Conditional depletion of the RNA polymerase I subunit PAF53 reveals that it is essential for mitosis and enables identification of functional domains

Our knowledge of the mechanism of rDNA transcription has benefited from the combined application of genetic and biochemical techniques in yeast. Nomura's laboratory (Nogi, Y., Vu, L., and Nomura, M. (1991) Proc. Natl. Acad. Sci. U.S.A. 88, 7026–7030 and Nogi, Y., Yano, R., and Nomura, M. (1991) Proc. Natl. Acad. Sci. U.S.A. 88, 3962–3966) developed a system in yeast to identify genes essential for ribosome biogenesis. Such systems have allowed investigators to determine whether a gene was essential and to determine its function in rDNA transcription. However, there are significant differences in both the structures and components of the transcription apparatus and the patterns of regulation between mammals and yeast. Thus, there are significant deficits in our understanding of mammalian rDNA transcription. We have developed a system combining CRISPR/Cas9 and an auxin-inducible degron that enables combining a “genetics-like”approach with biochemistry to study mammalian rDNA transcription. We now show that the mammalian orthologue of yeast RPA49, PAF53, is required for rDNA transcription and mitotic growth. We have studied the domains of the protein required for activity. We have found that the C-terminal, DNA-binding domain (tandem-winged helix), the heterodimerization, and the linker domain were essential. Analysis of the linker identified a putative helix–turn–helix (HTH) DNA-binding domain. This HTH constitutes a second DNA-binding domain within PAF53. The HTH of the yeast and mammalian orthologues is essential for function. In summary, we show that an auxin-dependent degron system can be used to rapidly deplete nucleolar proteins in mammalian cells, that PAF53 is necessary for rDNA transcription and cell growth, and that all three PAF53 domains are necessary for its function.

Ribosome biogenesis is essential for homeostasis and cell growth, mitotic or hypertrophic. This complex process is energetically expensive, consuming up to 60% of a yeast cell's RNA synthetic capacity (1,2). As such, it is subject to multiple levels of regulation. The rate-limiting step in this complex process is transcription of the pre-ribosomal RNA (rRNA) genes (rDNA) by RNA polymerase I (pol I) 2 (3)(4)(5). Abnormalities in the rates of rRNA synthesis and ribosome biogenesis cause a broad range of human diseases, the ribosomopathies (6 -10). Furthermore, the pleomorphic nucleoli of tumor cells, the sites of rDNA transcription, are characteristic of neoplasia and reflect dysregulation of rRNA synthesis in cancer (11)(12)(13)(14)(15). rDNA transcription requires a unique set of transcription factors not used by the other nuclear polymerases, and we are just beginning to understand the biochemistry and functional interactions of these mammalian factors. Many of the genes whose protein products are involved in ribosome biogenesis were included in CRISPR screenings of essential genes (16 -19) and in high-throughput screening of mammalian cells for proteins that affect nucleolar number (20,21). Although these procedures reveal proteins involved in ribosome biogenesis, they do not provide for biochemical analysis of function. Nomura's laboratory (22,23) demonstrated a screen for mutants deficient for transcription by yeast RNA polymerase I. These "rrn" mutants identified genes encoding unique subunits of pol I (22,24), as well as genes encoding other components of the committed template. The development of an in vitro transcription system (25)(26)(27)(28)(29) led to the combination of genetic and biochemical approaches that enhanced our understanding of the mechanism of rDNA transcription in yeast.
A fully-functional molecule of yeast RNA polymerase I consists of 15 subunits. This total includes the core pol I, a heterodimer of RPA49 -RPA34.5 and RRN3. Five of the core subunits are shared with the other two polymerases and two are shared with pol III (30,31).
Two of the pol I subunits, yeast RPA34.5 and RPA49, form a heterodimer with still poorly-defined roles in rDNA transcription. The heterodimer of RPA49/RPA34.5 is easily dissociable from the polymerase, and the association of the mammalian orthologues PAF53/PAF49 with pol I is subject to growth-related regulation (32)(33)(34)(35).
Although Saccharomyces cerevisiae RPA49 is not essential for viability (36), deletion of yeast RPA49 results in colonies that grow at 6% of the WT rate at 25°C (36). Similarly, when the Schizosaccharomyces pombe orthologue, RPA51, was deleted (37), specific rDNA transcription was reduced 70% (with no effect on nonspecific polymerase activity). Deletion of the other partner in the heterodimer, RPA34.5, has a minor effect on growth or rRNA synthesis, but it results in a polymerase that loses the RPA49 subunit upon purification (38). Biochemical purification of pol I results in two fractions, one of which does not contain either RPA49 or RPA34.5 (39). This is complemented by the observation that a large fraction of the polymerase particles prepared for cryo-EM are free pol I enzymes that "either lacked the RPA49/RPA34.5 subcomplex or displayed a flexible clamp-stalk region" (40). Interestingly, most of the interactions between the heterodimer and pol I in yeast appear to be mediated by the RPA49 subunit (41). The N termini of RPA34.5 and RPA49 are required for heterodimerization. The heterodimer stimulates polymerase nuclease activity and has a triple ␤-barrel domain similar to the core of TFIIF and the pol III heterodimer of C37/C53 (42)(43)(44). The C terminus of yeast RPA49 contains a domain with dual-winged helices (tandemwinged helix, t-WH) (42) that is capable of DNA binding and resembles a similar element in TFIIE (42) and the pol III subunit RPC34 (45,46). Mutations within the tWH of RPA49 result in increased sensitivity to 6-azauracil and mycophenolic acid, defects in transcription elongation (47), and lower levels of recruitment of pol I and Rrn3 at the promoter (36,47).
The mammalian orthologues of yeast RPA49 and RPA34.5, referred to as PAF53 and PAF49, respectively, are essential for promoter-specific transcription (32,33). Yamamoto et al. (32) reported that PAF53 and PAF49 were associated with a fraction of the core pol I molecules that are active in transcription. Hannan et al. (48) confirmed this observation and estimated that 60% of the polymerase molecules in a rat hepatoma cell line contained PAF53. Yamamoto et al. (32) also reported on the growth-dependent nucleolar localization of PAF49, a result subsequently confirmed by our laboratory (34,35). Most recently, several CRISPR/Cas9-based screenings of the mammalian genome identified PAF53 and PAF49 as being "essential" genes (16,17), which was confirmed when we found that we could not isolate cell lines that failed to express PAF53 (50) following CRISPR/CAS9 mutagenesis.
The assembly of pol I-specific "polymerase-associated factors" (Rrn3 and the heterodimer of PAF49 and PAF53) with pol I is a necessary step in rDNA transcription (25, 28, 32-34, 37, 47, 51-74). Although genetic studies in yeast and KO studies in mammalian cells demonstrate that the PAF53/PAF49 complex is essential for cellular physiology (17,42,50,75), their roles in rDNA transcription are unknown. Moreover, because there are significant differences in both the structures and components of the transcription apparatus and the patterns of regulation between mammals and yeast, studies on yeast rDNA transcription leave significant holes in our understanding of mammalian rDNA transcription. As transcription by pol I is essential for cell growth or viability (76 -79), it has been impossible to generate cell lines in which these essential genes are knocked out (17,50). Although shRNA and siRNA will knock down the levels of the proteins, this can take days, depending upon the half-life of the protein, and this allows compensatory mechanisms to occur. To facilitate the study of rDNA transcription in mammalian cells, we sought to develop a system that would allow us to rapidly knock down the protein levels and replace them with mutants.
Auxins (AUX), such as indole-3-acetic acid (IAA), promote the interaction between SCF TIR1 (TIR1), an auxin-dependent E3 ubiquitin ligase (80 -84), and the AUX or IAA family of transcription factors (85,86). In the presence of auxin, TIR1 associates with the AUX/IAA transcription factor. This induces the rapid ubiquitination and subsequent proteasome-dependent degradation of transcription repressors that contain an auxin-inducible degron (AID). The AID from the IAA17 protein of Arabidopsis thaliana has been used to program proteins for rapid degradation in yeast and animal cells (80,81,87).
We investigated the use of that system to study transcription by pol I in mammalian cells focusing on PAF53. Using CRISPR/ Cas9, we tagged the C terminus of the endogenous PAF53 genes with 43 amino acids (aa) from IAA17 (aa 68 -111) using homology-directed repair and a donor fragment with short homology arms (ϳ125 bp). When cells that expressed TIR1 from Oryza sativa (81) (tagged with a nuclear localization signal (NLS-TIR1)) and PAF53-AID were treated with IAA, the PAF53-AID was eliminated within 1 h. This resulted in the inhibition of rDNA transcription and caused cell cycle arrest. Using cell proliferation as the phenotype, we investigated the domains of PAF53. In silico analysis of mammalian PAF53 revealed structures strikingly similar to that of yeast RPA49. These included the N-terminal dimerization domain, a linker, and the C-terminal tandem winged helix. We have previously confirmed the functional homology of the N-terminal dimerization domain (88) to yeast RPA49, and we now demonstrate that the C-terminal tWH has DNA-binding activity similar to that of the yeast domain. We found that deletion of the N-terminal dimerization domain significantly inhibited the ability of ectopic PAF53 to rescue cell division. Interestingly, deletion of the C-terminal tWH had an intermediate effect on rescue, whereas deletion of the linker inactivated PAF53. We identified a DNA-binding motif in the linker and found that DNA binding by the linker was required for both PAF53 and RPA49 activity. Finally, our data imply that the binding of PAF53 to pol I requires dimerization with PAF49 a result that would be different from that found in yeast.

Results
Our first experiments were designed to determine whether a nucleolar protein could be targeted for degradation by TIR1. As shown in Fig. 1, we found that when amino acids 1-229 of Arabidopsis thaliana IAA17 were linked to H2B tagged with YFP and an AID (Fig. 1B), the protein was degraded when HEK293 cells were cotransfected with a vector expressing TIR1 and treated with IAA. Subsequently, we found that the same AID could target PAF53 expressed transiently (Fig. 1C). We also found that the 43-amino acid degron (aa 68 -111) of IAA17 would drive auxin-dependent degradation in HEK293 cells (Fig.  1C). We then examined the possibility of using an Aux/IAA PAF53 is essential for cell growth 13-amino acid domain II consensus sequence (aa 82-94) that has been shown to work in plants (82). That domain did not work in our cells (Fig. 1C, lanes 5 and 6). These experiments also serve to demonstrate that activation of TIR1 by IAA did not target the endogenous PAF53 (Fig. 1, lower band in C and D). Subsequently, we have used the 43-amino acid degron as the AID in our constructs. Nishimura et al. (80) reported that the AID tag could be on either the N or C termini of proteins and still support auxin-induced degradation. As shown in Fig. 1D, we confirmed that in mammalian cells expressing TIR1, the AID can be on either terminus of the target protein, and as discussed below, it can be internal.
Based on these observations, we established cell lines that constitutively expressed TIR1. However, we found that these cell lines did not degrade AID-tagged PAF53 as rapidly as predicted (data not shown) from the transient transfection experiments. We considered two models. Either we did not have the same levels of TIR1 expression in the stable cell lines as in the transiently-transfected cell lines or the TIR1 was not localizing to the nucleus/nucleolus. We examined the sequence of O. sativa TIR1 and found it did not contain a nuclear localization signal (NLS). A monopartite NLS (PAAKRVKLD) was added to the N terminus of TIR1, and cells that constitutively expressed NLS-TIR1 were selected with hygromycin. One clone, 569, that demonstrated the rapid degradation of transiently-transfected PAF53-AID was chosen for further cloning experiments. Clone 569 was then subject to CRISPR-driven homologydirected repair to introduce the AID onto the C terminus of the endogenous PAF53 genes using the strategy summarized in Fig.  2A. After selection and cloning by limiting dilution, the clones were screened by Western blottings for expression of the chimeric PAF53 to select for cells homozygous for AID-PAF53 expression (Fig. 2B). The two forms of PAF53 seen in Fig. 2B, lane 1, are the result of tagging only one allele, and lane 2 represents the results of tagging both alleles. Furthermore, recombination was confirmed by PCR for the recombinant insert (Fig.  2C). The PCR products were cloned, and four clones were sequenced. The sequences were identical and did not demonstrate errors in recombination (data not shown).
Clones that only expressed the AID-tagged PAF53 were selected, expanded, and stored. We found that the anti-PAF53 antibody recognized two bands in some of the clones (Fig. 2D). The consensus sequence from domain II is shown. B, H2B-GFP-IAA17 is targeted for auxin-dependent degradation in HEK293 cells that express TIR1. HEK293 cells constitutively expressing TIR1 were transfected with a plasmid (pcDNA3.1) coding for H2B-GFP-IAA17 for 24 h and then treated with 500 M IAA in water. Amino acids 1-229 of A. thaliana IAA17 were linked to H2B tagged with YFP and an AID. C, amino acids 1-229 and 68 -111 can program PAF53 for auxin-dependent degradation, but the degron core, amino acids 82-94, does not. In lanes 1-4, ectopic PAF53 was visualized with the anti-PAF53 antibody. In lanes 5 and 6, the ectopic AID-PAF53 was FLAG-tagged and visualized with anti-FLAG antibody as the ectopic protein migrated too close to the endogenous protein to be clearly visualized with anti-PAF53 antibody. D, amino acids 1-229 program PAF53 for auxin-dependent degradation if placed on either the N or C termini of the protein. The TIR1 used in these experiments did not contain a nuclear localization signal. The analyses presented were carried out after 3 h of auxin treatment (500 M of IAA (ϩ auxin) or water (Ϫ auxin)). Western blottings for PAF53 were carried out as described previously (48).

PAF53 is essential for cell growth
The major band, migrating at ϳ58 kDa, was at the molecular mass predicted for the AID-tagged PAF53 (P-A). The upper band migrated with a mass predicted if it were the complete PAF53-AID-F2A-NPT chimeric protein (P-A-N), ϳ87 kDa, due to incomplete F2A-dependent cleavage. When IAA was added to these cells, we observed a rapid degradation, within 1 h, of both bands that reacted with anti-PAF53 antibodies (Fig.  2D, lanes 2 and 4). This confirmed that they were the bands predicted and demonstrated that the presence of an AID in the middle of a protein could target the protein for degradation.
As introduced, the role(s) of yeast RPA49 in rDNA transcription are not clear. When mammalian cells were screened for genes "required for proliferation and survival," Wang et al. (17) reported that the PolR1E (PAF53) gene was essential. This result was confirmed by Bertomeu et al. (16) and others (50). Thus, we sought to determine the specific effect of knocking down PAF53 using the AID as both physiological and biochemical results should be apparent very rapidly.
As shown in Fig. 3A, treatment of clone 763 that expressed both TIR1 and AID-tagged PAF53 for 3 h with IAA resulted in the inhibition of rDNA transcription as demonstrated by metabolic labeling of total cell RNA. This experiment has been reproduced at least three times, and we observed at least 99% inhibition of synthesis of the 47S pre-rRNA in all experiments. The inhibition of rDNA transcription can induce a phenomenon referred to as "nucleolar stress" or "ribosome stress." The cellular responses to this stress can include cell cycle arrest or cell death (77-79, 89 -94). When clone 763 cells were exposed to IAA, they appeared to complete one round of division and then arrest (Fig. 3B). The cell cycle arrest was stable for up to 9 days with no apparent increase in cell death. To confirm that the cell-cycle arrest was due to the knockdown of PAF53, we sought to determine whether the ectopic expression of the protein would rescue the arrest. Clone 763 cells were transfected with pCDNA3.1 driving the expression of mouse PAF53 and a vector driving the expression of eGFP 2 days before the cells were treated with IAA. Control cells were transfected with an empty vector and the vector driving eGFP. Subsequently, the number of viable cells (trypan blue exclusion) was measured in each of the three groups. As shown in Fig. 3, expression of Clone 763 cells depleted of PAF53 did not demonstrate a net proliferation. This could be the product of cell cycle arrest or a steady-state relationship between cell division and cell death. To examine this question, we examined the cells for trypan blue dye exclusion. We did not observe a significant increase in trypan blue staining following treatment with IAA, indicating that we were not causing cell death (data not shown). The inhibition of rDNA transcription has been shown to cause cell cycle arrest at the M/G 1 boundary. However, the effect of rapidly targeting rDNA transcription has not been previously studied. Previous studies have used Cre-dependent recombination to knock down RRN3 and UBF (95,96), two components of the rDNA transcription apparatus (97). This process takes much longer than an hour and can cause apoptosis (96). As shown in Fig. 4, 24 h after the depletion of PAF53, there was no alteration in the distribution of cells through the cell cycle. However, after 2 days, we observed a decrease in the percentage of the cells in G 1 and an increase in the percent of cells in S phase (Fig. 4B). The increase in the S/G 1 ratio was significant after 48 h (Fig.  4C). Interestingly, we did not observe a significant accumulation of cells in either the G 1 or G 2 checkpoints.
As shown in Fig. 5, yeast RPA49 can be considered to consist of three domains: an N-terminal dimerization domain that mediates the interaction with RPA34.5, a C-terminal t-WH that has DNA-binding capacity, and a linker domain between the two (88) (SWISS-MODEL Q9GZS1 and see also PDB entry 3NF1). Both SWISS-MODEL and I-TASSER predict structures for mouse and human PAF53 similar to that of yeast RPA49 despite Ͻ25% sequence identity (98 -102). We have previously demonstrated that the N-terminal domains of mouse PAF53 and PAF49 are required for dimerization and are predicted to contain the triple barrel ␤-structure found in the RPA49/ RPA34.5 heterodimer and in TFIIF (42, 44, 88, 103). Interestingly, Forty eight hours after clone 763 cells, which expressed both TIR1 and AID-tagged PAF53, were transfected with empty vector or vector coding for ectopic PAF53, they were treated with 500 M IAA for 3 h. Following treatment, [ 3 H]uridine was added to the media for 30 min, and total RNA was isolated and analyzed as described under "Experimental procedures" (113). B, cells cultured in 30-mm wells were treated with 500 M IAA at 0 time. All cells were transfected at Ϫ48 h with either empty vector (control) or pcDNA3.1 expressing WT mouse PAF53 (ϩPAF53). Cells were then harvested at the times indicated, and the number of cells was determined as described (113). The growth curves were fitted with a second order polynomial. n ϭ 3, Ϯ S.D. C, expression of ectopic mouse PAF53 is maintained over the 7-day period of the experiments described in A. Cells were treated with IAA as indicated and harvested immediately after the onset of treatment or at the times indicated. The cells were lysed in HEPES lysis buffer, and the expression of endogenous or ectopic PAF53 was determined by Western blot analysis with anti-PAF53 antibody.

PAF53 is essential for cell growth
I-TASSER and SWISS-MODEL (Q9GZS1) predict similar structures for the C-terminal domains of the yeast (NP_014151.1) and mammalian orthologues (human NP_071935.1 or AAH14331.1) despite their being only 22% identical. Although it has been demonstrated that the t-WH of yeast RPA49 has DNA-binding activity, the same has not been reported for the predicted t-WH of the mammalian orthologues. As shown in Fig. 6A, we determined that human PAF53 was capable of binding to DNA. Moreover, we found that the N-terminal dimerization domain was not required for DNA binding (Fig. 6B, lanes 2 and 3, aa 109 -435). The t-WH domain, amino acids 216 -435, also bound DNA (Fig. 6B, lanes 4  and 5). However, when the N-terminal WH was deleted, the remainder lost DNA-binding activity (Fig. 6B, lanes 6 and 7). Thus, both the yeast and mammalian orthologues contain N-terminal dimerization domains and C-terminal DNA-binding domains.
We then sought to determine which, if any, of the domains of the protein were essential for cell cycle progression. Two days prior to treatment with IAA, cells were transfected with vectors expressing various deletion mutants of mouse PAF53 (Fig. 7A). We found that a construct lacking the N-terminal dimerization domain (PAF53 aa 109 -435) failed to restore cell division (Fig.  7D). The N-terminal dimerization domain (PAF53 aa 1-160) itself did not rescue cell division either (Fig. 7C). Surprisingly, we observed partial rescue with a construct that did not contain the C-terminal tWH, but it did contain most of the dimerization domain and the linker (PAF53 aa 1-222) (Fig. 7B). These three constructs would appear to be properly folded, as deletion of the dimerization domain did not inhibit DNA binding by the t-WH (Fig. 6B, lanes 4 and 5) and deletion of the t-WH did not inhibit dimerization (88). Furthermore, all three constructs were expressed for the duration of the experiment (Fig. 7, E and  F). Thus, the linker region between the dimerization and tWH domains appears to be a functional domain.
When we examined the structure of the linker, we noted that both mammalian PAF53 and yeast RPA49 contain a helixturn-helix motif (HTH). It has been reported that the linker spans the cleft in pol I (31, 104). We hypothesized that the HTH might come in contact with the template and that the linker might participate in transcription due to DNA-binding activity. When we expressed His-PAF53, aa 1-222, and tested it in a

PAF53 is essential for cell growth
DNA-binding assay, we found that the purified, recombinant protein had DNA-binding activity (Fig. 8). To examine the role of the putative HTH in DNA binding, we mutated each of the helices individually (Fig. 9, A and B) and tested equal amounts of the recombinant proteins for DNA-binding activity. As shown in Fig. 9C, mutation of either of the helices significantly inhibited DNA binding. Thus, mammalian PAF53 has two domains with DNA-binding activity, the tWH and the helix-turn-helix of the "linker." Subsequently, we carried out rescue experiments using PAF53 constructs that contained similar mutations of the HTH. As shown in Fig. 10, cells expressing PAF53 with the HTH mutated failed to divide following IAA treatment.
To understand whether the PAF53 HTH is functionally conserved across species, we used yeast growth assays to examine the effect of an HTH deletion in the yeast PAF53 orthologue RPA49, and we compared its growth to several well-characterized RPA49 mutant variants. Consistent with previous reports, transforming the RPA49 deletion strain with the WT RPA49 expression vector or a mutant variant lacking the nonessential N-terminal dimerization domain significantly restored cell growth at 30°C similarly to WT RPA49 (Fig. 11) (38,42,47). Likewise, expression of an RPA49 variant lacking the C-terminal tWH failed to rescue growth and remained slow growing as observed previously (Fig. 11) (36, 42, 47). Finally, we tested the RPA49 variant lacking the HTH, which retained a slow growth phenotype (Fig. 11), indicating that the HTH is important for RPA49 function in yeast. The slow growth of the HTH deletion was further exacerbated at 18°C (Fig. 11). This is a common characteristic of yeast strains with defects in ribosome biogenesis (105). Interestingly, the mutants with deletions of the HTH domain grew less well at 30°C than cells transfected with empty vector (pRS424). Together, these results are consistent with our PAF53 cell-viability assays, further emphasizing the functional importance of the HTH in pol I transcription across species.

Discussion
We have established that the TIR1 auxin-dependent degron system can be used to rapidly deplete nucleolar proteins. Our data demonstrate the rapid (Ͻ1 h) auxin-dependent degradation of a nucleolar protein PAF53 and the feasibility of using this approach to study the function of those proteins in mammalian cells following depletion or replacement with mutant forms of the protein. In our initial investigations of the feasibility of this system, we found that we had to optimize the system in order to apply it to nucleolar proteins.
We found that the degradation of a nuclear protein was facilitated when we added a nuclear localization signal to osTIR1. We also found that the minimally-conserved core of domain II of the IAA17 degron was insufficient to target degradation and that the 44-aa degron we used could be placed on the N or C terminus of the target or internally.
The yeast RPA34.5 and RPA49, and their mammalian homologues, PAF49 and PAF53, are structurally similar. Fig. 5 compares the structures of yeast RPA49 and the structure predicted for human PAF53 by I-TASSER (101,102,106). The three domains, dimerization, linker, and tWH, of RPA49 are indicated (88).
As discussed, yeast RPA49 is not essential for cell division (see Fig. 11, pRS424 empty vector). When we depleted PAF53, our HEK293 cells failed to divide. A priori, we expected to see arrest at either the G 1 /M or G 2 /S interfaces. However, we did not. There was a slight accumulation of cells in S but no distinct arrest in either G 1 or G 2 . This is in contrast to the patterns Figure 6. Putative tWH structure of mouse PAF53 has DNA-binding activity. A, SWISS-MODEL predictions for the structures of the PAF53 constructs used in these experiments. The predicted structure generated by SWISS-MODEL terminates at amino acid 420. B, purified, full-length mouse PAF53 has DNA-binding activity. DNA-binding assays (2 pmol of DNA) with increasing amounts of PAF53 (0.1-0.5 g) are presented in lanes 2-4 of the right panel in B. A Coomassie-stained gel of the protein used is presented in the left panel in A. C, deletion mutagenesis of PAF53 demonstrates that the PAF53 lacking the N-terminal dimerization domain has DNA-binding activity and that mutagenesis of the tWH inhibits DNA binding. The indicated mutants were used in the bandshift assays in two different concentrations (ϩ, 0.25 g, and ϩϩ, 0.5 g) in the assays. No shift was seen when 2 g of fragment 326 -435 was used (data not shown). A Coomassie-stained gel of the GSTtagged proteins used is presented in the right panel in C. Intervening lanes containing different amounts of the purified proteins were removed for the composite image. The amounts of the proteins used in the EMSA were normalized to the Coomassie-stained bands. The conditions for the EMSA were as described under "Experimental procedures." PAF53 is essential for cell growth observed when RRN3 and UBF were knocked down (95,96). In those instances, the investigators observed accumulation of the cells in either G 1 (RRN3) or G 2 (UBF) concomitant with a minor inhibition of the distribution of cells in S or its abrogation. Our observation requires further examination. One possibility for this result would be whether the cells were proceeding through the cell cycle at a greatly reduced rate.
The interaction of PAF53 with PAF49 requires the first 160 amino acids of PAF53 (88) and amino acids 41-86 of PAF49 (88). We have previously reported that when PAF49, the mammalian homologue of RPA34.5, is expressed ectopically it binds to pol I in the absence of ectopic PAF53 (88). Our observation that a form of PAF53, which lacks the heterodimerization domain (PAF53 aa 109 -435), does not rescue cell growth is  109 -435 , inhibits the ability of PAF53 to rescue cell proliferation. The dotted blue lines indicate the cell number that was observed when the cells were treated with IAA and transfected with an empty vector. E and F, Western blottings for PAF53 demonstrate that all three of the deletion mutants tested were expressed for the duration of the experiment. B-D, each point represents a separate biological replicate. The error bars indicate the mean Ϯ S.D. ** and ****, significance was determined by a two-tailed t test, as 00.01-0.001 and Ͻ0.0001, respectively.

PAF53 is essential for cell growth
strongly supportive of a model in which a significant portion of the interaction of PAF53 with pol I depends upon its N terminus and perhaps the dimerization domain. Specifically, the interaction of mammalian PAF53 with PAF49 is necessary for PAF53 function.
The tWH domain of RPA49 has been demonstrated to be a DNA-binding domain, and models of rDNA transcription suggest that the tWH of RPA49 may have multiple roles in rDNA transcription, including the stabilization of the open complex by binding upstream DNA (42,107). We have found that the putative tWH of mammalian PAF53 has DNA-binding activity and that the C terminus of PAF53 is required for full activity. This suggests an essential role for the tWH in mammalian cells. Engel et al. (108) suggest that the tWH domain may move with upstream DNA facilitating the transition from initially transcribing the complex to elongating the complex and stabilizing that structure. This would be consistent with the finding that the tWH domain enhances processivity in vitro (42).
As shown in Fig. 5C, we found that PAF53 1-222 was nearly 80% as effective as intact PAF53 in its ability to support cell division. In contrast, a mutant that lacks a significant fraction of the linker (aa 1-160) was essentially inactive in this assay. When combined with the finding that the dimerization domain itself was necessary to rescue cell growth fully, these results would argue for an essential role for the linker.
Experiments on yeast RPA49 had not predicted a role for the "linker." The observation that in silico analysis of the linker of mammalian PAF53 predicts a structure similar to that observed for the yeast argues for a conserved function. The predicted structure for mammalian PAF53 contains an HTH domain, a potential DNA-binding domain, that lies very close to the DNA just downstream of where the bubble would form in the closed complex (PDB 5W66). Furthermore, the construct PAF53 1-160 clips the C-terminal helix within this HTH. We then found that this region has DNA-binding activity and that mutagenesis of the helices, alone or in combination (data not shown), inhibited DNA-binding by PAF53  . Likewise, deletion of the HTH in yeast resulted in a molecule that fails to rescue cell growth, consistent with a conserved and essential role for RPA49 in pol I transcription. Interestingly, the structure of the linker has not always been clearly discerned in structures of pol I (104), possibly due to dynamics in its interactions with core pol I or the template. It has been proposed that the linker spans the activesite cleft and may in fact play a role in the process of proceeding from initiation to escape (104, 109). We are determining the role of the HTH in transcription, e.g. the transition from initiation to elongation.
The placement of the linker vis à vis the template in PDB code 5W66 led us to model a potential interaction with the DNA template. As shown in Fig. 12, the linker can be modeled to interact with the DNA template in the closed complex and with one strand of DNA in the open complex. As RPA49 is a single-stranded nucleic acid-binding protein, it is possible that the interaction of the linker with the template participates in melting, whereas the interaction of the tWH facilitates elongation.
Our results demonstrate that that all three domains of PAF53 are necessary for mitotic cell growth (1, 110 -112). The role(s) each of these domains play during rDNA transcription will be the subject of future studies. Most particularly, we found a requirement for both the linker and the heterodimerization domains of PAF53 (88). This latter finding suggests that PAF53 function may be dependent upon the interaction with PAF49. This is in contrast to what is observed in S. cerevisiae and bears further scrutiny. Yeast expressing a mutant RPA49 in which the dimerization domain has been deleted grew nearly as well as WT at 30 and 18°C (Fig. 11).
Previously, our lab has demonstrated that inhibiting rRNA synthesis in cancer cells causes cell death while normal cells arrest (113), and ribosome biogenesis has become a target for cancer chemotherapy (14, 114 -116). Understanding the role PAF53 plays in pol I transcription could provide novel drug targets that could be utilized in cancer therapy.

Cell culture, transfection, selection, and analysis
HEK293 cells (ATCC) were cultured as recommended in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and antibiotic/antimycotic from Invitrogen. Cells were routinely passed at 1:4 dilutions every 3rd day and were not used after the 18th passage. Transfection of 60% confluent HEK293 cells was carried out as described previously (34,88) using polyethyleneimine (117). After 8 h, the medium was changed. When selecting for stable transfections, the selection antibiotic was added 48 h post-transfection. The sequences encoding NLS-TIR1 were cloned in pcDNA3.1/Hygro(ϩ), and selection for expression of NLS-TIR1 required hygromycin (100 g/ml). When the goal was to recombine into the PAF53 gene, the cells were given puromycin 24 h following transfection. Forty eight hours later, the medium was changed to medium free of puromycin, but containing G418 (500 g/ml) to select for recombinants. To determine whether specific constructs rescued the ability of cells to divide, the cells were cotransfected with a vector expressing the PAF53 construct to be tested and a vector expressing eGFP to mark the transfected cells. Two days later, 500 M IAA was added to the culture PAF53 is essential for cell growth medium. IAA (Abcam) was made fresh by dissolving in water immediately before use. After the indicated time, the number of green fluorescent cells in the population was determined by FACS analysis. At least 10,000 cells were counted for each analysis, which was carried out in triplicate for each time point of an experiment. Cell counts, trypan blue exclusion, and cell cycle analysis were carried out as described (113,118). Metabolic labeling with [ 3 H]uridine was done as described previously (113). Briefly, clone 763 cells were transfected with empty vector or vector expressing mouse PAF53. 48 h later, the cells were treated with IAA or vehicle. Three hours later, [ 3 H]uridine was added to the medium. After 30 min, the cells were harvested, and RNA was isolated. The isolated RNA was fractionated by agarose gel electrophoresis, stained with ethidium bromide, photographed, and transferred to polyvinylidene difluoride. The filters were dried and impregnated with Enhance (PerkinElmer Life Sciences) and subjected to fluorography as described previously (113).

Yeast growth assays
The yeast RPA49 deletion strain used in this study was BY27858/NOY2305 (MATa ade2-1 ura3-1 trp1-1 leu2-3,112  his3-11,15 can1-1 rpa49⌬::His3MX6) and was obtained from the Nomura collection available from the Genetic Resource (NBRP, Japan). This strain was transformed with various highcopy number TRP1 marker RPA49 expression plasmids to test their effect on yeast growth. The entire RPA49 gene locus starting 360 bp upstream of the ATG start codon and 143 bp down-   , lacking the C-terminal tWH domain, were expressed in Rosetta DE3 (Novagen), purified by IMAC, analyzed by SDS-PAGE (inset), and assayed for DNA-binding activity using the same fragment as in Fig. 6. The Coomassie-stained SDS-PAGE is a composite picture from which intervening lanes containing different amounts of the purified proteins were removed.

PAF53 is essential for cell growth
stream of the TAG stop codon was cloned into the 2 micron pRS424 -TRP1 marker plasmid, and the indicated deletion mutations were generated by QuikChange PCR mutagenesis. Yeast cell viability assays were performed on glucose complete (GluC) plates lacking tryptophan (Trp). Cells were cultured overnight in GluC-Trp media at 30°C until saturated. Equivalent amounts of the saturated overnight culture were pelleted and washed two times in sterile, deionized H 2 O. Cells were resuspended in 1 ml of fresh sterile, deionized H 2 O and prepared using 5-fold serial dilutions. Cells were spotted onto GluC-Trp plates and grown at 30 and 18°C to examine yeast cell growth.

Constructs
The WT and deletion clones of mouse PAF53 used in these experiments were described previously (34,88) and is NCBI Reference Sequence NP_073722.1. In these clones, PCR was used to insert a FLAG-tag onto the N terminus of WT mouse PAF53 cloned in pCDNA3.1 downstream of the cytomegalovirus promoter. The deletion clones, 1-160 and 1-220, were cloned in pEBG-GST (Addgene). The vector expressing eGFP was described previously (34). All constructs were confirmed by sequencing. The vector expressing AID-H2B-YFP was obtained from Addgene.

gRNA Design
We used CHOPCHOP (http://chopchop.cbu.uib.no/) 3 as described (119) to design gRNA. Bsmb1 linkers were added to two gRNAs targeting exon 12 of human PAF53, TAATCTTC-CTCCGCTTTGCCAGG and TAGACGCATGCTTTCCAG-ACAGG (the PAM is underlined), and the oligonucleotides were then annealed following a standard protocol and ligated into the vector (120, 121), plentiCRISPR version 2 (Addgene (122)). MilliporeSigma supplied the oligonucleotides. The constructs were confirmed by sequencing. The use of plenti-CRISPR, version 2, which is constructed around a 3rd generation lentiviral backbone, allows for the simultaneous infection/ transfection of the vector for the expression of Cas9 and gRNA and for selection for puromycin resistance.

Homologous recombination
An oligonucleotide was designed that contained 125 nucleotides of PAF53-coding sequence in-frame with DNA that codes for amino acids 68 -111 of A. thaliana IAA17. This was followed by GSG and the sequence coding for an F2A self-cleaving peptide (123-125) (VKQTLNFDLLKLAGDVESNPGP) that was upstream of the sequence coding for neomycin phosphotransferase, a stop codon, and the next 125 nucleotides of 3Ј-noncoding PAF53 exon (Fig. 2). The oligonucleotide was synthesized by Genewiz and cloned into pUC57Kan. Cells were transfected with plentiCRISPR, version 2, coding for the gRNAs that target exon 12 along with the plasmid containing the oligonucleotide for recombination. Twenty four hours following transfection, puromycin was added to the culture medium (6 g/ml). After another 48 h, the medium was changed to medium containing G418 (500 g/ml) to select for recombination. After 72 h, the surviving cells were subject to cloning by limiting dilution. After cloning, the success of recombination was confirmed by PCR, as described previously, using the forward and reverse primers, 5Ј-CAGTCATGTTGAGGGGTCT-CTCCAGTTCTTCTG-3Ј and 5Ј-GCCCTTGAAGAACTCG-TCAAGAAGGC-3Ј, as indicated in Fig. 2 (50) and Western blottings for PAF53. The PCR products were cloned in PCR blunt II TOPPO (Invitrogen). Both strands of four separate clones were sequenced. Analysis demonstrated that the sequence of all four clones were identical.

Western blotting, protein purification, and EMSA
Cell lysis and Western blotting was carried out, as described previously (126), using antibodies to PAF53 (48). The anti-PAF53 antibodies were either raised to recombinant PAF53 in Figure 10. PAF53 function requires the linker HTH to function. Forty eight h prior to treatment with IAA, cells were cotransfected with vectors expressing eGFP and either WT PAF53 or the indicated HTH mutants of mouse PAF53 or mouse PAF53  . Cells were then treated with IAA (500 M) to deplete PAF53. Five days after IAA, cells were harvested and GFPϩ cells counted. The growth of the cells transfected with the indicated forms of PAF53 is presented relative to the growth of cells transfected with empty vector in matched experiments. Each point represents a separate biological replicate. ****, significance was determined by a one-way ANOVA. The error bars indicate the mean Ϯ S.D. Figure 11. Spot assays of yeast transformed with various high-copy pRS424 RPA49 expression vectors. The rpa49⌬ deletion strain was transformed with the indicated RPA49 expression vectors or with empty pRS424. These strains were tested for growth by plating 5 l of serial diluted cells onto GluC-Trp plates incubated at either 30 or 18°C.
PAF53 is essential for cell growth our laboratory (48) or obtained from Proteintech Group and recognizes both human and mouse PAF53. The antibody from our laboratory has been validated previously (35,48), and the depletion of endogenous PAF53 results in a loss of immunoreactivity for both proteins. Recombinant proteins were expressed in either BL21 or Rosetta (Novagen) cells, purified as described previously using either His-, FLAG-, or GST-tags (34,55,88), and analyzed by SDS-PAGE and Coomassie Brilliant Blue R staining. To quantitate the amount of protein used in the electrophoretic mobility shift analysis (EMSA), the purified proteins were electrophoresed in parallel with BSA standards. The gels were then stained with Coomassie Brilliant Blue R and scanned with a Chemi-Doc MP imaging system (Bio-Rad), and the amount of recombinant protein was determined by plotting a standard curve for the BSA (49). EMSA was carried out as described previously (55). Binding was carried out in 10% glycerol, 50 mM KCl, 5 mM MgCl 2 , and 1 mM DTT using fluorescently-tagged DNA, nucleotides Ϫ30 to ϩ10 of the rat rDNA promoter, and no competitor. Gel electrophoresis was carried using Tris (25 mM)/glycine (0.19 M) buffer at pH 8.3. The gels were pre-electrophoresed for 30 min at 60 V before loading and running in the dark. Western blottings and shifts were visualized with a ChemDoc MP (Bio-Rad).

Statistical analysis
All experiments were reproduced at least three times with three technical replicates each time. Quantitative results that required comparisons between groups were subject to statistical analysis using two-tailed Student's t test for two groups or one-way ANOVA followed by Dunnett's multiple comparison test to determine significant differences among more than two groups. Data met the assumptions of the tests (i.e. normal distribution, similar variance). Normality of our data was determined via a D'Agostino-Pearson omnibus normality test and a Shapiro-Wilk normality test. To deter-mine whether the variance differed between groups, an F test was performed when comparing two groups, and a Bartlett's and Brown-Forsythe test was used to compare the variance of more than two groups.