The Cyclin-dependent Kinase-activating Kinase (CAK) Assembly Factor, MAT1, Targets and Enhances CAK Activity on the POU Domains of Octamer Transcription Factors*

Octamer binding transcription factors (Oct factors) play important roles in activation of transcription of various genes but, in some cases, require cofactors that interact with the DNA binding (POU) domain. In the present study, a yeast two-hybrid screen with the Oct-1 POU domain as a bait identified MAT1 as a POU domain-binding protein. MAT1 is known to be required for the assembly of cyclin-dependent kinase (CDK)-activating kinase (CAK), which is functionally associated with the general transcription factor IIH (TFIIH). Further analyses showed that MAT1 interacts with POU domains of Oct-1, Oct-2, and Oct-3 in vitro in a DNA-independent manner. MAT1-containing TFIIH was also shown to interact with POU domains of Oct-1 and Oct-2. MAT1 is shown to enhance the ability of a recombinant CDK7-cyclin H complex (bipartite CAK) to phosphorylate isolated POU domains, intact Oct-1, and the C-terminal domain of RNA polymerase II, but not the originally defined substrate, CDK2. Phosphopeptide mapping indicates that the site (Ser385) of a mitosis-specific phosphorylation that inhibits Oct-1 binding to DNA is not phosphorylated by CAK. However, one CAK-phosphorylated phosphopeptide comigrates with a Cdc2-phosphorylated phosphopeptide previously shown to be mitosis-specific, suggesting that, in vitro, CAK is able to phosphorylate at least one site that is also phosphorylated in vivo. These results suggest (i) that interactions between POU domains and MAT1 can target CAK to Oct factors and result in their phosphorylation, (ii) that MAT1 not only functions as a CAK assembly factor but also acts to alter the spectrum of CAK substrates, and (iii) that a POU-MAT1 interaction may play a role in the recruitment of TFIIH to the preinitiation complex or in subsequent initiation and elongation reactions.

The octamer element (ATGCAAAT) has been shown to be one of the major determinants for B cell-specific promoter and enhancer function both in vivo (1,2) and in vitro (3,4). Surprisingly, the same element has been implicated as a critical motif in a number of other genes that include ubiquitously expressed small nuclear RNA genes, the cell cycle-regulated histone H2B gene, and VP16-dependent herpes simplex virus (HSV) 1 immediate early genes (reviewed in Refs. 5 and 6). Cognate octamer binding factors include the ubiquitous Oct-1 and a B cell-restricted Oct-2, which both contain a conserved DNA-binding domain designated the POU domain (7,8). However, activation of octamer-containing promoters through Oct-1 requires additional promoter-specific cofactors. For example, the B cell-specific coactivator OCA-B, the ubiquitous PTF, and the HSV-encoded VP16 are necessary for the maximum activation of immunoglobulin promoters, small nuclear RNA genes, and HSV immediate early genes, respectively. These additional factors interact with the POU domains of Oct-1 and Oct-2 (9 -13).
To identify other factors that modulate Oct-1 activity, we cloned additional POU domain-interacting proteins using a yeast two-hybrid screen. One of several isolates was identified as MAT1, a subunit of CDK-activating kinase (CAK). CAK was originally identified as a kinase that activates CDKs in vitro by phosphorylating a conserved threonine residue in the T-loop of CDKs (14). CAK consists of MAT1 along with CDK7 and cyclin H; MAT1 acts to assemble the latter two subunits to form a stable active kinase containing all three subunits (15)(16)(17). Some populations of CAK exist as part of the general transcription factor TFIIH (18 -20). CAK-containing TFIIH phosphorylates the C-terminal domain (CTD) of the largest subunit of RNA polymerase II, and this phosphorylation has been implicated in promoter clearance and transcriptional elongation (Refs. 18 -21; Ref. 22 and references therein; reviewed in Ref. 23). Here, we show that MAT1 directly binds to the POU domains of several Oct factors and, consequently, enhances their phosphorylation by CAK.

EXPERIMENTAL PROCEDURES
Expression Vectors-The expression vector for 6HisT7MAT1 was constructed as follows. First, 6HisT-pET21b was constructed by ligating the larger fragment of NdeI-PstI-digested 6HisT-pET11d (24) and the smaller fragment of NdeI-PstI-digested pET21b (Novagen). The coding region of MAT1 was amplified by polymerase chain reaction with primers (5Ј-CCTTGAATTCCATGGACGATCAGGGTTG-3Ј (upstream) and 5Ј-CCGGGTCGACTTATAAATGGTTAACTGGGCT-3Ј (downstream)) from the I.M.A.G.E. Consortium cDNA clone (identification number 230976), which contains the whole coding region of MAT1. The amplified fragment was digested with EcoRI and SalI and cloned into the EcoRI and SalI sites of 6HisT-pET21b. The resulting vector (pHT7MAT1) was used to express MAT1 containing N-terminal hexahistidine and T7 epitope tags. A hybrid gene encoding the GAL4 DNAbinding domain (amino acids 1-147), the hemagglutinin epitope tag, and the POU domain of Oct-1 was constructed in the yeast bait vector pCENCYH as follows. First the larger PvuI fragment of the yeast multiple copy vector pAS2, which contains the GAL4 DNA-binding domain (25), was ligated with the larger PvuI fragment of pRS314, which contains CEN6/ARSH4 and TRP1 (26), to regenerate the ␤lactamase gene. The resulting plasmid, pCENCYH is a low copy number plasmid in yeast. The POU domain (amino acids 279 -438) of Oct-1 was amplified by polymerase chain reaction from pBSoct1 (27) with an NdeI site in the N terminus and a termination codon followed by a BamHI site in the C terminus. This NdeI-BamHI fragment was cloned into the NdeI-BamHI sites of pCENCYH to obtain pAI11.
Yeast Two-hybrid Screening-pAI11 was transformed into the Y190 yeast strain (25). The expression of the GAL4-(1-147)-POU-1 hybrid protein was verified by Western blotting with anti-hemagglutinin monoclonal antibody 12CA5 (Babco) (data not shown). This yeast was transformed by a B cell cDNA library in pACT (30) and plated on SC medium lacking tryptophan, leucine, and histidine (containing 25 mM 3-aminotriazole). His ϩ colonies exhibiting ␤-galactosidase activity using the filter lift assay (31) were restreaked onto SC medium plates lacking tryptophan and leucine. 39 colonies that exhibited ␤-galactosidase activity reproducibly were further characterized. Library plasmids were recovered by transforming leucine-deficient Escherichia coli MH4 (leuB Ϫ , ⌬(lac)X74, galU, galK, hsr Ϫ , hsm ϩ , strA) with total DNA prepared from these colonies. Transformants were identified on minimal medium lacking leucine and containing ampicillin. To ensure that the correct cDNAs were identified, isolated library plasmids were individually cotransformed with pAI11 into Y190, and only four library plasmids were found to reproduce ␤-galactosidase activity. To check the specificity of interaction of these clones with the POU domain, Y190 was cotransformed with the library plasmids and pLAM5Ј, which encodes a human lamin C fused to a GAL4 DNA-binding domain (CLON-TECH). Since the four positive plasmids failed to give ␤-galactosidase activity, the sequences of the inserts were determined by dideoxynucleotide sequencing using Sequenase 2.0 according to the manufacturer's instructions (U.S. Biochemical Corp.).

Glutathione S-Transferase (GST) Fusion Protein Expression and in Vitro Binding
Assays-GST and GST-POU domain fusion proteins (10) were expressed in E. coli and purified by glutathione-Sepharose (Pharmacia Biotech Inc.). The amount of proteins bound to the resin was determined by SDS-PAGE using BSA (Boehringer Mannheim) as a standard.
[ 35 S]Methionine-labeled 6HisT7MAT1 and luciferase were transcribed and translated in vitro from pHT7MAT1 and the luciferase control DNA, respectively, by using the TnT system (Promega) according to the manufacturer's instructions. Six micrograms of GST-POU proteins or 8 g of GST protein bound to 10 l of glutathione-Sepharose in 210 l of BC100 buffer (20 mM Tris-HCl (pH 7.9), 20% (v/v) glycerol, 0.2 mM EDTA, 100 mM KCl, 0.25 mM PMSF, 0.1% (v/v) Nonidet P-40, 10 mM 2-mercaptoethanol) containing 0.2 mg/ml BSA were used for each binding experiment. One microliter (each) of reticulocyte lysate containing expressed proteins was added to initiate binding. After 1 h of incubation at room temperature, the resin was washed extensively with BC100 buffer at 4°C. Proteins retained on the resin were analyzed by SDS-PAGE after elution with SDS-loading buffer.
Similarly, 3.7 g of GST-POU-1 or GST-POU-2 proteins or 7.2 g of GST protein bound to 3 l of glutathione-Sepharose in 70 l of BC100 buffer containing 0.2 mg/ml BSA were used for binding experiments with 30 l of HeLa nuclear extract (8.4 mg of protein/ml) in BC100 buffer. After 17 h of incubation at 4°C, resins were washed extensively with BC100 buffer. Retained proteins were eluted with SDS-loading buffer, separated by SDS-PAGE, blotted onto a polyvinylidene difluoride membrane, probed with appropriate antibodies, and visualized by the ECL system (Amersham Corp.). Expression and Purification of Recombinant Proteins-6HisT7MAT1 was expressed in E. coli BL21(DE3)/pLysS after induction by isopropyl ␤-D-thiogalactopyranoside. Cells were disrupted in binding buffer (20 mM Tris-HCl (pH 7.9 at 4°C), 500 mM NaCl, 0.1% (v/v) Nonidet P-40, 20% (v/v) glycerol, 0.25 mM PMSF) by sonication, and the insoluble fraction containing most of 6HisT7MAT1 was collected by centrifugation. Proteins were solubilized in binding buffer containing 6 M guanidine hydrochloride and incubated at room temperature with nickel-nitrilotriacetate resin (Qiagen) preequilibrated with the same buffer. A complex of CDK7 and cyclin H (bipartite CAK) was prepared as follows. 20 flasks (150 cm 2 ) of monolayer high five cells (Invitrogen), maintained in Grace's medium supplemented with 10% fetal bovine serum with 70 -80% confluency, were infected simultaneously with recombinant baculoviruses that produce CDK7 or cyclin H for 48 h at 27°C. To harvest, the cells were centrifuged at 180 ϫ g for 12 min followed by washing one time with ice-cold PBS. The cell pellet was then resuspended in 20 ml of hypotonic buffer (10 mM Tris-HCl (pH 7.4), 25 mM NaCl, 1 mM EDTA, 5 mM dithiothreitol, 0.1 mM PMSF, 1 g/ml antipain, and 1 g/ml leupeptin) and lysed with 20 strokes of a Dounce homogenizer. After centrifugation at 38,000 ϫ g at 4°C, the resulting supernatant (60 mg of protein) was directly loaded onto a Q-Sepharose column (1.0 ϫ 1.3 cm, 1.5 ml) preequilibrated with buffer A (25 mM Tris-HCl (pH 7.5), 1 mM EDTA, 0.01% Nonidet P-40, 1 mM dithiothreitol, 10% glycerol, 0.1 mM PMSF, 0.2 g/ml antipain, and 0.1 g/ml leupeptin) plus 0.025 M NaCl. After washing the column with 30 ml of the equilibration buffer, the bound protein was eluted with 40 ml of a linear gradient from 0.025 to 0.4 M NaCl in buffer A. The peak of CDK7-cyclin H complex, monitored by silver staining following SDS-PAGE, was eluted around 0.18 M NaCl. The pooled CDK7-cyclin H fractions were then diluted 2-fold with buffer A prior to chromatography onto a heparin-Sepharose column (1.0 ϫ 2.5 cm, 2 ml) preequilibrated with buffer A plus 0.1 M NaCl. After washing the column with 100 ml of the equilibration buffer, the bound protein was eluted with 40 ml of a linear gradient from 0.1 to 1 M NaCl in buffer A. The CDK7-cyclin H complex was eluted at around 0.35 M NaCl with a yield of ϳ2 mg with more than 90% purity as judged by silver staining following SDS-PAGE.
Kinase Assays and Phosphopeptide Analysis-In addition to the buffer components introduced by the addition of kinase and substrates, the components of the kinase buffer were 50 mM Tris-HCl (pH 7. Phosphopeptide analysis was performed as described by Roberts et al. (33).

Identification of MAT1 as an Oct-1 POU Domain-interacting
Protein by a Yeast Two-hybrid Screen-To isolate possible cofactors modulating Oct-1 activity, screening with the yeast two-hybrid system was employed (34). Since several cofactors are already known to interact with Oct-1 through the POU domain (POU-1), this domain was used as a bait for the screening. Yeast cells that express the GAL4-(1-147)-POU-1 fusion protein were transformed with a GAL4 activation domaintagged human B cell cDNA library. Out of 11.3 ϫ 10 7 transformants, about 150 were His ϩ and lacZ ϩ . Out of these, 38 colonies were selected, and plasmids containing cDNAs were recovered by transforming E. coli. Four plasmids could reproduce the ␤-galactosidase activity by cotransforming the yeast with the GAL4-(1-147)-POU-1 plasmid. The inability of these plasmids to direct activation of the lacZ reporter gene by itself or with a GAL4-(1-147)-lamin plasmid (an unrelated control GAL4 plasmid) suggested that the interactions were specific. Therefore, these four plasmids were sequenced and were found to encode four different proteins (designated PIP-1, -2, -3, and -4 for POU-interacting proteins). The nucleotide sequence of PIP-4 revealed that this clone contains a sequence encoding approximately 2 ⁄3 of the carboxyl terminus of the recently cloned CAK assembly factor, MAT1. MAT1 is also a subunit of the general transcription factor TFIIH ( Fig. 1A Vitro-To show that the interaction detected in the yeast twohybrid assay is due to a specific interaction between MAT1 and the Oct-1 POU domain (POU-1), in vitro binding studies were performed. The POU domains of Oct-1, Oct-2, and Oct-3 were fused to GST (GST-POU-1, GST-POU-2, and GST-POU-3, respectively), expressed in bacteria, and immobilized on glutathione-Sepharose beads for analysis of MAT1 interactions. A full-length MAT1 clone was obtained from an expressed sequence tag clone and subcloned for expression in the reticulocyte lysate system. MAT1 was synthesized in vitro in the presence of [ 35 S]methionine and incubated with the different POU domain affinity resins. After washing, the bound proteins were eluted with SDS and analyzed by SDS-PAGE. As shown in Fig.  2A, MAT1 was preferentially retained by the GST-POU-1, GST-POU-2, and GST-POU-3 affinity resins, while no detectable MAT1 was retained by the resin containing GST alone. It has been argued that inclusion of ethidium bromide in the binding assay allows a distinction between a bona fide proteinprotein interaction and interactions due to contaminating nonspecific bridging DNA (36). Binding of MAT1 to GST-POU proteins was unaffected by the presence of up to 100 g/ml ethidium bromide, strongly arguing in favor of bona fide direct protein-protein interactions between MAT1 and POU domains (Fig. 2B). Moreover, the inability of a nonrelated control protein, luciferase, to be retained by any of the resin-bound GST proteins further indicates the specificity of these interactions ( Fig. 2A).  , the segment that is predicted to form a coiled-coil structure (amino acids 92-180), and the segment that is responsible for CAK assembly (amino acids 50 -309) (15,17,35) are shown. A direct repeat found in MAT1 is shown by two arrows. The cDNA clone isolated by the yeast two-hybrid screen contains the fragment between nucleotide residues 381 and 1286 of the MAT1 cDNA (35)

MAT1 Specifically Enhances Phosphorylation of POU Domains and the CTD of RNA Polymerase II by CDK7-Cyclin
H-Since MAT1 is a subunit of CAK, it was of interest to test whether CAK can phosphorylate interacting POU domains. For this purpose, bipartite CAK containing stoichiometric amounts of CDK7 and cyclin H was prepared from baculovirus-coinfected insect cells, and the influence of MAT1 on the phosphorylation of histidine-tagged POU-1 (6HisPOU-1) by bipartite CAK was examined. The addition of MAT1 to the bipartite CAK enhanced the kinase activity on 6HisPOU-1 in a concentration-dependent manner, up to 30-fold (Fig. 3A, lanes 7-10). Maximum phosphorylation was observed when an approximately stoichiometric amount (10 ng) of MAT1 was added to the bipartite CAK (Fig. 3A, lane 10), suggesting that recombinant MAT1 is fully active and incorporated into the bipartite CAK. When MAT1 levels were increased above the stoichiometric amount, the kinase activity decreased (data not shown); this presumably reflects binding of free MAT1 to 6HisPOU-1 and sequestration of the substrate from the tripartite CAK containing CDK7, cyclin H, and MAT1. Control assays showed that phosphorylation of 6HisPOU-1 was dependent upon bipartite CAK (Fig. 3A, lane 6 versus lane 10) as well as MAT1 and that autophosphorylation of CDK7 was only weakly stimulated by MAT1 (Fig. 3A, lanes 4, 5, and 7-10).
To examine whether bipartite CAK can also phosphorylate full-length Oct-1 and whether this activity is also enhanced by the addition of MAT1, bacterially expressed FLAG-tagged fulllength Oct-1 (FLAG-Oct-1) was used as a substrate. The addition of a stoichiometric amount of MAT1 to the bipartite CAK enhanced phosphorylation of FLAG-Oct-1 approximately 300fold (Fig. 3B), suggesting that MAT1 is indispensable for CAK activity on full-length Oct-1. Consistent with the effect of MAT1 on phosphorylation of POU-1 and Oct-1 by CAK, as well as the physical interaction results in Fig. 2, A and B, MAT1 also enhanced phosphorylation of GST-POU-2 and GST-POU-3 by bipartite CAK (Fig. 3C). GST alone was not phosphorylated by CAK under the same conditions (data not shown).
Two other known substrates for CAK, the C-terminal domain of the large subunit of RNA polymerase II and CDK2, were also tested. Phosphorylation of GST-CTD by bipartite CAK was increased about 12-fold by MAT1 (Fig. 3D), while phosphorylation of GST-CDK2, the original substrate used to identify and purify CAK, was not further enhanced by the addition of MAT1 (Fig. 3E). Consistent with this, MAT1 did not affect the bipartite CAK-mediated activation of cyclin A-CDK2 to phosphorylate histone H1 (data not shown). The observed effects of MAT1 on the phosphorylation of CTD and CDK2 by bipartite CAK agree with recent results reported by Yankulov and Bentley (37). These results indicate that although CAK can phosphorylate a variety of substrates, MAT1 can further stimulate this phosphorylation on only a subset of these targets. This indicates that MAT1 is acting in a substrate-specific manner.
The CAK Phosphorylation Sites on Oct-1 Are Similar to Those of Cdc2 Kinase-Earlier studies of Oct-1 during progression through the cell cycle revealed a complex temporal program of phosphorylation. In particular, Oct-1 is hyperphosphorylated during mitosis, and DNA binding of Oct-1 is blocked by mitosis-specific phosphorylation of Ser 385 in the POU homeodomain by an unidentified cell cycle-regulated kinase (33,38). Therefore, it was important to determine whether CAK can phosphorylate Ser 385 in vitro. Since protein kinase A was shown to phosphorylate Ser 385 in vitro (38), 6HisPOU-1 phosphorylated by protein kinase A was used as a marker in phosphopeptide analysis. As shown in Fig. 4A, thermolytic peptide maps from 6HisPOU-1 phosphorylated by CAK clearly differ from those generated with protein kinase A, indicating that CAK does not phosphorylate Ser 385 in vitro.
Somewhat unexpectedly, the tryptic peptide map of FLAG-Oct-1 phosphorylated by CAK (Fig. 4B, left panel) appeared very similar to the tryptic peptide map of FLAG-Oct-1 phosphorylated in vitro by a p13-associated Cdc2-related protein kinase (Fig. 4B, middle panel; Ref. 33). Moreover, analysis of a mixture of the two sets of tryptic phosphopeptides showed that six of the CAK-phosphorylated peptides comigrated with six of the Cdc2-phosphorylated peptides (Fig. 4B, right panel). The dependence on MAT1 of FLAG-Oct-1 phosphorylation by CAK (Fig. 3B) rules out the possibility that the overlapping CAK and Cdc2 phosphopeptide maps result from contaminating Cdc2 kinase present in our CAK preparation. Significantly, a prior analysis of the Cdc2-labeled phosphopeptides showed that one of the six common phosphopeptides (indicated by horizontal arrowheads in Fig. 4B) comigrates with an in vivo labeled mitosis-specific phosphopeptide (33). These results suggest that Oct-1 may be a physiological substrate of CAK or a CAKrelated kinase. Since the reported CAK recognition sites in CDKs ((R/M) X (Y/L)(S/T) XXV), as well as the consensus Cdc2 recognition site ((S/T) PX(R/K); Ref. 39), differ significantly from the CAK recognition site in the CTD (YSPTSPS; Ref. 40), the results presented here raise the possibility that these kinases may nonetheless have an overlapping specificity, at least in vitro.
TFIIH in Nuclear Extract Binds to POU Domains-Since MAT1 is also a subunit of TFIIH and since the CTD of the large subunit of RNA polymerase II is a known substrate for the CAK within TFIIH, we tested whether POU domains can bind to TFIIH. As described above, GST and GST-POU-1 proteins were immobilized on glutathione-Sepharose beads and employed as affinity resins. After incubation of the resins with HeLa nuclear extract, followed by extensive washes, bound proteins were separated by SDS-PAGE and visualized by immunoblot with appropriate antibodies. As shown in Fig. 5, various subunits of TFIIH (MAT1, CDK7, cyclin H, p62, and ERCC3) were retained by immobilized GST-POU-1 (lanes 3) but not by immobilized GST (lanes 4). The failure to observe any significant retention of TFIIE␣, RAP74, or TFIIB on GST-POU-1 (lanes 3) indicates the specificity of the TFIIH interaction. Similar interactions were observed when nuclear extract was incubated with GST-POU-2 and when chromatographic fractions enriched in TFIIH were incubated with GST-POU-1 and GST (data not shown). Inclusion of 100 g/ml ethidium bromide in the binding buffer did not affect the retention of TFIIH subunits, indicating that this retention is due to direct protein-protein interactions (data not shown).

DISCUSSION
The present results demonstrate that MAT1 can interact directly with isolated POU domains, that it can stimulate phosphorylation by CDK7-cyclin H of Oct-1 and isolated POU domains (including, apparently, at least one physiological site) as well as isolated CTD repeats, and that the POU domain selectively interacts with TFIIH relative to other general initiation factors. These results have implications for the regulation of CAK substrate specificity and function, the modulation of Oct-1 functions, and the modulation of basic initiation and elongation mechanisms.
Substrate Specificity and Function of CAK-The previously reported substrates of CAK were restricted to the CDKs, the CTD of the large subunit of RNA polymerase II, and TBP (reviewed in Refs. 41 and 42). It is notable, therefore, that the DNA-binding Oct factors also fall within this group and are phosphorylated (minimally) in the POU domains in a MAT1dependent manner. The much greater effect on CAK-mediated phosphorylation of Oct-1 (300-fold) than on CAK-mediated phosphorylation of the POU domain (30-fold) may reflect a conformational change that allows increased phosphorylation or it may reflect the increased number of phosphorylation sites available on the full-length molecule.
Bipartite CDK7-cyclin H and tripartite CDK7-cyclin H-MAT1 complexes have been purified from vertebrates (15)(16)(17). However, it is not clear whether both species exist in vivo or whether they arise artifactually during purification, although bipartite CAK can be reconstituted from recombinant proteins (43,44). In addition, two groups have identified an essential FIG. 4. Phosphopeptide analysis of POU-1 and Oct-1 phosphorylated by CAK. A, thermolytic peptide maps of 6HisPOU-1 phosphorylated in vitro by tripartite CAK and by protein kinase A (PKA). The origin of each sample is marked by a vertical arrowhead. B, tryptic phosphopeptide maps of FLAG-Oct-1 phosphorylated in vitro by tripartite CAK and by Cdc2. Equal amounts of radioactive phosphopeptides (about 1000 cpm) from tripartite CAK-phosphorylated and Cdc2-phosphorylated Oct-1 were mixed in the mixing experiment. Horizontal arrowheads indicate the phosphopeptide previously shown to comigrate with one of mitosis-specific tryptic peptides (33) and common to all three maps. The origin of each sample is marked by a vertical arrowhead.
FIG. 5. POU-1 interacts with TFIIH. HeLa nuclear extract was incubated with GST-POU-1 (lanes 3) and GST (lanes 4) immobilized on glutathione-Sepharose beads. Bound proteins were subjected to SDS-PAGE and analyzed by immunoblot with antibodies against proteins shown on the left. Five percent (lanes 1) and 2.5% (lanes 2) of input nuclear extract were used as controls. In each panel the segments corresponding to lanes 1-3 and lane 4 were from the same autoradiogram. The same filter was used for each of the Western blots and was stripped prior to each probe. yeast protein that functions as a CAK and lacks polypeptides homologous to vertebrate CDK7, cyclin H, and MAT1 (45,46), while another group has reported that the yeast kinase KIN28, which is homologous to CDK7, functions only in transcription and not as a CAK (47). This has led to the proposal that the CDK7-cyclin H-MAT1 complex identified as the predominant CAK in vertebrate cell extracts may function primarily during transcription (as part of TFIIH) and not as a physiological CAK (45). Our finding of the Oct family of transcription factors as new substrates of CAK further supports this view and, minimally, suggests a novel MAT1-regulated function of CDK7cyclin H in transcription regulation. Our finding that MAT1 greatly enhances phosphorylation by CDK7-cyclin H further supports the idea that MAT1 may provide a regulatory mechanism for selective substrate phosphorylation. A similar suggestion was made by Yankulov and Bentley (37) on the basis of their observation (also confirmed here) that MAT1 switches the substrate specificity of CDK7-cyclin H to favor the RNA polymerase II CTD over CDK2.
Modulation of the Function of Oct Factors-Since the POU domains of Oct factors comprise the DNA-binding domains, the observed MAT1-dependent POU domain phosphorylation events could regulate site-specific DNA binding on cognate target genes. However, while the DNA binding ability of Oct-1 is blocked during mitosis by phosphorylation (38), the site critical for this inhibition (Ser 385 in the POU homeodomain of Oct-1) is not phosphorylated by CAK in vitro. On the other hand, the observed phosphorylation by CAK could modulate the function of coactivators known to interact with Oct-1 or Oct-2 through the POU domain. These include the HSV-encoded VP16 (6), the B cell-specific OCA-B implicated in immunoglobulin promoter function (5,10), the ubiquitous PTF implicated in small nuclear RNA gene transcription (11), and a presumptive coactivator for histone H2B transcription (10). It also has been reported that Oct-1 and Oct-2 facilitate preinitiation complex formation and that Oct-2 may act at an early stage in this process (48,49). Consistent with this conclusion, Zwilling et al. (50) showed that TBP can bind to the POU domains of Oct-1 and Oct-2. On the other hand, our finding that POU domains interact directly with MAT1 raises the possibility that Oct factors may also affect a late stage of preinitiation complex formation by recruiting TFIIH to octamer-containing promoters. This possibility is further supported by our observation that TFIIH, but not TFIIE or TFIIF, selectively binds Oct-1 and Oct-2 POU domains. This result is consistent with earlier reports of direct activator interactions with TFIIH (22,51,52).
Possible Modulation of Initiation and Postinitiation Events-Since a TFIIH helicase is essential for an early step in the initiation process (Refs. 53 and 54; reviewed in Ref. 55), an Oct-1-MAT1 interaction also could facilitate initiation per se. Significantly, however, some promoters may require CTD phosphorylation by the CAK component of TFIIH for promoter clearance (Refs. 56 and 57; reviewed in Ref. 55) and/or enhanced elongation events (22,58). It has been speculated that such events might result in the release of constraining interactions of the CTD with TBP (59), TFIIE (60), or possibly other cofactors (reviewed in Refs. 23 and 61); TFIIH-mediated phosphorylation of TBP, TFIIE, and TFIIF has also been reported (37,41,62). By analogy, the phosphorylation of Oct factors by TFIIH might facilitate postinitiation events by releasing Oct-TFIIH interactions; alternatively, POU interactions with MAT1 could regulate other TFIIH interactions and functions.
A final important question (see above) is whether the CAK functions that result from MAT1 interactions with Oct-1 take place in the context of free CAK or (as observed for phospho-rylation of TFIIE␣, RAP74, and the intact CTD within RNA polymerase II (see Ref. 41 and references therein)), in the context of TFIIH.