We have provided strong evidence that BUR1/BUR2 is required for high-level Ser2 phosphorylation in vivo, and that it contributes a proportion of Ser2P independently of the major Ser2 CTD kinase CTK1, particularly near promoters. We found that impairing BUR1 activity by eliminating BUR2 or chemically inhibiting the bur1-as product reduces Ser2P with an attendant increase in hypophosphorylated RPB1 in bulk Pol II. The bur2Δ and bur1-as mutations also decrease Ser2P occupancy in the ARG1 coding sequences without a commensurate reduction in occupancy of elongating Pol II itself. BUR1/BUR2 makes a smaller contribution than CTK1 to the overall level of Ser2P in bulk Pol II. However, bur2Δ, bur1-as and ctk1Δ have comparable effects in reducing Ser2P at the 5′ end of ARG1, whereas ctk1Δ evokes a stronger decrease in Ser2P at the 3′ end of the gene. Moreover, bur2Δ, bur1-as and ctk1Δ all conferred strong reductions in Ser2P in the pool of bulk Pol II hyperphosphorylated on Ser5 (immunoprecipitated with H14 antibodies). These results suggest that BUR1/BUR2 makes a substantial contribution to Ser2P in Pol II molecules already phosphorylated on Ser5 and located near the promoter, whereas CTK1 is responsible for the majority of Ser2P and is the predominant Ser2 CTD kinase distal to the promoter.
Although a decrease in Ser2P at PMA1
was noted previously in bur1-23
cells, it was attributed to decreased occupancy of Pol II itself at this gene. In addition, while Ser2P in bulk Pol II was reduced in a bur2Δ
strain, this was judged to be an indirect consequence of lower levels of elongating Pol II (Keogh et al., 2003
). Our findings suggest that BUR1/BUR2's contribution to Ser2P is not limited to its role in producing elongating Pol II molecules as substrates for CTK1. First, Pol II occupancy in the ARG1
ORF is not reduced in our bur2Δ
strain, such that the Ser2P:RPB3 ratio at this gene is diminished in this mutant. Second, the Ser2P content of Pol II molecules phosphoryated on Ser5P, which should represent Pol II molecules engaged in elongation, is reduced in both bur1-as
mutants. Most importantly, inactivation of the bur1-as
mutant in ctk1Δ
cells eliminated the residual Ser2P in bulk Pol II that remains after eliminating CTK1 alone, proving that BUR1 can promote Ser2P formation independently of CTK1.
Combining these findings with the sequence similarity of BUR1 to P-TEFb, and genetic similarities between BUR1
, leads us to propose that BUR1/BUR2 functions as a Ser2 CTD kinase in vivo. Nevertheless, a proportion of the decline in Ser2P evoked by bur2Δ
mutations, at least at promoter-distal locations, probably results indirectly from reduced phosphorylation by CTK1. This follows from our finding that ctk1Δ
eliminates ≈90% of the detectable Ser2P in bulk Pol II, yet inactivating BUR1/BUR2 eliminates much more than 10% of the total Ser2P. One possibility is that Ser2 phosphorylation of a limited number of CTD repeats by BUR1/BUR2 near the promoter can enhance the ability of CTK1 to phosphorylate Ser2 throughout the CTD, and counteract the Ser2P phosphatase, as elongation proceeds downstream. This hypothesis is difficult to test in vitro because steady-state kinetic analysis of CTK1 phosphorylation of the full-length CTD is problematic (Jones et al., 2004
), and it would impossible at present to reconstitute the extent or pattern of partial CTD phosphorylation by BUR1/BUR2 that prevails in vivo. It is also possible that phosphorylation of other, unknown substrates of BUR1/BUR2 indirectly stimulates Ser2P phosphorylation by CTK1 in vivo.
Our findings that bur2Δ and bur1-as produce a substantial decrease in Ser2P at the 5′ end of ARG1, and in Pol II hyperphosphorylated on Ser5, fit with our observations that BUR1/BUR2 occupancy peaks near the ARG1 and GAL1 promoters, and our discovery that recruitment of BUR1/BUR2 is stimulated by the Ser5 CTD kinase KIN28. We discovered that the C-terminal half of BUR1 contains a CID capable of binding Ser5P peptides in vitro. Removing this CID from BUR1 by the −CΔ mutation preferentially reduces BUR1/BUR2 occupancies at the 5′ end of the ARG1 ORF, so that they peak in the 3′ end of the ORF. The −CΔ and -7Ala substitutions in the BUR1 CID also reduce the level of Ser2P in bulk Pol II. Hence, we propose that binding of the BUR1 CID to Ser5P, generated by KIN28, stimulates BUR1/BUR2 recruitment to the 5′ end of the gene (), enhancing its ability to phosphorylate Ser2 early in the elongation cycle (). We envision a cascade of CTD phosphorylation, wherein Ser5 phosphorylation by KIN28 enhances Ser2 phosphorylation by BUR1/BUR2 in the same or adjoining CTD repeats of promoter-proximal Pol II, which stimulates or gives way to Ser2 phosphorylation by CTK1 further downstream in the coding sequences ().
Schematic model of Ser5P-stimulated BUR1/BUR2 recruitment and differential contributions of BUR1/BUR2 and CTK1 to Ser2P and H3-K36Me3 formation at 5′ and 3′ ends of a coding sequence
In addition to stimulating BUR1/BUR2 recruitment, Ser5 phosphorylation by KIN28 might also enhance the ability of BUR1/BUR2 to phosphorylate Ser2, as shown previously for CTK1 (Jones et al., 2004
). Our purified BUR1/BUR2 does not phosphorylate Ser5-phosphorylated or unphosphorylated CTD peptides of 4 heptad repeats under conditions where it phosphorylates a GST-CTD substrate with the native 26 repeats (data not shown). While this is an interesting observation for future study, it precluded our ability to determine if Ser5P enhances Ser2 phosphorylation by BUR1/BUR2 in vitro.
mutations confer 6-AU sensitivity and are synthetically lethal with spt5-194
, but bur1-CΔ
is resistant to 6-AU and confers only a slight growth defect in the spt5-194
background (Keogh et al., 2003
). This suggests that decreasing BUR1/BUR2 recruitment to the Ser5-phosphorylated CTD by the -CΔ
mutation has a modest effect on elongation, which agrees with our finding that Ser2P levels are only moderately reduced in the bur1-CΔ
mutant. However, bur1-CΔ
is synthetically lethal with ctk1Δ
, which can now be explained by proposing that the moderate decrease in Ser2P levels conferred by bur1-CΔ
is intolerable in the absence of CTK1. The same explanation can be extended to the bur1-as
mutant, which produces a moderate decrease in Ser2P levels in CTK1
cells and strongly impairs growth only in the ctk1Δ
Considering that BUR1
is essential, it may seem surprising that the bur1-as
mutant has almost no growth defect at high concentrations of NM-PP1, even though much lower NM-PP1 concentrations eliminate bur1-as/BUR2 kinase activity in vitro. We presume that NM-PP1 does not completely inhibit the bur1-as
product in vivo, and that low-level kinase activity is sufficient for viability. Supporting this idea, substitution of Thr-240 in the BUR1 activation loop nearly destroys BUR1 kinase activity in vitro (Keogh et al., 2003
; Yao and Prelich, 2002
), but has little effect on growth in otherwise WT cells (Yao and Prelich, 2002
) and confers only slow growth in the ctk1Δ
background (Keogh et al., 2003
). On the other hand, point mutations in conserved residues of the BUR1
kinase domain that likewise abolish kinase activity in vitro are lethal, presumably because they completely eliminate kinase function in vivo.
Our conclusion that BUR1/BUR2 plays an important role in Ser2 phosphorylation of promoter-proximal Pol II provides new insights into the observation that BUR2 promotes H3-K36Me3 formation, especially near the 5′ ends of constitutively expressed genes (Chu et al., 2007
). We made a similar observation for induced ARG1
, finding that H3-K36Me3 occupancy is reduced by inactivation of the bur1-as
product near the 5′ end of the gene (Fig. S8
). Ser2 phosphorylation by CTK1 stimulates H3-K36 trimethylation by SET2 in downstream coding sequences (Kizer et al., 2005
; Krogan et al., 2003b
; Li et al., 2003
; Xiao et al., 2003
). Hence, our proposal that BUR1/BUR2 phosphorylates Ser2 on promoter-proximal Pol II molecules provides a possible explanation for the role of BUR2 in stimulating H3-K36Me3 formation near promoters (). Considering that SET2 contains a CID that interacts preferentially with CTD peptides doubly phosphorylated on Ser2 and Ser5 (Kizer et al., 2005
), BUR1/BUR2-mediated Ser2 phosphorylation of CTD repeats already phosphorylated on Ser5 should enhance SET2 function near the promoter. The effect of bur2Δ
in reducing H3-K36Me3 formation was not limited to the 5′ ends of the PYK1
genes (Chu et al., 2007
), which fits with our finding that bur2Δ
reduces Ser2P levels throughout the ARG1
ORF, and only makes a proportionately greater contribution at the 5′end.
Our ChIP analysis suggests that the occupancies of myc-BUR1, myc-BUR2 and myc-CTK1 do not vary greatly across the ARG1
gene, with BUR1/BUR2 moderately exceeding CTK1 at the 5′ end owing to the BUR1 CID (cf. results in , , and ). Thus, the fact that BUR1/BUR2 and CTK1 make roughly equal contributions to Ser2P formation at the 5′ end of ARG1
might be explained quite simply by proposing that BUR1/BUR2 and CTK1 have similar kinase activities, as well as similar occupancies, near the promoter. On the other hand, the much greater contribution of CTK1 to Ser2P at the 3′end of ARG1
seems to imply a change in kinase activity between the 5′ and 3′ ends of ARG1
for BUR1/BUR2, CTK1, or both. For example, CTK1 could become more active as elongation proceeds, perhaps owing to the stimulatory effect of BUR1/BUR2 on CTK1 function deduced from our experiments, or to some other modification of the CTD. There is evidence that transient accumulation of H2B-Ub impedes CTK1 recruitment during GAL1
induction (Wyce et al., 2007
), leading us to consider whether this mechanism could have a role in reducing CTK1 activity near the ARG1
promoter. However, it was reported that H2B-Ub accumulates transiently across the GAL1
ORF, not only at the 5′end. And as noted above, CTK1 occupancy is not substantially lower at the 5′ versus 3′ end of ARG1
(), at least after 30 min of induction when our ChIP measurements were made.
Alternatively, BUR1/BUR2 might become less functional as transcription proceeds downstream. It could be proposed that a reduction in Ser5P distal from the promoter would diminish BUR1/BUR2 activity towards the 3′ end. This may be unlikely, however, because Ser5P remains quite high at the 3′ end of ARG1
). In fact, it is uncertain whether Ser5P levels, or only the reactivity of RPB1 to Ser5P-specific (H14) antibodies, declines as elongation proceeds (Phatnani et al., 2004
). In addition, Ser2P formation by CTK1 is highly stimulated by Ser5P (Jones et al., 2004
), so that loss of Ser5P would likewise reduce CTK1 function towards the 3′ end of the gene. Uncovering the mechanisms responsible for the greatly different contributions of BUR1/BUR2 and CTK1 to Ser2P formation between the 5′ and 3′ ends of ARG1
remains an important goal for future research.