|Home | About | Journals | Submit | Contact Us | Français|
The cyclin T/CDK9 complex, also called positive transcription elongation factor b (P-TEFb) phosphorylates the C-terminal domain of the large fragment of the RNA polymerase II. This action is a hallmark of the transition from transcription initiation to elongation. P-TEFb is itself modified by phosphorylation and ubiquitination. Recently, the core components of P-TEFb, cyclin T1 and CDK9, were identified as novel substrates of histone acetyltransferases. Here, we review how posttranslational modifications regulate the activity of the P-TEFb complex and discuss how acetylation of the complex optimizes transcription elongation in the context of other posttranslational modifications.
Transcription occurs when the RNA polymerase II (RNAPII) moves away from the promoter of a gene and begins generating an mRNA, leaving behind several general transcription factors that are essential for the formation of the pre-initiation complex. At certain genes, the enzyme moves about 20–40 nucleotides along the DNA template, then pauses and waits for signals that promote the formation of a fully mature elongation complex.1 The formation of this complex depends critically on the kinase activity of P-TEFb, a heterodimeric complex composed of cyclin T proteins and CDK9. Promoter-proximal pausing of RNAPII is caused by the association of two negative elongation factors, DRB-induced stimulating factor (DISF) and the negative elongation factor (NELF), and is considered a checkpoint to assess whether the polymerase is correctly prepared for productive elongation.2,3 Pausing has also emerged as a major control mechanism to regulate expression of rapidly induced genes in response to stimuli, such as heat shock, developmental cues, and the activation of T cells and macrophages.4-10
P-TEFb phosphorylates a structurally flexible C-terminal domain (CTD) within the largest fragment of the initiated but elongation-incompetent RNAPII complex (Fig. 1). The CTD contains characteristic heptad repeats with the consensus amino acid sequence YSPTSPS, in which serines 2, 5 and 7 (S2, S5, and S7) can be dynamically phosphorylated.11 Phosphorylation of S5 occurs during transcription initiation, while phosphorylation of S2 marks the mature elongation complex and is mediated by P-TEFb. P-TEFb also phosphorylates DSIF and NELF, relieving their negative effect on transcription elongation.2,3,12,13 In addition, P-TEFb plays important roles in the synthesis, processing and transport of mRNA14 and is an essential cellular cofactor of the human immunodeficiency virus (HIV).15 The HIV transactivator Tat interacts with the cyclin T1 subunit of P-TEFb and recruits P-TEFb directly to elongating HIV transcripts.15,16
The activity of P-TEFb also regulates a hierarchy of histone modifications in a CTD-dependent and -independent manner.17 Although promoter regions where transcription initiates are generally nucleosome-free, nucleosomes can obstruct the path of the elongating RNAPII. During transcription elongation, histones undergo multiple posttranslational modifications, including acetylation, methylation, phosphorylation and ubiquitination.18 The activity of P-TEFb was linked to monoubiquitination of histone H2B and trim-ethylation of histone H3 (K4 and K36).17 A combined mark on histone H3 (acetylated K9 and phosphorylated S10) and H4 (acetylated K16) was recently identified as the signal that recruits P-TEFb to actively transcribed genes through the interaction with its binding partner Brd4.20
Here, we focus on the emerging role of posttranslational modifications of the P-TEFb complex itself as a novel mechanism to regulate the P-TEFb kinase activity and transcription elongation. Modifications described for members of the P-TEFb complex include phosphorylation, ubiquitination, and acetylation and establish a novel level of regulation in the intricate interplay between modifications at the RNAPII complex, chromatin, and the regulation of transcription elongation.
The core component of P-TEFb is a heterodimer formed of cyclin T proteins and CDK9. Cyclin T1 is a 87-kDa protein found in ~80% of human P-TEFb complexes.21 Other complexes contain cyclin T2A and B, both splice variants of the same gene.22 Cyclin T2A and T2B are expressed widely in tissues of adult mice, and their expression is critical for embryogenesis as cyclin T2-/- mice die in utero.23 Notably, only P-TEFb-containing human cyclin T1 proteins can bind the HIV Tat protein to support HIV transcription.16 Human cyclin T1 contains a Tat recognition motif (TRM) of 18 amino acids, located just downstream of the N-terminal cyclin box that interacts with CDK924 (Fig. 2A). The cyclin-box and the TRM are also involved in the interaction with the inhibitory protein Hexim1 and 7SK small nuclear (sn) RNA.19,25-27 Other domains of cyclin T1 include a highly structured helical domain (amino acids 379–430), which in cyclin T1, but not cyclin T2, is predicted to form a coiled-coil domain. Downstream of the coiled-coil domain lies the interaction domain for the cofactor Brd4 (amino acids 426–516) and a histidine-rich domain (amino acids 506–530), which can directly interact with the polymerase CTD.28,29 The C-terminal PEST domain of cyclin T1 (amino acids 707–726) acts as an interaction domain for F-box proteins, determining substrate recognition by SCF E3 ligases (i.e., p19skp1, CDC53/cullin and F-box containing protein).75
The main form of CDK9 is a 42-kDa protein with ~40% overall sequence identity with CDK2 (Fig. 2B). In structural studies, CDK9 shows a typical kinase fold, comprising the N-terminal lobe (residues 16–108), which consists mainly of a beta-sheet with one alpha-helix, and a C-terminal lobe (residues 109–330) composed of alpha-helices.31 The protein contains a phosphorylated T loop structure (amino acids 168–197) that is conserved among CDK proteins and controls access of ATP and substrate to the enzyme.32,33 A 55-kDa isoform of CDK9 contains a 117-amino acid extension within the N-terminal region.34,35 The two forms of CDK9 are transcribed from two different promoters in the CDK9 gene. Their relative abundances have been studied in macrophages after treatment with lipopolysaccharides.34
Early findings indicated that the majority of cyclin T1/CDK9 is not present at intra-cellular sites containing RNAPII.36 More than half of cellular P-TEFb is estimated to be sequestered in a large inactive ribonucleoprotein complex, from where active cyclinT1/CDK9 can be rapidly mobilized (i.e., in response to stress).25,37-39 Within this inactive complex, cyclin T1 interacts with 7SK snRNA, a very abundant and evolutionally conserved transcript of 331 nucleotides.37,39 Cyclin T1 interacts with the 3'-terminal hairpin in 7SK snRNA, while the 5'-terminal hairpin is occupied by Hexim1.27,40 Hexim1 was originally identified as a protein induced when vascular smooth muscle cells were treated with hexamethylene bis-acetamide (HMBA).41,42 Binding to 7SK snRNA transforms Hexim1 into a potent inhibitor of the P-TEFb kinase activity.25,27,38 Other proteins found in the 7SK ribonucleo-protein complex include Hexim2, which has redundant activities in P-TEFb binding and inhibition with Hexim1 but is expressed in different tissues, methylphosphate capping enzyme (MePCE, previously identified as BCDIN3), P-TEFb interaction protein for 7SK snRNA stability (PIP7S; a lupus-associated antigen-related protein) and heterogeneous ribonuclear protein (hnRNP) A1, A2, R, Q and RNA helicase A.43
Hexim1 binds to 7SK snRNA via a highly basic region (BR; amino acids 150–177), which is also part of the nuclear localization motif of Hexim1 (Fig. 2C).27,44 The protein also binds directly to cyclin T1 via the C-terminal domain of Hexim1 (amino acids 181–359).25,27,38 In this C-terminal domain, a highly conserved PYNT motif (amino acids 202–205) is involved in the interaction with cyclin T1.27 Two acidic regions (AR1 and AR2) (amino acids 211–249) interact electrostatically with the basic region in the absence of 7SK snRNA, thereby blocking Hexim1 interaction with P-TEFb.44
Two coiled-coil regions (CR) (amino acids 279–352) mediate homodimerization of Hexim1.26,45 CR1 (amino acids 279–315) also binds cyclin T1 directly through electrostatic interactions involving E290, E295 and E302 in CR1.26,45 Because Hexim1 homodimerizes through its coiled-coil regions, it was proposed that two Hexim1 molecules and one 7SK snRNA associate with two cyclin T1/CDK9 heterodimers to form the inactive P-TEFb complex.45,46 Another study found a 1:1:1 ratio for CDK9, cyclin T1 and Hexim1 molecules in the large P-TEFb complex.47
After dissociation from the 7SK ribonucleoprotein complex, the cyclin T1/CDK9 heterodimer can bind to the BET proteins Brd2, Brd4 and BrdT.48 These proteins contain two bromodomains (amino acids 58–169 and 349–461) and a downstream extraterminal (ET) domain (amino acids 600–678). While bromodomains are bona fide recognition motifs for acetylated lysines, the function of the ET domain is not known.49 BET proteins contain a highly conserved C-terminal helical structure (amino acids 1209–1400) that interacts with P-TEFb and the human papilloma virus E2 protein.48,50 Brd4, which was originally identified as a component of the mammalian Mediator complex, has emerged as a critical tethering factor of P-TEFb to actively transcribed genes, including the HIV promoter.28,48,51,52 The bromodomains in Brd4 bind acetylated lysine residues in histones H3 and H4.53-56 The combination of acetylated K9 in histone H3 and acetylated K16 in histone H4 together with phosphorylated S10 in histone H3 serves as a specific recognition code for Brd4 within actively transcribed genes.20
The Brd4 bromodomains can also interact with P-TEFb directly at amino acids 426–516 in cyclin T1.28,52 The second bromodomain in Brd4 interacts in vitro with acetylated K390 in cyclin T1, but the in vivo relevance of this interaction remains unclear.57
Cyclin T1, CDK9 and Hexim1 are subject to reversible phosphorylation. Human cyclin T1 was originally identified as a cellular phosphoprotein that associates specifically with the transactivation domain of the HIV Tat protein.16 Autophosphorylation experiments with purified cellular P-TEFb showed that cyclin T1 is likely a substrate of CDK9.58 A C-terminally truncated cyclin T1 protein was not phosphorylated in in vitro auto-phosphorylation assays, indicating that CDK9 phosphorylates residues beyond amino acid 298 of cyclin T1.31,59 The precise sites of phosphorylation in cyclin T1 and their biological significance remain unknown (Fig. 2A).
Hexim1 was recently identified as a target of the AKT/protein kinase B within the phosphoinositide-3-kinase (PI3K) signaling pathway.60 Phosphorylated residues were mapped to T270 and S278, both residing in the cyclin T-binding domain of Hexim1 critical for the interaction with P-TEFb (Fig. 2C). Hexim1 phosphorylation by PI3K/AKT was activated in response to treatment with HMBA and led to the dissociation of cyclin T1/CDK9 from the large inactive P-TEFb complex activating HIV transcription in latently infected T cells.
Treatment with HMBA also regulates phosphorylation of CDK9.19,46,61 Known phosphorylation sites in CDK9 include T29,62 S175,19,52 T186,19,31,61 and S347, T362 and T363,31 (Fig. 2B). T186 is located at the tip of the conserved T-loop in CDK9 (homologue to T162 in CDC2 and T160 in CDK2) and is specifically dephosphorylated after treatment with HMBA or ultraviolet irradiation when P-TEFb dissociates from the 7SK ribonucleoprotein complex.61,67 T186-specific phosphatases are the calcium-sensitive and calmodulin-activated serine/threonine phosphatase PP2B and the alpha subunit of protein phosphatase 1 as well as manganese- or magnesium-dependent protein phosphatase 1A and 1B (PPM1A and PPM1B).61,67 Both PP2B and PP1α are required for successful dephosphorylation of CDK9; it is believed that PP2B induces conformational changes in the large P-TEFb complex that allow subsequent dephosphorylation of T186 by PP1α.61 PPM1A and PPM1B were identified based on a candidate approach; they also dephosphorylate the T loop of CDK2 and CDK6 during cell cycle progression.63,64,67
The kinase that phosphorylates T186 in CDK9 is unknown. In in vitro phosphorylation assays, T186 in CDK9, together with S347, S362 or S363, is a target for the CDK9 autophosphorylation activity, but whether the site is also autophosphorylated in vivo remains unknown.31 Interestingly, activation of purified CD4+ T cells rapidly increases phosphorylation of T186 in cellular CDK9 potentially by activating the T186 kinase.65
Phosphorylation of the T loop is considered essential for maximal CDK activation; it induces a conformational change of the T loop, allowing entry of the substrate and ATP into the CDK catalytic pocket.66 However, CDK9 phosphorylated at T186 is mainly found within the inactive P-TEFb complex.61 There it could serve to keep CDK9 poised for rapid activation upon dissociation from Hexim1 and 7SK RNA.19,61 Structural studies show that phosphorylated T186 in CDK9 plays less of an organizational role than comparable residues in other CDKs since it does not make contact with the cyclin component.31 Nevertheless, it can promote an active conformation of the T loop. In vitro assembly of the 7SK ribonucleoprotein complex depends on phosphorylation of T186, underlining the critical role of this modification for the inactivation of P-TEFb.46,61,67 Indeed, the finding that T186 undergoes dephosphorylation during P-TEFb activation after HMBA or ultraviolet treatment supports the model that phosphorylation of T186 in CDK9 plays a specific, yet unknown role in the assembly or maintenance of the inactive P-TEFb complex.
In contrast, phosphorylation of S175 is regarded a positive mark in CDK9 and promotes interaction between P-TEFb with the cofactor Brd4.52 The kinase that phosphorylates S175 in CDK9 and the mechanisms of how phosphorylation of CDK9 enhances binding of P-TEFb to Brd4 are not known. Since Brd4 is not known to contact CDK9 directly, phosphorylation of S175 might induce conformational changes in P-TEFb, allowing Brd4 to bind cyclin T1 more efficiently. S175 in CDK9 is conserved in CDK7, and phosphorylation of this residue is important for CDK9-mediated phosphorylation of the RNAPII CTD.19,52,68,69 Mutation of S175 to alanine inhibited CDK9 kinase activity as expected, but mutation to aspartic acid, mimicking a phosphorylated serine, did not have the expected activating effect.19,52 Both mutations interfered with the binding of P-TEFb to Brd4 and weakened the in vivo association of CDK9 with the HIV promoter.52
Autophosphorylation of CDK9 appears to be dynamically regulated during the transition from transcription initiation to elongation, pointing to a positive role of CDK9 autophosphorylation during transcription elongation. Activation of auto-phosphorylation depends on release of the general transcription factor TFIIH, which in in vitro autophosphorylation assays directly suppresses CDK9 autophosphorylation via its XPB component.70 Although the sites of TFIIH-sensitive autophosphorylation in CDK9 have not been mapped, mass spectrometry of in vitro autophosphorylated P-TEFb identified T186 and S347 together with T362 or T363 as autophosphorylated residues.31 CDK9 autophosphorylation at the HIV promoter is induced by the HIV Tat protein when TFIIH is released.70 This process may support binding of the Tat protein to HIV transcripts and could allow CDK9 to enter the complex formed by Tat, TAR RNA and cyclin T1.59,71 Autophosphorylated residues important for inclusion of CDK9 in in vitro formed Tat/TAR/P-TEFb complexes were mapped to amino acids 345–358 in the C-terminus of CDK9, a region containing the confirmed autophosphorylation site S347.31,59 Interestingly, these residues are not conserved among other CDK enzymes.
In contrast, a homology search with other CDK enzymes identified T29 as a target of autophosphorylation in CDK9.62 This site is homologous to T14 and T15 in CDC2 and CDK2, respectively, and similar to the phosphorylation of these sites, phosphorylation of T29 inhibits CDK9 kinase activity.62,72,73 Surprisingly, although the Tax transactivator of the human T-lymphotropic virus type 1 (HTLV-1) relies on P-TEFb to activate HTLV-1 transcription, purified Tax enhanced autophosphorylation of T29 and diminished CDK9-mediated phosphorylation of RNAPII in in vitro kinase assays.74 The same effect was observed when purified Brd4 was incubated with P-TEFb. This is unexpected because over-expression of Brd4 increased cellular levels of RNAPII phosphorylated at S2 in previous studies.28 Consistent with the negative role of T29 autophosphorylation on P-TEFb activity, CDK9 phosphorylated at T29 was found to be associated with the transcriptionally silenced HIV provirus.74
CDK9 and Hexim1 are regulated by polyubiquitination.75,76 In pulse-chase experiments, endogenous and overexpressed CDK9 proteins degraded rapidly in cells with a half life of approximately 50 min, while cyclin T1 remained relatively stable.75 Proteasome inhibitor treatment stabilized CDK9, and P-TEFb was shown to interact with SCF core components, p19SKP1, CDC34, cul-1, and the F-box protein determining substrate specificity, p45SKP2.75,77,78 However, CDK9 does not interact directly with p45SKP2; instead, cyclin T1 recruits p45SKP2 through its C-terminal PEST sequence to CDK9 while not being a target of ubiquitination itself.75,76 Indeed, the expression of CDK9 increased in mouse embryonic fibroblasts from p45SKP2-/- mice, whereas expression of cyclin T1 was unchanged.75,76
Interestingly, transactivation of the HIV LTR was increased after co-expression of Tat and p45SKP2, underscoring the requirement of ubiquitination for the coactivator role of P-TEFb in HIV transcription.76 Consistent with a direct role of ubiquitination in regulating P-TEFb activity, ubiquitination of CDK9 by Skp2 facilitated the formation of the ternary complex between P-TEFb, Tat and TAR RNA. The sites of ubiquitination in CDK9 remain unknown.
A similar effect of polyubiquitination was recently described for Hexim1. Hexim1 was identified as a target of human double minute-2 (HDM2), a p53-specific E3 ubiquitin ligase.79 The sites of ubiquitination were mapped to K150-152 and K159-161 located within the basic region of Hexim1. Again, ubiquitination of Hexim1 did not lead to proteasome-mediated protein degradation, but rendered Hexim1 a more efficient inhibitor of P-TEFb-dependent transcription.79
While phosphorylation and ubiquitination are common modifications shared with other CDK/cyclin complexes, acetylation was first identified in CDK9 and cyclin T1. Reversible acetylation is a relatively “new” modification for nonhistone proteins. However, a growing number of nonhistone substrates for histone acetyltransferases (HATs) and histone deacetylases (HDACs) have been characterized. A recent proteomics screen identified 1750 acetylated proteins in human cells.80 Acetylated proteins were found in the nucleus, cytoplasm and mitochondria, and were involved in diverse biological function, prominently in RNA splicing and cell cycle. Interestingly, acetylation sites were frequently found in regions with ordered secondary structure contrary to phosphorylation, which is thought to occur mainly in unstructured regions of proteins. RNA binding domains were the most highly represented domain architecture in acetylated proteins.
Consistent with these criteria, four acetylation sites (K380, 386, 390 and 404) were identified in cyclin T1, which are located in the highly structured predicted coiled-coil region of the protein and negatively influence 7SK snRNA-binding properties of P-TEFb81 (Fig. 2A).
Acetylation of these residues was mediated by the HAT activity of p300 and dissociated Hexim1 and 7SK snRNA from P-TEFb. Accordingly, acetylated cyclin T1 was exclusively found in the active P-TEFb complex but not in the inactive 7SK ribonucleoprotein complex. Cyclin T1 acetylation is important for the full transcriptional activity of P-TEFb at NFκb target genes, including interleukin-8 and the HIV LTR.81 In contrast, Tat-mediated transactivation of the HIV LTR does not require cyclin T1 acetylation, thus supporting the concept that HIV Tat has evolved to recruit nonacetylated P-TEFb directly from the large inactive P-TEFb complex to support continuous high level HIV transcription.82
While the exact mechanisms how acetylation disrupts the 7SK nRNA ribonucleoprotein complex and liberates P-TEFb remain unclear, it is striking that acetylated residues occur in those positions of the predicted coiled-coil structure that are well positioned for protein-protein interactions (Fig. 3A). This suggests a potential role for the acetylation sites in cyclin T1 in the regulation of homo- or heterodimerization of cyclin T1. Since acetylation of different residues could positively or negatively influence these interactions, it remains unclear whether acetylation destabilizes the inactive complex or contributes positively to the formation of active P-TEFb. In addition, acetylation of the central coiled-coil region may induce conformational changes in the neighboring 7SK snRNA- or Brd4-interacting domains, supporting the dissociation of 7SK snRNA or the recruitment of Brd4, respectively (Fig. 3A). Direct interaction of acetylated K390 in cyclin T1 with the second bromodomain of Brd4 has been reported, but the functional relevance of this interaction remains to be examined.57
Recently, acetylation of another cyclin, cyclin A, was described, supporting a wider role of acetylation in cyclin biology.83 Here, acetylation occurred together with ubiquitination and is required for the efficient proteasome-mediated degradation of cyclin A. In addition, cyclin A acetylation may play a role in regulating the enzymatic activity of kinase complexes that interact with cyclin A during the degradation process.83
Acetylation sites in CDK9 were identified by two independent groups. Fu et al. found that the inactive P-TEFb complex interacts with the nuclear receptor core-pressor (N-CoR) complex and speculated that P-TEFb is a target of deacetylation by the N-CoR-associated histone deacetylase HDAC3.84 They identified two residues in CDK9 that are acetylated by p300, K44 and 48, with a preference for K44, and show that K44 is also deacetylated by HDAC3 and HDAC1 (Fig. 2B). Acetylation-deficient CDK9 (K44R) or CDK9 isolated from cells overexpressing HDAC1 or HDAC3 was strongly impaired in its ability to perform in vitro phosphorylation of the CTD of RNAPII. In addition, the transcriptional activity of K44-mutant CDK9 fused to the Gal4 DNA-binding domain or wildtype Gal4-CDK9 expressed in the presence of overexpressed HDAC3 was markedly decreased, supporting the model that acetylation of K44 activates P-TEFb activity.
Sabò et al. showed that members of the GCN5-related N-acetyltransferase (GNAT) family (GCN5 and PCAF) acetylate K44 and K48 in vitro and in cells, with a preference of GCN5 for K48.85 Acetylation by GCN5 decreased the interaction of CDK9 with 5'-fluorosulfonylbenzoyl-5'-adenosine, an ATP analogue, consistent with the model that acetylation of K48 negatively regulates P-TEFb activity (Fig. 2B). Indeed, computer modeling of the recently published crystal structure of CDK9 shows that an acetyl group added to K48 directly interferes with ATP binding while the acetyl group on K44 points away from the ATP binding site (Fig. 3B). K48 is highly conserved among CDK enzymes, and acetylation of this site may well be involved in regulating access of ATP to other CDK enzymes. Notably, acetylation of CDK9 has no effect on the interaction of P-TEFb with Hexim1 or 7SK snRNA, in agreement with the model that this modification regulates P-TEFb activity independently from the inhibitor function of Hexim1.84,85
The remarkable complexity of posttranslational modifications regulating P-TEFb activity underscores the critical role of this transcription elongation factor in biology (Fig. 4). The amount of active P-TEFb is highly restricted in cells; more than half of cellular P-TEFb is estimated to reside in a large inactive complex containing 7SK snRNA and Hexim1 and is not available at active promoters. P-TEFb modifications serve to regulate P-TEFb activity in Hexim1-dependent and -independent manners. Phosphorylation of T270 and S298 in Hexim1 by PI3K/AKT, dephosphorylation of T186 in CDK9 by PP1α/PP2B and PPM1A/PPM1B, and acetylation of cyclin T1 at K380, K386, K390 and K404 trigger the release of active P-TEFb from Hexim1 and 7SK snRNA (Fig. 4). Other modifications, such as acetylation of CDK9 at K44 and K48, phosphorylation of CDK9 at T29, or S175 or ubiquitination of CDK9 at unknown sites, regulate P-TEFb activity via Hexim1-independent mechanisms. Acetylation of K48 restricts ATP binding to CDK9, while ubiquitination can destabilize CDK9 but also augment its transcriptional activity at the HIV promoter. Phosphorylation of S175 in CDK9 enhances interaction of P-TEFb with Brd4, while autophosphorylation at T29 inhibits CDK9 activity through an as-yet unknown mechanism. Today we do know little about how these signals cooperate in P-TEFb to achieve targeted and transient phosphorylation of the RNAPII CTD at S2. However, it has become clear that through the enzymes mediating these modifications the P-TEFb activity is connected to an emerging network of extra-cellular signals that trigger transcription elongation in response to Ca2+ influx, cell cycle progression, apoptosis or carcinogenesis. Unraveling this network will bring important clues to how transcription elongation of central genes is turned on or off during development, stress or activation-induced signaling.
We apologize to colleagues whose work could not be cited due to space constraints. We thank John Carrol and Alisha Wilson for graphics, Gary Howard for editorial and Veronica Fonseca for administrative assistance. We thank Matija Peterlin, Olivier Bensaude and Qiang Zhou for helpful discussions and members of the Ott Lab for continuous support. This work was supported by NIH/NIAIDS R01 AI083139 and funds from the Gladstone Institutes.