Characterization of the residues phosphorylated in vitro by different C-terminal domain kinases.

The C-terminal part of the largest subunit of eukaryotic RNA polymerase II is composed solely of the highly repeated consensus sequence Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7. This domain, called the C-terminal domain (CTD), is phosphorylated mostly at serine residues during transcription initiation, but the precise role of this phosphorylation remains controversial. Several protein kinases are able to phosphorylate this sequence in vitro. The aim of this work was to define the positions of the amino acids phosphorylated by four of these CTD kinases (transcription factor (TF) IIH-kinase, DNA-dependent protein kinase, and the mitogen-activated protein kinases ERK1 and ERK2) and to compare the specificity of these different protein kinases. We show that TFIIH kinase and the mitogen-activated protein kinases phosphorylate only serine 5 of the CTD sequence, whereas DNA-dependent protein kinase phosphorylates serines 2 and 7. Among the different CTD kinases, only TFIIH kinase is appreciably more active on two repeats of the consensus sequence than on one motif. These in vitro results can provide some clues to the nature of the protein kinases responsible for the in vivo phosphorylation of the RNA polymerase CTD. In particular, the ratio of phosphorylated serine to threonine observed in vivo cannot be explained if TFIIH kinase is the only protein kinase involved in the phosphorylation of the CTD.

A characteristic feature of eukaryotic RNA polymerase II is the carboxyl-terminal domain (CTD) 1 of its largest subunit, composed of multiple repeats of the sequence Tyr 1 -Ser 2 -Pro 3 -Thr 4 -Ser 5 -Pro 6 -Ser 7 . The number of repeats differs according to the species. The RNA polymerase II CTD contains 26 repeats in yeast (1), 45 in Drosophila (2,3), and 52 in mammals (4). The consensus sequence is highly conserved, and this domain is essential for cell viability (2,3,5). Partial deletions of the CTD alter the regulatory properties of distinct promoters in different ways. The CTD has been shown to interact with a multisubunit complex containing the TATA-binding protein, which is an integral part of the transcription initiation complex (6).
The CTD motif is mainly composed of phosphorylatable amino acid residues, and the RNA polymerase II CTD is actually highly phosphorylated in vivo, mostly at serine and to a lesser extent at threonine and tyrosine (7)(8)(9)(10). This phosphorylation appears to play a role in transcription initiation (11)(12)(13), but its precise function remains to be established. The in vivo phosphorylation sites are not known, but this issue has been approached indirectly by comparing the ratio of phosphorylated serine and threonine. The predominance of serine phosphorylation in vivo (with a serine/threonine ratio of ϳ10:1) has been explained by phosphorylation at positions 2 and 5: the serine at position 2 is replaced by a threonine in one-fifth of the repeats, and position 5 has only one threonine out of 51 (9). Moreover, yeast strains in which CTDs have been modified by the substitution of these two serines are nonviable (14).
In the past 5 years, many protein kinases from various organisms have been described as being able to phosphorylate the CTD in vitro (15)(16)(17)(18)(19)(20)(21)(22). Some of them are now well identified. ␦-Kinase from rat (20) (also called factor b in yeast (16) and TFIIH or BTF2 in humans (22)) is a multisubunit transcription factor that contains DNA-dependent helicase activity, DNA repair activity, and CTD kinase activity (see Ref. 23 for review). A subcomplex called TFIIK is responsible for the CTD kinase activity and is composed of the Cdk-related protein kinase MO15/Cdk7 in mammals (KIN28 in yeast), associated with cyclin H (CCL1) and MAT-1 (24 -27, 67, 68). The kinase-cyclin pair MO15-cyclin H (KIN28-CCL1) is already known as a cyclindependent kinase (CAK (Cdc2-activating kinase)) in vitro (28,29). Other protein kinases have also been described as able to phosphorylate the CTD. DNA-dependent protein kinase (DNA-PK) is composed of a catalytic subunit (DNA-PK cs ) and a regulatory component corresponding to the Ku autoimmune antigen (30,31). It acts as a CTD kinase when stimulated by linear double-stranded DNA and by several transcriptional activators (30,32). Recently, DNA-PK has also been shown to phosphorylate the RNA polymerase I transcriptional apparatus and to inhibit RNA polymerase I transcription. Moreover, this protein kinase plays a major role in DNA repair processes and recombination of immunoglobulin gene locus in the cells of the immune system (33). MAP kinases are induced by mitogenic stimuli and by heat shock. They are able to phosphorylate the CTD among their many known in vitro substrates (34).
The aim of this study was to determine the sites phosphorylated in vitro in the CTD by TFIIH kinase, DNA-PK, and MAP kinases. Comparison of the results obtained in vitro with what is known about CTD phosphorylation in vivo gives some clues to the nature of the protein kinases involved in the in vivo phosphorylation of the RNA polymerase CTD. To determine which of the three serines is phosphorylated by the different CTD kinases, we synthesized a set of peptides containing one or two CTD motifs in which each serine was successively replaced by an alanine. We devised new electrophoretic conditions to be able to separate the phosphorylated peptides. Similar experiments were performed with another set of peptides to determine the influence of amino acids surrounding the phosphorylated site and to compare the specificity of the different CTD kinases. Using a similar method, we were recently able to distinguish the different CTD kinase activities that are induced by stress and heat-shock treatment (35).
Phosphorylation reactions were performed with 4 l of human ERK1 (agarose-conjugated ERK1; Upstate Biotechnology, Inc.) for 90 min at 30°C, with 0.4 units/ml murine recombinant ERK2 (from Prof. P. Cohen, University of Dundee, Dundee, United Kingdom) for 30 min at 30°C, with ϳ40 ng of TFIIH kinase (TSK SP-5-PW fraction (36)) for 2.5 h at 30°C in the presence of 1.6 mg/ml polyvinyl alcohol and 5 mg/ml bovine serum albumin, or with 2 l of DNA-PK (from Dr. G. Smith, Wellcome/Cancer Research Campaign Institute, Cambridge, United Kingdom) in the presence of 200 g/ml salmon sperm DNA for 30 min at 30°C. The times used for in vitro reactions were chosen to be in initial rate conditions. Peptides were synthesized by Dr. O. Siffert (Organic Chemistry Laboratory, Institut Pasteur). Reactions were stopped by adding 1 volume of Laemmli sample buffer (37) containing 5% ␤-mercaptoethanol.
Hepta-1 Peptide Analysis and Modified Peptides-Hepta-1 peptides were too small to be separated from the radiolabeled ATP front by the usual electrophoretic conditions. We acidified the buffer by replacing the Tris/glycine buffer with a phosphate buffer (38) as we previously described (35). Reaction samples were subjected to electrophoresis on 10% denaturing phosphate buffer gels at pH 6.0. Gels were fixed in ethanol/acetic acid/trichloroacetic acid/water (3:1: 1:5, v/v), dried, and submitted to autoradiography with an intensifying screen at 4°C. Signals were quantified with a Fuji BAS reader, and measurements made with PC BAS.

Phosphoamino Acid Analysis
The SPTTPSY peptide was phosphorylated by TFIIH kinase and analyzed as described above. The radiolabeled spot was cut out of the gel and washed rapidly in 500 l of water. The phosphorylated peptide was eluted in 0.5 M ammonium acetate at 37°C with gentle agitation overnight. Peptide hydrolysis and phosphoamino acid analysis were performed as described previously (39).

TFIIH Kinase and the MAP Kinases ERK1 and ERK2
Phosphorylate Serine 5 of the CTD Motif-Serizawa et al. (40) previously showed that TFIIH kinase phosphorylates the RNA polymerase II CTD at serine residues. To determine which of the three serines was phosphorylated by TFIIH kinase, we assayed a set of peptides in which serines were replaced by alanines. Fig. 1A shows that phosphorylation was completely lost when serine 5 was replaced by an alanine, whereas the two other serine substitutions had no effect on phosphorylation. Thus, only serine 5 of the CTD motif appears to be phosphorylated by TFIIH kinase. These results are in agreement with those reported by Roy et al. (25). Identical results were obtained with ERK1 ( Fig. 1B) and ERK2 (data not shown).
Phosphorylation by TFIIH Kinase and by ERK1 and ERK2 Reveals Differences in the Recognition of the Hepta-2 Peptide-Within the entire CTD, all the repeats are contiguous, and all (but one) serines at positions 2 and 7 are close to the extremity of another repeat. Therefore, the absence of phosphorylation of these serines by TFIIH kinase (or MAP kinases) might be due to the absence of the amino acids surrounding these residues in the CTD. To rule out this possibility, we performed experiments with similar alanine replacements in hepta-2 peptides (Ser 2 -Pro 3 -Thr 4 -Ser 5 -Pro 6 -Ser 7 -Tyr 8 -Ser 9 -Pro 10 -Thr 11 -Ser 12 -Pro 13 -Ser 14 -Tyr 1 ) (Fig. 2). These peptides were electrophoresed on a 22% denaturing polyacrylamide gel. Phosphorylation of the hepta-2 peptide by ERK1 or ERK2 revealed two separated bands with similar intensities. Phosphorylation efficiency was very low, and the peptide concentration that we used did not allow the characterization of multiple phosphorylations. The two bands correspond to phosphorylation in the left or right part of the peptide. Indeed, when serine 5 of the right part was replaced by an alanine, the upper band disappeared, and when the same substitution was made in the left side, there was no longer a lower band. Replacement of serine 9 had no effect on the phosphorylation. Therefore, as for hepta-1 peptides, only serine 5 of the CTD motif is phosphorylated in hepta-2 peptides by ERK1 and ERK2.
As far as TFIIH kinase is concerned, the two bands were also observed, and we obtained qualitatively the same effect for the substitutions (Fig. 2), although the relative intensities were different from those obtained by MAP kinase phosphorylation. The intensity of the upper band, which corresponds to phosphorylation of the serine present in the central position of the left motif, was very low compared with the other. This differ-FIG. 1. Determination of the serine phosphorylated by TFIIH kinase, MAP kinase, and DNA-PK within the CTD motif using hepta-1 peptides. A set of four peptides containing seven amino acids was chemically synthesized. Their sequences were derived from the CTD motif by successively replacing each serine by an alanine: SPTSPSY (peptide H1), APTSPSY (peptide A2), SPTAPSY (peptide A5), and SPTSPAY (peptide A7). Kinase assays were performed by adding TFIIH kinase, ERK1, ERK2, or DNA-PK to these synthetic peptides (1.6 mg/ml) as described under "Materials and Methods." For all the experiments, identical results were obtained for ERK1 and ERK2, which are reported as MAP kinase on all the figures. Phosphorylated peptides were separated by electrophoresis on denaturing phosphate buffer-polyacrylamide gels. Gels were dried and autoradiographed. ence was still evident with substituted peptides. The two sites are not equivalent, suggesting that TFIIH kinase recognizes sequence or conformational information in the right side of the phosphorylated site. Fig. 3 shows the result of a comparison of the efficiency of the different protein kinases to use one, two, or three repeats of the CTD motif. To be able to compare the phosphorylation of the three different peptides in a single experiment, we used the electrophoretic conditions normally used for the hepta-1 peptides. Under these conditions, only one spot is obtained for the hepta-2 peptide. For this experiment, the same peptide concentration was used for all the peptides. ERK1 and ERK2 phosphorylated the hepta-3 and hepta-2 peptides 3-and 2-fold better than the hepta-1 peptide, respectively, whereas TFIIH kinase phosphorylated the hepta-2 and hepta-3 peptides 13-and 20-fold better than the hepta-1 peptide, respectively. The hepta-3/hepta-2 ratio was therefore identical for the three kinases. This indicates that two repeats increase the phosphorylation efficiency of TFIIH kinase.
Determination of the Consensus Sequence Recognized by TFIIH and MAP Kinases-Another set of hepta-1 peptides was used to determine the involvement of each amino acid in the recognition by TFIIH and MAP kinases. At first glance, the comparison shows that the specificity of these different kinases is very similar. The hepta-1 peptide without any substitution was considered as the reference (Fig. 4A). C-terminal proline replacement by an alanine (SPTSASY) led to the abolition of peptide phosphorylation, a feature of the proline-dependent protein kinase family. Likewise, no significant peptide phosphorylation was observed when the N-terminal proline was displaced or removed (PSTSPSY, STSPSY, and STPSPSY). In conclusion, the two prolines of the motif are absolutely essential for TFIIH and MAP kinases. Phosphorylation was decreased by half when the tyrosine was replaced by an alanine. Thus, the tyrosine is important, but not essential. Replacement of the threonine at position 4 by a leucine or an arginine significantly increased the phosphorylation of the peptide, whereas glutamic acid had an opposite effect. Confirming the positive effect of a leucine at position 4, the N-terminal proline became nonessential when threonine 4 was replaced by a leucine (SALSPSY) (see "Discussion").
Previous studies showed that serine is a better phosphate acceptor than threonine for ERK1 and ERK2 (41). Fig. 4A (upper panel) shows that threonine at position 5 is an even poorer phosphate acceptor for TFIIH kinase than for MAP kinases, revealing a significant difference between TFIIH and MAP kinases. The weak intensity of TFIIH kinase phosphorylation suggests another hypothesis. Due to the substitution of serine 5, serines 2 and/or 7 might become minor phosphate acceptors. Phosphoamino acid analysis of the phosphorylated peptide revealed only the presence of phosphothreonines, showing that serines 2 and 7 cannot be phosphorylated by TFIIH kinase and that limited phosphorylation of the SPTTPSY peptide is due to phosphorylation of threonine at position 5.
Serines Phosphorylated by DNA-dependent Protein Kinase-The same set of peptides used for TFIIH and MAP kinases was assayed to characterize the sites phosphorylated by DNA-PK in the CTD motif (Fig. 1C). None of the substitutions completely abolished the peptide phosphorylation. This suggested that more than one serine is phosphorylated by DNA-PK. However, serine 2 (APTSPSY) or serine 7 (SPTSPAY) replacement decreased peptide phosphorylation significantly. On the other hand, when serine 5 (SPTAPSY) was substituted, DNA-PK was even more efficient than with hepta-1. Another peptide was synthesized to determine the effect of the simultaneous substitution of serines 2 and 7 (peptide A2A7, APTSPAY). The introduction of these changes severely depressed phosphorylation by DNA-PK. We observed a residual phosphorylation after a long exposure that migrated like peptide A2A7 phosphorylated by ERK1, as if in the absence of its highest affinity sites, DNA-PK is able to weakly phosphorylate the serine located at position 5 (data not shown). In contrast, even after a long exposure, we never observed any phosphorylation of peptide A5 with the MAP kinases.
FIG. 3. Phosphorylation efficiency of TFIIH kinase, MAP kinase, and DNA-PK to use one, two, or three repeats of the CTD motif. TFIIH kinase, MAP kinase (MAPK), and DNA-PK were assayed for phosphorylation of hepta-3, hepta-2, or hepta-1 peptides (1.6 mg/ml). Phosphorylated peptides were separated on phosphate buffer-polyacrylamide gels at pH 6.0 and were quantified with a Fuji BAS reader. The amount of radioactivity incorporated in the hepta-1 peptide was arbitrarily set to 100 to compare the relative activities of the different kinases. au, arbitrary units.
Since the sites phosphorylated by DNA-PK were different from those phosphorylated by TFIIH and MAP kinases, two other modified hepta-2 peptides were synthesized. Four bands were observed by phosphorylating the hepta-2 peptide (Ser 2 -Pro 3 -Thr 4 -Ser 5 -Pro 6 -Ser 7 -Tyr 8 -Ser 9 -Pro 10 -Thr 11 -Ser 12 -Pro 13 -Ser 14 -Tyr 1 ) with DNA-PK (Fig. 5). They may correspond to phosphorylation of serines 2, 7, 9, and 14, the four potential phosphorylation sites contained in the hepta-2 peptide. When serine at position 9 was replaced by an alanine (peptide A9), one of the bands disappeared, and when two serines were replaced (peptide A7A14), only two bands remained. These results support the hypothesis that each of the four bands corresponds to a dipeptide phosphorylated at only one site and that positions 2, 7, 9, and 14 are actually phosphorylated by DNA-PK.
Influence of Substitution at Several Positions on Phosphorylation by DNA-PK-Peptides identical to those used Fig. 4 were assayed with DNA-PK. A striking observation is that no substitution led to a complete loss of phosphorylation (Fig. 6). In half of the cases, the peptides were phosphorylated at least as well as the hepta-1 peptide. The replacement of the C-terminal proline (peptide A6) did not affect peptide phosphorylation. The same was true for the N-terminal proline with an increase in phosphorylation for peptide P4 (sequence STPSPSY); therefore, DNA-PK is not a proline-dependent protein kinase. Tyrosine replacement produced a decrease in phosphorylation; thus, this amino acid is important, but not essential. The pep-tide that contains arginine, a basic amino acid, at position 4 (SPRSPSY) was not a good substrate for DNA-PK. The two radiolabeled spots corresponding to this peptide that were separated on polyacrylamide gels may correspond to phosphorylation of either serine 2 or serine 7 (the amount of phosphorylated peptide reported on the histogram is the addition of the two values); the same is probably true for the other peptides, but they are not separated on the gel. In contrast, an acidic residue at position 4 (SPESPSY) led to very efficient peptide phosphorylation (Fig. 6, lower panel). No decrease in phosphorylation was observed when threonine 4 was substituted by a leucine, which is a neutral amino acid without a phosphate acceptor group, showing that this threonine is probably not phosphorylated by DNA-PK. Substitution of serine 5 by a threonine (SPTTPSY) decreased peptide phosphorylation. This result is surprising because this serine is not supposed to be phosphorylated by DNA-PK. Changes in charges or in the steric environment may explain the indirect effects on phosphorylation of serines 2 and 7.

Models Concerning CTD Functions
Despite all the studies on the CTD and its phosphorylation, the role of this domain of eukaryotic RNA polymerase II is still unclear. A model frequently evoked for RNA polymerase II transcription initiation is the recruitment of a hypophosphorylated RNA polymerase II to the preinitiation complex and phosphorylation during the transition from initiation to elongation (11)(12)(13). Possible pausing at an early phase of transcription, dependent on proximal activating sequences (42), might necessitate another phosphorylation for the release of RNA polymerase. Therefore, a first phosphorylation might be necessary for initiation and a second for relief of pausing. These two phosphorylations might be carried out by different Ser/Thr or Tyr protein kinases. By using immunofluorescence microscopy, Weeks et al. (43) showed on polytene chromosomes that some specific genes are transcribed by hyperphosphorylated polymerases, whereas hsp70 mRNAs are elongated by a mixture of hypo-and hyperphosphorylated forms. Similar results were obtained by in vivo protein-DNA cross-linking assays (44). The CTD phosphorylation state during transcription elongation may thus be different for the different types of gene. Experiments in yeast suggest the involvement of several protein kinases in CTD phosphorylation in vivo. Lee et al. (45) have FIG. 4. Nature of the site phosphorylated by TFIIH and MAP kinases. A third set of peptides was synthesized. Amino acids surrounding the phosphorylated residue were substituted or displaced in the hepta-1 motif to determine their importance in the recognition of the peptide. A, the different peptides were assayed with TFIIH (upper panel) or MAP kinase (MAP K; lower panel) at 1.6 mg/ml. Phosphorylated peptides were analyzed as described for Fig. 1. B, phosphoamino acid analysis was performed on SPTTPSY phosphorylated by TFIIH kinase. Panel 1, cold phosphoserine (Pser) and phosphothreonine (Pthr) standards were added to the amino acid hydrolysate and revealed by ninhydrin after chromatography on a cellulose plate. Panel 2, the same experiment was submitted to autoradiography. 5. Effect of substitution of serines at position 9 and 14 of a hepta-2 CTD peptide on phosphorylation by DNA-PK. The phosphorylation of hepta-2 and modified hepta-2 synthetic peptides (0.4 mg/ml) was performed using DNA-PK as described under "Materials and Methods." Taking into account the serines phosphorylated by DNA-PK in hepta-1 peptides (Fig. 1), serines 9 and 14 of the hepta-2 peptide were successively or simultaneously substituted by alanines. Phosphorylated peptides were analyzed as described for Fig. 2. characterized the CTK1 gene encoding a yeast nuclear CTD kinase, which presents some homologies to cdc2/CDC28. CTK1 gene disruption decreases in vivo phosphorylation of RNA polymerase II without abolishing it (45). The different sensitivity of CTD phosphorylation to protein kinase inhibitors suggests that it is the same in higher eukaryotes: CTD phosphorylation in quiescent cells is decreased by the protein kinase inhibitor 5,6-dichloro-1-␤-D-ribofuranosylbenzimidazole, whereas it is unaffected in serum-treated cells or heat-shocked cells (34,46). The number of protein kinases able to phosphorylate the CTD in vitro is still growing. A CTD kinase has been located in the yeast holoenzyme. This kinase is encoded by a new gene called SRB10, regulated by the cyclin SRB11 (47). Recently, another CTD kinase has been identified as a component of the positive transcription elongation factor of Drosophila. It has been suggested that this protein kinase could phosphorylate the CTD to release the RNA polymerase II from an abortive elongation state (48). In addition, McCracken et al. (69) showed recently that the CTD is required for efficient RNA processing. CTD phosphorylation might also be implicated in mRNA splicing. All these observations suggest that depending on the signal that induces specific gene transcription, the step in the transcription cycle or mRNA splicing, different enzymes might be involved in CTD phosphorylation. These protein kinases may have different site specificities in the CTD motif.

In Vivo Phosphorylation of the RNA Polymerase II CTD
Little is known about the exact phosphorylation state of RNA polymerase II in vivo. A shift on SDS-polyacrylamide gels revealed multiple CTD phosphorylation, and two-dimensional paper electrophoresis of in vivo labeled CTD showed a predominant phosphorylation of serine compared with threonine (ratio of ϳ10:1). This low level of threonine phosphorylation cannot be explained by threonine 4 phosphorylation because this position is highly conserved, and its phosphorylation would not lead to a correct ratio. However, it can be explained by phosphorylation at positions 2 and 5, which present some nonconsensus threonines instead of serines (9).

Specificity of the Different Protein Kinases
That Phosphorylate the CTD Motif TFIIH and MAP Kinases Phosphorylate the Same Site in the CTD Motif-TFIIH kinase, ERK1, and ERK2 phosphorylate the same residue in the repeated motif of the CTD. Surprisingly, the influence of each amino acid of the motif seems very similar for TFIIH and MAP kinases. They are proline-directed protein kinases. The consensus sequence deduced from CTD peptides (PX(S/T)P) is identical to the sequence determined for MAP kinases in the myelin basic protein and epidermal growth factor receptor (41,49,50). For both kinases, peptide phosphorylation is strongly enhanced by the replacement of threonine 4 by a leucine. Moreover, the N-terminal proline, which is essential for CTD sequence phosphorylation, is no longer necessary in the presence of a leucine at position 4 (SALSPSY). When the site phosphorylated by MAP kinases in whole proteins does not present a proline at position Ϫ2, a hydrophobic amino acid at position Ϫ1 is very often observed, confirming a strong influence of this residue on peptide phosphorylation (51). One of the differences we found between MAP and TFIIH kinase specificities in vitro is the more stringent requirement of the latter for a serine at the phosphorylated position. The second concerns the difference between hepta-1 and hepta-2 peptide phosphorylation efficiencies. We observed a small and progressive increase in phosphorylation of peptides containing one, two, and three CTD motifs by ERK1 and ERK2. This is in agreement with the data showing that the entire RNA polymerase II CTD is better phosphorylated than a small peptide by most of the CTD protein kinases. However, we showed that TFIIH kinase phosphorylates the hepta-2 peptide 13-fold better than the hepta-1 peptide, whereas the difference between the hepta-3 and hepta-2 peptides is low and identical to that observed for the other CTD kinases. Despite an identical phosphorylation site, these experiments show that MAP and TFIIH kinase recognition sites are not equivalent.
The Cdk2 Sequence Recognized by TFIIH Kinase Is Different from the Consensus Sequence Determined with CTD Peptides-TFIIH kinase is a multisubunit complex containing several enzymatic activities. The CTD kinase catalytic subunit of TFIIH kinase corresponds to the protein kinase MO15/Cdk7 complexed with cyclin H (24 -27, 67, 68). MO15 is considered to be a Cdc2-like protein (52,53). We showed in this study that TFIIH kinase is also a proline-directed protein kinase like Cdc2. However, definition of a general consensus sequence for TFIIH is problematic. It was previously shown that this MO15cyclin H complex, also called CAK, was implicated in p34 cdc2 and p33 cdk2 activation by phosphorylation of threonine 161 for p34 cdc2 and threonine 160 for p33 cdk2 , whose surrounding sequences are almost identical (RVYT*HEVVTLWYR) (54).  Fig. 4 was assayed with DNA-PK for this experiment. The SPRSPSY peptide value was quantified by adding the two radiolabeled spots corresponding to this peptide. The SPESPSY peptide requires a long migration to be separated from the free radiolabeled ATP. The quantification of this peptide was not done with the gel shown in this figure. au, arbitrary units.
There are no prolines, either at position Ϫ2 or at position ϩ1. One could imagine that a subunit supplementary to MO15 and cyclin H would allow a change in specificity. However, this hypothesis seems unlikely because Serizawa et al. (67) showed that TFIIH kinase (identical to the factor we used) phosphorylates both the RNA polymerase II CTD and Cdk2. One similarity between CTD kinase and Cdc2 kinase is their relative efficiency in phosphorylating serine and threonine. We showed that a serine-containing peptide is a better substrate for TFIIH kinase than a threonine-containing peptide. An identical conclusion was reached with Cdk2 when Thr 160 was replaced by a serine (54).
DNA-PK Is Able to Phosphorylate Two Serines in the CTD Motif-Several proteins such as HSP90, Sp-1, p53, c-Jun, and SV40 large T antigen are in vitro substrates for DNA-PK. Studies of the sites phosphorylated by DNA-PK in these proteins defined a common minimal consensus sequence, which is Q(S/T) or (S/T)Q)(55-60), but these sequences are not present in the CTD, even in the non-consensus motifs. However, it has been previously noted that certain proteins such as c-Fos are also efficient substrates, but are apparently not phosphorylated on Q(S/T) or (S/T)Q motifs (61).
We showed in this study that in contrast to TFIIH and MAP kinases, DNA-PK does not phosphorylate serine 5 of the CTD motif, but rather the serines at positions 2 and 7. The substitution of each amino acid surrounding the phosphorylated serines did not lead to a complete loss of phosphorylation, suggesting that none of these amino acids is absolutely necessary. All the substitutions showed that prolines are not important for CTD phosphorylation by DNA-PK.
For DNA-PK, the replacement leading to the most striking increase is the substitution of threonine 4 by a glutamic acid. The positive influence of glutamic acid was previously shown for p53 and c-Jun. One of the sites phosphorylated in vitro by DNA-PK is ES 7 Q for p53 and ES 249 QE for c-Jun (59,60). The successive replacement of each of these glutamic acid residues in the c-Jun sequence decreased phosphorylation. The influence of glutamic acid 251 on serine 249 phosphorylation may be similar to the influence of glutamic acid 4 on serine 2 phosphorylation in the CTD sequence. This observation is consistent with the idea that acidic residues contribute to the recognition by DNA-PK, but such sequences are never found in the RNA polymerase II C-terminal domain, even in the non-consensus motifs.
Tyrosine May Play a Role in DNA-Substrate Binding-Tyrosine 1 replacement by an alanine significantly decreases peptide phosphorylation. Tyrosine may be important for enzymesubstrate binding or for DNA-substrate binding necessary for good recognition by DNA-PK. Most of the potential substrates phosphorylated by DNA-PK bind DNA, and the colocalization of the enzyme and the substrate on the same DNA fragment appears important. A unique CTD motif with a tyrosine at each end is able to bind DNA, but the case of a simple hepta-1 peptide has not been tested (62). More recently, West and Corden (14) showed that tyrosine replacement by a phenylalanine in yeast CTD is lethal for the cell. However, interpretation of the latter data remains difficult.

Physiological Significance of the in Vitro Phosphorylation Studies
The approach that we have chosen is open to criticism since it relies on the assumption that the peptides are recognized with the same efficiency when free in solution or inserted into a polypeptide chain. However, recent studies have determined the optimal consensus sequence phosphorylated by protein kinases by comparing their efficiency in degenerate peptide li-braries (63). In the case of the CTD, this approach appears more valid as its highly repeated nature suggests that this domain has a repetitive three-dimensional structure. Indeed, this suggestion is supported by recent NMR studies (64). Therefore, it is likely that the conformation of 1-or 2-fold repeats of the motif is probably similar to the conformation of the same motif in the whole RNA polymerase II subunit.
The CTD kinases that we chose for this study are good candidates to participate in CTD phosphorylation in vivo. Cytoplasmic MAP kinases are activated by a broad range of agents and migrate into the nucleus only in the presence of inducers. Previous studies have shown that DNA-PK is present in preinitiation complexes and phosphorylates the CTD of endogenous RNA polymerase II (65). The sites phosphorylated by DNA-PK defined in this study are not in full agreement with the serine/threonine ratios already published for this protein kinase (66): the CTD phosphorylated by DNA-PK contains approximately equal amounts of phosphoserines and phosphothreonines. However, positions 2 and 7 indeed contain the highest number of substitutions of serines by threonines (64 Ser/14 Thr). Moreover, this ratio can decrease if DNA-PK preferentially phosphorylates the C-terminal end of the CTD, where the number of substitutions is higher, or if DNA-PK is more specific for threonines (a possibility that has not yet been tested). TFIIH kinase is a good candidate for in vivo CTD phosphorylation because of its presence in the holoenzyme. Moreover, we show in this study that its recognition site corresponds to two adjacent motifs, indicating a good specificity for CTD repeats.
All these protein kinases are located very close to the transcription apparatus, but we cannot conclude about their roles in CTD phosphorylation in vivo. If the CTD is phosphorylated by a single enzyme, neither TFIIH kinase nor DNA-PK and the MAP kinases can explain the in vivo ratio of serine to threonine (1:10) (9). In contrast to the former model, several protein kinases might be involved, inducing different CTD modifications. In both cases, a systematic determination of the specificity of each CTD kinase for the CTD motif remains useful.