Assembly of a double-stranded RNA synthesizing complex: RNA-DEPENDENT RNA POLYMERASE 2 docks with NUCLEAR RNA POLYMERASE IV at the clamp domain

In plants, transcription of selfish genetic elements such as transposons and DNA viruses is suppressed by RNA-directed DNA methylation. This process is guided by 24 nt short-interfering RNAs (siRNAs) whose double-stranded precursors are synthesized by DNA-dependent NUCLEAR RNA POLYMERASE IV (Pol IV) and RNA-DEPENDENT RNA POLYMERASE 2 (RDR2). Pol IV and RDR2 co-immunoprecipitate, and their activities are tightly coupled, yet the basis for their association is unknown. Here, we show that RDR2 stably associates with Pol IV’s largest catalytic subunit, NRPD1 at three sites, all within the clamp module. The clamp is a ubiquitous feature of DNA-dependent RNA polymerases that opens to allow DNA template entry and closes to encase the DNA-RNA hybrid adjacent to the RNA exit channel. The clamp also provides binding sites for polymerase-specific subunits or regulatory proteins, thus RDR2 binding to the Pol IV clamp is consistent with this theme. Within RDR2, the site of interaction with NRPD1 is very near the catalytic center. The locations of the NRPD1-RDR2 contact sites suggest a model in which transcripts emanating from Pol IV’s RNA exit channel align with the template cleft of RDR2, facilitating rapid conversion of terminated Pol IV transcripts into double-stranded RNAs. Significance Statement Short interfering RNAs (siRNAs) play important roles in gene regulation by inhibiting mRNA translation into proteins or by guiding chromatin modifications that inhibit gene transcription. In plants, transcriptional gene silencing is guided by siRNAs derived from double-stranded (ds) RNAs generated by coupling the activities of DNA-dependent NUCLEAR RNA POLYMERASE IV and RNA-DEPENDENT RNA POLYMERASE 2. We show that the physical basis for Pol IV-RDR2 coupling is RDR2 binding to the clamp domain of Pol IV’s largest subunit. The positions of the protein docking sites suggest that nascent Pol IV transcripts are generated in close proximity to RDR2’s catalytic site, enabling rapid conversion of Pol IV transcripts into dsRNAs.


Abstract
In plants, transcription of selfish genetic elements such as transposons and DNA viruses is suppressed by RNA-directed DNA methylation. This process is guided by 24 nt short-interfering RNAs (siRNAs) whose double-stranded precursors are synthesized by DNA-dependent NUCLEAR RNA POLYMERASE IV (Pol IV) and RNA-DEPENDENT RNA POLYMERASE 2 (RDR2). Pol IV and RDR2 co-immunoprecipitate, and their activities are tightly coupled, yet the basis for their association is unknown. Here, we show that RDR2 stably associates with Pol IV's largest catalytic subunit, NRPD1 at three sites, all within the clamp module. The clamp is a ubiquitous feature of DNA-dependent RNA polymerases that opens to allow DNA template entry and closes to encase the DNA-RNA hybrid adjacent to the RNA exit channel. The clamp also provides binding sites for polymerase-specific subunits or regulatory proteins, thus RDR2 binding to the Pol IV clamp is consistent with this theme. Within RDR2, the site of interaction with NRPD1 is very near the catalytic center. The locations of the NRPD1-RDR2 contact sites suggest a model in which transcripts emanating from Pol IV's RNA exit channel align with the template cleft of RDR2, facilitating rapid conversion of terminated Pol IV transcripts into double-stranded RNAs.

Introduction
In eukaryotes, noncoding RNAs guide transcriptional gene silencing to keep transposons, repeated elements and viruses in check, thus defending against genome instability (1-3). These noncoding RNAs include siRNAs and scaffold RNAs to which the siRNAs basepair, an interaction that enables the recruitment of chromatin modifying complexes to the associated chromosomal locus (4)(5)(6). In most eukaryotes, noncoding silencing RNAs are dependent on DNA-dependent RNA Polymerase II (Pol II). However, plants evolved two Pol II-derived DNAdependent RNA polymerases to specialize in noncoding RNA synthesis (7)(8)(9)(10), NUCLEAR RNA POLYMERASE IV (Pol IV) and NUCLEAR RNA POLYMERASE V (Pol V) (7), that play non-redundant roles in a process known as RNA-directed DNA methylation (11,12). Pol IV transcribes chromosomal DNA to generate relatively short transcripts of 25-45 nt (13,14).
An RNA-dependent RNA polymerase, RDR2 (15) then uses the Pol IV transcripts as templates to synthesize complementary strands, yielding double-stranded RNAs (dsRNAs) (16). Resulting dsRNAs are then cleaved by DICER-LIKE 3 (DCL3) to produce short-interfering RNAs (siRNAs) that can be 24 nt or 23 nt in length (16). The 24 nt siRNAs become stably associated with ARGONAUTE 4 (AGO4), or related Argonaute family members, yielding RNA-Induced Silencing Complexes (RISCs) (17)(18)(19). RISCs then find their target loci via siRNA basepairing with noncoding scaffold RNAs transcribed by NUCLEAR RNA POLYMERASE V (Pol V) (20,21). Protein-protein interactions between AGO4 and the C-terminal domain (CTD) of the Pol V largest subunit (22) or between AGO4 and the Pol V-associated protein, SPT5L (23,24) also contribute to RISC recruitment to sites of Pol V transcription. Once recruited, RISCs facilitate recruitment of the de novo DNA methyltransferase, DRM2 (25) as well as histone-modifying activities that act in coordination with DNA methylation (11,12,26). Collectively, these activities generate chromatin states that are refractive to promoter-dependent gene transcription, thus accounting for gene silencing.
Using purified Pol IV, RDR2 and DCL3, the siRNA biogenesis phase of the RNAdirected DNA methylation pathway has been recapitulated in vitro (16). These experiments revealed that Pol IV and RDR2 enzymatic activities are coupled, with Pol IV's distinctive termination mechanism somehow facilitating the direct hand-off, or channeling, of transcripts to RDR2 (27). In keeping with their coordinated activities, Pol IV and RDR2 co-immunoprecipitate in Arabidopsis thaliana (28,29) ands Zea mays (the maize RDR2 ortholog is MOP1 (30).
However, it is not known whether the two enzymes interact directly or via one or more bridging molecules.
Using several independent methods, we show that recombinant RDR2 directly interacts with the largest catalytic subunit of Pol IV, NRPD1. By testing NRPD1 and RDR2 deletion constructs, synthetic peptide arrays, and chemical crosslinking and mass spectroscopy, we identify three intervals within the predicted clamp module of NRPD1 that interact with sequences near the RDR2 catalytic center. Based on homology modeling to related polymerases, we propose that Pol IV-RDR2 docking places the Pol IV RNA exit channel in close proximity to the RDR2 active site, enabling efficient dsRNA synthesis.

Recombinant RDR2 has RNA-dependent RNA polymerase activity
Using a baculovirus vector, we expressed in insect cells a recombinant gene that encodes wild-type RDR2 enginered to have a V5 epitope tag and His tag at the N-terminus. A derivative active site mutant (RDR2-asm; Figure 1A) was similarly expressed. Nickel affinity and gel filtration chromatography yielded highly purified RDR2 (Fig 1A, lanes 3 and 5) whose identities were verified by immunoblotting using anti-V5 or anti-RDR2 antibodies (Fig 1A, bottom). Upon analytical gel filtration chromatography, both proteins elute as single peaks with estimated masses of ~140 kD (Fig 1B, see insert), in good agreement with their predicted monomeric masses of ~134 kD.
To test for RNA-dependent RNA polymerase activity, RDR2 and RDR2-asm were provided a 32 P end-labeled 37 nt single-stranded (ss) RNA template and all four nucleotide triphosphates. Conversion of ssRNA into double-stranded (ds) RNA was then monitored by native gel electrophoresis and autoradiography ( Figure 1C). Using RDR2, dsRNA products are detected within 1 minute. In contrast, the RDR2-asm mutant showed no detectable activity, even after 60 minutes (lane 10). Reaction products of RDR2 are resistant to the ssRNA-specific ribonuclease, RNase ONE TM but are sensitive to the dsRNA-specific ribonuclease, RNase V1 (see lanes 11 and 12). Collectively, these results show that recombinant RDR2 can catalyze dsRNA synthesis using an ssRNA template, in vitro.

Pol IV-RDR2 complex reconstitution using recombinant RDR2
Pol IV and RDR2 co-purify and co-immunoprecipitate (co-IP) in both Arabidopsis thaliana and maize (28)(29)(30). We tested whether Pol IV-RDR2 association can be recapitulated in vitro using recombinant RDR2 ( Figure 2). For this experiment, we made use of an A. thaliana transgenic line in which NRPD1 bearing a FLAG epitope tag at its C-terminus (NRPD1-FLAG) rescues an nrpd1-3 null mutant (8). This allows Pol IV to be affinity purified using anti-FLAG resin (see cartoon of Figure 2A). Using polyclonal antibodies recognizing native NRPD1 or native RDR2, both proteins are detected upon NRPD1-FLAG IP (Figure 2A lane 2), in agreement with prior studies (29). If NRPD1-FLAG IP is used to affinity-capture Pol IV from a rdr2 null mutant background, RDR2 is not detected, as expected (lane 3). RDR2 is also not detected in Pol V or Pol II fractions affinity purified by virtue of FLAG-tagged NRPE1 (the largest subunit of Pol V) or NRPB2 (the second-largest subunit of Pol II)(lanes 4 and 5) (8), in keeping with RDR2's specific association with Pol IV.
Pol IV, Pol V and Pol II were next IPed following incubation with recombinant RDR2 ( We next tested whether the Pol IV-RDR2 complexes assembled using recombinant RDR2 can carry out the coupled enzymatic reactions that generate dsRNA from a DNA template strand (16). In this assay, diagrammed in Figure 2B, Pol IV transcription is initiated using an RNA primer hybridized to a T-less (no thymidines) 51 nt DNA template strand. When the elongating Pol IV encounters a 28 nt non-template strand of DNA, basepaired to the template strand, Pol IV transcribes only ~12-16 nt into the double-stranded region and then terminates, yielding transcripts of ~34-37 nt (16). Importantly, Pol IV termination in this manner is required for RDR2 to engage the Pol IV transcript and synthesize the complementary RNA strand (16).
Because no thymidines are present in the DNA template, α-32 P-ATP is not incorporated into the initial Pol IV transcript. However, uracils incorporated into the Pol IV transcripts template the incorporation of α-32 P-ATP into the RNA second strands that are synthesized by RDR2. Thus α-32 P-ATP incorporation into RNAs of ~34-37 nt is indicative of coupled Pol IV-RDR2 transcription (16).
Lane 1 of Figure 2B shows reaction products generated by Pol IV associated with wildtype RDR2 (see immunblots at bottom), with prominent 34-37 nt body-labeled RDR2 transcripts being readily apparent. Pol IV purified from an rdr2 null mutant does not generate these transcripts (lane 2). The ladder of short RNA transcripts near the bottom of lane 1 are RDR2 transcripts generated using the 16 nt RNA primer as a template (29). Haag et al. showed that RDR2 associated with Pol IV generates these short primer transcripts but RDR2 isolated from a pol iv mutant does not (29). In keeping with these observations, recombinant RDR2 does not generate the short primer transcripts ( Figure 2B

RDR2 physically interacts with the largest catalytic subunit of Pol IV
Having demonstrated that recombinant RDR2 associates with Pol IV to form a functional complex capable of dsRNA synthesis, we sought to identify the basis for Pol IV-RDR2 association. As an initial test, we performed a far-western blot in which FLAG-tagged Pol II, Pol To test whether RDR2 can interact with NRPD1, we designed a synthetic transgene encoding NRPD1 fused at its C-terminus to a FLAG epitope tag, expressed the protein in insect cells using a baculovirus vector, and purified the protein to near homogeneity ( Figure 3B). Upon incubation with V5-tagged RDR2, the proteins co-IP, using either anti-V5 or anti-FLAG resin, indicating that NRPD1 and RDR2 do, indeed interact ( Fig 3C, lanes 3 and 6).
As a third test of RDR2's ability to interact with NRPD1, we performed reciprocal yeast two-hybrid interaction experiments ( Figure 3D) using NRPD1 or RDR2 as either bait (when fused to the Gal4 DNA binding domain, DBD of pDEST32) or prey (when fused to the Gal4 activation domain, AD of pDEST22). Either combination of NRPD1 and RDR2, as bait or prey, resulted in HIS3 expression, enabling colony growth on media lacking histidine and containing 25 mM 3 amino-1,2,3-triazole (3AT) to increase the stringency of HIS selection (Fig 3D, lanes 5 and 6).

The N-terminal region of NRPD1 interacts with RDR2
Full-length NRPD1 is 1453 amino acids in length. To search for regions that interact with RDR2, we engineered six recombinant expression vectors, each encoding ~33 kDa portions of NRPD1 ( Figure S1A), each overlapping its neighbors by ~5 kDa, and each having a C-terminal FLAG epitope tag. The polypeptides were generated in vitro using a bacterial transcription- To further test the ability of the NRPD11-300 polypeptide to interact with RDR2, we engineered a recombinant gene that expresses the NRPD11-300 polypeptide in E. coli and affinity purified the polypeptide by virtue of its C-terminal FLAG tag ( Fig S1B). Following incubation with V5-tagged RDR2, reactions were then IPed using anti-V5 or anti-FLAG resin. IPed proteins were then subjected to SDS-PAGE and immunoblotting, using anti-V5 antibody to detect RDR2 and anti-FLAG antibody to detect NRPD11-300 ( Figure 4B). In each test, RDR2 and NRPD11-300 co-IPed (see Figure 4B, lanes 3 and 6).

A site flanking the RDR2 active site interacts with NRPD1
To identify regions of RDR2 that interact with NRPD1, we engineered constructs encoding seven overlapping polypeptides of ~25 kDa, each bearing an HA epitope tag, and expressed the polypeptides in vitro using a bacterial transcription-translation system ( Figure   S2A). Following incubation with FLAG-tagged NRPD11-300, IP was conducted using anti-FLAG resin and affinity-captured proteins were subjected to SDS-PAGE and immunoblotting using anti-HA or anti-FLAG antibodies ( Figure 4C). The polypeptide corresponding to RDR2 amino acids 771-970, which includes the conserved aspartate triad of the active site, co-IPs with NRPD11-300 ( Fig 4C, lane 7). The partially overlapping polypeptides corresponding to amino acids 617-816 or 925-1133 did not co-IP with NRPD11-300. Collectively, these results suggest that RDR2 sequences between amino acids between 816 and 925 interact with NRPD1.
To confirm the ability of RDR2771-970 to interact with FLAG-tagged NRPD11-300 , we expressed RDR2771-970 , fused to an HA epitope tag, in E. coli and affinity-purified the protein using anti-HA resin ( Figure S2B). We then incubated the protein with FLAG-tagged NRPD11-300 and performed IP using anti-FLAG or anti-HA resin. These tests showed that RDR2771-970 and NRPD11-300 co-IP, regardless of which partner is affinity captured ( Figure 4D, lanes 3 and 6).

Fine mapping Pol IV-RDR2 interaction sites using peptide arrays
Having identified regions of NRPD1 and RDR2 that interact, we next searched for peptides within these regions that might account for the interactions. Twenty peptides that collectively comprise the amino acids of RDR2771-970 and thirty peptides that comprise the NRPD11-300 sequence were synthesized and arrayed by dot blotting onto nitrocellulose ( Figures 5A and B).
Peptides were typically 15 amino acids in length and overlapped their neighbors by 5 amino acids. The RDR2 peptide array was incubated with NRPD11-300 -FLAG and the NRPD1 peptide array was incubated with RDR2771-970-HA. Filters were washed to remove unbound probe proteins and then incubated with HRP-conjugated antibodies recognizing the HA or FLAG tags of the NRPD11-300 or RDR2771-970 polypeptides. Filters were again washed, then assayed for HRP-catalyzed chemiluminescence to screen for the presence of immobilized probe polypeptides (see cartoon of Figure 5A).
Three contiguous peptides of RDR2, peptides 10-12, interact with NRPD11-300 ( Figure   5A). If shared sequences of the non-interacting, overlapping peptides 9 and 13 are excluded, the sequence 866 DVTLEEIHKFFVDYMISDTLGVIST 890 emerges as the candidate NRPD1 contact region. This sequence is close to the active center, beginning 32 amino acids downstream from the 830 DLDGD 834 motif that coordinates a magnesium ion at the site of phosphodiester bond formation (see upper diagram of Figure 5C).
Importantly, lysine 874 of RDR2 is present within peptides 10 and 11 of the peptide array (see Figure 5A) and is the only lysine within NRPD1-interacting peptides 10-12. This region of RDR2 is depicted as a pink rectangle in the diagram of Figure 5C. Likewise, two of the three RDR2-interacting regions of NRPD1 previously defined in the peptide array experiment included lysines that were crosslinked to RDR2, specifically K131, which is present in peptides 12 and 13, and K254 and K255, present within peptides 25 and 26 ( Figures 5C and 5B).
Moreover, crosslinks formed between lysines within NRPD11-300 (green arcs in Figure 5C) indicate that RDR2 interacting region 3, which is seemingly far apart from interacting regions 1 and 2, must actually be in close proximity to these regions in the folded polypeptide because the distance given that the reactive groups of the BS3 crosslinker is only 11.4 Å. Collectively, these results suggests that NRPD11-300 peptides 9-10, 12-13 and 24-26 may comprise a RDR2-binding surface.

Pol IV-RDR2 interaction sites are conserved in plants
The A. thaliana genome encodes six RNA-dependent RNA polymerases (RdRPs), but only RDR2 interacts with Pol IV. The NRPD1 interaction site within RDR2 is deduced to involve the sequence DVTLEEIHKFFVDYMISDTLGVIST, as discussed above. This sequence has little similarity to the corresponding sequences of RDRs 1,3,4,5 or 6 ( Figure 6A). By contrast, the nearby active site region is highly conserved, especially between RDR2 and its closest paralogs, RDR1 and RDR6. The conservation at the active site extends to Neurospora crassa,QDE-1, for which structural information is available (31)(see below).
RDR2 and NRPD1 co-purify in A. thaliana, a dicotyledenous (dicot) plant and maize, a monocotyledonous plant, whose last common ancestor existed approximately 150-200 million years ago. Figure 6B compares the NRPD1-interaction site of RDR2 and its orthologs in two dicots (Arabidopsis and Glycine), two monocots (Zea and Oryza) and two gymnosperms, Pinus and Cycas, who last shared a common ancestor with monocots and dicots ~400 million years ago. Substantial sequence conservation is apparent. Likewise, alignment of the three RDR2 interaction sites of Arabidiopsis NRPD1 with orthologs of the other species also reveals blocks of identical or similar amino acids ( Figure 6C). Collectively, the high degree of sequence conservation at NRPD1 and RDR2 interactions sites suggest that Pol IV-RDR2 complexes may assemble in the same way throughout the plant kingdom.

Discussion
Our results show that Pol IV and RDR2 form a stable complex via interactions between  (Figure 7). Genetic experiments have shown that substitutions at these invariant amino acids in Pol II confer temperature sensitive or lethal phenotypes (38). In Pols I, II and III, Zn8 and all preceeding sequences contribute to their structurally similar clamp cores whereas the conserved CxxC motif of Zn 6 contributes to their clamp heads. Given the conservation at these sites, it is likely that the same is true for Pol IV.
In yeast Pol II, Zn6 coordination involves the highly conserved CxxC motif and two cysteines (also highlighted in red in Figure 7) located 38 and 57 amino acids downstream, respectively. However, in NRPD1 the highly conserved Zn6 CxxC motif is followed, 17-20 amino acids downstream, by a second CxxC motif (highlighted in red and marked by asterisks in Figure 7) that is not found in the largest subunits of Pols I, II, III. This second, Pol IV-specific CxxC motif is part of RDR2-interaction site 2. Thus, RDR2 interaction sites 1 and 2 each include CxxC motifs, and both motifs are highly conserved ( Figure 6C). These CxxC motifs in NRPD1 have the potential to coordinate a zinc atom, thereby generating a Pol IV-specific zinc finger within the RDR2 interacting region. However, our data suggest that formation of this putative zinc finger is not critical for RDR2 docking given that peptides that include the individual CxxC motifs (peptides 10 and 12) bind RDR2 in the peptide array experiment, as did flanking peptides, 9 and 13, which lacked the CxxC motifs. These results suggest that RDR2 can bind independently to either side of the putative zinc finger in a sequence-specific manner, consistent with the conservation of sequences adjacent to the CxxC motifs (see Figure 6C). Moreover, a tyrosine substitution at the first cysteine of the CxxC motif at the beginning of RDR2 interaction site 2 (C118Y, defining mutant allele nrpd1-50) did not disrupt RDR2 interaction, despite severly impairing Pol IV activity in vivo and in vitro (39). Thus the putative zinc finger may play a role in Pol IV transcription independent of RDR2 interaction.

Cloning, expression, and purification of active or catalytically dead RDR2
An RDR2 cDNA expression construct encoding the wild-type protein fused to V5 epitope and 6xHis tags, was described previously (13). An active site mutant (RDR2-asm) was generated by site-directed mutagenesis using primers listed in Supplementary Table 1. The RDR2 and RDR2-asm genes were expressed in insect cells as BaculoDirect TM expression vectors (Thermo Fisher Scientific). Recombinant proteins were affinity purified using nickel-nitrilotriacetic acid (Ni-NTA) agarose (Qiagen) and elution with imidazole, followed by FPLC using Heparin-

Sepharose and Superdex 200 columns. Purity was assessed by SDS-PAGE and staining with
Coomassie Blue. See Supplemental information for additional details.

Analytical size-exclusion chromatography
To estimate RDR2's mass in solution, RDR2 and protein standards (GE Healthcare) were subjected to FPLC using a Superdex 200 10/300 GL column at 4˚C using 50 mM HEPES-KOH pH7.5, 150 mM NaCl as the running buffer and a flow rate of 0.3 ml/min). Protein peaks were detected by UV absorbance at 280 nm, and SDS-PAGE with staining by Coomassie Blue.

RDR2 activity assay
Conversion of single-stranded (ss) RNA into dsRNA was carried out in 50 μl reactions containing 200 ng RDR2, 25 nM template RNA (a 37 nt RNA 5' end-labeled using T4 kinase and gamma-32 P-ATP), 25 mM HEPES-KOH pH 7.5, 2 mM MgCl2, 0.1 mM EDTA, 0.1% Triton X100, 20 mM ammonium acetate, 3% PEG 8000, 0.1 mM each of ATP, GTP, CTP and UTP and 0.8 U/μL RiboLock RNase inhibitor (Thermo Fisher Scientific). Reactions were incubated at room temperature and stopped by adding EDTA to 10 mM. For RNase treatments, 1 unit of RNase ONE TM (Promega) or 0.01 unit of RNase V1 (Thermo Fischer Scientific) was added to reaction products and incubated for 15 minutes at 37 °C. Reactions were stopped by adding SDS to 0.1% (w/v). Reaction products were subjected to TRIzol extraction, precipitated with ethanol, resuspended in nuclease-free water, adjusted to 1X DNA loading dye (Thermo Fisher Scientific), and subjected to non-denaturing gel electrophoresis using a 15% polyacrylamide gel and 0.5X TBE (Tris-Borate-EDTA) running buffer. The gel was transferred to blotting paper, covered with plastic wrap and exposed to X-ray film at -80°C for 1 hour.

Reconstitution of Pol IV-RDR2 complexes
Pol IV, Pol V, and Pol II were affinity-purified by virtue of their FLAG-tagged NRPD1, NRPE1 or NRPB2 subunits as described previously (29). To reconstitute Pol IV-RDR2 complexes, Pol IV expressed in an rdr2 null mutant background was immobilized on anti-FLAG M2 resin then incubated with 1 µM recombinant RDR2 for 30 mins at 4 °C in 50 mM HEPES-KOH pH7.5, 150 mM NaCl. The resin was then washed three times with 100 µl of the same buffer to remove unbound proteins then subjected to SDS-PAGE and immunoblotting using anti-FLAG-HRP (Sigma Aldrich) or antibodies recognizing native NRPD1, NRPE1, or NRPB2. Anti-V5-HRP antibody (Invitrogen) was used to identify recombinant RDR2.

Assay for dsRNA synthesis by Pol IV-RDR2 coupled reactions
Pol IV isolated in association with native RDR2, or Pol IV isolated from an rdr2 null mutant (29) and reconstituted with recombinant RDR2, were tested for dsRNA synthetic ability as in Transcription reactions were then passed through PERFORMA spin columns (Edge Bio) and adjusted to 0.3 M sodium acetate (pH 5.2). 15 μg Glycoblue TM (Thermo Fisher Scientific) was added, and RNAs precipitated with 3 volumes of isopropanol at -20˚C. Following centrifugation, pellets were washed with 70% ethanol and resuspended in 5 μl water. 5 μl of 2X RNA loading dye (New England Biolabs) was added, and the samples heated 5 min at 75⁰ C. RNAs were resolved on 15% polyacrylamide, 7M urea gels. Gels were transferred to filter paper, vacuum dried, and subjected to autoradiography or phosphorimaging.

Cloning, expression, and purification of NRPD1
The NPRD1 open reading frame, fused to a C-terminal FLAG tag and codon optimized for insect cell expression, was synthesized by GenScript Ò and cloned into the baculovirus expression vector pKL-10xHis-MBP-SED-3C. The vector was first transfected into Sf9 insect cells to produce recombinant baculovirus particles, then High Five TM cells for NRPD1 over-expression.
The NRPD1-Maltose Binding Protein (MBP) fusion protein was affinity purified using Amylose resin (New England Biolabs) and eluted with 20 mM maltose. The fusion protein was next subjected to anti-FLAG affinity purification, with in-column Prescission Protease digestion used to cleave the linker between MBP and NRPD1. Following extensive washing to remove free MBP, NRPD1 was eluted using an excess of FLAG peptide. See Supplemental information for a more detailed protocol.

Cloning, over-expression and purification of NRPD11-300 and RDR2771-970
NRPD1 amino acids 1-300 fused to His and FLAG tags, and RDR2 amino acids 771-971 fused to His and HA tags, were expressed in E coli using pET28 vectors. The proteins were then purified using nickel-NTA column chromatography and elution with imidazole. Purity was assessed by SDS-PAGE and immunoblotting using anti-FLAG-HRP or anti-HA-HRP antibodies. mM NaCl. Proteins were eluted by addition of 6x SDS loading dye, and heating at 95°C for 2 minutes, then subjected to SDS-PAGE. Immunoblot analysis was used to detect the proteins using anti-FLAG-HRP to detect NRPD1 or anti-V5-HRP to detect RDR2 antibodies. The same methods were used to test for co-IP of full-length RDR2 with NRPD11-300 or for co-IP of NRPD11-300 with RDR2771-970.

Yeast two-hybrid interaction assay
Saccharomyces cerevisiae strain MaV203 was transformed according to the Matchmaker protocol (Thermo Fisher) with plasmids expressing NRPD1 or RDR2 fused to the GAL4 activation domain or GAL4 DNA-binding domain. Transformants were selected on SC-Leu-Trp plates for 3 days at 30 °C. To test for interactions, each strain was replica plated onto SC,-Trp,-Leu and SC,-Trp,-Leu,-His, +3AT agar. Growth after 3 days at 30 °C was then assessed.

Expression of NRPD1 and RDR2 polypeptides
20-33 kD subregions of NRPD1 and RDR2 polypeptides were expressed in vitro using a PURExpress ® In Vitro Protein Synthesis Kit (New England Biolabs). Briefly, the desired intervals of the synthetic genes were amplified using the primers in Supplementary Table 1 and then used as templates for in vitro transcription-translation reactions.

Peptide array protein interaction assays
Custom peptide libraries corresponding to NRPD1 amino acids 1-300 or RDR2 amino acids 771-971 were obtained from Genscript and dot-blotted (10 ng) onto nitrocellulose. Filters were submerged in 5% (w/v) nonfat dry milk in 1X TBST (Tris buffered saline, 0.1 % Tween 20), 30 min at room temperature to block free protein binding sites. The blocking solution was then discarded. NRPD1 peptide arrays were then probed with ~200 ng of RDR2771-971-HA in 8 ml of 50 mM HEPES-KOH pH 7.5, 100 mM NaCl at room temperature for 1 hour. The RDR2 peptide array was probed with ~200 ng of NRPD11-300-FLAG. Filters were then washed twice with 8 ml of 1X TBST , then incubated with anti-FLAG-HRP or anti-HA-HRP in 8 ml 1X TBST at room temperature for 1 hour. The filters were washed twice with 8 ml 1X TBST, then developed using a Pierce TM Enhanced Chemiluminescence Western Blotting Kit (Thermo Fisher Scientific).

Crosslinking-Mass Spectroscopy
Purified RDR2771-970 and NRPD11-300 proteins were pre-incubated then crosslinked using 0.1 mM Bissulfosuccinimidyl suberate (BS3). Crosslinked protein complexes were reduced using 10 mM TCEP (Tris(2-carboxyethyl)phosphine), alkylated with 20 mM iodoacetamide, and digested with 12.5 ng/μL each of Trypsin and Chymotrypsin for 16 hr. Resulting peptides were resolved by HPLC using a C18 column using an acetonitril gradient and subjected to electrospray ionization and anlyzed using an Orbitrap Fusion Lumos mass spectrometer. Data are available at the ProteomeXchange Consortium via the PRIDE partner repository (40) with the dataset identifier PXD020170. See Supplemental information for a detailed protocol.

Molecular modeling of NRPD1 and RDR2
A structural model for NRPD1 was generated using Phyre2 protein structure prediction software (41), in intensive mode, based on homologies to six Pol II or Pol I largest subunits. The RDR2 model is based on two fungal QDE-1 RNA-dependent RNA polymerases. Resulting models were then superimposed onto the structures of budding yeast Rpb1 (the Pol II largest subunit; PDB:3HOV) (42) or Neurospora crassa QDE-1 (PDB:2J7N) (31) crystal structures using the MatchMaker tool of Chimera (43). Superimposed models were then imported to ChimeraX (44) and solvent-excluded surfaces of RPB1 and QDE-1 were visualized and colored. See Supplemental information for additional details.

Sequence alignments
Amino acid sequence alignments were performed using CLUSTAL W. Conserved sequences were highlighted using BOXSHADE V3.31.

Acknowledgments
VM dedicates this work in loving memory of his father, Dr. Umesh Chandra Mishra. We thank the Drosophila Genome Resource Center at Indiana University Bloomington for access to their insect cell culture facilities. We thank Tsuyoshi Imasaki for his efforts in engineering the baculovirus vector used for NRPD1 overexpression.     A. Far-western blot test for RDR2 interacting proteins. Pols II, V and IV assembled using FLAG-tagged NRPB2, NRPE1 or NRPD1, respectively were IPed using anti-FLAG resin.
Following SDS-PAGE and electroblotting, the filter was incubated with recombinant V5-tagged RDR2. After washing, the filter was incubated with anti-V5 antibody to detect immobilized  A. Testing NRPD1 amino acid intervals for interaction with RDR2. Six overlapping polypeptides that collectively represent the 1453 amino acids of full-length NRPD1 were designed, each having a C-terminal FLAG tag. The polypeptides were then expressed in vitro using a cell-free transcription-translation system (see Figure S1A) and incubated with V5-tagged recombinant RDR2. Anti-V5 immunoprecipitation was then used to pull down RDR2 and any associated proteins, as depicted in the cartoon. IPed proteins were resolved by SDS-PAGE and subjected to immunoblotting using anti-FLAG and anti-V5 antibodies. Immunoblots were probed with anti-FLAG and anti-V5.
C. Testing RDR2 amino acid intervals for interaction with NRPD11-300. Seven overlapping polypeptides of ~22 kDa that collectively represent the amino acid sequence of full-length RDR2 (1133 amino acids) were designed, each having a C-terminal HA epitope tag. The recombinant RDR2 polypeptides were then expressed in vitro using a cell-free transcription-translation system (see Figure S2A). The RDR2 polypeptides were then incubated with NRPD11-300 followed by anti-FLAG IP to pull down NRPD11-300 and any associated proteins, as depicted in the cartoon at top. IPed proteins were resolved by SDS-PAGE and subjected to immunoblotting using anti-FLAG and anti-HA antibodies. Immunoblots were probed with anti-HA and anti-FLAG. The diagram at bottom shows that RDR2771-970 includes the enzyme's active site .