Molecular Cell, Vol. 18, 399–402, May 13, 2005, Copyright ©2005 by Elsevier Inc. DOI 10.1016/j.molcel.2005.04.017
Activation: Right on Target
Michael R. Green*
Howard Hughes Medical Institute
Program in Gene Function and Expression and
Program in Molecular Medicine
University of Massachusetts Medical School
Worcester, Massachusetts 01605
or reinitiation) or, as discussed below, by recruiting chro-
matin-modifying activities that facilitate transcription.
Activator-mediated stimulation of PIC assembly re-
sults, at least in part, from a direct interaction between
the transcriptional AD and one or more components of
the transcription machinery, termed the “target” (Figure
1A). Unresolved questions include, which transcription
components are the direct targets of activators? How
many transcription components function as targets?
What is the specificity of activator-target interactions?
Are activator-target interactions essential or redundant?
Based primarily upon in vitro protein-protein interac-
tion experiments, often involving an isolated compo-
nent of a multisubunit complex, a wide variety of
factors have been proposed to be the direct targets of
ADs. Consistent with this notion, many components
can stimulate transcription after fusion to a DBD in
artificial recruitment experiments (Ptashne and Gann,
1997). What has been elusive is the convincing demon-
stration that a protein able to interact with a transcrip-
tional AD is an authentic, functional target in vitro and,
more importantly, in vivo. Below I discuss a selection of
relatively recent studies in which targets of eukaryotic
transcriptional ADs have been clearly identified.
The Tra1 Subunit of the SAGA Complex Is the Target
of Multiple Yeast Activators
In a recent issue of Molecular Cell, Fishburn et al. (2005)
have devised an elegant protein crosslinking procedure
to identify factors that interact with a transcriptional AD
in a functional PIC (Figure 1B). The transcriptional AD
under investigation was that of the well-characterized
yeast activator Gcn4. A dithiothreitol (DTT)-cleavable
crosslinking reagent was125I-labeled and attached to
the Gcn4 AD through a cysteine residue. The Gcn4 de-
rivative was then used to direct assembly of a func-
tional PIC on a promoter containing a single Gcn4 bind-
ing site. The promoter DNA was attached to a magnetic
bead, enabling the facile purification of the PIC. After
induction of crosslinking by irradiation with ultraviolet
light, the complex was treated with DTT to transfer the
125I-label to a Gcn4 AD-interacting protein, and125I-
tagged polypeptides were detected by SDS-PAGE and
Using this approach, Fishburn et al. (2005) identified
three crosslinked polypeptides: Tra1, a component of
two yeast multisubunit complexes, SAGA and NuA4;
Gal11, a component of the multisubunit Mediator com-
plex; and TAF12, a component of both SAGA and TFIID.
Based upon analysis of extracts from mutant yeast
strains and Gcn4 AD derivatives, the authors deduced
that only one of the three crosslinked polypeptides, the
Tra1 present in the SAGA complex, contributed sub-
stantially to Gcn4-mediated transcription activation.
Consistent with these results, a previous study in which
protein-protein interactions were monitored by cross-
linking and affinity chromatography found that Gal4,
Gcn4, and several other yeast activators could interact
with Tra1 in vitro (Brown et al., 2001).
The results of Fishburn et al. (2005) fit remarkably
The unambiguous identification of the direct targets
of eukaryotic transcriptional activators has been a
major challenge in the field. Recently, the authentic
targets of several yeast and mammalian activators
have been determined, and the results of these
studies have important implications for our under-
standing of transcriptional activation mechanisms.
Transcription of eukaryotic structural genes is regu-
lated by promoter-specific activator proteins (activa-
tors), which are generally sequence-specific DNA bind-
ing proteins. In addition to a DNA binding domain
(DBD), a typical activator also contains a separable ac-
tivation domain (AD) that is required for the activator to
stimulate transcription (Ptashne and Gann, 1997).
Activators function through an elaborate set of gene-
ral (or basic) transcription factors (GTFs) that are nec-
essary and can be sufficient for accurate transcription
initiation. These GTFs include RNA polymerase II and a
variety of auxiliary components including TFIIA, TFIIB,
TFIID, TFIIE, TFIIF, and TFIIH. The GTFs function by as-
sembling on the core promoter to form a preinitiation
complex (PIC) that directs RNA polymerase II to the
transcription start site. A third group of factors, termed
“coactivators,” are operationally defined as compo-
nents required for activator-directed (“activated”) tran-
scription but dispensable for activator-independent
(“basal”) transcription (Roeder, 2005).
Several lines of evidence indicate that an important
aspect of activator function is to increase PIC assem-
bly. For example, assays that directly measure associa-
tion of GTFs with the promoter have shown that activa-
tors stimulate PIC assembly both in vitro (Lin and
Green, 1991) and in vivo (Kuras and Struhl, 1999; Li et
al., 1999). This conclusion is supported by the results
of so-called artificial recruitment experiments in which
a subunit of the transcription machinery is fused to a
DBD, and the resultant fusion protein is tested for its
ability to activate transcription. In many cases, such fu-
sion proteins artificially direct the assembly of a PIC
complex, resulting in transcription. Transcription acti-
vation by these artificial fusion proteins is thought to
mimic the mechanism of action of a natural activator
(Ptashne and Gann, 1997). Activators can also function
by mechanisms other than stimulating PIC assembly,
such as promoting a step in the transcription process
subsequent to PIC assembly (e.g., initiation, elongation,
Figure 1. Activator-Mediated Stimulation of Preinitiation Complex Assembly
(A) General model.
(B) Experimental strategy of Fishburn et al. (2005) for identifying the target of the Gcn4 activation domain (AD).
well with studies on another well-characterized activa-
tor, the yeast Gal4 protein. Using fluorescence reso-
nance energy transfer (FRET) as an assay to monitor
protein-protein interactions in living yeast cells, Tra1
was found to interact with the Gal4 AD (Bhaumik et al.,
2004). These FRET results, coupled with a series of ex-
periments that analyzed PIC assembly and transcrip-
tion (see below), were used to demonstrate that the
Tra1 subunit of the SAGA complex was the direct in vivo
target of Gal4. Significantly, Gal4 only interacted with
Tra1 when it was in the SAGA complex and selectively
recruited SAGA and not NuA4 to the promoter.
Distinct Subunits of the Mediator Complex Are
Targets of Diverse Mammalian Activators
Mediator is an evolutionary-conserved complex that
contains approximately 25–30 subunits and has multi-
ple roles in transcriptional regulation (Conaway et al.,
2005). Based upon its ability to support activated, but
not basal, transcription in vitro, Mediator is often re-
ferred to as a coactivator. However, this designation is
a misnomer because Mediator is a component of the
PIC, and several Mediator subunits are required for
transcription of almost all genes. Mediator has been
proposed to be an activator target, and several studies
have now conclusively shown that distinct subunits of
the mammalian Mediator complex are essential and se-
lective targets of different transcriptional ADs.
In one case, a series of biochemical experiments
found that the transcriptional AD of the adenovirus E1a
protein interacted with the mammalian Mediator sub-
unit MED23 (originally called Sur2). Med23−/−embry-
onic stem cells were specifically unable to support
transcription directed by the E1a AD and the AD of the
mammalian Elk-1 protein, which also interacts with
MED23 (Stevens et al., 2002). Likewise, extracts lacking
MED23 are selectively defective for supporting tran-
scription activation by the E1a and Elk-1 ADs (Cantin
et al., 2003).
A related but separate set of studies focused upon
another subunit, TRAP220, of the mammalian Mediator
complex. In a series of biochemical experiments,
TRAP220 was found to bind to several steroid hormone
receptors in a ligand-dependent manner. Trap220−/−
primary embryonic fibroblasts display selective defects
in supporting transcription directed by several steroid
hormone receptors; a particularly striking example is
peroxisome proliferator-activated receptor γ2, a nuclear
receptor that is essential for adipogenesis (Ge et al.,
2002). Collectively, the results of the studies discussed
above identify two mediator subunits, MED23 and
TRAP220, as selective and essential targets of specific
mammalian transcriptional ADs.
How Does the Activator-Target Interaction Stimulate
The results of these studies are fundamental to our
understanding of how activators stimulate transcrip-
tion. As expected, the E1a-MED23 (Cantin et al., 2003)
and Gal4-Tra1 (Bhaumik et al., 2004) interactions were
shown to be required for stimulation of PIC assembly.
Interestingly, interaction between MED23 and Elk-1
also stimulates transcription of the mouse Egr1 gene at
a step after PIC assembly (Wang et al., 2005). In the
Gal4 study (Bhaumik et al., 2004), a detailed pathway
of sequential recruitments was delineated on the yeast
GAL1 promoter. First, the Gal4-Tra1 interaction recruits
SAGA to the upstream activating sequence (UAS),
where Gal4 itself is bound. SAGA, in turn, recruits the
Mediator complex to the UAS. The UAS bound Media-
tor is required for recruitment of the GTFs to the core
promoter and assembly of the PIC. Although SAGA
contains Gcn5, a histone-acetyl transferase (HAT),
Gcn5 is not required for GAL1 transcription (discussed
in Bhaumik et al. ). Gcn5 is also not expected to
be required in the Fishburn et al. (2005) assay because
transcription was monitored on a naked (nonchromatin)
Thus, for Gal4 (and presumably also for Gcn4) SAGA
functions as an “adaptor” that enables the transcrip-
can occur by artificial recruitment of many components
of the PIC. By contrast, similar artificial tethering of
chromatin-modifying components activates transcrip-
tion very poorly, at best (see, for example, Georgako-
poulos et al. ). Second, global perturbation of
chromatin structure has a surprisingly limited effect on
genome-wide activation of transcription (see, for exam-
ple, Wyrick et al. ). Third, expression-profiling
studies reveal that well-studied yeast chromatin-modi-
fying activities, such as the Swi/Snf and RSC chroma-
tin-remodeling complexes and the HAT Gcn5, are re-
quired for expression of only a small percentage of
genes (Holstege et al., 1998; Angus-Hill et al., 2001).
It is important to bear in mind that the low percentage
of yeast genes that are dependent on a single chroma-
tin-modifying activity may be explained by redundancy
among the different activities. Consistent with this pos-
sibility, yeast chromatin-modifying activities have been
found to be recruited to promoters that do not require
them for activity. For example, Gal4-mediated tran-
scription of the GAL1 gene does not require Gcn5, al-
though Gcn5 is recruited to the GAL1 promoter as part
of the SAGA complex (Bhaumik et al., 2004). In higher
eukaryotes, chromatin structure is more condensed
than in yeast and, as a result, genome-wide transcrip-
tion in higher eukaryotes may be more dependent on a
chromatin-modifying activity. A precedent for this idea
is that although Swi/Snf and Gcn5 are normally dis-
pensable for GAL1 transcription, both are required dur-
ing mitosis when the chromatin becomes more con-
densed (Krebs et al., 2000).
Genetic or pharmaceutical inhibition of chromatin-
modifying activities has clearly shown their requirement
for activation of specific genes (Narlikar et al., 2002).
In these cases, the chromatin-modifying activity may
enable the DNA bound activator to interact with its
target and stimulate PIC assembly. Alternatively, a
chromatin-modifying activity may be required for bind-
ing of the activator to the promoter, but not for the DNA
bound activator to stimulate transcription. Chromatin
modification may also be important for aspects of tran-
scription other than activation. For example, Gcn5 HAT
activity is not required for the steady-state level of acti-
vated PHO5 transcription but increases the rate of acti-
vation (Barbaric et al., 2001). Chromatin modification
may also play a role in “transcriptional memory”, a
somewhat ill-defined phenomenon relating to the long-
term maintenance of a transcriptional state.
How Many Targets in Total?
Previous experimental observations have suggested
that specific activators will have multiple, functionally
redundant targets. For example, in vitro protein-protein
interaction experiments have shown that a single acti-
vator can interact with multiple components of the tran-
scription machinery. Likewise, in artificial recruitment
experiments, a wide variety of transcription compo-
nents can stimulate PIC assembly and transcrip-
tion and, thus, could potentially function as targets
(Ptashne and Gann, 1997). However, in contrast to this
expectation, in the studies reviewed above only a single
essential and, thus, nonredundant, target was identified
for each activator.
The fact that Tra1 and Mediator subunits have now
repeatedly been found to be activator targets raises the
Figure 2. Different Activator-Target Interactions Can trigger Similar
tional AD to recruit Mediator. The results of these col-
lective studies reveal how different activator-target
interactions can trigger similar activation pathways
(Figure 2). In addition to the protein interaction network
described above for Gal4, several other transcription
complex assembly pathways have been proposed in-
cluding independent activator-mediated recruitment of
SAGA and Mediator, and direct interaction between the
SAGA subunit Spt3 and TATA-box binding protein (TBP)
(discussed in Bhaumik et al. ).
In addition to this “direct” mechanism for stimulating
PIC assembly, activators have also been proposed to
stimulate transcription through interaction with chro-
matin-modifying activities (Narlikar et al., 2002). Such
modifying activities include ATP-dependent remodeling
complexes, which use energy to noncovalently modify
chromatin structure, and histone-modifying complexes,
which add or remove covalent groups (e.g., acetyl
groups, methyl groups, and phosphates) from histone
tails. In addition, there are inhibitory chromatin-modify-
ing activities, such as histone deacetylases, which can
establish repressive chromatin structures that block the
accessibility of activators and PIC components. The
existence of activities that increase or decrease histone
acetylation is consistent with the longstanding obser-
vation that the level of histone acetylation correlates
with transcription. For the vast majority of genes in
yeast and higher eukaryotes, the relative contributions
of chromatin modifications versus activator-PIC com-
ponent interactions for transcriptional activation have
not been determined.
In all of the studies reviewed above, a component of
the PIC was found to be an essential activator target.
Although it may be premature to draw strong conclu-
sions based upon this limited number of examples, it
is tempting to speculate that at least one interaction
between an activator and a PIC component may be re-
quired for transcription activation. This possibility is
supported by several independent observations that
when taken together raise the possibility that modifica-
tion of chromatin structure may be necessary but, in
general, will not be sufficient for transcription activa-
tion. For example, as discussed above, transcription
Conaway, J.W., Florens, L., Sato, S., Tomomori-Sato, C., Parmely,
T.J., Yao, T., Swanson, S.K., Banks, C.A., Washburn, M.P., and Con-
away, R.C. (2005). FEBS Lett. 579, 904–908.
Fishburn, J., Mohibullah, N., and Hahn, S. (2005). Mol. Cell 18,
Ge, K., Guermah, M., Yuan, C.X., Ito, M., Wallberg, A.E., Spiegel-
man, B.M., and Roeder, R.G. (2002). Nature 417, 563–567.
Georgakopoulos, T., Gounalaki, N., and Thireos, G. (1995). Mol.
Gen. Genet. 246, 723–728.
Holstege, F.C., Jennings, E.G., Wyrick, J.J., Lee, T.I., Hengartner,
C.J., Green, M.R., Golub, T.R., Lander, E.S., and Young, R.A. (1998).
Cell 95, 717–728.
Krebs, J.E., Fry, C.J., Samuels, M.L., and Peterson, C.L. (2000). Cell
Kuras, L., and Struhl, K. (1999). Nature 399, 609–613.
Li, X.Y., Virbasius, A., Zhu, X., and Green, M.R. (1999). Nature 399,
Lin, Y.S., and Green, M.R. (1991). Cell 64, 971–981.
Narlikar, G.J., Fan, H.Y., and Kingston, R.E. (2002). Cell 108, 475–
Ptashne, M., and Gann, A. (1997). Nature 386, 569–577.
Roeder, R.G. (2005). FEBS Lett. 579, 909–915.
Stevens, J.L., Cantin, G.T., Wang, G., Shevchenko, A., and Berk,
A.J. (2002). Science 296, 755–758.
Wang, G., Balamotis, M.A., Stevens, J.L., Yamaguchi, Y., Handa, H.,
and Berk, A.J. (2005). Mol. Cell 17, 683–694.
Wyrick, J.J., Holstege, F.C., Jennings, E.G., Causton, H.C., Shore,
D., Grunstein, M., Lander, E.S., and Young, R.A. (1999). Nature 402,
possibility that of the numerous components of the
transcription machinery (>100 polypeptides) only a lim-
ited number may serve as actual targets. Although the
majority of yeast genes are not SAGA dependent (Hol-
stege et al., 1998), the role of Tra1 as an activator target
may be more widespread because Tra1 is also present
in the NuA4 complex. Some SAGA-independent genes,
such as ribosomal protein genes, are NuA4 dependent,
and Tra1 could be the target of the relevant activators
in the NuA4 complex. Differences in Tra1 conforma-
tion in the SAGA and NuA4 complexes could provide
a basis for selective interaction with transcriptional
ADs. Yeast Tra1 protein is homologous to the human
transformation/transcription domain-associated protein
(TRRAP), which has been reported to interact with the
ADs of several oncogenic transcription factors such as
c-Myc and E2F. Thus, Tra1 and TRRAP may function as
direct targets of many transcriptional ADs in yeast and
Although the total number of actual targets may
be more limited than originally suspected, the recent
studies confirm that there is more than simply one “uni-
versal” target. This is directly relevant to our under-
standing of how multiple activators cooperate with one
another to activate transcription synergistically—a phe-
nomenon that occurs in higher eukaryotic promoters,
which are composed of binding sites for multiple acti-
vators. One model for cooperativity is that the different
activators act at distinct steps; for example, one
activator could stimulate PIC assembly whereas an-
other activator(s) could promote transcriptional elonga-
tion or recruit an essential chromatin-modifying activity.
An alternative model is that the multiply bound activa-
tors simultaneously interact with different PIC compo-
nents, synergistically increasing PIC assembly and
transcription. The crosslinking study of Fishburn et al.
(2005) helps establish feasibility for this model by
showing that a single DNA bound activator can interact
with multiple components of the PIC.
The definitive identification of a set of activator targets
represents an important advance in our understanding
of eukaryotic transcription activation mechanisms. To
confirm and generalize the conclusions from these
studies will demand the analysis of a more comprehen-
sive set of yeast and mammalian activators. Several of
the experimental approaches used in these studies,
such as the in vitro crosslinking and in vivo FRET as-
says, are general methods that can be used to identify
the targets of other yeast and mammalian activators.
Angus-Hill, M.L., Schlichter, A., Roberts, D., Erdjument-Bromage,
H., Tempst, P., and Cairns, B.R. (2001). Mol. Cell 7, 741–751.
Barbaric, S., Walker, J., Schmid, A., Svejstrup, J.Q., and Horz, W.
(2001). EMBO J. 20, 4944–4951.
Bhaumik, S.R., Raha, T., Aiello, D.P., and Green, M.R. (2004). Genes
Dev. 18, 333–343.
Brown, C.E., Howe, L., Sousa, K., Alley, S.C., Carrozza, M.J., Tan,
S., and Workman, J.L. (2001). Science 292, 2333–2337.
Cantin, G.T., Stevens, J.L., and Berk, A.J. (2003). Proc. Natl. Acad.
Sci. USA 100, 12003–12008.