Analysis of the Large Inactive P-TEFb Complex Indicates That It Contains One 7SK Molecule, a Dimer of HEXIM1 or HEXIM2, and Two P-TEFb Molecules Containing Cdk9 Phosphorylated at Threonine 186*

Positive transcription elongation factor b (P-TEFb) regulates eukaryotic gene expression at the level of elongation, and is itself controlled by the reversible association of 7SK RNA and an RNA-binding protein, HEXIM1 or HEXIM2. To further understand how P-TEFb is regulated, we analyzed the stoichiometry of all the known components of the large, inactive P-TEFb complex. Mutational analyses of a putative coiled coil region in the carboxyl-terminal portion of HEXIM1 revealed that the protein is a dimer in solution and remains a dimer after binding to 7SK. Although a HEXIM1 dimer contains two potential RNA binding motifs and ulti-mately recruits two P-TEFb molecules, it associates with only one molecule of RNA. The first 172 nucleotides of the 330-nucleotide 7SK are sufficient to bind HEXIM1 or HEXIM2, and then recruit and inhibit P-TEFb. Deletion of the first 121 amino acids of HEXIM1 allowed it to inhibit P-TEFb partially in the absence of 7SK RNA. Mutation of a conserved tyrosine (Tyr 271 in HEXIM1) to alanine or glutamate or mutation of a conserved phenylalanine ( H2 ). E , recombinant P-TEFb (Cdk9/cyclin T1) was resolved on a SDS gel, and the band containing Cdk9 was excised followed by partial trypsin in-gel digestion. The resulting peptides were harvested and analyzed by reverse phase liquid chromatography ( HPLC )-tandem mass spectrometry ( LC-MS / MS ). One peptide ion eluted with a retention time (RT) of 11.9 min with a very high signal to noise ratio ( S / N ). F , the MS survey scan at the point of 11.9 min, indicated the presence of the peptide ion of m / z 665.6. The ion ranges from m / z 665.0 to 668.0 because of the naturally occurring isotopic distribution and this mass range was used in F to extract ion current signal. A partial region ( m / z 500–1000) is shown for simplicity, although the entire scan range was from m / z 300 to 1600. G , MS/MS scan of the precursor ion m / z 665.6, which was fragmented into multiple labeled product ions ( b and y ions) led to the identification of a phosphopeptide with a modification site on Thr 186 according to the mass shift ( (cid:4) 80 Da) caused by phosphorylation. A predominant product ion ([M-H 3 PO 4 (cid:4) 2H] 2 (cid:4) ) was generated during fragmentation by loss of a phosphate group (as phosphoric acid, 98 Da), which is a common characteristic of Ser/Thr phosphopeptides.

Positive transcription elongation factor b (P-TEFb) regulates eukaryotic gene expression at the level of elongation, and is itself controlled by the reversible association of 7SK RNA and an RNA-binding protein, HEXIM1 or HEXIM2. To further understand how P-TEFb is regulated, we analyzed the stoichiometry of all the known components of the large, inactive P-TEFb complex. Mutational analyses of a putative coiled coil region in the carboxyl-terminal portion of HEXIM1 revealed that the protein is a dimer in solution and remains a dimer after binding to 7SK. Although a HEXIM1 dimer contains two potential RNA binding motifs and ultimately recruits two P-TEFb molecules, it associates with only one molecule of RNA. The first 172 nucleotides of the 330-nucleotide 7SK are sufficient to bind HEXIM1 or HEXIM2, and then recruit and inhibit P-TEFb. Deletion of the first 121 amino acids of HEXIM1 allowed it to inhibit P-TEFb partially in the absence of 7SK RNA. Mutation of a conserved tyrosine (Tyr 271 in HEXIM1) to alanine or glutamate or mutation of a conserved phenylalanine (Phe 208 ) to alanine, aspartate, or lysine, resulted in loss of inhibition of P-TEFb, but did not affect formation of the 7SK⅐HEXIM⅐P-TEFb complex. Analysis of T-loop phosphorylation in Cdk9 indicated that phosphorylation of Thr 186 , but not Ser 175 , was essential for kinase activity and for recruitment of P-TEFb to the 7SK⅐HEXIM complex. A model illustrates what is currently known about how HEXIM proteins, 7SK, and P-TEFb assemble to maintain an activated kinase in a readily available, but inactive form.
Cyclin-dependent kinases (Cdks) 1 are key regulators of a variety of cellular processes, such as cell cycle progression, transcription, and neuronal differentiation. The kinase activity of Cdks in turn is tightly regulated. Association with a cyclin partner and phosphorylation of the T-loop is needed for activa-tion of Cdks. They are also subject to negative regulation via phosphorylation or through interaction with a family of Cdk inhibitory proteins (1)(2)(3)(4).
P-TEFb plays a key role in RNA polymerase II elongation control (5)(6)(7). It is comprised of one of two isoforms of Cdk9 (8,9) and one of three cyclins, T1, T2 (10), or K (11) in human. One of the major targets of the kinase activity of P-TEFb is the carboxyl-terminal domain of the largest subunit of RNA polymerase II (12), and this phosphorylation of the carboxyl-terminal domain by P-TEFb occurs during transcription elongation (13). P-TEFb controls gene expression by regulating the fraction of RNA polymerase II molecules that generate full-length mRNAs (6). In addition to its normal cellular role, P-TEFb has been shown to be recruited by the viral transactivator Tat to the promoter to enhance viral transcription, which is required for efficient HIV-1 replication (8, 14 -18). P-TEFb is uniquely regulated by the reversible association of a small nuclear RNA, 7SK (19,20), and HEXIM proteins (21)(22)(23)(24). Glycerol gradient analyses of cell lysates indicate that two forms of P-TEFb exist in the cell: a large inactive form containing 7SK and HEXIM proteins, and a smaller active form comprised of just P-TEFb subunits (9,19,21,23,24). When cells are treated with P-TEFb inhibitors, such as 5,6-dichloro-1-␤-Dribofuranosylbenzimidazole, or other agents that block transcription elongation, the large form is converted into the small active form (21). This form of P-TEFb regulation is physiologically significant, because it has been shown that all signals that trigger cardiac hypertrophy converge at the critical step of activating P-TEFb through dissociation of 7SK and HEXIM. This activation causes increased cellular transcription, and an increase in the size of cardiomyocytes (25)(26)(27). Several studies have uncovered some of the important interactions in the 7SK⅐HEXIM1⅐P-TEFb complex. Two regions of HEXIM1 have been characterized. The region centered upon KHRR (amino acids 152-155) is involved in binding of 7SK, and contains nuclear localization signals (21,28,29). A second region centered upon PYNT (amino acids 202-205) is involved in interaction with P-TEFb (23,29). In addition, regions involved in interactions have been narrowed down to amino acids 1-254 of 726 of cyclin T1, all of Cdk9, and nucleotide 1-175 of 7SK (19,21,22). Furthermore, phosphorylation of the T-loop of Cdk9 has been implicated in activation of P-TEFb, and is required for the formation of the 7SK⅐HEXIM1⅐P-TEFb complex (30).
In this report, we analyzed the stoichiometry of the 7SK⅐HEXIM1⅐P-TEFb complex. We identified residues critical for inhibition of the kinase activity of P-TEFb, and formation of the 7SK⅐HEXIM1⅐P-TEFb complex. These findings provide mechanistic insight into how P-TEFb kinase activity is controlled by 7SK RNA and HEXIM proteins.

Expression, Purification, and Mutagenesis of HEXIM and P-TEFb
Proteins-The His-tagged HEXIM1 wild type and mutant proteins described under "Results " were expressed in Escherichia coli BL21(DE3) cells by overnight induction with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside at 18°C. Purification on Ni-NTA resin was carried out as previously described (23). PCR-based site-directed mutagenesis was carried out with Pfu Ultra HF DNA polymerase (Stratagene) according to the manufacturer's instructions. Baculoviruses expressing human cyclin T1, wild type and mutant Cdk9s, as described under "Results, " were generated using the BaculoDirect Baculovirus Expression System (Invitrogen) according to the manufacturer's instructions. Purification of wild type and mutant P-TEFb was carried out as previously described (23). For the co-expression experiment, FLAG-tagged, wild type HEXIM1 was cloned into pACYCDuet-1 vector (Novagen), and Histagged ⌬N was cloned into pET21a (Novagen).
Glycerol Gradient Analysis-HeLaS3 cells, at 90% confluence in a T-75 flask, were transfected with pFLAG-CMV2-HEXIM1-2LR plasmid using Lipofectamine 2000 (Invitrogen). After 6 h the transfection media was removed, and the cells were cultured in Dulbecco's modified Eagle's medium-F12 with 10% fetal bovine serum under standard conditions for 48 h. Cells were scraped, spun down at 2,000 rpm, and then lysed for 15 min on ice in 150 mM NaCl, 2 mM MgCl 2 , 10 mM HEPES, 1 mM EDTA, 1 mM dithiothreitol, 1% phenylmethylsulfonyl fluoride, EDTA-free Complete protease inhibitor mixture from Roche, and 0.5% Nonidet P-40, and the lysates were clarified by centrifugation for 10 min at 14,000 rpm prior to fractionation on 5-ml 5-45% glycerol gradients in the same buffer conditions used during lysis, except that Nonidet P-40 was omitted. Gradients were run at 45,000 rpm for 16 h in a Beckman SW-Ti55 rotor before being fractionated.
Immunoprecipitations and Western Blotting-HeLa cells stably expressing FLAG-Cdk9 (provided by Dr. Zhou, University of California at Berkeley) or transiently transfected with pFLAG-CMV2-HEXIM1 were cultured in Dulbecco's modified Eagle's medium-F12 with 10% fetal bovine serum under standard conditions (37°C in 5% CO 2 ). Cell lysates and glycerol gradient analyses were performed as described above. Gradient fractions 4 and 9 (out of 16 fractions) were incubated with EZview TM Red ANTI-FLAG® M2 Affinity Gel (Sigma) for 2 h at 4°C. The beads were washed 3 times with 10 bead volumes of phosphate-buffered saline and 0.2% Tween 20 prior to suspension in SDS-PAGE loading buffer. Western blotting was carried out as previously described (23).
Electrophoretic Mobility Shift Assay (EMSA) and Native Gel Analysis-12-l Reactions were carried out in 25 mM HEPES, pH 7.6, 15% glycerol, 60 mM KCl, 0.1 mM EDTA, 5 mM dithiothreitol, 0.01% Nonidet P-40, 1 g of bovine serum albumin, 300 ng of poly(I)-poly(C) (Amersham Biosciences), and included 500 pg of radiolabeled 7SK RNA, recombinant P-TEFb comprised of Cdk9 and cyclin T1, and recombinant HEXIM1 or HEXIM2 as indicated. 7SK and poly(rI)-poly(rC) were heated individually for 5 min at 75°C and cooled on ice for another 5 min, prior to addition. Reactions were incubated at room temperature for 20 min and resolved on a 3.5% polyacrylamide (19 to 1, acrylamide: bisacrylamide ratio) gel in 0.5ϫ Tris glycine at 4°C for 1.5 h at 6 watts. The dried gel was subjected to autoradiography. For native gel analysis, the indicated HEXIM1 proteins in 60 mM HGKEDP were resolved on a 6% gel in 0.5ϫ Tris glycine at 4°C for 2 h at 6 watts. Proteins were visualized by silver staining.
Kinase Assay-16-l Kinase reactions containing purified recombinant wild type or mutant P-TEFb with Drosophila RNA polymerase II or human DSIF as the substrate were carried out in 30 mM KCl, 20 mM HEPES, pH 7.6, 7 mM MgCl 2 , 30 M ATP, 1.3 Ci of [␥-32 P]ATP (Amersham Biosciences), 1 g of bovine serum albumin per reaction, and the indicated amounts of wild type or mutant HEXIM1 proteins. T7-transcribed 7SK RNA was added last to the preincubation after it was heated for 5 min at 75°C and then cooled on ice for another 5 min. All reactions were incubated for 10 min at 23°C prior to the addition of ATP. The kinase reactions were then incubated for 20 min at 30°C and then stopped by the addition of SDS-PAGE loading buffer. Reactions were resolved by 9% SDS-PAGE. The dried gel was subjected to autoradiography and quantified with a Packard InstantImager.
Identification of Cdk9 Phosphorylation Sites by Mass Spectrometry-Recombinant P-TEFb (His-Cdk9/cyclin T1, ϳ2 g) was purified using Ni-NTA resin, separated by 12.5% SDS-PAGE, and stained with Coomassie Blue G-250. The Cdk9-containing band was excised from the gel and split into two pieces, one of which was fully trypsinized as described (31), and the other was incubated with trypsin for only 1 h at room temperature to allow partial digestion. The tryptic peptides were extracted and analyzed as previously reported (32) using an LCQ-DECA XP-Plus ion trap mass spectrometer (Thermo Finnigan, San Jose, CA).

RESULTS
HEXIM1 Is a Dimer-Sequence alignment of HEXIM1 and HEXIM2 proteins from different species revealed a relatively conserved region at their COOH termini, containing several invariant leucine residues (23). When this region of human HEXIM1 (amino acids 284 -318) was plotted on a helical wheel, a pattern of leucine residues characteristic of a leucine zipper coiled coil motif emerges (Fig. 1A). Because this motif is frequently involved in protein-protein interactions, we first examined if HEXIM1 could oligomerize via this region. Two proteins containing single mutations with either leucine 287 or 294 substituted with arginine (L287R, L294R), and a double mutant with both Leu 287 and Leu 294 replaced by arginine (2LR) were produced. All three mutant proteins had relatively similar mobility compared with wild type HEXIM1 when analyzed by SDS-PAGE (Fig. 1B). However, when analyzed on a native gel, L287R and 2LR had higher mobility than wild type and L294R (Fig. 1C). It is unlikely that the differences are because of the small fractional changes in charge. This is supported by two observations. Wild type and L294R have similar mobilities, but different charge, and L287R and L294R have the same charge, but different mobilities. It is likely that the differences in mobility are because of the disruption of oligomerization in the L287R and 2LR mutants. To confirm the hypothesis that the carboxyl-terminal domain allowed oligomerization, two truncation mutants were generated. A control protein, ⌬N, lacked the amino-terminal 120 amino acids, whereas ⌬C, lacked the carboxyl-terminal 77 amino acids that contained the entire leucine zipper motif (Fig. 1A). ⌬N had much higher mobility than ⌬C on the SDS-PAGE gel (Fig. 1B). In contrast, this difference in mobility was reversed on the native gel ( Fig.  1C), strongly suggesting the leucine zipper motif mediates oligomerization of HEXIM1.
The function of these mutants was compared with wild type HEXIM1 using a mobility shift assay and a kinase assay. HEXIM1 formed the expected complex with 7SK that was further shifted by association with P-TEFb (Fig. 1D). The ⌬C mutant, assayed at two concentrations, formed a complex with double the mobility of the wild type HEXIM1 complex, but was still able to recruit P-TEFb. Because ⌬C cannot dimerize, we assume that the predominate shift seen at the highest concentration contains a HEXIM1 monomer. The less severe mutant, 2LR, formed a 7SK complex with slightly lower mobility than ⌬C, and at a higher concentration also formed a complex with mobility similar to the wild type complex (Fig. 1D). Evidently, it binds as a monomer at low concentration, but can dimerize at higher concentrations. The 2LR dimer may be stabilized by interactions with 7SK. Both 2LR⅐7SK complexes were able to recruit P-TEFb (Fig. 1D). The effects of the HEXIM1 proteins on the kinase activity of P-TEFb were assayed in the absence and presence of 7SK. As expected, without 7SK the wild type and leucine zipper mutants had no effect on P-TEFb kinase activity (Fig. 1E). Supporting an idea stated earlier that removing the amino-terminal region might unmask P-TEFb binding or inhibitory regions of HEXIM1 (29), ⌬N inhibited P-TEFb activity down to 40% without 7SK. In the presence of 7SK all proteins inhibited P-TEFb, but the inhibition by proteins with reduced oligomerization (L287R, 2LR, and ⌬C) was less robust (Fig. 1E). The ability of ⌬N to inhibit P-TEFb was further stimulated by 7SK. Evidently the unmasking of the inhibitory regions by deletion of the first 120 amino acids is not complete, and is further stimulated by binding of 7SK. Evidence that HEXIM1 is a dimer came from a co-expression experiment. A FLAG-tagged, wild type HEXIM1 and a Histagged ⌬N were co-expressed in E. coli using two different plasmids with compatible replication origins ( Fig. 2A). Cell lysates were first subjected to Ni-NTA chromatography and FLAG-tagged wild type HEXIM1 co-eluted with His-tagged ⌬N, demonstrating that the two HEXIM1 proteins could stably interact (Fig. 2B). The material that eluted from the Ni-NTA column was further purified by FLAG affinity chromatography. Again, His-tagged ⌬N and FLAG-tagged wild type HEXIM1 co-eluted. Unlike the material that eluted from the Ni column, which contained more His-tagged protein than FLAG-tagged protein, equal amounts of each protein eluted from the FLAG antibody column (Fig. 2B). This 1 to 1 ratio indicated that HEXIM1 was at least a dimer. Because similar results would have been obtained if HEXIM1 formed a trimer, tetramer, or higher order oligomer, the oligomerization state was determined by comparing the mobilities of the two homo-oligomers with the hetero-oligomer on a native gel. If HEXIM1 dimerizes, then only one band with intermediate mobility should be given by the hetero-oligomer. IF HEXIM1 trimerizes, then two bands with intermediate mobility should be seen (one with 2 large and 1 small protein and one with 2 small and one large pro-tein). Silver staining of the native gel demonstrated that the hetero-oligomer gave only one band of intermediate mobility between the two homo-oligomers, clearly indicating that HEXIM1 forms dimers (Fig. 2C). To determine whether HEXIM1 binds to 7SK as a dimer, the same logic was applied to the results of a mobility shift assay. The mobility of 7SK complex with the heterodimer gave a single band that was intermediate between the two homodimers (Fig. 2D). In addition, all three complexes were able to recruit P-TEFb (Fig. 2D). Therefore, HEXIM1 maintains its dimerization state during association with 7SK and P-TEFb.
Oligomerization of HEXIM1 was also analyzed in vivo using HeLa cells transfected with FLAG-tagged HEXIM1 expressing constructs. Lysates from cells expressing FLAG-tagged wild type HEXIM1 were subjected to glycerol gradient sedimentation and fractionation to separate free HEXIM1 from HEXIM1 in the 7SK⅐HEXIM1⅐P-TEFb complex. Immunoprecipitation experiments were carried out with anti-FLAG antibodies using fraction 4 containing free HEXIM1 or fraction 9 containing HEXIM1 in the large complex. Using anti-FLAG antibodies, endogenous HEXIM1 protein co-precipitated with the FLAG HEXIM1 from both fractions (Fig. 3A). Thus we conclude that HEXIM1 oligomerizes regardless of its association with 7SK and P-TEFb in vivo. Cdk9 was present in both fractions, but, as expected, it only immunoprecipitated using FLAG antibodies with HEXIM1 in fraction 9 containing the large P-TEFb complex. HeLa cells were also transfected with a plasmid containing the FLAG-tagged 2LR mutant, and the cell lysate analyzed by glycerol gradient sedimentation. The distribution of FLAGtagged 2LR mutant was predominantly in fractions 1-3, in contrast to the distribution of the endogenous HEXIM1, which was fractions 2-4 (Fig. 3B). The difference in sedimentation likely reflects the loss of oligomerization by the 2LR mutant. FIG. 2. HEXIM1 is a dimer in vitro. A, diagram of FLAG-tagged full-length HEXIM1 (F) and His-tagged HEXIM1 mutant lacking the amino-terminal 120 amino acids (H) that were expressed in E. coli as described under "Experimental Procedures" and in the text. B, purification of hybrid HEXIM1 from lysates of cells co-expressing both proteins by Ni-NTA chromatography (Ni), followed by FLAG affinity chromatography (FLAG). Material eluting from each column was analyzed by SDS-PAGE and silver staining. C, silver-stained native gel of FLAGtagged wild type HEXIM1 (F/F), His-tagged ⌬N (H/H), and HEXIM1 hybrid (F/H). D, EMSA of 7SK with F/F, H/H, or F/H in the absence or presence of P-TEFb as indicated. Although we have not rigorously proven that HEXIM1 is a dimer, rather than a trimer or higher oligomer in vivo, the sedimentation of wild type free HEXIM1 (40 kDa) is consistent with it being a dimer (80 kDa) because it sediments slower than free P-TEFb (120 kDa).
A similar immunoprecipitation experiment was carried out on cells expressing FLAG-tagged Cdk9. Anti-FLAG antibodies brought down only FLAG-tagged Cdk9 from a glycerol gradient fraction containing the free form of P-TEFb (Fig. 3C). However, FLAG-tagged Cdk9 and untagged, endogenous Cdk9, as well as HEXIM1, were pulled down from a fraction containing the large form of P-TEFb (Fig. 3C). These results suggest that free P-TEFb does not dimerize in the cell, but that the large complex contains two P-TEFb molecules.
The 7SK⅐HEXIM⅐P-TEFb Complex Contains One 7SK Molecule-We examined the stoichiometry of 7SK in complexes formed with HEXIM proteins. A 330-nucleotide full-length 7SK and a truncated RNA containing the first 172 nucleotides of 7SK were synthesized by in vitro transcription. The abilities of the full-length and truncated 7SK to support HEXIM1-mediated inhibition of P-TEFb in vitro were analyzed. Neither RNA inhibited P-TEFb alone, but when mixed with equal molar amounts of HEXIM1 both RNAs produced a dose-dependent inhibition down to about 10% (Fig. 4A). Evidently, only the first half of 7SK is needed for a functional interaction with HEXIM1 and P-TEFb. Both RNAs were able to form complexes with HEXIM1 and subsequently recruit P-TEFb (Fig. 4B). To determine whether more than one 7SK would bind to a HEXIM1 dimer, the two versions of 7SK were mixed together before adding HEXIM1 and then analyzed by EMSA. The shifts that were generated were identical to the shifts seen with the individual RNAs (Fig. 4B). Identical results were obtained when HEXIM1 was replaced by a functionally similar protein, HEXIM2 (23) (Fig. 4C). Because there was no detectable intermediate complex we conclude that there is only one 7SK in the complex.
Tyrosine 271 and Phenylalanine 208 of HEXIM1 Are Critical for Inhibition of P-TEFb Kinase Activity-The sequence alignment of HEXIM proteins across many species revealed five invariant tyrosine residues corresponding to amino acids 167, 203, 271, 274, and 291 of human HEXIM1. Previously, we showed that Tyr 203 is located in the region of HEXIM1 important for the recruitment of P-TEFb to the 7SK⅐HEXIM1 complex (29). To examine potential roles of the other tyrosines, mutant HEXIM1 proteins containing Y167E, Y271E, Y274E, or Y291E were generated and tested for their ability to inhibit P-TEFb kinase activity in the presence of 7SK. Only Y271E lost most of its inhibitory properties (data not shown). To examine the role of Tyr 271 in detail, Y271A and Y271F mutants were also generated, and the abilities of the three mutants to inhibit P-TEFb in the presence of 7SK were compared with wild type HEXIM1. Y271A and Y271E did not support robust inhibition of P-TEFb, but Y271F behaved the same as the wild type (Fig.  5A). Interestingly, all four proteins were able to form complexes with 7SK and recruit P-TEFb (Fig. 5B). These results indicate that it is critical for inhibition of P-TEFb to maintain an aromatic residue at position 271, but that this residue is not essential for the recruitment of P-TEFb to the 7SK⅐HEXIM1 complex. A similar analysis of a conserved phenylalanine, Phe 208 , was carried out. F208A, F208D, and F208K mutants were generated, and the purified proteins were used in a kinase assay and electrophoretic mobility shift assay to determine their ability to inhibit P-TEFb and to recruit P-TEFb to a 7SK⅐HEXIM1 complex. All three mutants displayed a significantly reduced ability to inhibit P-TEFb in the presence of 7SK, but all mutants, except F208K, maintained their ability to recruit P-TEFb to the 7SK⅐HEXIM1 complex (Fig. 5, C and D). We conclude that both Tyr 271 and Phe 208 are critical for inhibition of P-TEFb, but have only a minimal influence on the ability of HEXIM1 to associate with P-TEFb.
Phosphorylation of Threonine 186 of Cdk9 Is Required for P-TEFb Kinase Activity and Formation of the 7SK⅐HEXIM1⅐P-TEFb Complex-To examine the role of phosphorylation of the T-loop of Cdk9, recombinant P-TEFb proteins containing cyclin T1 and wild type, S175A, S175D, T186A, or T186E mutations in Cdk9 were purified individually (Fig. 6A). Because Cdk9, but not cyclin T1 was His-tagged, and equal molar amounts of Cdk9 and cyclin T1 were purified, none of the mutations had an effect on association of the cyclin. Increasing amounts of each purified protein were subjected to a direct kinase assay using the large subunit of DSIF (SPT5) as a substrate. The wild type kinase and the two S175 mutants had similar high levels of activity (Fig. 6B). P-TEFb containing Cdk9 with a T186A mutation had less than 10% of the activity of the wild type kinase. This leads to the assumption that Thr 186 phosphorylation occurs to some extent during expression of P-TEFb in baculovirus-infected insect cells. Introduction of a negative charge to mimic phosphorylation (T186E) increased the kinase activity about 3-fold from what was seen with the T186A mutant (Fig.  6B). This suggests that phosphorylation of Thr 186 of Cdk9 is required for efficient kinase activity of P-TEFb. The mobility shift assay was then used to look at association of the P-TEFb mutants with 7SK⅐HEXIM1 (Fig. 6C) or 7SK⅐HEXIM2 (Fig.  6D). A dose-dependent shift was seen for the wild type and Ser 175 mutants (Fig. 6, C and D). P-TEFb mutants containing the T186A or T186E displayed a greatly reduced affinity for 7SK⅐HEXIM1 and 7SK⅐HEXIM2 (Fig. 6, C and D). At the highest level of P-TEFb added, a second P-TEFb shift was visible with 7SK⅐HEXIM1 complexes and was clearly present with 7SK⅐HEXIM2 complexes. This supports the finding that two P-TEFb molecules were found in the large 7SK⅐HEXIM1⅐P-TEFb complex in vivo (Fig. 3C). All these results indicate that phosphorylation of Thr 186 , but not Ser 175 is needed for kinase activity and for association with the 7SK⅐HEXIM complex.
Tandem mass spectrometry (MS/MS) was used to confirm that Cdk9 in recombinant P-TEFb is phosphorylated on Thr 186 . Cdk9 protein was trypsinized and then fractionated by capillary reverse phase high performance liquid chromatography. Eluted peptides were ionized and transferred into an on-line mass spectrometer, where they were further separated based on mass to charge ratio (m/z). The peptide ions detected were then selected sequentially and fragmented to generate specific MS/MS spectra containing sequence information. This is possible because MS/MS is a high-resolution separation tool to physically isolate peptide ions according to their m/z value despite the presence of many other co-eluting peptides. After a data base search, one Cdk9 phosphopeptide with a miscleavage (NS 180 QTNRYT 186 NR) was identified (Fig. 6F). Partial tryptic digestion increased the yield of this peptide. Assignment of product ions clearly indicated that the phosphorylation site was Thr 186 not Ser 180 (Fig. 6G), although the latter is conserved in Cdk7. The ion signal of this phosphopeptide was remarkably high (Fig. 6F) and was comparable with other major peptide ion signals (data not shown), suggesting Thr 186 is a major phosphorylation site on Cdk9. Although negative MS/MS data is not conclusive, we did not detect any phosphopeptide containing modified Ser 175 , suggesting that it may not contribute significantly to Cdk9 activation. DISCUSSION In this report, we provide evidence that HEXIM1 is a dimer both when it is free and when it is complexed with 7SK and P-TEFb, and that the 7SK⅐HEXIM1⅐P-TEFb complex contains one 7SK molecule and two P-TEFb molecules. We also identified conserved residues Tyr 271 and Phe 208 as critical for the inhibition of P-TEFb kinase activity, but not the formation of the 7SK⅐HEXIM1⅐P-TEFb complex. Furthermore, we demonstrated that phosphorylation of Thr 186 in the T-loop of Cdk9 is required for the P-TEFb kinase activity, and for P-TEFb to associate with the 7SK⅐HEXIM1 complex.
Of the six functional domains in HEXIM proteins (Fig. 7A), three are involved in formation of the 7SK⅐HEXIM1⅐P-TEFb complex. Previously, an RNA binding domain containing the sequence KHRR and an adjacent nuclear localization signal were identified (21,28,29). Also a number of residues in a third domain, the P-TEFb binding domain, were identified (29). Mutations in HEXIM1 (Y203A, Y203D, T205A, T205D, Y203D/ T205D) and the residue in HEXIM2 corresponding to Thr 205 in HEXIM1 (T143A, T143D) blocked the recruitment and inhibition of P-TEFb by the 7SK⅐HEXIM complex (23,29). During the preparation of this manuscript, Yik et al. (24) reported that HEXIM1 and HEXIM2 could form homo-and hetero-oligomers in transfected HeLa cells (24), but the domain responsible was not identified. We found that the oligomerization domain re- sided in a putative leucine zipper coiled coil domain in the carboxyl-terminal region of HEXIM1. Our results demonstrate that HEXIM1 is a dimer in vitro when it is free or in complex with 7SK and does not change upon binding P-TEFb. Our results in vivo indicate that free and complexed HEXIM1 is an oligomer. Although our results suggest that HEXIM1 is also a dimer in vivo, trimerization or higher order oligomerization cannot be ruled out. Yik et al. (24) showed that HEXIM1 and HEXIM2 can oligomerize when HEXIM2 is overexpressed in HeLa cells. This hetero-oligomerization cannot be merely driven by the concentrations of the two proteins, because we found that HEXIM1 and HEXIM2 partition differently into free and large P-TEFb complexes in two cell lines (23) and this would not be expected if homo-and hetero-oligomerization were equally favored. Oligomerization may play a role in stabilizing HEXIM1 in vivo. During transient expression of HEXIM1 proteins in HeLa cells, the levels of dimerization defective mutants (L287R and 2LR) were much lower than wild type or mutant HEXIM1s that supported dimerization in vitro. 2 Our studies here have identified residues of HEXIM1 that may be involved in the mechanism of inhibition of P-TEFb. Mutation of a conserved Tyr 271 in the SETYER region of HEXIM1 greatly reduced the ability of HEXIM1 to inhibit P-TEFb in the presence of 7SK, but did not affect the recruitment of P-TEFb into the 7SK⅐HEXIM1 complex. Similar results were obtained with mutations of a highly conserved phenylalanine at position 208 (F208A, F208D, F208K). The biochemical phenotype of these mutations distinguishes them from mutations in the PYNT region that block recruitment of P-TEFb. Although separated in the linear sequence, Tyr 271 and Phe 208 may be part of a P-TEFb inhibitory domain that is discussed below.
With all the information available now, it is possible to construct a more detailed model of how P-TEFb is controlled by association with 7SK and HEXIM proteins (Fig. 7B). From our experiments expressing P-TEFb with Cdk9 mutants in insect cells, assembly of a P-TEFb heterodimer can occur in the absence of T-loop phosphorylation (Fig. 6A). P-TEFb het-2 Q. Li, unpublished data. FIG. 6. Phosphorylation of Thr 186 in Cdk9 is required for P-TEFb kinase activity and the formation of the 7SK⅐HEXIM1⅐P-TEFb complex. A, purified P-TEFb kinases containing wild type Cdk9 (WT), or Cdk9 with mutations of Ser 175 (S175A, S175D) and Thr 186 (T186A, T186E). Protein concentrations were normalized as revealed by silver staining of the SDS-PAGE gel. B, effects of these mutations on the kinase activity of P-TEFb using human DSIF as the substrate. Label is incorporated into the large DSIF subunit (SPT5). Three different concentrations of each P-TEFb (ϫ1, 3, and 10) were used. C, EMSA analysis using 7SK and HEXIM1 with the indicated P-TEFb mutants. The first lane (0) is 7SK plus HEXIM1. Note that the region of the gel in which the free RNA runs was deleted from the image. D, same as C except HEXIM1 was replaced by HEXIM2 (H2). E, recombinant P-TEFb (Cdk9/cyclin T1) was resolved on a SDS gel, and the band containing Cdk9 was excised followed by partial trypsin in-gel digestion. The resulting peptides were harvested and analyzed by reverse phase liquid chromatography (HPLC)-tandem mass spectrometry (LC-MS/MS). One peptide ion eluted with a retention time (RT) of 11.9 min with a very high signal to noise ratio (S/N). F, the MS survey scan at the point of 11.9 min, indicated the presence of the peptide ion of m/z 665.6. The ion ranges from m/z 665.0 to 668.0 because of the naturally occurring isotopic distribution and this mass range was used in F to extract ion current signal. A partial region (m/z 500 -1000) is shown for simplicity, although the entire scan range was from m/z 300 to 1600. G, MS/MS scan of the precursor ion m/z 665.6, which was fragmented into multiple labeled product ions (b and y ions) led to the identification of a phosphopeptide with a modification site on Thr 186 according to the mass shift (ϩ80 Da) caused by phosphorylation. A predominant product ion ([M-H 3 PO 4 ϩ 2H] 2ϩ ) was generated during fragmentation by loss of a phosphate group (as phosphoric acid, 98 Da), which is a common characteristic of Ser/Thr phosphopeptides. erodimers were formed with wild type Cdk9 or with Cdk9 carrying T-loop mutations T186A, T186E, S175A, or S175D, but only wild type and S175A and S175D mutants were completely active and the T186E mutant was partially active (Fig.  6B). This and the MS/MS analysis strongly points toward Thr 186 , not Ser 175 , as the critical site of activating phosphorylation. Our results differ somewhat from those obtained in the Zhou laboratory (30) who found that Cdk9 expressed in HeLa cells containing T186A, T186E, and S175A mutations had no activity, and that a S175D mutant had full kinase activity. Using recombinant P-TEFb in electrophoretic mobility shift assays, we found that only kinases that were activated (wild type, S175A, and S175D) associated efficiently with the 7SK⅐HEXIM complex containing either HEXIM1 or HEXIM2 (Fig. 6, C and D). P-TEFb containing the partially activating mutation T186E had a low but detectable binding to the 7SK⅐HEXIM complex. HEXIM1 and, presumably, HEXIM2 are normally found as dimers that do not interact with or inhibit P-TEFb (Figs. 2 and 3). The binding to the first 172 residues of a single molecule of 7SK results in a conformational change that enables HEXIM to bind to and inhibit P-TEFb (Fig. 4). Because removal of the first 120 amino acids of HEXIM1 enable it to inhibit P-TEFb in the absence of 7SK (Fig. 1E), we propose that the amino-terminal domain is an autoinhibitory domain that masks the P-TEFb binding domain, and that binding of 7SK results in a conformational change that unmasks the P-TEFb binding domain. The presence of an autoinhibitory domain was previously suggested to explain the ability of a HEXIM1 mutant lacking the first 180 amino acids to associate with Cdk9 with or without 7SK (29). During that study we were unable to demonstrate 7SK independent inhibition of P-TEFb for HEXIM1 (181-359). 2 Perhaps that large deletion damaged the ability of HEXIM1 to inhibit P-TEFb without significantly reducing its ability to bind to P-TEFb. Finally, our results indicate that the large inactive form of P-TEFb contains one 7SK RNA, a HEXIM dimer, and two P-TEFb molecules. If no other proteins are present then the total mass of the large complex would be 430 kDa. This is in close agreement with the size estimates from glycerol gradients (19).
It is possible that HEXIM1 shares mechanistic details with the Cdk2⅐cyclin A inhibitor, p27 kip . Like p27 kip , HEXIM1 binds to the cyclin partner of the Cdk it inhibits (29). Binding of p27 kip to cyclin A positions several aromatic residues (tyrosine and phenyalanine) in the catalytic cleft of Cdk2 and blocks the interaction of Cdk2 and ATP, thereby, inhibiting kinase activity (33). We found that mutation of Tyr 271 to phenylalanine (Y271F) allowed continued function of HEXIM1, but substitution of alanine or glutamate significantly reduced the inhibitory function without significantly affecting the interaction with P-TEFb. HEXIM1 Phe 208 mutants also blocked inhibition without disrupting the interaction with P-TEFb. Our results are consistent with Tyr 271 and/or Phe 208 blocking ATP binding to Cdk9. Another similarity between HEXIM1 and p27 kip is that both bind to the phosphorylated (activated) form of the kinases they inhibit. The function of HEXIM1 may be more similar to the p27 kip family than to the INK4 family that bind to Cdks and weaken the interaction of the cyclin and Cdk and distort that active site (34).
Although some details of the interactions of the components of the large inactive P-TEFb complex are known, it is still not clear how the activated P-TEFb stored there is released. Control of release of P-TEFb by removal of the phosphate from Thr 186 is counter-intuitive because even though this might cause dissociation, the kinase would be inactive. We favor a model in which the association of HEXIM proteins with 7SK is regulated. Dissociation of 7SK might allow the inhibitory domain of HEXIM proteins to remask the P-TEFb binding domain, and allow P-TEFb to be free and active. Further studies are needed to uncover factors that are responsible for the regulated assembly and disassembly of the large inactive P-TEFb complex.