P-TEFb Containing Cyclin K and Cdk9 Can Activate Transcription via RNA*

Different positive transcription elongation factor b (P-TEFb) complexes isolated from mammalian cells contain a common catalytic subunit (Cdk9) and the unique regulatory cyclins CycT1, CycT2a, CycT2b, or CycK. The role of CycK as a transcriptional cyclin was demonstrated in this study. First, CycK activated transcription when tethered heterologously to RNA, which required the kinase activity of Cdk9. Although this P-TEFb could phosphorylate the C-terminal domain (CTD) of RNA polymerase II (RNAPII)in vitro, in contrast to CycT1 and CycT2, CycK did not activate transcription when tethered to DNA. Interestingly, when the C termini of CycT1 and CycT2 or only the histidine-rich stretch from positions 481 to 551 in CycT1 were added to CycK, the extended chimeras activated transcription equivalently via DNA. Moreover, these transcriptional effects required the CTD of RNAPII in cells. Thus, CycK functions as P-TEFb only via RNA, which suggests the presence of cellular RNA-bound activators that require CycK for their transcriptional activity.

The elongation step is critical for transcription by RNA polymerase II (RNAPII). 1 Many cellular factors have been identified for their roles in this process. They include P-TEFb and the negative transcription elongation factor (N-TEF) (1)(2)(3). In this scheme, RNAPII can initiate but not elongate because of its interaction with N-TEF (2,4). Most likely, N-TEF is composed of the 5,6-dichloro-1␤-D-ribofuranosylbenzimidazole (DRB) sensitivity-inducing factor (DSIF) and the negative elongation factor (NELF). DSIF contains SPT4 (14 kDa) and SPT5 (160 kDa) (5,6). NELF contains four subunits, of which the NELF-E/RD subunit contains an RNA recognition motif (RRM) (7). P-TEFb is then recruited to the transcription complex, where its kinase subunit phosphorylates the C-terminal domain (CTD) of RNAPII (8,9) and N-TEF (10,11), allowing the elongation of transcription to proceed. P-TEFb was first iden-tified as the factor that was required for the reconstitution of DRB sensitivity in Drosophila melanogaster (8). Over the past 2 years, different P-TEFb complexes have been identified. These heterodimers contain a catalytic subunit, Cdk9, and a regulatory subunit, which can be CycT1, CycT2, or CycK (12)(13)(14). CycT2 exists in two forms, CycT2a and CycT2b, because of alternative splicing. CycT1, CycT2a, CycT2b, and CycK share extensive sequence similarity in their cyclin boxes at the N terminus from positions 1 to 250, which contain the Cdk9 binding domains. CycT1, CycT2a, and CycT2b also contain long C-terminal extensions, but their sequences diverge significantly. However, CycK is relatively small. It contains a short C-terminal domain of 107 residues from positions 251 to 357.
Recent findings revealed that P-TEFb, which contains CycT1 and Cdk9, is the key cellular factor that supports Tat transactivation and HIV replication (2,15,16). The human but not the murine P-TEFb supports the effects of Tat (12,(17)(18)(19)(20). Tat is expressed early in the replicative cycle of HIV and is essential for viral gene expression and replication (21). It recognizes the 5Ј-bulge in the transactivation response (TAR) stem loop RNA, which is located at the 5Ј-end of all viral transcripts. Tat binds CycT1, and together, they form the combinatorial surface that interacts with the TAR RNA with high affinity and specificity (12). The obligate partner of CycT1, Cdk9, then phosphorylates the CTD of RNAPII. Thus, Tat promotes HIV transcription at the step of elongation rather than initiation (22). The key step in these effects of Tat is the recruitment of P-TEFb to RNAPII. These findings also raised the possibility that such a system may exist in higher eukaryotes where Tat homologs may recruit P-TEFb to RNA and activate transcription of cellular genes.
CycK was first identified as a protein that could restore progression through the cell cycle and was most closely related to human cyclins C and H (23). Reports have also suggested that CycK associates with a potent Cdk kinase activity (CAK) in vitro. Recently, the kinase partner of CycK was identified as Cdk9 (14). This complex between CycK and Cdk9 could also function as a CTD kinase in vitro (14). However, the role of CycK in transcriptional regulation has not been defined in vivo, and the question of whether CycK is solely a CAK or is also a transcriptional cyclin has not been answered. Additionally, this complex appears to play no role in Tat transactivation. Nevertheless, the understanding of other P-TEFb complexes should give us a more complete picture of the function of these transcription elongation factors.
In this study, the ability of CycK to activate transcription when recruited to a complete promoter via RNA or DNA was examined. First, we tethered CycK to RNA. Later, we fused the C termini of CycT1 or CycT2 (more precisely the histidine-rich stretch from CycT1 to the C terminus of CycK) and then tethered these hybrid proteins to DNA. We discovered that CycK can function as a transcriptional cyclin only via RNA, and the * This work was supported by Grant RO1 AI49104-01 from the National Institutes of Health and Grant R00-SF-006 from the Universitywide AIDS Research Program (UARP) (to M. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  histidine-rich stretch from CycT1 confers its ability to function via DNA in vivo.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Plasmids coding for the HIV-1 long terminal repeat where TAR was replaced by stem loop IIB (SLIIB) from the Rev (regulator of expression of virion genes) response element (RRE) linked to the CAT reporter gene (pRRESCAT), Rev⅐Cdk9 chimera (pRevCdk9) and kinase-negative Rev⅐Cdk9D167N (pRevCdk9D167N) fusion proteins had been described previously (24,25). The plasmid effector pRev contains the cDNA copy of the HIV-1 Rev gene under control of the cytomegalovirus immediate early promoter, which was present in the parental pBC12/CMV plasmid. A full-length CycK cDNA EcoRI fragment and an EcoRI fragment of the CycT1 cDNA sequence encoding amino acids 1-280 were cloned into the pRev vector and fused in-frame with the C terminus of Rev to construct pRevCycK and pRevCycT1(1-280), respectively. Plasmid pGal (pSG424) has been described before (26). It contains the coding sequence of GAL4-(1-147) immediately followed by a multiple cloning site. GAL4-(1-147) binds specifically to the upstream activator sequences, but does not activate transcription (27). CycT1 fragments from positions 301 to 726 and 481 to 551 and the CycT2 fragment from positions 301 to 730 as well as the full-length CycK fragment were cloned into pGal to construct pGalCycT1(301-726), pGalCycT1(481-551), pGalCycT2(301-730), and pGalCycK, respectively. All these expressed hybrid proteins fused in-frame with the C terminus of the Gal4 protein. A full-length CycK fragment was then cloned into pGalCycT1(301-726), pGalCycT1(481-551), and pGalCycT2(301-730). CycK fragments were fused in-frame with the C terminus of the Gal4 protein to construct pGalCycKCycT1(301-726), pGalCycKCycT1(481-551), and pGalCycKCycT2(301-730), respectively.
Cell Culture, Transfection, and CAT Assay-293T cells were seeded at a density of 10 5 cells/60-mm dish in Dulbecco's modified Eagle's medium and 10% fetal calf serum supplemented with antibiotics. After incubation for 48 h, cells were transfected with plasmid DNA using Lipofectamine (Invitrogen, Carlsbad, CA). All transfections were balanced for a total of 5 g of DNA with the empty plasmid vector.
Cell lines RajiLS*wt and RajiLS*⌬5 contain LS*wt (HAwt, expressing HA epitope-tagged wild-type RNAPII) and LS*⌬5 (HA⌬5, expressing an HA epitope-tagged deletion mutant RNAPII with only five repeats of the CTD (YSPTSPS)) (28). These were kind gifts from Dr. Dirk Eick. All stable cell lines were grown in RPMI 1640 supplemented with 10% fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine in the presence of 1 mg/ml G418, and 0.1 g/ml tetracycline (28). For the expression of wild-type and ⌬5 CTD, 2 ϫ 10 7 cells were washed three times with 40 ml of RPMI 1640 supplemented with 1% fetal calf serum and subsequently resuspended in 20 ml of growth medium without G418 and tetracycline. After 24 h, ␣-amanitin (2 g/ml final concentration, Roche Molecular Biochemicals, Indianapolis, IN) was added to the medium to inhibit the endogenous RNAPII (28). At 24 h after ␣-amanitin addition, stable cell lines were transfected by electroporation (900 microfarads, 250 V) (28). All transfections were balanced for a total of 30 g of DNA with the empty plasmid vector. Transfected cells were maintained in RPMI 1640 without G418 and tetracycline but in the presence of ␣-amanitin for 48 h. At 48 h after transfection, cells were lysed, and CAT activities were measured with a liquid scintillation assay as described previously (29).
Immunoprecipitations, in Vitro Kinase Assays, and Western Blotting-293T cells were mock-transfected or transfected with pGalCycK by Liptofectamine. At 48 h after transfection, cells were lysed (50 mM HEPES-KOH, pH 7.6, 150 mM NaCl, 0.1% Triton X-100, 5 mM EDTA, 5 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1 mM NaF, 0.1 mM NaVO 4 , 10 g/ml aprotinin, 1 g/ml leupeptin), and the supernatants were immunoprecipitated with the ␣-Gal antibody (Upstate Biotechnology, Lake Placid, NY). Immunoprecipitations, which were bound to protein A-Sepharose beads, were washed three times with lysis buffer and then two times with CTD kinase buffer (20 mM Tris, pH 7.6, 50 mM KCl, 5 mM MgCl 2 , 2.5 mM MnCl 2 , 10 mM dithiothreitol). Approximately 25 ng of the eluted GST⅐CTD fusion proteins (29) were used in each kinase reaction. Reactions were supplemented with 10 Ci of [␥-32 P]ATP and 50 M unlabeled ATP or 50 M unlabeled ATP, by itself, in a final reaction volume of 50 l. Reactions were incubated for 1 h at 30°C. For radiolabeled kinase reactions, samples were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) and then exposed to film. For unlabeled kinase reactions, samples were analyzed by Western blotting. After SDS-PAGE, the gel was electroblotted onto Hybond-P membrane (Amersham Biosciences) and probed with H14 or H5 antibodies against the phosphorylated CTD at serine 5 or serine 2 (BabCO, Berkeley, CA).

CycK Can Activate Transcription via RNA, Which
Requires the Kinase Activity of Cdk9 -To determine whether CycK can activate transcription via RNA, we used a heterologous RNA tethering system, where CycK was fused to Rev from HIV. Rev binds with high affinity and specificity to the stem loop IIB (SLIIB) of Rev response element (RRE) RNA, which was grafted onto the double-stranded stem in TAR (pRRESCAT) (24). Previously, we demonstrated that CycT1 and Cdk9 activated transcription when fused to Rev (25). Importantly, the kinase-negative Cdk9 (Cdk9D167N) protein was inactive in this system (25). Additionally, a short version of human CycT1, the mutant CycT1-(1-280) protein, which contained N-and C-terminal helices and two cyclin boxes but lacked residues from positions 281 to 726, could rescue Tat transactivation in murine cells (25). Thus, the truncated CycT1-(1-280) protein could activate transcription when recruited to TAR by Tat.
CycK contains extensive sequence similarity with CycT1 and CycT2 in this region but then diverges in its short C terminus. In Fig. 1, we compared the activities of the hybrid Rev⅐CycK and Rev⅐CycT1-(1-280) proteins. These two chimeras activated transcription 10-fold from pRRESCAT in 293T cells (Fig. 1,  lanes 3 and 4). This activity was one-half that of the hybrid Rev⅐Cdk9 protein (Fig. 1, lane 5), which is consistent with previous results where complexes between the full-length CycT1 and Cdk9 proteins had higher activities on the CTD substrate than those between CycK and Cdk9 in vitro (14). The transcriptional activity of the Rev⅐CycK fusion protein could be inhibited efficiently by DRB at concentrations between 10 and 20 M (Fig. 2A). The co-expression of the kinase-negative Cdk9 protein also blocked the transcriptional activity of the hybrid Rev⅐CycK protein on pRRESCAT (Fig. 2B). The kinase-negative Cdk9 when fused to Rev also failed to activate transcription (Fig. 1C, lane 6). These results indicate that CycK can activate transcription when tethered to RNA, which requires the kinase activity of Cdk9.
The Addition of C-terminal Sequences from CycT1 and CycT2 or Only the Histidine-rich Stretch from CycT1 Rescues the Ability of CycK to Activate Transcription via DNA-To examine the ability of the specific P-TEFb complex that contains CycK and Cdk9 to activate transcription via DNA, we fused CycK to the DNA binding domain of Gal4 (Gal). When co-expressed with the pG6CAT plasmid target with six Gal4 DNA binding sites (upstream-activating sequences), this chimera failed to activate transcription in 293T cells (Fig. 3, lane 3). Previously, we demonstrated that the full-length CycT1 but not mutant CycT1 proteins containing C-terminal deletions after position 590 could activate transcription via DNA (30). As CycT1, CycT2, and CycK share the highly conserved N-terminal cyclin boxes, we examined whether these C-terminal extensions of CycT1 and CycT2 could confer upon CycK the ability to activate transcription via DNA. Indeed, when the Gal⅐CycK fusion protein was extended with the C-terminal domains of CycT1 from positions 301 to 726 (Gal⅐CycK⅐CycT1-(301-726)) and from positions 301 to 730 in CycT2 (Gal⅐CycK⅐CycT1-(301-730)), these tripartite chimeras activated transcription from pG6CAT equivalent to the chimeras between Gal and full-length CycT1, Gal⅐CycT1-(1-726) (Fig. 3, lanes 5, 7, and 10). Importantly, the fusion proteins between these C termini and Gal, lacking the cyclin boxes, had no effect in this assay (Fig. 3, lanes 4 and 6). Additionally, when only the minimal C-terminal domain from positions 481 to 551 in CycT1 (CycT1-(481-551)), which contained the histidine-rich stretch was fused with the hybrid Gal⅐CycK protein, this Gal⅐CycK⅐CycT1-(481-551) fusion pro- tein also activated transcription to levels similar to the Gal⅐CycK⅐CycT1-(301-726) chimera (Fig. 3, lane 9). Again, this fragment had no activity in the absence of the cyclin boxes (Fig.  3, lane 8). This result is in agreement with our previous study, where the histidine-rich stretch in the C-terminal domain of CycT1 bound the unphosphorylated CTD of RNAPII (30). A structurally and functionally homologous sequence is also found in the C terminus of CycT2 from positions 550 to 590. We conclude that CycK lacks the binding site for RNAPII. When this sequence is provided, this P-TEFb complex can also activate transcription via DNA.
The Complex between the Hybrid Gal⅐CycK Protein and Cdk9 Is an Active CTD Kinase in Vitro-To prove that the Gal⅐CycK fusion protein contains a CTD kinase activity, complexes of the Gal⅐CycK chimera and Cdk9 were purified by immunoprecipitation from 293T cells and assayed for their ability to phosphorylate the hybrid GST⅐CTD protein in vitro. Anti-Gal antibody was used for the immunoprecipitation (Fig. 4). As shown in Fig.  4, the complex between the hybrid Gal⅐CycK protein and Cdk9 could phosphorylate the GST⅐CTD fusion protein only from cells that expressed the chimera (Fig. 4, compare lanes 1 and  2). We also determined the specificity of this CTD kinase using H14 and H5, two antibodies that recognize phosphorylated serine 2 and serine 5 in the heptapeptide repeats from the CTD, respectively. Indeed, the complex between the hybrid Gal⅐CycK protein and Cdk9 could phosphorylate both serines at positions 2 and 5 in the heptapeptide repeats of CTD (Fig. 4,  lanes 4 and 6). Thus, the complex between the Gal⅐CycK fusion protein and Cdk9 is an active CTD kinase, but it cannot activate transcription via DNA in vivo.
Transcriptional Activities of Mutant Gal⅐CycK⅐CycT1-(301-726), Gal⅐CycK⅐CycT2-(301-730), and Gal⅐CycK⅐CycT1-(481-551) Chimeras Require the CTD of RNAPII-To confirm that the CTD is required for transcriptional effects of Gal⅐CycK⅐CycT1-(301-726), Gal⅐CycK⅐CycT2-(301-730), and Gal⅐CycK⅐CycT1-(481-551) chimeras, we examined the activities of these fusion proteins in Raji cells, which expressed the ␣-amanitin-resistant wild-type RNAPII (52 heptapeptide repeats, HAwt) or the deletion mutant RNAPII protein that contained only 5 heptapeptide repeats (HA⌬5) (Fig. 5). Only cells that expressed HAwt supported the activity of our tripartite chimeras (Fig. 5, lanes 8, 9, and 10). In Raji cells where HA⌬5 was expressed, no activation was observed with any of our proteins in the presence of ␣-amanitin, which was added for the duration of the assay (Fig. 5, lanes 1-6, see "Experimental Procedures"). Taken together, these data indicate that the complex between CycK and Cdk9, similar to other P-TEFb complexes, requires the CTD of RNAPII for function. DISCUSSION In this study, we have demonstrated that CycK forms an active P-TEFb complex with Cdk9 and promotes transcription via RNA in vivo. In sharp contrast, CycK could not activate transcription via DNA although it still functioned as a CTD kinase in vitro (14). Because CycT1 and CycT2 can activate transcription via DNA and the histidine-rich stretch in CycT1 binds the CTD, the transcriptional activity of the hybrid Gal⅐CycK protein could be rescued by extending it with the C termini from CycT1 or CycT2. More importantly, the histidinerich stretch from CycT1 alone could also rescue the activity of the Gal⅐CycK fusion protein via DNA. These extended tripartite fusion proteins required the CTD of RNAPII for activity. Thus, the C termini of these cyclin subunits from P-TEFb play important regulatory roles and dictate their substrate specificities.
It is to be noted that our studies were performed in vivo rather than in vitro. In in vitro transcription systems, P-TEFb is already present in the preinitiation complex, and the complex between CycK and Cdk9 functions via DNA (14). This finding could be due to very different stoichiometries of DNA templates and transcription complexes or compositions of nuclear extracts. Additionally, exogenously added, abundant P-TEFb could bind the CTD or RNAPII without the help of other proteins. Nevertheless, all other studies point to the obligatory recruitment of P-TEFb by RNA-or DNA-bound activators or their coactivators in cells. Thus, Tat, CIITA, NF-B, androgen receptor, and c-Myc all recruit P-TEFb to the transcription complex (12,(31)(32)(33)(34). Only then does P-TEFb interact with the CTD of RNAPII and N-TEF, leading to their phosphorylation (8 -11). This modification results in the transition from initiation to elongation of eukaryotic transcription. Importantly, FIG. 4. The complex between Gal⅐Cyck fusion protein and Cdk9 functions as a CTD kinase in vitro. In vitro kinase assays were performed with the complex between the Gal⅐CycK chimera and Cdk9, which was immunoprecipitated by the anti-Gal antibody from 293T cells (lanes 2, 4, and 6) and control mock-transfected cells (lanes 1, 3, and 5). The hybrid GST⅐CTD protein was used as the template. Radiation label was incorporated from [␥-32 P]ATP in the reaction (lanes  1 and 2). The phosphorylated CTD bands are indicated by the arrows. H5 and H14 are antibodies that recognize the phosphorylated serine 2 and serine 5 in the heptapeptide repeats from the CTD, respectively. GST⅐CTD with phosphorylated serine 2 (lane 4) or serine 5 (lane 6) was detected by Western blotting as the arrows indicate. CTDo is the hyperphosphorylated form of CTD; CTDa is the underphosphorylated form of CTD. CycK lacks this CTD-interacting domain. Because nascent RNA moves from the catalytic pocket in RNAPII along the CTD (35), Cdk9 bound to CycK can still phosphorylate this substrate. DNA presentation is qualitatively different, where P-TEFb must first find and bind the CTD for Cdk9 to phosphorylate the RNAPII. The histidine-rich stretches in CycT1 and CycT2 perform this function. This finding explains why all three cyclins function via RNA, but only CycT1 and CycT2 can activate transcription via DNA.
The theme of cyclins directing the activity of their associate kinases is not new. For example, the specificity of Cdk2 is governed by its associated cyclin A, especially by its hydrophobic MRAIL sequence, which is required for the binding and recognition of target proteins that contain the RXL motif (36). Likewise, CycK, CycT1, and CycT2 contain their highest sequence similarity in their cyclin boxes, where they bind Cdk9. Their C termini are very divergent. CycT1 and CycT2 share little besides the histidine-rich stretch, and only CycT1 possesses the TRM and PEST sequences (12)(13). Additionally, only the complex between CycT1 and Cdk9 can activate HIV-1 transcription (12). The TRM sequence is required for this function (12). A recent study demonstrated the PEST sequence in CycT1 is required for the interaction between CycT1 and SCF (SKP2), which then targets Cdk9 for ubiquitination and degradation by the proteasome (37). Moreover this P-TEFb complex only phosphorylates CTD at serine 2 during HIV-1 transcription (38). In this study, the complex between CycK and Cdk9 could phosphorylate both serines 2 and 5. Thus, this P-TEFb might have a broader specificity. Supporting this notion, CycK was first associated with a CAK activity (23). Other differences in the modes of action and target specificities of these different P-TEFb complexes will be revealed in future studies with many distinct activators and by following its genetic inactivation in the mouse.
Our studies with CycK also suggest the tantalizing possibility that other cellular activators, i.e. Tat homologs, exist that function via RNA. They are expected to bind CycK and recruit this P-TEFb complex to positions downstream from the site of initiation of transcription. Indeed, a strategy using the yeast three-hybrid screening system to detect RNA-protein interactions identified 70 RNA sequences that functioned independently of an exogenous activation domain on their associated proteins (39). Moreover, these RNA sequences needed to be positioned near the promoter for their effects (39). It is likely that these RNA species could fold into structures that directly recruited RNA-bound activators (39). Finally, CycK could also participate in cotranscriptional processing to maintain the hyperphosphorylated state of RNAPII before polyadenylation and maturation of primary transcripts. Such a cotranscriptional role for P-TEFb is also hinted at by recent studies, which suggest that some complexes between CycT1 and Cdk9 associate with 7SK RNA (40,41), which has been colocalized with U1 snRNA in the nucleus (42).