Structure and mechanism of the plant RNA polymerase V

In addition to the conserved RNA polymerases I to III (Pols I to III) in eukaryotes, two atypical polymerases, Pols IV and V, specifically produce noncoding RNA in the RNA-directed DNA methylation pathway in plants. Here, we report on the structures of cauliflower Pol V in the free and elongation conformations. A conserved tyrosine residue of NRPE2 stacks with a double-stranded DNA branch of the transcription bubble to potentially attenuate elongation by inducing transcription stalling. The nontemplate DNA strand is captured by NRPE2 to enhance backtracking, thereby increasing 3′-5′ cleavage, which likely underpins Pol V’s high fidelity. The structures also illuminate the mechanism of Pol V transcription stalling and enhanced backtracking, which may be important for Pol V’s retention on chromatin to serve its function in tethering downstream factors for RNA-directed DNA methylation. Description Chromatin retention of Pol V A plant atypical RNA polymerase, Pol V, specifically synthesizes the scaffold long noncoding RNA for recruiting downstream effectors to chromatin to mediate DNA methylation. Xie et al. report the structure of a cauliflower Pol V in elongation conformation. Distinct from Pol II, the second subunit of Pol V, NRPE2, specifically stacks with the upcoming double-stranded DNA branch to attenuate transcription and captures the nontemplate DNA strand of the transcription bubble to enhance the backtracking, suggesting a chromatin retention mechanism underlying the scaffold function of Pol V. —DJ Structure of plant Pol V elucidates the mechanism underlying transcription stalling and enhanced backtracking.


Structure determination of cauliflower Pol V
To investigate the Pol V transcription mechanism, we obtained a monoclonal antibody against a C-terminal peptide of cauliflower (Brassica oleracea var. botrytis) NRPE1, a distinctive Pol V subunit. We used this antibody to purify Pol V from cauliflower inflorescence ( Fig. 1A and fig. S1, A and B), a tissue enriched in dividing cells and DdRPs (35,36). All 12 subunits of Pol V were confirmed in the purified material by mass spectroscopy (MS) ( fig. S1, B and C). Although Arabidopsis Pol V was shown to only transcribe the bipartite scaffold of a template DNA (DNA T ) plus an RNA primer (12), the purified B. oleracea Pol V (BoPol V) showed substantial transcription activity toward both a bipartite scaffold and a transcription bubble (fig. S1, D and E). We determined the cryo-electron microscopy (cryo-EM) structures of BoPol V in both apo and transcription bubble-containing elongation complex (EC) conformations at 3.57-and 2.73-Å resolution, respectively (Fig. 1, figs. S2 to S4, and tables S1 and S2). Overall, the two structures were similar with a superimposition root mean square deviation of 0.8 Å. Despite being detected by MS, the NRPE4 and NRPE7 subunits could not be traced in the density, whereas the other 10 subunits could be traced and modeled in both states. In Pol II, the RPB4-RPB7 subcomplex interacts with the clamp domain of RPB2 and the C terminus of RPB1 (37). These two regions showed weak density in our Pol V structure, which suggests loose binding and/or flexible conformations of NRPE4-7, potentially explaining why they were missing in our structure.
Overall structure of BoPol V Both Pols IV and V evolved from Pol II (38). In Arabidopsis, Pol V has distinctive subunits NRPE1, NRPE5, and NRPE7, whereas NRPE (D)2 and NRPE(D)4 are shared by Pols V and IV but are different from those in Pol II. All other small subunits are shared by Pols II, IV, and V (27). The overall structure of Pol V adopts the classic Pol architecture and resembles other Pols, especially Pol IV ( fig. S5) (13,30,31,33). Like Pol IV, Pol V lacks the binding surface for Pol II transcription factors, such as the initiation-related factor TFIIB and cleavagerelated factor TFIIS (fig. S6, A and B) (39,40). This is consistent with the lack of these factors in our MS data and indicates that Pol V acts via a distinctive regulatory mechanism compared with Pol II. Despite being encoded by different genes, NRPE5 resembles RPB5 of Pol II and NRPD5 of Pol IV ( fig. S6C). By contrast, NRPE9, which is common to Pols II, IV, and V, displays notable structural differences in the different Pols. Although the NRPE9 Nterminal jaw domains of Pols II, IV, and V occupy similar positions, the C-terminal zinc ribbon domain of NRPE9 occupies a specific position to interact with NRPE1, which is different from that seen in Pol II, in which RPB9 interacts with RPB2 ( fig. S6D) (13,30). Compared with Pol IV, the Pol V NRPE9 zinc ribbon domain resides in a similar position but has ã 40°rotation ( fig. S6E) (13). The funnel domain of Pol V NRPE1 is~50 residues shorter than that of Pol IV NRPD1, resulting in a reduced binding interface between NRPE1 and the NRPE9 zinc ribbon domain (figs. S6F and S7).
In the EC structure, the last 3′-end RNA nucleotide is linked to the RNA ( fig. S4Q). It pairs with the DNA T and occupies the +1 position ( Fig. 2A), representing a pretranslocation conformation (41). In the active center of DdRPs, two Mg 2+ ions, metals A and B, are involved in RNA substrate and incoming nucleoside triphosphate (NTP) binding and catalysis (fig. S8A) (41). Because we did not observe NTP binding in our structure, only one Mg 2+ ion at the metal A binding site was observed to be coordinated by Asp 449 , Asp 451 , and Asp 453 of NRPE1, forming the catalytic site, consistent with a previous biochemical and genetic study (25) (Fig. 2A and fig. S8B). Like other Pols, the active center of Pol V has two essential structural elements: the bridge helix (BH) and the trigger loop (TL) ( Fig. 2A). The Pol V BH resembles the Pol II BH, despite multiple sequence variations (29, 41) ( Fig. 2B and fig. S7).   The Pol V TL is shorter than that in Pol II and also exhibits sequence variations ( fig. S7). In contrast to the Pol II TL which shows conformational change in the transition from the apo form to the EC form (29,41) (Fig. 2C), the TL in both the Pol V apo and EC structures adopts a similar compacted conformation (Fig. 2D), which suggests less conformational flexibility of the Pol V TL. Compared with Pols I and III, although their BHs adopt conformations similar to that of Pol V ( fig. S8, C and D), the TLs of Pols I and III are flexible such that they are partially disordered in both the apo and EC forms (fig. S8, E and F) (31,33,42). Given that the TL conformational dynamics is required for NTP substrate binding to promote the transcription reaction, the less-flexible Pol V TL may decrease the NTP incorporation, resulting in lower activity (12,41).

Interactions between NRPE2 and the transcription bubble
Except for some terminal nucleotides, the entire transcription bubble, including the unpaired region of the nontemplate DNA (DNA NT -UPR), which is often disordered in Pol II structures (43,44), can be fully traced and modeled in our Pol V EC structure (Fig. 3, A and B). This is likely due to the extensive interactions between DNA NT -UPR and NRPE2 (Fig. 3B), the shared second subunit of Pols IV and V. Overall, DNA NT -UPR is clamped between the lobe and fork domains of NRPE2 (Fig. 3C). At the downstream double-stranded DNA (dsDNA) branching site of the transcription bubble, the first unpaired nucleotides, G19′ of the DNA NT and A16 of the DNA T , form a pseudo-pair (Fig. 3D), probably mimicking the last DNA base pair in natural transcription. The NRPE2 fork loop residue Tyr 495 -which is conserved in NRPE(D)2 across multiple species, but not in the NRPA2, NRPB2, and NRPC2 clades (Fig. 3E)-specifically inserts to the DNA branching position to stack with the upcoming A16-G19′ pseudo-pair while also forming a hydrogen bond with the phosphate group of the DNA T G17 (Fig. 3D). The specific stacking and hydrogen-bonding interactions by the bulky side chain of Tyr 495 may help Pol V to stabilize the transcription bubble, potentially allowing Pol V pausing to help it serve as a scaffold to recruit downstream effectors, and not only to produce and release RNA transcripts. Moreover, bases G14′, A16′, T17′, and T18′ of DNA NT -UPR are anchored, through extensive interactions, by a series of surface base-binding pockets on NRPE2 (Fig.  3 the steric clashes with Pol I and III and a loss of specific interactions upon modeling into Pol II ( fig. S9) (13,30,31,33), suggesting that this is a feature specific to the Pol IV-V clade.

BoPol V attenuates transcription elongation
To investigate the biochemical relevance of our structures, we performed in vitro transcription assays using purified BoPol V and transcription bubble substrates with different gaps between the RNA primer 3′ end and the downstream dsDNA branching site, which are stacked by Tyr 495 in our structure (Fig. 4A, scaffolds 2 to 5). Pol V showed substantial transcription activity, with an accumulation of transcript corresponding to the unpaired region of DNA T (DNA T -UPR) (Fig. 4B). By contrast, the bipartite substrate lacking the DNA NT did not show accumulation of a specific transcript (Fig. 4B), which demonstrates transcription pausing at the downstream DNA branching site, consistent with NRPE2 Tyr 495 obstructing Pol V translocation at this site. Low levels of longer transcripts were also observed with the transcription bubble substrates (Fig. 4B), which indicates that Pol V has a weak ability to open downstream paired DNA. By contrast, purified cauliflower Pol II (fig. S10) produced longer transcripts without an accumulation of DNA T -UPR transcription product (Fig. 4C), suggesting that transcription pausing at the branch site is a specific feature of Pol V, or more generally of the Pol IV-V clade, consistent with Tyr 495 being conserved in NRPE(D)2 but not in that of the other clades (Fig. 3E). Our results suggest that Pol V has relatively high transcriptional activity on the DNA T -UPR but is less efficient in opening paired DNA, likely because of the conserved bulky Tyr 495 stacking with, and blocking access to, downstream paired DNA. Because the natural template is always fully paired, Pol V may constitutively exhibit low transcription activity, plausibly contributing to transcriptional pausing and thereby enhancing Pol V retention on chromatin to support its function in tethering chromatin factors to promote RdDM.

NRPE2-DNA NT -UPR interactions enhance backtracking
Transcription pausing is a key step in the induction of transcription backtracking and subsequent 3′-5′ cleavage for proofreading (39,45). Consistent with transcription pausing, abundant bands corresponding to the RNA-backtracked 3′-5′ cleavage product were observed in our Pol V activity assays ( Fig. 4 and fig. S1, D and E). In the presence of Mg 2+ and absence of NTP, Pol V almost exclusively showed RNA primer cleavage activity without transcription elongation (Fig. 4C). EDTA inhibited this cleavage, which confirmed a Mg 2+ -dependent cleavage mechanism, similar to that of Pol II (Fig. 4C) (39). Addition of NTP and Mg 2+ trig-gered BoPol V transcription elongation while maintaining strong cleavage (Fig. 4C). By contrast, under the same reaction conditions, BoPol II showed stronger transcription elongation and lower cleavage activity (12, 28) (Fig. 4C). For the DdRP proofreading function to achieve high fidelity (45), 3′-5′ cleavage is required, which is consistent with the hypothesis that enhanced 3′-5′ cleavage activity underpins the higher fidelity of Pol V (28).
Excluding the A16-G19′ pseudo-pair, the Pol V transcription bubble contains a DNA NT -UPR of 10 nucleotides (nt) with the downstream branching site-proximal bases T18′, T17′, A16′, and G14′ specifically captured by NRPE2 (Fig. 3  intact transcription bubble, breaking the DNA NT at the downstream branch site or removing the DNA NT -UPR (Fig. 4A, scaffolds 6 and 8) resulted in similar patterns that decreased both transcription and cleavage activities (Fig. 4D). This is likely because of the loss of interaction with the downstream reaction center in both cases. DNA NT breakage at the upstream branch site (Fig. 4A, scaffold 7) resulted in almost full transcription activity and partial cleavage activity, relative to the intact bubble (Fig. 4D). Moreover, the shortening of DNA NT while only keeping the downstream paired region decreased both the elongation and cleavage activities as well ( fig. S11). Overall, these data suggest that the downstream branch and DNA NT -UPR conformation both influence transcription and cleavage activity.
To investigate the NRPE2-DNA NT -UPR interaction that is specific to Pol V (Fig. 3, B and F to I), we eliminated individual DNA NT -UPR nucleotides by replacing them with an apurinicapyrimidinic (AP) site while retaining DNA NT -UPR integrity ( fig. S12A). The loss of DNA NT -UPR bases mildly reduced cleavage activity, with downstream bases having a greater effect than upstream ones ( fig. S12B), which is consistent with downstream DNA NT -UPR nucleotides having more interactions with NRPE2 (Fig. 3B). By contrast, all substitutions retained transcription activity similar to that of the unmodified substrate ( fig. S12B). Nevertheless, all of the different individual substitutions only weakly affected cleavage activity. We also simultaneously substituted five contiguous downstream DNA NT -UPR nucleotides with AP sites (Fig. 4A, scaffold 5X), which yielded a loss of cleavage activity but did not affect the elongation (Fig. 4E). By contrast, this multiple substitution showed no effect on either transcription or cleavage using BoPol II (Fig. 4F), which suggests that the importance of these base interactions is specific to Pol V, or likely the Pol IV-V clade. These results suggest that the interaction of the DNA NT -UPR with NRPE2 is mainly responsible for enhancing Pol V cleavage activity. The 3′-5′ cleavage of all Pols requires transcription bubble backtracking to feed RNA into the active site on the largest subunit, that is, NRPE1 in Pol V. Therefore, the NRPE2-DNA NT -UPR interaction promotes increased backtracking to feed RNA into NRPE1, resulting in enhanced Pol V 3′-5′ cleavage activity, a mechanism different from the intrinsic backtracking that is common in Pols.

Discussion
Distinct from Pols I to III, whose principal function is producing transcripts for release, Pols IV and V have evolved specialized transcription features adapted for RdDM. For example, Pol IV generates a single-stranded RNA 3′ end as a substrate for RDR2 through backtracking (13), and Pol V likely limits its transcription rate to prolong its chromatin occupancy, thereby providing a scaffolding function for the recruitment of chromatin factors. We speculate that the Pol IV-V-clade DdRPs evolved the shared NRPE(D)2 subunit so that they can stall transcription through the conserved NRPE(D)2 tyrosine residue and enhance backtracking through the NRPE(D)2-DNA NT -UPR interactions. In the case of Pol V, the strong backtracking and transcription stalling may regulate the equilibrium between forward and backward steps, which delays termination and release of transcripts and thus leads to Pol V-long noncoding RNA complex retention on chromatin to promote its scaffolding function. This may also explain the short and relatively uniform length of Pol IV transcripts (35 to 50 nt) (8,13,17). Pol IV likely backtracks to feed RNA into RDR2 through an interpolymerase channel that likely bypasses cleavage (13), which may also explain its lower fidelity (28). Given NRPE(D)2-induced transcription stalling and high backtracking activity, Pol IV-RNA is unable to undergo long extension before backtracking. Pol IV-RNA may be captured by RDR2 once the nascent transcript is long enough to trigger the dsRNA synthesis by RDR2 and subsequent termination of Pol IV transcription (13).