A Dual Interface Determines the Recognition of RNA Polymerase II by RNA Capping Enzyme^*

Identifying Structural Determinants for Pol IIO-CE Recognition

As pointed out in the Introduction, the CTD-P-alone model for CE recruitment does not fully explain the precision and specificity of CE targeting. As described in the preceding section, CTD-P alone cannot be responsible for the stoichiometric complex either. A recruitment model solely dependent on CTD-P interactions also fails to account for the recent finding that addition of the CTD to Pol III or the T7 RNA polymerase is not sufficient to direct RNA processing to the CTD-bearing polymerases (52). It is, therefore, necessary to elucidate additional structural determinants required for the preferential binding of CE to Pol IIO as opposed to other forms of RNA polymerases.

As a first step toward this end, we mapped the interaction using the subcomplexes of CE and Pol II that are structurally stable when isolated in solution. These included the Ceg1 (guanylyltransferase) and Cet1 (triphosphatase) subunits of the yeast CE, the Rpb4-Rpb7 subcomplex (consisting of Pol II subunits Rpb4 and Rpb7), and the 10-subunit Pol II lacking Rpb4-Rpb7. In experiments described in the supplemental Data, Section 2.1, we determined that the stable and stoichiometric interaction with CE requires only the 10-subunit core of Pol IIO (supplemental Fig. S2C); the Rpb4-Rpb7 subcomplex is dispensable for this interaction. Moreover, a stable complex was observed between the 12-subunit Pol IIO and Ceg1 but not Cet1 (supplemental Fig. S2, A and B), suggesting that Cet1 is not critical for the interaction with the polymerase.

To further probe the structural features of Pol IIO that might be involved in recruiting CE, we copurified the 12-subunit Pol IIO with recombinant Ceg1-Cet1 by SEC and crystallized the complex (supplemental Fig. S3A). Similar attempts using the 10-subunit Pol IIO did not yield crystals at all. SDS-PAGE analysis revealed Ceg1, Cet1, and Pol II subunits in the crystals (supplemental Fig. S3B). The crystals diffracted weakly, and native anomalous data were eventually collected to 4.7 Å at the zinc edge (supplemental Table S2). Crystallographic phases were experimentally determined using the multi-crystal approach that proved effective for resolving the electron density of Pol II (58). The experimental map of the cocrystal showed Pol II density that matched the known atomic model (59) (Fig. 2A), validating the experimental phase values for the complex. We did not refine the Pol II model against the cocrystal data due to the low resolution. Additional protein-like density not ascribed to Pol II was observed adjacent to the multihelical Foot domain of Rpb1 (Fig. 2, A and B). As a control, diffraction data from Pol IIO crystals obtained in the absence of CE were analyzed, but no significant density was observed adjacent or near to the Rpb1 Foot domain (not shown). These data indicated that the extra density was not due to a folded portion of Pol II. The possibility that the CTD-P on Pol IIO might fold up and account for the additional density was also diminished by the fact that the CTD-P tail of Pol IIO was as susceptible to chymotrypsin digestion as the unstructured CTD of Pol IIA (supplemental Fig. S3C).

Open in a separate window

FIGURE 2.

X-ray diffraction analysis of Pol IIO-CE complex. A, shown is an electron density map of the Pol IIO-CE complex. The map (green, contoured at 1.0 σ) was experimentally phased by using anomalous scattering from the zinc ions bound in the Pol II molecule and, hence, was not biased by any existing model. The refined 12-subunit model (gray and black) of Pol II (59) was readily fitted into the density. The density (CE density) adjacent to the Rpb1 Foot domain (brown and red) is considered to be that of the OB domain of Ceg1 based on the docking experiment shown in supplemental Fig. S3D. B, a stereo-plot of the electron density (blue, 1.0 σ) attached to the surface of Foot (brown and red) is shown. The contact surface is contributed by secondary structural elements H2, H4, and LP (H4-H5) of the Foot domain.

To account for the additional density, we attempted to manually dock domains within the CE structure (18) into the complex. We found that neither the NT domain of Ceg1 subunit nor the Cet1 subunit could be docked into this density, whereas several differently orientated models of the oligonucleotide binding (OB) domain of Ceg1 could be placed inside the density (supplemental Fig. S3D), indicating a local rotational disorder in the cocrystal. A difference Fourier (2F_o − F_1c) map phased by the unrefined Pol II model revealed density also at the same location (not shown), but the density was much weaker than that based on the experimental phases. This was taken as an additional indication for the disorder. This structural disorder could reflect potential different modes of CE binding to Pol IIO in the absence of a transcription bubble and other factors. Nevertheless, the location of the additional experimental density adjacent to a surface on the Foot domain (Fig. 2B) indicated that this Pol II domain might interact with CE and most likely via the OB domain.

The Rpb1 Foot is an independently folded domain in the Pol II structure (50). The Foot domain is conserved among Pol IIs from different species (supplemental Fig. S4A), but it is one of the two regions that vary the most when comparing sequences of the largest subunits from Pol I, Pol II, and Pol III (supplemental Fig. S4, B–D). The structural volume of this domain differs noticeably between Pol I and Pol II (60). These observations suggest a model wherein the Foot domain, in concert with the CTD-P, presents surface determinants for the specific recognition of Pol II by CE. We designed biochemical and functional experiments to investigate this model. We designated the crystallographic interface between the CE OB and Pol II Foot domain as PCI1 (polymerase CE Interface 1) (Fig. 2B) and the previously defined CE interface on CTD-P (51) as PCI2. Based on the Pol II structure, the polymerase component of PCI1 consists of the secondary structural elements H2, H4, and LP (H4-H5), the loop between H4 and H5 (Fig. 2B) of the Foot domain.

Biochemical Characterization of the Pol IIO-CE Interfaces

When we monitored CE binding to different subdomains of Pol II (supplemental Fig. S5A) using GST-mediated affinity chromatography, we observed weak interactions between CE and an isolated Foot domain, in agreement with the crystallographically observed PCI1. Furthermore, we confirmed the Foot as the only non-CTD site for CE recognition (supplemental Fig. S5B; details are in supplemental Data, Section 2.2). No interactions were found between an isolated Cet1 and the Pol II domains (supplemental Fig. S5C), consistent with the result described above that Ceg1, but not Cet1, was essential for binding.

To evaluate the role of the CTD-based PCI2, we produced MBP-CTD, a construct of intact yeast CTD fused with MBP (maltose-binding protein). We phosphorylated it to produce MBP-CTD-P by the same procedure as for generating Pol IIO. SEC results showed that MBP-CTD-P did not form a stable complex with CE (Fig. 3, A and B). We reasoned that formation of the stable complex might require the presence of both PCI1 and PCI2 and, hence, produced an MBP-Foot fusion protein. This construct was combined in trans with MBP-CTD-P and incubated with CE. Each of the proteins in this mixture migrated independently through the SEC column (Fig. 3C). The GST-Foot protein was tested as well and produced similar results (not shown). Therefore, in contrast to Pol IIO, neither PCI1 nor PCI2 alone nor their copresence as separate proteins could support formation of a stoichiometric complex with CE.

Open in a separate window

FIGURE 3.

Separated Foot and CTD-P domains do not support stable interactions with CE. A, shown is a SEC profile of the recombinant CTD-P with the peak fraction indicated by the upward arrow. B, CE separates from the recombinant CTD-P in SEC (upward arrows). The downward arrow approximately marks a potential peak position for a stable complex. C, copresence of the CTD-P and Foot domains in trans does not support formation of a stable complex with the CE. The staggered profiles of each protein (upward arrows) indicate their separation. A complex formed among the proteins would create a comigration peak (approximated by the downward arrow) ahead of the individual peaks. Note that the CTD is unfolded and its migration is abnormally slow. A contaminating protein is marked by an asterisk. MK, an M_r marker.

To combine PCI and PCI2 in cis, we expressed a soluble MBP fusion protein with the Rpb1 module consisting of tandem domains of the Foot, nucleic acid cleft, and CTD (FCC) in their natural order (Rpb1 861–1733) (Fig. 4A). This protein, named BFCC (B for Pol II) module herein, was partially purified using the MBP-amylose affinity chromatography and then phosphorylated in the same way as for Pol II (Fig. 4B). The impurities in the preparation proved to be breakdown products as they all bound to the amylose resin; they likely formed high molecular weight aggregates as indicated by their SEC elution positions, possibly via the Cleft domain, which is mostly buried in the Pol II tertiary folding but exposed in this construct. The phosphorylated (BFCCo) but not the unphosphorylated (BFCCa) module formed a stable and stoichiometric complex with CE (Fig. 4, C and D). It is worth noting that the profile of unbound Ceg1-Cet1 complex (Fig. 4D) does not spread into the fractions that correspond to elution positions of the bound CE. This characteristic is persistent throughout the SEC runs including that shown in Fig. 1C. Because this binding was phosphorylation-dependent and the breakdown products in the BFCC preparation were not responsive to the CTD phosphorylation (hence CTD-less) as determined by mobility shift in SDS-PAGE (Fig. 4B), the stable complex in Fig. 4C was deemed to have resulted from interactions with the BFCCo protein (top band). The breakdown products common to the BFCCa and BFCCo preparations apparently did not influence the results in this assay. The positive interactions detected using only the BFCCo construct also suggests that the multihelical Foot domain is properly folded within this module. These observations demonstrate that the stable stoichiometric interactions between CE and intact Pol IIO can be recapitulated when the Foot and CTD-P domains are linked in cis as they occur in the BFCCo module. As such, the BFCC module offered a means to analyze each component of the dual interface using mutagenesis.

Open in a separate window

FIGURE 4.

Linked Foot and CTD-P domains recapitulate the stable interaction between Pol IIO and CE. A, shown is the BFCC module (magenta) in the Pol II structure. The CTD is not crystallographically determined but is represented by the wavy line to indicate its unfolded state. B, time-course of MAPK2 phosphorylation of the BFCC module fused to MBP. Incubation durations are indicated on top. BFCCa and BFCCo denote the unphosphorylated and phosphorylated forms, respectively. Contaminating proteins (asterisks) in the partially purified samples proved not responsive to the kinase. C, CE forms a stable 1:1 complex with the phosphorylated BFCC module (MBP-BFCCo) as assayed in SEC. Asterisks indicate contaminating proteins not responsive to the phosphorylation. D, CE separates from the unphosphorylated BFCC module (MBP-BFCCa) in SEC. E and F, SEC experiments demonstrating that AFCCo and CFCCo, the phosphorylated modules with a Foot domain from Pol I and Pol III, respectively, do not form a stable complex with CE.

To assess the functional significance of PCI1, we replaced the Foot domain in BFCC with its counterparts from Pol I and Pol III to produce the AFCC and CFCC (A for Pol I; C for Pol III) modules, respectively. These domain-swapped constructs were partially purified and phosphorylated (supplemental Fig. S6A) but failed to form stable complexes with CE (Fig. 4, E and F), indicating that the Foot domains of Pol I and Pol III cannot functionally substitute for that of Pol II and that the Pol II Foot is required for stable association with CE. This biochemical result offers an explanation for the observation that a CTD-tailed mutant Pol III is not able to impart RNA processing to its transcripts in vivo (52). Introduction of the analogous domain-swapped mutations into yeast RPB1 conferred lethality (see below), confirming that the Rpb1 Foot domain is essential for Pol II function in vivo. The possibility that other Pol II subunits (e.g. Rpb5 and Rpb6) might be involved in the interaction with CE was ruled out through the additional experiments described in the supplemental Data, Section 2.2.

A Minimum CTD Length in Pol IIO Is Required for Stable Interaction with CE

We probed PCI2 by evaluating the minimal length of CTD that is capable of maintaining interactions with CE. Mutant BFCC proteins were produced by terminating the CTD domain at the carboxyl ends of the 16th, 14th, 12th, and 10th heptad repeats (Fig. 5A). Again, the degradation products evident in these preparations were not responsive to the phosphorylation as judged by mobility in the SDS-PAGE (supplemental Fig. S6A). In SEC performed at 20 °C, CE formed stable complexes with the construct of 16CTD in a phosphorylation-dependent manner (Fig. 5, B and C). Characteristically, again, the profile of unbound CE did not spread into positions of the complexed CE (Fig. 5C). Complexed CE was also found with BFCCo(14CTD) in the SEC profile (Fig. 5D). However, we were not able to discern the formation of a stable complex with BFCCo(12CTD) (Fig. 5E) or the 10CTD module (not shown), as no CE was detectable by Coomassie staining. However, a slight shift (0.5–1 fraction, 0.15 ml/fraction) of the CE profile toward the peak of BFCCo(12CTD) was noticed, which might suggest a certain level of unstable interactions with the 12CTD module. Overall, these results indicate that CE requires at least 14 linear heptads on the polymerase to form a stable complex at 20 °C in vitro in this assay. It should also be noted that the last 10 nonconsensus residues from the very C terminus of yeast CTD are probably not involved in CE binding, as they were present in the 10CTD construct (Fig. 5A), and no interaction were detected (data not shown). These biochemical data correlated well with phenotypes of the strains (shown below) bearing the same CTD truncations in Rpb1; namely, that the 12CTD allele could not support yeast growth at 18 °C and only allow very slow growth at 30 and 37 °C and the 10CTD allele is inviable at all the temperatures tested.

Open in a separate window

FIGURE 5.

A minimum CTD length of 13-to-14 heptad repeats is required for stable interaction with CE in vitro. A, the protein sequences of truncated CTDs in the BFCC modules used in this analysis are shown. The numbers of heptad repeats are indicated on top, and the binding results from SEC experiments are given below each sequence. Note that the italicized amino acid ID in the 10CTD construct corresponds to the last 10 nonconsensus residues found at the C terminus of yeast CTD. B, the BFCCo with 16 CTD heptads forms a stable complex with CE in SEC. The peak profile of the BFCCo-CE complex is marked below the SDS gel. The elution fraction number increases in the direction of the arrow (top). Asterisks indicate the contaminating proteins that were not responsive to the phosphorylation as shown in supplemental Fig. S6A. A M_r marker (250, 130, 95, 72, 55 and 36 kDa, from the top) was loaded in the left-most lane for B–E. C, the BFCCa with 16 CTD heptads does not form a stable complex with CE in SEC. Note the profile of the unbound CE peak (marked below the gel) did not spread into the migration positions of the complex. D, the BFCCo with 14 CTD heptads forms a stable complex with CE in SEC. Both the Ceg1 and Cet1 subunits of CE were found in the peak fractions corresponding to the complex (marked below the gel). E, the BFCCo with 12 CTD heptads does not form a stable complex with CE in SEC. No proteins of CE could be seen in the fractions corresponding to the complex.

These observations indicate that a CTD containing 13–14 heptad repeats on the yeast Pol II is required and sufficient for the stable interaction with CE. This notion is consistent with earlier findings that the S. pombe CE requires 12–14 CTD heptads for interaction in a two-hybrid assay (61) and that human Pol II with 15 heptads from either the N- or C-terminal half of the CTD supports capping in vivo (62). When considered together with the C. albicans structure mentioned above in which the PCI2 encompasses ∼2.5 CTD heptads (51), our data suggest a possibility that the yeast CE might interact with 2–3 consecutive heptads between registers 11–14 on the native Pol IIO molecule; an alternative possibility is that formation of the complex might require 2–3 consecutive heptads at a certain linker distance from the body of the polymerase – a relaxed form of the former model. Such a restriction of CE interaction to a specific segment within the CTD could be a mechanism that explains why CTD di-heptads appear to be functional units of the CTD (63).

Functional Significance of the Pol IIO-CE Interface

To assess the importance of PCI1 and PCI2 for yeast growth, we used the plasmid shuffle strategy (64) to substitute a wild-type (WT) RPB1 in the yeast with a copy of rpb1 carrying mutations in PCI1 or PCI2. Both the WT and mutant alleles were under the control of the native promoter of RPB1 (“Experimental Procedures”). As shown in Fig. 6A, when the Rpb1 Foot was substituted by A-Foot and C-Foot, its counterparts from Pol I and Pol III (sequences aligned in supplemental Fig. S4B), or by Δ-Foot, a di-glycine linkage between Rpb1 872 and 1059, cells were not viable (Fig. 6A, −Leu +FOA). These results are consistent with the observed loss of CE interactions with the corresponding AFCCo and CFCCo modules in the SEC assays (Fig. 4, E and F). The C-Foot and Δ-Foot mutations displayed a dominant negative growth effect over WT RPB1 allele born on the URA3 plasmid that was used for shuffling (Fig. 6A, −Leu). This phenomenon first manifested in the persistently low frequencies of transformation of the LEU2-marked plasmids carrying these alleles (Fig. 6B). A simple explanation for the dominance is that these mutations generate defective Pol II complexes that blocked or interfered with transcription by the normal Pol II complexes that comprise a WT Rpb1.

Open in a separate window

FIGURE 6.

Phenotypes of the PCI1 and PCI2 mutants. A, growth effects due to Foot swaps and deletion (A-Foot, C-Foot, and Δ-Foot) as tested with 20-fold serially diluted cultures are shown. FOA is toxic to URA3-containing cells and, thus, allows the LEU2-linked alleles to be assayed in the absence of the WT gene. The LEU2-linked RPB1 (WT) served as the positive control, and the empty LEU2 shuffle plasmid (vector) served as the negative control. B, frequencies of the LEU2 plasmids carrying a Foot-swap, Foot deletion, or PCI2 (CTD shortening) mutant rpb1 to transform the shuffle strain that contains RPB1 on the URA3 plasmid are shown. Constant amounts of competent cells and plasmid DNA (key below) were used. C, locations of the mutated residues in Rpb1 Foot domain are shown. Mutated PCI1 residues in the FtH4, FtH2, and FtLP elements are colored red, magenta, and yellow, respectively. D, the PCI1 and PCI2 mutations impair yeast growth. The shuffle assays were performed at 37 °C (not shown), 30 °C (upper left), and 18 °C (lower left). The former two conditions yielded similar phenotypes. The slow-growth defects of PCI1 worsened at 18 °C, indicating cold sensitivity. The PCI2 mutations also slowed yeast growth (right panels). Shortening the CTD to 12 heptads severely impaired growth at 30 °C and prevented growth at 18 °C; the 10CTD allele could not support growth at all temperatures. E, the PCI2 mutations dominantly impair yeast growth at lower temperature (18 °C). The LEU2-marked plasmids carrying rpb1 with the PCI1 and PCI2 mutations were each transformed into a WT strain derived from W303. The selection and growth of LEU⁺ strains were carried out on SC−Leu plates, and growth rates were assessed using 20-fold serial dilution.

To evaluate the importance of Pol II surface residues that are located within PCI1, we made three groups of alanine substitutions in the secondary elements H2, H4, and LP (H4-H5) of the Foot domain (Fig. 2B) and named them FtH2 (K924A/L928A/E931A/K938A), FtH4 (K984A/L988A/K991A/D992A/E995A), and FtLP (K1003A/E1005A/R1012A) (Fig. 6C). These rpb1 alleles supported viability but resulted in slow growth at 30 °C, and the defects became more pronounced at 18 °C (Fig. 6D, left panels).

The PCI2 mutant with a CTD identical to that in the 10CTD module used in the SEC experiment was not able to support growth at all temperatures tested (Fig. 6D, right panels), the 12CTD mutant grew extremely slowly at 30 °C and was inviable at 18 °C, the 14CTD and 16CTD alleles were apparently wild type at 30 °C, but the 14CTD mutant exhibited cold sensitivity. These observations are similar to those made previously by Young and co-workers (65) and Corden and co-workers (66) and support the conclusion that a minimum of 11–12 CTD repeats is required for budding yeast viability. This minimum CTD length required for viability is nearly identical to the aforementioned threshold length, 13–14 repeats, for forming a stable complex with CE in vitro. Combined with the biochemical results, our data pinpoint the impairment of PCI2 and loss of CE interactions as another detrimental consequence of CTD-shortening beyond a minimum length in addition to the disruption of Mediator interactions, 3′-processing, and termination (2, 11, 38).

To test for dominant effects of the non-lethal mutations over an endogenous RPB1, we transformed the LEU2-marked PCI1 and PCI2 alleles (under the native RPB1 promoter) into a WT strain (a derivative of W303). As shown in Fig. 6E, the PCI1 surface mutants (FtH4, FtH2, and FtLP) did not affect growth in the WT background. The 14CTD and 12CTD alleles displayed dominant negative effects at the lower temperature (18 °C), which reflected the low frequency and small colony size in the initial transformation (Fig. 6B). Similar dominant effects had been noticed for S5A substitutions in the CTD and for the complete removal of CTD (67). These mutational studies confirm that the PCI1 and PCI2 sites identified in vitro contribute to essential functions in vivo.

A Specific CTD Segment Involved in the Assembly of Pol II Cotranscriptional Capping Complex

Our data suggest a possibility that CE might interact with a particular segment of 2–3 haptads within the CTD-P when forming a specific complex with Pol IIO. This proposal is consistent with the binding of ∼2.5 heptads by C. albicans CE in the cocrystal structure (51); it could also offer a physical explanation for the observation that CTD functions are confined within di-heptads (63). Factors involved in CTD-linked RNA processing and other nuclear events could use similar modes to interact with the phosphorylated CTD. With the data available, however, it remains a distinct possibility that a longer CTD would afford CE with many alternative binding sites beyond register 11. Resolution of this issue will require further experimentation. It is tempting to suggest that PCI1 might spatially coordinate productive CE binding on the CTD, raising the possibility that higher order complexes between factors and CTD may have some organization along the length of CTD in the absence of primary sequence differences in CTD repeats. A spatial constraint of CTD by bridging the interactions between PCI1 and PCI2 via CE might in turn be important for the productive binding or regulation of other CTD-targeting factors and would represent a role for CE beyond its enzymatic functions.

Definition of Dynamic Regulatory Complexes by Combining Weak Interactions

The relatively weak interactions observed for isolated PCI1 and PCI2 may reflect a dynamic property of the system that may be important for cotranscriptional capping. Conceivably, this process is highly specific and temporally sensitive to ensure that CE actions are minimized toward other classes of RNAs abundant in the nucleus. In this light, a stable complex formed based on combining the individually weak PCI1 and PCI2 seems well suited for this task, as dissolution of this complex would occur rapidly upon perturbation of either interface. For instance, PCI1 and/or PCI2 may separate as a result of the CE conformational change that accompanies the capping reaction; likely, the action by the Ser-5-specific CTD phosphatase Ssu72 may weaken the PCI2 interactions at a post-capping step (70), although the mechanism that determines the order of these events is unknown. It also appears that this principle, namely combining weak interactions alongside post-translational modifications such as phosphorylation, may universally govern the dynamic assembly and disassembly of multiprotein regulatory complexes that function in transcription and possibly other pathways.

Relationship of the Dual Interface to Other Aspects of CE Recruitment

Our model is consistent with the majority of published data that implicate CTD phosphorylation in CE recruitment. In addition, the interaction based on PCI1 offers an alternative explanation for the residual coupling activities for capping (a 4-fold enhancement) by the CTD-less Pol IIA (71) and for the inability of a CTD-tailed human Pol III to instigate processing to its transcripts (52). However, our data do not address the reported inhibition of Pol II due to direct interactions with the triphosphatase Cet1 (56).

Our model highlights the binding of CE via the OB domain of Ceg1 with certain disorder, which would leave the NT domain of Ceg1 free to interact with CTD-P to form PCI2. We believe this and additional disorders may have prevented the visualization of electron density for the NT domain and the entire Cet1 subunit, both present in the crystal (supplemental Fig. S3B). The absence of the NT domain in our structure is consistent with the known flexibility between NT and OB domains in guanylyltransferases (19). Also, the absence of Cet1 is consistent with the observation that Cet1 is flexibly attached to Ceg1 OB domain via a linker peptide (Cet1 241–268) (18).

Because our Pol IIO-CE complex does not contain a template and DNA/RNA hybrid, it represents at best a core structure of a complete elongation complex that may ordinarily include Spt4-Spt5. Given the functional involvement of Spt5 in early elongation and the salubrious contributions of Spt5 C-terminal region to CE recruitment as seen in S. pombe (46), the Spt5 C-terminal region might interact with parts of Cet1 and/or the Ceg1 NT domain to help orient and position the entire CE in a productive assembly on the polymerase.