ABSTRACT
Previously, we showed that the viral transactivator proteins E1A and VP16
specifically interact with a cellular CTD kinase activity
in vitro
. We now report that E1A and VP16 complexes contain human CDK8, a newly identified member of the cyclin-dependent kinase family that has been shown to be a component of the RNA
polymerase II (RNAP II) holoenzyme complex. The presence of CDK8 in the E1A- and VP16-containing complexes is specific for a functional activation domain
of these viral transactivators, strongly suggesting that this association is
relevant for the transactivation function of E1A and VP16. We show that CDK8 is
associated with CTD kinase activity and that CDK8 co-fractionates with E1A- and VP16- associated CTD kinase activity over several chromatography
columns. Therefore, CDK8 is likely responsible for the E1A- and VP16-associated CTD kinase activity. Gel filtration chromatography
indicates that the E1A- and VP16-associated CTD kinase activity has a molecular size of
~
1.5 MDa and contains cyclin C and the human homolog of SRB7 in addition to CDK8.
This implies that E1A and VP16 associate with the RNAP II holoenyzme. We also
looked at the transcriptional activity of CDK8 and found that CDK8 can function
as a transcriptional activator when fused to the DNA binding domain of GAL4.
Surprisingly, the ability of GAL4-CDK8 to activate transcription in this assay was not dependent on the
kinase activity of CDK8, since a kinase-deficient mutant of CDK8 stimulated transcription nearly as well as wild-type GAL4-CDK8. This suggests that CDK8 may play a role in
transcription that is distinct from its ability to function as a CTD kinase.
Transcription of eukaryotic genes is a complex and highly regulated process that is comprised of several discrete steps, each of which is subject to regulation and requires the presence of multiple protein factors.
Recent work has shown that transcription is mediated by a large multi-protein complex termed the RNA polymerase II (RNAP II) holoenzyme that was
originally discovered in yeast (
1
-
3
). In addition to the multi-subunit RNA polymerase II (RNAP II) enzyme, the yeast holoenzyme contains
general transcription factors (GTFs), which are necessary for RNAP II to
initiate transcription at specific sites, and a mediator complex, which is
required for proper response to transcriptional regulatory factors. The
mediator complex contains proteins, known as SRB (suppressor of RNA polymerase
B) proteins, that were originally identified as dominant extragenic suppressors
of a mutant form of RNAP II in which the largest subunit contained partial deletions of the C-terminal domain (CTD) (
4
,
5
). In mammalian cells, a large multisubunit complex containing RNAP II and GTFs has been
described (
6
). More recently, RNAP II holoenzyme complexes have been isolated from human
cells that also contain homologs of the yeast SRB proteins (
7
,
8
).
The CTD of RNAP II contains an unusual, highly repetitive structure consisting
of tandem repeats of the consensus sequence Tyr-Ser-Pro-Thr-Ser-Pro-Ser (
9
,
10
). Although its precise role in transcription remains unclear, the function of
the CTD is thought to be mediated through protein-protein interactions with SRB proteins and other components of the
transcription complex. These interactions are likely regulated by
phosphorylation of the CTD. The CTD is a target for phosphorylation by a number
of protein kinases, including a TFIIH-associated kinase (
11
), the cyclin-dependent kinases CDC2 and CDK2 (
12
,
13
) and several partially purified activities (
13
-
15
). Two of these kinases are associated with the RNAP II holoenzyme: the TFIIH-associated kinase and SRB10/11. The catalytic component of TFIIH is CDK7 (also known as MO15) (
16
-
19
). Addition of TFIIH containing wild-type CDK7 reconstitutes transcription in a TFIIH-dependent
in vitro
system, while a kinase-deficient form of TFIIH supports only abortive initiation, suggesting that
the kinase activity of CDK7 is required for transcription at a stage after the
formation of the first phosphodiester bond (
20
). The SRB10/11 proteins are a CDK/cyclin pair that are found in RNAP II holoenzymes isolated from
Saccharomyces cerevisiae
(
21
). Holoenzymes lacking SRB10/11 kinase activity are deficient in CTD phosphorylating activity and show defects in transcription
in vivo
, although they are not defective for transcription using a purified
in vitro
system (
21
). The SRB10 protein is identical to UME5 (
22
), SSN3 (
23
) and ARE1 (
24
), proteins identified in independent genetic screens. SRB10 is also
structurally related to the human CDK8 protein (originally named K35) (
25
). CDK8, isolated in a search for human protein kinases that might play a role in cell cycle control (
26
), encodes a protein of 464 amino acids with a predicted molecular mass of 53
kDa (
25
). CDK8 associates with cyclin C
in vitro
and
in vivo
. Cyclin C is the closest known mammalian relative of SRB11. Like their yeast
counterparts, both CDK8 and cyclin C are found associated with the human RNAP
II holoenzyme (
8
).
It is likely that cellular CTD kinases are involved in mediating the action of
at least some viral transactivator proteins as several especially potent viral transactivator proteins can specifically target cellular CTD kinases. For example, the herpes simplex virus VP16 protein is
known to interact with TFIIH (
27
,
28
) and the human immunodeficiency virus Tat proteins specifically interact with
the Tat-associated kinase (TAK) (
13
,
29
,
30
). More recently, we have identified a cellular CTD kinase activity that
specifically interacts with the activation domains of VP16 and the adenovirus
E1A proteins
in vitro
(
31
). The VP16- and E1A-associated kinase activities were shown to be related to each other
but distinct from TFIIH and TAK (
31
).
In this report, we show that E1A and VP16 associate with a large complex,
presumably the RNAP II holoenzyme, that contains CDK8, cyclin C and the human
homolog of SRB7. We also show that CDK8 is associated with CTD kinase activity
and that CDK8 co-fractionates with the E1A- and VP16-associated CTD kinase activity over several chromatography
columns. This suggests that the E1A- and VP16-associated CTD kinase is most likely CDK8. We further demonstrate
that CDK8 is capable of activating transcription when artificially targeted to
the promoter. Surprisingly, the ability of CDK8 to stimulate transcription is
not dependent on the kinase activity of CDK8. However, our results indicate
that CDK8 possesses a weak activation domain at its C-terminus, a region not conserved among other CDKs. Possible roles for CDK8
in transcription are discussed.
Bacterial expression plasmids used to express fusions of glutathione-
S
-transferase (GST) to the activation domains of the E1A and VP16 proteins (
31
) or the Tat proteins (
32
) have been described previously. The construct used for expression of GST-CTD was a gift from W. Dynan and colleagues (
33
). For construction of GAL4-CDK7, CDK7 sequences were PCR amplified from a CDK7-encoding plasmid (obtained from D.O. Morgan), inserted into a modified version of the plasmid pGEM3Z+ (Promega) to generate
pGEM-CDK7 and the sequence was confirmed by di-deoxy sequencing. The fragment from pGEM-CDK7, encoding the CDK7 sequences was then cloned into pSG424 (
34
), which encodes the DNA binding domain (a.a. 1-147) of GAL4 directed by the SV40 early promoter. For construction of
GAL4-CDK8, CDK8 sequences were excised from a CDK8/Bluescript expression vector (
25
) and inserted into pSG424. The D173A mutant was constructed using the
Transformer Site Directed Mutagenesis kit (Clontech). To construct GAL4-D173A, the region of CDK8 containing the mutation was removed from the D173A/Bluescript expression vector and inserted into the corresponding region of GAL4-CDK8. The sequence of the D173A mutant was verified by di-deoxy sequencing. The GAL4-CDK8NX mutant contains the
Nco
I to
Xba
I fragment of CDK8 (amino acid residues 318-464) in the pSG424 vector. To generate the FLAG-CDK8 fusions, CDK8 sequences obtained from the GAL4-CDK8 plasmid were inserted into pFLAG-CMV-2 (Kodak), which codes for an eight amino acid tag 5' of the multiple cloning site, yielding
pFLAG-CDK8. The construct pFLAG-D173A was generated by digestion of GAL4-D173A to obtain a fragment of CDK8 containing the mutation
which was then inserted into pFLAG-CDK8 such that the mutation replaced the wild-type sequence.
All transfections were performed using a calcium phosphate transfection
procedure (
35
). For analysis of transactivation by GAL4 fusions, HeLa cells in DMEM + 10% fetal bovine serum (Gibco) were seeded onto 6
cm dishes one day prior to transfection to yield 50% cell confluency. Each plate was transfected with 2 [mu]g GAL4 parental or fusion plasmid + 2 [mu]g pG6(-83)HIVLTR[Delta]TAR- CAT (
36
) + 0.5 [mu]g pSV-[beta]gal (
37
) + 5 [mu]g salmon sperm DNA. At ~48 h post-transfection, cells were harvested for CAT and [beta]-galactosidase assays. Results of CAT assays were
quantified on a Betascope 603 blot analyzer (Betagen). Reaction conditions were
adjusted so that all results used for quantitation of CAT activity had <50% acetylation of the total chloramphenicol. Likewise, all measurements used for quantitation of [beta]-galactosidase activity had OD
420
values of <0.4.
Immunoprecipitations were performed as described (
38
). Immune complexes were washed three times in EBC buffer (50 mM Tris-HCl pH 8.0, 120 mM NaCl, 0.5% NP-40 and 5 mM DTT) and resuspended in Laemmli sample buffer.
Immunoblot analysis was carried out as described previously (
31
) using enhanced chemoluminescence for detection of immunoreactive proteins. The rabbit anti-CDK8 (
25
) and anti-CDK7 (
39
) polyclonal antibodies were used at a dilution of 1:5000, the anti-cyclin C antibody (
40
) was used at a dilution of 1:2000, and the anti-hSRB7 antibody (
7
) was used at a dilution of 1:400. One-dimensional peptide analysis using V8 protease (Pierce) was performed as
described (
38
).
Fractionation of HeLa cell nuclear extracts over phosphocellulose P11 and Resource Q columns was described previously (
31
). The peak of the E1A- and VP16-associated kinase activities eluted from the Resource Q column at
270-300 mM KCl. This peak was concentrated using a Centricon-30 column and applied to a Superose 6 FPLC column (Pharmacia)
equilibrated with 50 mM Tris pH 8.0, 10% glycerol, 0.2 mM EDTA, 100 mM KCl, 0.5
mM DTT and 0.5 mM PMSF. Proteins were eluted with equilibration buffer at 0.2
ml/min and fractions (400 [mu]l) were analyzed directly by immunoblot analysis or for CTD kinase activity.
The molecular size of the complex was determined by extrapolation from
molecular weight standards.
Kinase assays were performed as described previously (
31
). Briefly, fusions of viral transactivator proteins expressed in bacteria as
GST-fusions were purified as described (
13
,
29
) and bound to HeLa cell nuclear extracts (
31
) for 60 min at 4oC with gentle rocking. Complexes were washed three times with EBC buffer
containing 0.03% SDS, followed by one wash with Kinase Buffer (50 mM Tris pH
7.4, 10 mM MgCl
2
and 5 mM DTT). Kinase assays were performed by adding 25 [mu]l of a mix containing Kinase Buffer, 2.5 mM MnCl
2,
200 ng GST-CTD, 5 [mu]M ATP and 5 [mu]Ci [[gamma]-
32
P]ATP (NEN, 3000 Ci/mmol) to the bead complexes and incubating for 60 min at
room temperature. Complexes were pelleted briefly, denatured in Laemmli sample buffer, and resolved by 9% SDS-PAGE.
Previously, we reported that a CTD kinase activity associates with the viral
transactivator proteins E1A and VP16
in vitro
(
31
). The interaction of a CTD kinase with E1A or VP16 was dependent on a
functional activation domain of these proteins, suggesting that the interaction
is relevant for the transactivation function of E1A and VP16. We also showed
that the kinase activities that associate with E1A and VP16 are biochemically related to each other but distinct from two
other known CTD kinase activities, TFIIH and TAK. Recently, the cloning of a novel member of the cyclin- dependent kinase family, CDK8, was reported (
25
). Given the sequence similarity of human CDK8 to the
S.cerevisiae
SRB10 protein, a kinase involved in CTD phosphorylation (
21
), we investigated whether human CDK8 is associated with CTD kinase activity and whether it interacts with E1A or VP16.
To test whether CDK8 is associated with CTD kinase activity, a CTD kinase assay
was performed as described in Materials and Methods using GST-CTD as a substrate. Hyperphosphorylation of the GST-CTDa substrate leads to the generation of the more slowly
migrating GST-CTDo form; the appearance of the CTDo form is indicative of CTD kinase
activity (
41
). An anti-CDK8 antibody (
25
) was used to immunoprecipitate CDK8 from a HeLa cell nuclear extract. The
immune complexes were washed and a CTD kinase assay was performed. As seen in Figure
1
A, the CTDo form was generated in reactions using the anti-CDK8 antibody but not the preimmune serum. There was also an increase in
the level of phosphorylation of the CTDa form. This experiment indicates that
CDK8 is associated with CTD kinase activity. Results presented below using a
kinase-deficient form of CDK8 (Fig.
4
) further suggest that CDK8 possesses CTD kinase activity.
To further characterize the E1A- and VP16-associated CTD kinase activities, we partially purified the viral-associated kinase activities. HeLa cell nuclear extracts were fractionated over a
phosphocellulose P11 column and the 0.5 M KCl eluate was then passed over a Resource Q ion
exchange column (
31
). The peak of E1A- and VP16-associated CTD kinase activity coincided precisely with peak levels
of CDK8 as determined by immunoblot analysis (not shown). These fractions also
contained the highest levels of cyclin C and the human homolog of the yeast
SRB7 protein (hSRB7), a component of the RNAP II holoenzyme mediator
subcomplex. The majority of CDK7, another CTD kinase present in the holoenzyme,
eluted from the Resource Q column at a lower salt concentration than the peak
of E1A- and VP16-associated CTD kinase activity, although low levels were detected in
the peak fractions (not shown).
We have previously proposed a model in which some viral transactivators function
by recruiting a CTD kinase to the promoter (
13
,
31
). If CDK8 is important in mediating the response to E1A and VP16, we reasoned
that CDK8 might activate transcription in the absence of E1A or VP16 if it was
artificially targeted to the promoter. For this purpose, the cDNA clone of CDK8
(
25
) was inserted into the pSG424 vector (
34
) which encodes the GAL4 DNA binding domain (a.a. 1-147) driven by the SV40 early promoter. The GAL4-CDK8 plasmid was transfected into HeLa cells along with a reporter plasmid, G6(-83)HIVLTR[Delta]TAR (
36
), that contains six GAL4 binding sites upstream of a promoter based on the HIV-1 LTR that is fused to the chloramphenicol acetyltransferase (CAT) gene.
As shown in Figure
3
, the GAL4-CDK8 plasmid produced a 40-fold increase in expression from the reporter plasmid relative to a
construct containing the GAL4 DNA binding domain only (after normalization to [beta]-galactosidase activity to control for transfection efficiencies). A similar result was obtained using a different
reporter plasmid, G5E1B-CAT-SP (
42
), which contains an Sp1 site upstream of the E1B TATA element (
43
). Transactivation was dependent on the presence of the GAL4 DNA binding sites
as GAL4-CDK8 did not increase expression from a reporter construct that lacked
the GAL4 DNA binding sites (
43
).
The sequence similarity between human CDK8 and the yeast SRB10 protein, a
component of the yeast mediator complex, suggested that CDK8 may play a role in
CTD phosphorylation and transcriptional regulation. The work reported in this
study is consistent with such a role for CDK8. First, we showed that CDK8 is
tightly associated with CTD kinase activity. CTD kinase activity is present in CDK8-containing immune complexes (Fig.
1
A) and CDK8 co-fractionates with CTD kinase activity over three chromatography columns
(Fig.
2
and data not shown). The result that the kinase-deficient FLAG-D173A mutant does not phosphorylate the CTD (Fig.
4
) further supports the idea that CDK8 is a CTD kinase. Recent work from E. Lees
and colleagues has also shown that a CDK8/cyclin C complex displays kinase
activity for the CTD (
40
).
Drosophila
CDK8 has also been shown to be associated with CTD kinase activity (
44
). However, it is still formally possible that CDK8 associates with another
kinase that phosphorylates the CTD.
Second, we showed that two viral transactivators, E1A and VP16, interact with a CDK8-containing complex (Figs
1
B and C). We currently do not know whether the interaction between CDK8 and the
viral activator proteins is a direct physical interaction. It is possible that
E1A and VP16 directly contact another component of the holoenzyme complex, so
that CDK8 is detected in association with E1A and VP16 as a result of direct
binding to another holoenzyme protein. However, we have demonstrated that the interaction is specific for a functional activation domain of E1A and VP16. Transactivation-deficient mutants of E1A and VP16 fail to associate with CDK8, strongly suggesting that this interaction with the
CDK8-containing complex is important for the transactivation function of these
viral transactivators. CDK8 does not, however, interact with another strong viral transactivator protein, Tat. This is perhaps not too surprising, considering that Tat, unlike
E1A and VP16, interacts with a downstream RNA element and therefore may contact
different components of the transcription complex. Tat interacts with another cellular CTD kinase, TAK,
which has been proposed to mediate Tat transactivation (
13
).
We have also shown that the CDK8-containing complex targeted by E1A and VP16 is a large complex (~1.5 MDa) that also contains cyclin C and hSRB7 (Fig.
2
). CDK7 fractionates differently than the other proteins and is not detected in
fractions that have the highest levels of E1A- and VP16-associated kinase activity. The yeast counterparts of CDK8, cyclin C
and hSRB7 have been found in the mediator subcomplex of the holoenzyme, while
CDK7 is not thought to reside in the mediator subcomplex (
1
). This suggests that E1A and VP16 may target the mediator subcomplex of the
holoenyzme. This has previously been demonstrated for VP16 by Young and
colleagues (
3
), but to our knowledge has not been shown for E1A. Therefore, targeting of the
mediator complex may be a general mechanism used by activators in recruitment
of the holoenzyme to the promoter. Our results indicate that this may not be
universal mechanism, however, since the HIV Tat protein does not appear to interact with the CDK8-containing complex (Fig.
1
C and data not shown).
Finally, we showed that CDK8 can activate transcription from a reporter plasmid when artificially targeted to the promoter (Fig.
3
). Furthermore, the C-terminal region of CDK8 alone can act as a weak activator of
transcription. The C-terminal 87 amino acids of CDK8 contains 14% glutamine, with a stretch of
five continuous glutamines between amino acid residues 378 to 464. Interestingly,
Drosophila
CDK8 contains 65% glutamine in this region, with two continuous stretches of
glutamines from positions 373 to 400 and 423 to 441. This is suggestive of a
glutamine-rich type of activation domain at the C-terminus of CDK8. This region is not present in other CDKs.
Although the kinase activity of CDK8 appears to be dispensable for the ability of CDK8 to activate transcription in this assay, it is likely
that the kinase activity of CDK8 is required
in vivo
. Liao
et al.
(
21
) have found that a kinase-deficient mutant of the yeast counterpart of CDK8, SRB10, results in a
large decrease in galactose-induced transcription
in vivo
. In our artificial system, GAL4-CDK8 probably functions by recruiting the holoenzyme to the promoter but
once present at the promoter, it is possible that other CTD kinases in the RNAP
II holoenzyme complex, such as endogenous CDK8 or perhaps CDK7, could
compensate for the loss of GAL4-CDK8 kinase activity. It also remains possible that the kinase activity
of GAL4-CDK8 may be necessary under certain conditions not assayed here, such as
in response to some activators or in response to growth regulatory signals.
Although CDK8 can clearly function as a transcriptional activator, it is not
nearly as powerful as E1A. In experiments similar to that shown in Figure
3
, GAL4-E1A produced at least a 20-fold higher level of activation of expression from the reporter construct than the GAL4-CDK8 construct (
43
). One interpretation of this result is that while recruitment of the CDK8 complex by E1A is
important for the transactivation function of E1A, other protein-protein interactions, such as that with TFIID (
45
) are required for full transactivation by E1A.
How does CDK8 activate transcription? It is possible that CDK8 can function in
recruitment of the RNAP II holoenzyme complex to the promoter. This likely
occurs when CDK8 is artificially targeted to the promoter through a fusion with
a DNA binding domain. This may also occur under more physiological conditions
through an association with DNA-bound activators such as E1A and VP16. The finding that GAL4-CDK8 activates transcription is consistent with results in the
yeast system, where components of the yeast mediator complex have been shown to
activate transcription when fused to a DNA binding domain (
23
,
46
). Ptashne and colleagues have proposed that recruitment of the RNAP II
holoenzyme complex to the promoter through a single activator-holoenzyme contact is sufficient for gene activation (
46
). Our results are consistent with this proposal. The finding that the C-terminus of CDK8 alone can activate transcription, albeit less efficiently
than the full length protein, suggests that this region may contact other
components of the holoenzyme complex. The failure of CDK7 to activate
transcription as a GAL4 fusion suggests that not all interactions with
holoenzyme components result in gene activation.
Finally, it is possible that CDK8 plays a dual role in transcription. CDK8 may function through protein-protein interactions to recruit the RNAP II holoenzyme complex to the
promoter and it may also act catalytically by phosphorylating the CTD or possibly other components of the RNAP II holoenzyme complex.
We thank D. Chao, E. Lees, D.O. Morgan and R. Young for plasmids and antibodies.
This work was supported by NIH grant AI35381 (A.P.R), NIH predoctoral training
grant AI07483 (M.O.G.), the Curtis Hankamer Basic Research Fund (C.H.H.), and
the Baylor Center for AIDS Research (C.H.H.). E.A.N. and J-P. T were supported by a grant (to E.A.N.) from the Swiss National Science
Foundation (31-33615.92).
+
Present address: Department of Molecular Biology Sciences II, University of
Geneva, 30 Quai Ernest-Ansermet, CH-1211 Geneva 4, Switzerland
REFERENCES
Return