriptional activation by retinoic acid receptor and
retinoid X receptor. Mol. Cell. Biol. 22, 1446–1459.
Lewis, B.A., and Reinberg, D. (2003). The mediator coactivator
complex: functional and physical roles in transcriptional regulation.
J. Cell Sci. 116, 3667–3675.
Lewis, B.A., Kim, T.K., and Orkin, S.H. (2000). A downstream ele-
ment in the human beta-globin promoter: evidence of extended
sequence-specific transcription factor IID contacts. Proc. Natl.
Acad. Sci. USA 97, 7172–7177.
Lohnes, D., Kastner, P., Dierich, A., Mark, M., LeMeur, M., and
Chambon, P. (1993). Function of retinoic acid receptor gamma in
the mouse. Cell 73, 643–658.
Lohnes, D., Mark, M., Mendelsohn, C., Dolle, P., Dierich, A., Gorry,
P., Gansmuller, A., and Chambon, P. (1994). Function of the retinoic
acid receptors (RARs) during development (I). Craniofacial and
skeletal abnormalities in RAR double mutants. Development 120,
Maldonado, E., Drapkin, R., and Reinberg, D. (1996). Purification of
human RNA polymerase II and general transcription factors. Meth-
ods Enzymol. 274, 72–100.
Malik, S., and Roeder, R.G. (2000a). Transcriptional regulation
through Mediator-like coactivators in yeast and metazoan cells.
Trends Biochem. Sci. 25, 277–283.
Malik, S., Gu, W., Wu, W., Qin, J., and Roeder, R.G. (2000b). The
USA-derived transcriptional coactivator PC2 is a submodule of
TRAP/SMCC and acts synergistically with other PCs. Mol. Cell 5,
Mangelsdorf, D.J., and Evans, R.M. (1995). The RXR heterodimers
and orphan receptors. Cell 83, 841–850.
McKenna, N.J., and O’Malley, B.W. (2002). Combinatorial control of
gene expression by nuclear receptors and coregulators. Cell 108,
Meisterernst, M., Stelzer, G., and Roeder, R.G. (1997). Poly(ADP-
ribose) polymerase enhances activator-dependent transcription
in vitro. Proc. Natl. Acad. Sci. USA 94, 2261–2265.
Mendelsohn, C., Lohnes, D., Decimo, D., Lufkin, T., LeMeur, M.,
Chambon, P., and Mark, M. (1994a). Function of the retinoic acid
receptors (RARs) during development (II). Multiple abnormalities at
various stages of organogenesis in RAR double mutants. Develop-
ment 120, 2749–2771.
Mendelsohn, C., Mark, M., Dolle, P., Dierich, A., Gaub, M.P., Krust,
A., Lampron, C., and Chambon, P. (1994b). Retinoic acid receptor
beta 2 (RAR beta 2) null mutant mice appear normal. Dev. Biol. 166,
Mo, X., Kowenz-Leutz, E., Xu, H., and Leutz, A. (2004). Ras induces
mediator complex exchange on C/EBP beta. Mol. Cell 13, 241–250.
Molinete, M., Vermeulen, W., Burkle, A., Menissier-de Murcia, J.,
Kupper, J.H., Hoeijmakers, J.H., and de Murcia, G. (1993). Overpro-
duction of the poly(ADP-ribose) polymerase DNA-binding domain
blocks alkylation-induced DNA repair synthesis in mammalian
cells. EMBO J. 12, 2109–2117.
Naar, A.M., Beaurang, P.A., Zhou, S., Abraham, S., Solomon, W.,
and Tjian, R. (1999). Composite co-activator ARC mediates chro-
matin-directed transcriptional activation. Nature 398, 828–832.
Oei, S.L., Griesenbeck, J., Schweiger, M., and Ziegler, M. (1998).
Regulation of RNA polymerase II-dependent transcription by poly-
(ADP-ribosyl)ation of transcription factors. J. Biol. Chem. 273,
Orphanides, G., LeRoy, G., Chang, C.H., Luse, D.S., and Reinberg,
D. (1998). FACT, a factor that facilitates transcript elongation
through nucleosomes. Cell 92, 105–116.
Rachez, C., and Freedman, L.P. (2001). Mediator complexes and
transcription. Curr. Opin. Cell Biol. 13, 274–280.
Rachez, C., Lemon, B.D., Suldan, Z., Bromleigh, V., Gamble, M.,
Naar, A.M., Erdjument-Bromage, H., Tempst, P., and Freedman, L.P.
(1999). Ligand-dependent transcription activation by nuclear re-
ceptors requires the DRIP complex. Nature 398, 824–828.
Ren, Y., Behre, E., Ren, Z., Zhang, J., Wang, Q., and Fondell, J.D.
(2000). Specific structural motifs determine TRAP220 interactions
with nuclear hormone receptors. Mol. Cell. Biol. 20, 5433–5446.
Ryu, S., Zhou, S., Ladurner, A.G., and Tjian, R. (1999). The tran-
scriptional cofactor complex CRSP is required for activity of the
enhancer-binding protein Sp1. Nature 397, 446–450.
Schreiber, V., Ame, J.C., Dolle, P., Schultz, I., Rinaldi, B., Fraulob,
V., Menissier-de Murcia, J., and de Murcia, G. (2002). Poly(ADP-
ribose) polymerase-2 (PARP-2) is required for efficient base exci-
sion DNA repair in association with PARP-1 and XRCC1. J. Biol.
Chem. 277, 23028–23036.
Shall, S., and de Murcia, G. (2000). Poly(ADP-ribose) polymerase-1:
what have we learned from the deficient mouse model? Mutat. Res.
Sharma, D., and Fondell, J.D. (2002). Ordered recruitment of his-
tone acetyltransferases and the TRAP/Mediator complex to thyroid
hormone-responsive promoters in vivo. Proc. Natl. Acad. Sci. USA
Simbulan-Rosenthal, C.M., Rosenthal, D.S., Luo, R., Samara, R.,
Espinoza, L.A., Hassa, P.O., Hottiger, M.O., and Smulson, M.E.
(2003). PARP-1 binds E2F–1 independently of its DNA binding and
catalytic domains, and acts as a novel coactivator of E2F–1-medi-
ated transcription during re-entry of quiescent cells into S phase.
Oncogene 22, 8460–8471.
Skerjanc, I.S. (1999). Cardiac and skeletal muscle development in
P19 embryonal carcinoma cells. Trends Cardiovasc. Med. 9, 139–
Sun, X., Zhang, Y., Cho, H., Rickert, P., Lees, E., Lane, W., and Rein-
berg, D. (1998). NAT, a human complex containing Srb polypeptides
that functions as a negative regulator of activated transcription.
Mol. Cell 2, 213–222.
Taatjes, D.J., and Tjian, R. (2004). Structure and function of CRSP/
Med2; a promoter-selective transcriptional coactivator complex.
Mol. Cell 14, 675–683.
Taatjes, D.J., Naar, A.M., Andel, F., 3rd, Nogales, E., and Tjian, R.
Molecular Cell Download full-text
(2002). Structure, function, and activator-induced conformations of
the CRSP coactivator. Science 295, 1058–1062.
Taatjes, D.J., Marr, M.T., and Tjian, R. (2004). Regulatory diversity
among metazoan co-activator complexes. Nat. Rev. Mol. Cell Biol.
Tulin, A., and Spradling, A. (2003). Chromatin loosening by poly-
(ADP)-ribose polymerase (PARP) at Drosophila puff loci. Science
Tulin, A., Stewart, D., and Spradling, A. (2002). The Drosophila het-
erochromatic gene encoding poly(ADP-ribose) polymerase (PARP)
is required to modulate chromatin structure during development.
Genes Dev. 16, 2108–2119.
Vaquero, A., Scher, M., Lee, D., Erdjument-Bromage, H., Tempst, P.,
and Reinberg, D. (2004). Human SirT1 Interacts with Histone H1
and Promotes Formation of Facultative Heterochromatin. Mol. Cell
Wada, O., Oishi, H., Takada, I., Yanagisawa, J., Yano, T., and Kato,
S. (2004). BRCA1 function mediates a TRAP/DRIP complex through
direct interaction with TRAP220. Oncogene 23, 6000–6005.
Wang, Z.Q., Auer, B., Stingl, L., Berghammer, H., Haidacher, D.,
Schweiger, M., and Wagner, E.F. (1995). Mice lacking ADPRT and
poly(ADP-ribosyl)ation develop normally but are susceptible to skin
disease. Genes Dev. 9, 509–520.
Wu, Q., Burghardt, R., and Safe, S. (2004). Vitamin D-interacting
protein 205 (DRIP205) coactivation of estrogen receptor alpha (ER-
alpha) involves multiple domains of both proteins. J. Biol. Chem.
Yamaguchi, H., Tanaka, K., Kitagawa, Y., and Miki, K. (1999). A
PEA3 site flanked by SP1, SP4, and GATA sites positively regulates
the differentiation-dependent expression of Brachyury in embryo-
nal carcinoma P19 cells. Biochem. Biophys. Res. Commun. 254,
Yu, W., Ginjala, V., Pant, V., Chernukhin, I., Whitehead, J., Docquier,
F., Farrar, D., Tavoosidana, G., Mukhopadhyay, R., Kanduri, C., et
al. (2004). Poly(ADP-ribosyl)ation regulates CTCF-dependent chro-
matin insulation. Nat. Genet. 36, 1105–1110.
Yuan, C.X., Ito, M., Fondell, J.D., Fu, Z.Y., and Roeder, R.G. (1998).
The TRAP220 component of a thyroid hormone receptor-associ-
ated protein (TRAP) coactivator complex interacts directly with
nuclear receptors in a ligand-dependent fashion. Proc. Natl. Acad.
Sci. USA 95, 7939–7944.