Nucleolar GTPase Bms1 displaces Ttf1 from RFB-sites to balance progression of rDNA transcription and replication

Abstract 18S, 5.8S, and 28S ribosomal RNAs (rRNAs) are cotranscribed as a pre-ribosomal RNA (pre-rRNA) from the rDNA by RNA polymerase I whose activity is vigorous during the S-phase, leading to a conflict with rDNA replication. This conflict is resolved partly by replication-fork-barrier (RFB)-sites sequences located downstream of the rDNA and RFB-binding proteins such as Ttf1. However, how Ttf1 is displaced from RFB-sites to allow replication fork progression remains elusive. Here, we reported that loss-of-function of Bms1l, a nucleolar GTPase, upregulates rDNA transcription, causes replication-fork stall, and arrests cell cycle at the S-to-G2 transition; however, the G1-to-S transition is constitutively active characterized by persisting DNA synthesis. Concomitantly, ubf, tif-IA, and taf1b marking rDNA transcription, Chk2, Rad51, and p53 marking DNA-damage response, and Rpa2, PCNA, Fen1, and Ttf1 marking replication fork stall are all highly elevated in bms1l mutants. We found that Bms1 interacts with Ttf1 in addition to Rc1l. Finally, we identified RFB-sites for zebrafish Ttf1 through chromatin immunoprecipitation sequencing and showed that Bms1 disassociates the Ttf1‒RFB complex with its GTPase activity. We propose that Bms1 functions to balance rDNA transcription and replication at the S-phase through interaction with Rcl1 and Ttf1, respectively. TTF1 and Bms1 together might impose an S-phase checkpoint at the rDNA loci.


Introduction
During the S-phase, eukaryotic pre-ribosomal RNA (pre-rRNA), the precursor for 18S, 5.8S, and 28S ribosomal RNAs (rRNAs), is actively synthesized by RNA polymerase I (Pol-I), leading to either a head-on or codirectional conflict between rDNA transcription and replication (Klein and Grummt, 1999;Akamatsu and Kobayashi, 2015). The head-on conflict is resolved by the replication fork barriers (RFBs) containing DNA sequences downstream of the RNA gene, whereas the codirectional conflict is resolved by the RFBs formed by an RNA-DNA hybrid called R-loop (Gerber et al., 1997;Kobayashi, 2003;Mirkin and Mirkin, 2007;Akamatsu and Kobayashi, 2015;Hamperl and Cimprich, 2016). In baker's yeast, the locus of rDNA contains~150 tandem repeats of the 35S pre-rRNA genes separated by 5S rRNA gene (Henras et al., 2015) and the headon RFBs are localized between the 35S pre-rRNA and 5S rRNA genes (Linskens and Huberman, 1988). The head-on RFB sequences serve as the docking site for protein factors such as Fob1 to resolve the collision between transcription and replication (Kobayashi, 2003;Mirkin and Mirkin, 2007;Akamatsu and Kobayashi, 2015;Hamperl and Cimprich, 2016). In mouse and human cells, the head-on RFB sequences, also called the 'Salboxes', are located downstream from the 47S pre-rRNA-coding region (Bartsch et al., 1987;Lopez-estrano et al., 1998;Akamatsu and Kobayashi, 2015) and Ttf1 is the factor to bind to the head-on RFB-sites to terminate rDNA transcription and mediate replication fork arrest (Bartsch et al., 1987;Evers et al., 1995;Diermeier et al., 2013;Akamatsu and Kobayashi, 2015). However, how TTF1 is displaced from the head-on RFBsites to allow the replication fork progression remains elusive.
Successful completion of cell cycle is controlled by the G1/S, S/G2, G2/M, and metaphase-to-anaphase transition checkpoints (Keaton, 2007;Malumbres, 2014;Saldivar et al., 2018). Considering the special genomic features of the rDNA loci, which usually harbor hundreds of copies of tandemly arrayed rDNA genes, and the role of TTF1 in resolving the head-on conflict between rDNA transcription and replication, we hypothesized that the on-and-off of TTF1 at the RFB-sites might act as a control of the S-to-G2 progression. We reveal here that nucleolar GTPase Bms1 directly displaces Ttf1 from the RFB-sites to facilitate the replication-fork progression, thus establishing a molecular mechanism for resolving the head-on confliction between transcription and replication at the rDNA loci at the S-phase.

Results
Loss-of-function of Bms1l upregulates 45S pre-rRNA transcription and alters nucleolar morphology The zebrafish bms1l sq163/sq163 homozygous mutant, which harbors L 152 to Q 152 substitution in Bms1l, confers a small liver phenotype due to cell cycle arrest, but not to cell apoptosis, characterized by reduced ratios of phospho-Histone H3-positive (pH3-positive) cells (Wang et al., 2012). We generated a new mutant allele bms1l zju1 via CRISPR-Cas9 technology. bms1l zju1 carries one base substitution and a 13-base pair insertion in exon2 that disrupts the bms1l open-reading frame (Supplementary Figure  S1A). Compared with the wild-type (WT) control, the bms1l zju1/ zju1 homozygotes displayed short trunk, small eyes, heart edema, and hardly detectable signals of fatty acid binding protein 10a (fabp10a, a liver marker), intestine fatty acid binding protein (ifabp, an intestine marker), and trypsin (an exocrine-pancreas marker) at 5 day-post-fertilization (5dpf) (Supplementary Figure  S1B-E), phenotypes more severe than that observed in the bms1l sq163/sq163 mutant (Wang et al., 2012). Allelism analysis revealed no or a small liver in the bms1l sq163/zju1 hemizygotes (Supplementary Figure S1F). Therefore, Bms1l plays an essential role in digestive organ development.
Loss-of-function of Bms1l arrests cells at the S-phase but with continuous genomic DNA over-replication The small liver in bms1l sq163/sq163 was attributed to cell cycle arrest (Wang et al., 2012). Surprisingly, immunostaining Figure 1 Loss-of-function of Bms1l upregulates the pre-rRNA transcription and increases the volume of nucleoli in hepatocytes. (A) qPCR analysis of total rRNA transcripts in 5dpf-old bms1l sq163/sq163 mutant and its siblings (the pool of bms1 þ/þ and bms1 sq163/þ , used thereafter) using two pairs of primers derived from 18S (P1) and 28S (P2) showing an upregulation of rRNA expression in bms1l sq163/sq163 mutant. Upper panel, a diagram showing the genomic structure of the zebrafish rDNA gene and the positions of P1 and P2. (B) qPCR analysis showing the upregulation of ubf, tif-IA, and taf1b expression in 5dpf-old bms1l sq163/sq163 mutant compared with the siblings. The qPCR values were normalized against GAPDH and expressed as fold change of expression. The values plotted represent mean ± SEM. **P < 0.01, ***P < 0.001. (C-F) Fibrillarin immunostaining and DAPI staining (green in C and red in E) showing the significant increase in the number of nucleoli in bms1l sq163/sq163 hepatocytes (D, three WT embryos, 1397 cells examined; three bms1l sq163/sq163 embryos, 1274 cells examined) and pancreatic cells (F, three WT embryos, 1015 cells examined; three bms1l sq163/sq163 embryos, 463 cells examined) when compared with WT, respectively, at 5dpf. The pancreatic region (pa) was outlined by a dashed line. Scale bar, 20 lm (C) and 10 lm (E). (G and H) DAPI staining (G) showing the significant increase in the size of nucleolus in bms1l zju1/zju1 hepatocytes compared with WT at 5dpf (H, three WT embryos, 112 cells examined; four bms1l zju1/zju1 embryos, 119 cells examined). Scale bar, 20 lm. (I) Western blotting showing the upregulation of Fibrillarin protein levels in 5dpf-old bms1l sq163/sq163 and bms1l zju1/zju1 mutants compared with their siblings. b-Actin: loading control. Insets in C, E, and G showing higher magnification of a representative nucleus (boxed). The values in D, F, and H plotted represent mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001; NS, no significance. showed a much higher ratio of PCNA-positive (an S-phase marker) hepatocytes in bms1l sq163/sq163 (94.9%) than in WT (46.4%), while without significant difference in the neural tube (NT) (3.1% in WT vs. 3.2% in bms1l sq163/sq163 ) at 5dpf (Figure 2A; Supplementary Figure S3A and B). An EdUincorporation experiment (for checking the activity of DNA biosynthesis, so does BrdU labeling) showed that the ratio of EdU-positive hepatocytes was strikingly higher in bms1l sq163/sq163 at both 4dpf (28.4% in bms1l sq163/sq163 vs. 13.2% in WT) ( Figure 2B; Supplementary Figure S3C) and 5dpf (19.0% in bms1l sq163/sq163 vs. 4.8% in WT) ( Figure 2C). Interestingly, while there was a significant difference in the NT at 4dpf (0.6% in bms1l sq163/sq163 vs. 1.3% in WT) ( Figure 2B), no significant difference (0.7% in bms1l sq163/sq163 vs. 0.8% in WT) was observed at 5dpf between WT and bms1l sq163/sq163 ( Figure 2C), which coincides with the dynamic expression patterns of bms1l during these developmental stages (Wang et al., 2012). We previously showed that bms1l transcripts are maternally deposited, which is expected to support early embryogenesis such as NT formation. The zygotic expression of bms1l was clearly enriched in the digestive organs but not in the NT at 2.5dpf and 4dpf (Wang et al., 2012), suggesting that the early development of digestive organs but not of the NT likely relies on the zygotic Bms1l. Therefore, Bms1l depletion is expected to have little effect on the NT development at 5dpf and the status of the cell proliferation in the NT can be used as the control to demonstrate that the cell proliferation abnormality in the mutant digestive organs is due to the loss-of-function of bms1l rather than a developmental delay.
Data above suggest that the bms1l sq163/sq163 hepatocytes are undergoing active DNA biosynthesis despite of cell cycle arrest. To confirm this, we performed an EdU/BrdU double-labeling pulse-chase experiment. Considering that the retention time of free EdU in the liver is <6 h (Supplementary Figure  S3D), we first injected EdU into the embryos at 4dpf and followed by BrdU injection 5 h later. Doubly injected embryos were fixed at 7 h after BrdU injection for staining of EdU and BrdU, respectively ( Figure 2D, drawing on the top). The scenario is that if a cell is labeled by EdU only (EdU-positive), this cell has been experiencing the S-phase after EdU injection but completes the S-phase before BrdU injection. On the other hand, if a cell is labeled by BrdU only (BrdU-positive), this cell enters the S-phase only after BrdU injection. For a cell labeled by both EdU and BrdU (EdU þ BrdU double-positive), there might be three possibilities: (i) an EdU-positive cell re-enters cell cycle, which can be labeled by BrdU; (ii) an EdU-positive cell has yet completed its S-phase when BrdU is injected; (iii) an EdUpositive cell is arrested at the S-phase but with continuing DNA biosynthesis. Therefore, we anticipated that by comparing the number of BrdU þ EdU double-positive cells vs. BrdU-onlypositive cells, we could determine the status of DNA biosynthesis in each cell during this period ( Figure 2D). Statistics showed that the fold of BrdU þ EdU double-labeled to BrdU-only-labeled cells was $2.2 and $14.8 in the bms1l sq163/sq163 and bms1l zju1/zju1 livers, respectively, drastically and significantly higher than that in the WT (0.42) ( Figure 2E). These data suggest that mutant hepatocytes are arrested at the S-phase characterized by continuous genomic DNA biosynthesis.
To determine whether continuous genomic DNA biosynthesis finally yields cells containing abnormal DNA contents, we microdissected liver buds from 5dpf and 6dpf WT and bms1l sq163/sq163 embryos in the Tg(fabp10a: RFP) genetic background where hepatocytes were specifically labeled by RFP (Her et al., 2003). Liver buds were disassociated and $86% of the dissociated cells were confirmed to be hepatocytes by costaining of RFP and the hepatocyte marker Bhmt (Yang et al., 2011;Gao et al., 2018;Supplementary Figure S3E). DNA content analysis of RFP-positive cells by flow cytometry (Guan et al., 2016) showed that the ratio of cells with DNA content !4n against total cells were $30.4% and $33.1% in 5dpf and 6dpf bms1l sq163/sq163 mutants, respectively, much higher than that in the siblings (the pool of bms1 þ/þ and bms1 sq163/þ , 10.5% and 8.7% at 5dpf and 6dpf, respectively) ( Figure 2F and G). Notably, the DNA content >4n in the mutant cells did not show distinct peaks, suggesting that these cells suffered from genomic DNA partial over-replication but not polyploidic.

Loss-of-function of Bms1l upregulates the transcriptome associated with the nucleolar activities
Next, we performed an RNA sequencing (RNA-seq) experiment to determine the effect of Bms1l depletion on the gene expression profiles in 3dpf-old embryos. The qualities of the RNA-seq data from three WT and three bms1l zju1/zju1 mutant samples were evaluated by the number of the clean reads (between 5.4G and 6.4G), mapping rates (>84% for all six samples) of the clean sequences to the zebrafish genome ((Danio_rerio.GRCz10.84 from ENSEMBL) and hierarchical clustering analysis (Supplementary Figure S4A and B). Cross comparison of gene expression using the DEseq method based on the reads per kilobase million mapped reads method (Wang et al., 2010;Wagner et al., 2012) identified 453 downregulated genes (log 2 À1, P < 0.05) and 390 upregulated genes (log 2 ! 1, P < 0.05) in the bms1l zju1/zju1 mutant embryos (Supplementary Figure 4C, Tables S1 and S2).
Loss-of-function of Bms1l activates DNA-damage response Consistent with the RNA-seq data (Supplementary Figure   4D), a qPCR analysis showed that the expression level of cyclin E1 but not CDK2 was significantly higher in bms1l sq163/sq163 and bms1l zju1/zju1 with respect to their siblings ( Figure 3B). Consequently, the level of Cyclin E1 protein was elevated in both mutants ( Figure 3C). Cyclin-dependent kinase 2 (CDK2) is essential for the G1/S-phase transition during cell cycle. Cyclin E1 is a member of the highly conserved cyclin family, which regulates the function of CDK2 by forming a complex with CDK2. The timing expression of Cyclin E1 plays a direct role in the initiation of DNA replication (Keaton, 2007;Malumbres, 2014;Saldivar et al., 2018). The upregulation of Cyclin E1 nicely coincides with the genomic DNA partial over-replication in bms1l sq163/sq163 and bms1l zju1/zju1 hepatocytes (Berger et al., 2005). Genomic DNA partial over-replication is expected to cause DNA-damage response (Alexander and Orr-Weaver, 2016). Indeed, protein levels of Chk2 and Rad51, two DNA-damage response factors, were highly elevated in both mutants at 5dpf ( Figure 3D and E). We also observed an obvious upregulation of p53 in both mutants ( Figure 3F), together with upregulation of p53-responsive genes D113p53, p21, mdm2, and cyclin G1 (Berghmans et al., 2005;Chen et al., 2005Chen et al., , 2009; Supplementary Figure S5A) and the D113p53 family proteins ( Figure 3F; Shi et al., 2015). Strikingly, the increased p53 and D113p53 were highly enriched in the nucleolus of the mutant hepatocytes (Supplementary Figure S5B). Introducing the p53 M214K mutation (Berghmans et al., 2005) to the bms1l sq163 mutant (i.e. p53 M214K/M214K bms1l sq163/sq163 doublehomozygous mutant) reduced the transcripts of the p53 target genes D113p53, p21, mdm2, and cyclin G1 ( Figure 3G). Interestingly, the sizes of liver and pancreas were only partially recovered (Supplementary Figure S5C-F), suggesting that Bms1l is a multifunctional protein.

Loss-of-function of Bms1l causes replication-fork stall
Next, we examined PCNA, Rpa2, and Fen1, hallmarks for the DNA replication fork progression (Shen et al., 2005) and found that all were highly upregulated in bms1l sq163/sq163 and bms1l zju1/zju1 at 5dpf ( Figure 4A). Strikingly, immunostaining analysis revealed the accumulation of Rpa2 in the nucleoli ( Figure 4B) and the ratio of nucleolar Rpa2-positive cells was significantly increased in both mutants ( Figure 4C). Immunotransmission electron microscopy (immuno-TEM) analysis of liver cells from 5dpf WT and bms1l sq163/sq163 embryos in the Tg(fabp10a: RFP) genetic background identified a large number of Rpa2-positive gold particles localized in the nucleolus in bms1l sq163/sq163 , in contrast to only a few in WT ( Figure 4D; Supplementary Figure S6). Compared with a significant increase of Rpa2-positive gold particles in the mutant nucleoli, no significant difference was observed in the nucleoplasm between WT and bms1l sq163/sq163 hepatocytes ( Figure 4E). Therefore, replication fork is stalled in bms1l sq163/sq163 , prominently in the nucleoli.

Ttf1 is accumulated in the nucleolus in bms1l mutants
The head-on conflict between rDNA transcription and replication during the S-phase in eukaryotes is resolved by the RFBs and the associating factors such as TTF1 in human and mouse ( Figure 5A; Gerber et al., 1997;Kobayashi, 2003;Mirkin and Mirkin, 2007;Akamatsu and Kobayashi, 2015;Hamperl and Cimprich, 2016). The upregulation of rDNA transcription and replication fork stall in bms1l sq163/sq163 and bms1l zju1/zju1 prompted us to explore whether Bms1l plays a role in facilitating replication-fork progression at the rDNA loci. The Myb-like DNA-binding protein Ttf1 is a key head-on RFB-site-binding factor at the rDNA loci (Bartsch et al., 1987;Gerber et al., 1997;Akamatsu and Kobayashi, 2015). Zebrafish genome contains two closely linked homologous ttf1 genes on chromosome 5, namely ttf1a and ttf1b (Supplementary Figure S7A-C). Sequencing analysis revealed that except for 2dpf ($40%), over 60% of the transcripts represented ttf1a from 3dpf to 5dpf (Supplementary Figure S7D). The identity of the endogenous Ttf1 was determined based on an antibody detecting both Ttf1a and Ttf1b proteins (Supplementary Figure S7E) and a MO targeting the translation start codon (TSC) ATG region (identical in ttf1a and ttf1b) (Supplementary Figure S7F and G) for knockdown of the endogenous Ttf1a/Ttf1b protein expression (Supplementary Figure S7H). Total Ttf1, as Bms1l, was expressed in embryos from 2dpf to 5dpf (Supplementary Figure  S7I).

ChIP-seq identification of zebrafish Ttf1-binding sites (RFB-sites)
Zebrafish genome contains two main rDNA loci, the maternal type (M-type rDNA, located on chromosome 4, whose transcripts were detected in the egg and female gonad) and the somatic type (S-type, located on chromosome 5, transcribed in embryos and different tissues/organs) (Locati et al., 2017;Tao et al., 2020). A chromatin immunoprecipitation sequencing (ChIP-seq) analysis (GEO accession number: GSE176455) revealed that Ttf1 bound to DNA sequences in 35 regions in the zebrafish genome (Supplementary Tables S5 and S6), including four homologous regions, namely peak-1 to peak-4, downstream of the somatic rDNA gene on chromosome 5 at 3dpf ( Figure 5E; Supplementary Figure S9A). Notably, the sequences in these four peaks do not contain an obvious consensus DNA sequence (AGGTCGACCAGATTANTCCG) for human Ttf1 (Supplementary Figure S9A; Bartsch et al., 1987Bartsch et al., , 1988, which might owe to the low identity (30%) between their Myblike DNA-binding domains in zebrafish Ttf1a and human TTF1.
We then expressed FLAG-tagged Ttf1a (cloned into the pCS2 þ expression vector) in 293 T cells and extracted the protein for incubation with synthesized peak-1 and peak-3 sequences (using the protein from the pCS2 þ vector-transfected cells as the control). ChIP-qPCR analysis showed that Ttf1a strongly bound to peak-1 and peak-3 sequences ( Figure 5F; Supplementary S9B). Furthermore, a ChIP-PCR analysis revealed that the occupancy of the endogenous Ttf1 on its binding sites was significantly enriched in bms1l sq163/sq163 and bms1l zju1/zju1 when . The ratio of the nucleolar Rpa2-positive cells was significantly higher in bms1l sq163/sq163 and bms1l zju1/zju1 mutants than in WT at 5dpf (C). Samples: two WT embryos, 544 cells examined; three bms1l 163/sq163 embryos, 845 cells examined; two bms1l zju1/zju1 embryos, 277 cells examined. Scale bar, 10 lm. (D and E) Representative images showing Immuno-TEM of Rpa2 in hepatocytes dissected from the WT and bms1l sq163/sq163 embryos at 5dpf (D). Statistical analysis showing a significant accumulation of Rpa2-positive gold particles in the nucleoli but not in the nucleoplasm in bms1l sq163/sq163 hepatocytes compared with WT (E). For each genotype, 17-20 nucleoli were examined. Each dot represents the ratio of gold particle numbers against area of the nucleolus or nucleoplasm in one cell. Scale bar, 0.2 lm. In C and E, **P < 0.01, ***P < 0.001; NS, no significance. Figure 5 Ttf1 is accumulated on its binding sites in bms1l mutants. (A) A drawing to illustrate the head-on conflict between rDNA transcription and replication during the S-phase in higher eukaryotes. Binding of Ttf1 to the RFB-site prevents the conflict. Once the transcription is completed, Ttf1 is proposed to leave the RFB-site to allow progression of the replication fork; however, how Ttf1 is dissociated from the RFB-site remains unknown. (B) Western blotting showing an increase of Ttf1 protein level in bms1l sq163/sq163 and bms1l zju1/zju1 but not in rcl1 À/À compared with their corresponding siblings at 5dpf. Tubulin: loading control. ?: unknown protein. (C) Signal intensity of nucleolar Ttf1 in the gut epithelia was significantly higher in bms1l sq163/sq163 and bms1l zju1/zju1 than in WT and rcl1 À/À . Samples: three WT embryos, 364 cells counted; three bms1l sq163/sq163 embryos, 183 cells counted; three bms1l zju1/zju1 embryos, 98 cells counted; three rcl1 À/À embryos, 176 cells counted. (D) Ratio of the nucleoli displaying intervening signals of Ttf1 and Rpa2 in the gut epithelia was significantly higher in bms1l zju1/zju1 than in WT. Samples: three WT embryos, 264 cells examined; three bms1l zju1/zju1 embryos, 140 cells examined. (E) ChIP-seq identification of Ttf1-binding sites (peak-1 to peak-4) downstream of the 3 0 -end of rDNA gene on chromosome 5 (Chr.V). (F) ChIP-qPCR analysis showing the strong binding of Flag-tagged Ttf1a to peak-1 and peak-3 but not to the 18S (left panel) or 28S (right panel) rDNA. (G) Endogenous Ttf1 showed significant enrichment on its binding sites (ChIP-peak: peak-1 to peak-4) in bms1l sq163/sq163 (163 MU) and bms1l zju1/zju1 (zju1 MU) than in WT at 3dpf. IgG and ChIP-18S: negative controls. (H and I) ChIP-qPCR analysis showing that, in the presence of Mg 2þ , the binding of purified Ttf1a to peak-3 depends on GTP but not ATP (H) and is correlated to the dosage of purified Ttf1a (I). Lower panel in I: Ttf1a was examined by western blotting. In C, D, and F-H, **P < 0.01, ***P < 0.001; NS, no significance. compared with that in WT at 3dpf ( Figure 5G). Finally, we purified Flag-tagged Ttf1a expressed in 293 T cells (Supplementary Figure S9C). Upon mixing, the binding of the purified Ttf1a to the peak-3 sequence was depended on the presence of GTP but not ATP ( Figure 5H) in a dosage-dependent manner ( Figure 5I). Therefore, zebrafish Ttf1 binds to specific DNA sites within the rDNA locus.
Bms1l interacts with Ttf1 and directly dissociates the Ttf1-RFB complex with its GTPase activity Coimmunoprecipitation (Co-IP) showed that Bms1l interacted with both Ttf1a and Ttf1b ( Figure 6A). Co-IP analysis of proteins extracted from 3dpf WT embryos showed that endogenous Ttf1, albeit less than Rcl1, was successfully pulled down by Bms1l ( Figure 6B; Supplementary Figure S10A). Interestingly, both endogenous and overexpressed Bms1l sq163 mutant proteins could interact with Ttf1 ( Figure 6B and C).
To determine whether Bms1l could dissociate the Ttf1-DNA complex, we first mixed the purified Ttf1a with the peak-3 DNA and GTP (10 mM). This mixture was then incubated with purified HA-tagged Bms1l-WT, Bms1l sq163 , and Bms1l-X proteins with or without 50 mM GTP ( Figure 6E; Supplementary Figure S10E), respectively. We found that although all three proteins were able to interact with Ttf1, only Bms1l-WT robustly displaced Ttf1a from the peak-3 DNA ( Figure 6E). Since Bms1l-X is a GTPaseinactivated mutant (Leipe et al., 2002) and Bms1l sq163 exhibited only 43% of the WT Bms1l activity ( Figure 6D), we concluded that dissociation of the Ttf1-DNA complex apparently depends on the Bms1l GTPase activity.

Discussion
Organogenesis is characterized by cell specification and proliferation. During cell proliferation, DNA replication at the rDNA loci is poised to at least two main challenges: (i) how to ensure completion and accuracy of replication, since pre-rRNA gene is tandemly repeated on the chromosome and (ii) how to resolve the conflict between rDNA transcription and replication, since Pol-I activity is vigorous during the S-phase (Kobayashi, 2003;Mirkin and Mirkin, 2007;Hamperl and Cimprich, 2016). Based on our data, we propose that Bms1 in both zebrafish and human mediates the communication between rDNA transcription and replication with its GTPase activity at the S-phase. In WT, Bms1l interacts with Rcl1 to initiate the pre-rRNA processing. When pre-rRNA processing is near completion (which is coupled with completion of transcription), Bms1 turns to interact with Ttf1 to displace the latter from the head-on RFB-sites to allow the head-on replication fork to proceed ( Figure 8A). When Bms1l GTPase activity is compromised, rDNA transcription is upregulated due to the impaired pre-rRNA processing ( Figure 8B). Meanwhile, Ttf1 remains on RFB-sites blocking the head-on replication fork progression. Partial over-replication of the genomic DNA triggers DNA-damage response by upregulation of the expression of factors such as Chk2 and Rad51 ( Figure 8B). Combination of cellular stresses, including DNAdamage response and accumulation of aberrantly cleaved rRNA transcripts, activate the p53 pathway ( Figure 8B). All of these are likely the reasons to cause the cell cycle arrest at the S-phase observed in bms1l sq163/sq163 and bms1l zju1/zju1 mutants and in Bms1-koncdown Hela cells. Therefore, nucleolar factors Bms1l and Ttf1 are essential for cell cycle progression at the rDNA loci during the S-phase.
It is surprising that GTP is needed for effective binding of Ttf1 to its target DNA. One possibility is that GTP might help to shift the conformation of Ttf1 to bind to DNA with a high affinity.
Resolving the structure of Ttf1 with or without GTP in the future will answer this intriguing question. We observed that, in the absence of Bms1l, mutant cells undergo continuous genomic DNA re-replication even though DNA-damage response and p53 pathway are activated, suggesting that Bms1l might also play a role, directly or indirectly, in the regulation of the G1-to-S transition, which is independent to its duet with Ttf1. It would be interesting in the future to check the status of ataxia-telangiectasia and Rad3-related, the key regulator of G1-to-S transition (Saldivar et al., 2018), and chromatin licensing and DNA replication factor 1 (Cdt1, a regulator of DNA re-replication) (Roukos et al., 2011) or other key cell cycle regulators in bms1l sq163/sq163 and bms1l zju1/zju1 mutants, which might allow us to understand more about the role of Bms1l in the G1/S transition.
Ribosomopathies are named for the category of diseases caused by defective ribosome function or production (Farley-Barnes et al., 2019;Huang et al., 2020). It is worth investigating whether this function of Bms1 is conserved in mammals. A thorough investigation of the full spectrum of Bms1linteracting proteins will not only allow us to ravel the role of Bms1l in ribosome biogenesis and regulation of cell cycle progression but also shed light on the understanding of ribosomopathies from the views of both defective ribosome production and cell cycle progression. (D) GTPase activity assay using purified proteins showing that Bms1l-X-HA is enzymatically inactive, while Bms1l sq163 -HA exhibits $43% of Bms1l-WT-HA activity. Free phosphate concentration in each reaction was obtained by A630 color absorbance at 5 and 60 min, respectively. The subtracted value between 60 and 5 min is presented as the relative GTPase activity. (E) ChIP-qPCR analysis for the effect of Bms1l on dissociation of the Ttf1a-DNA complex. Purified Ttf1a was mixed with the peak-3 DNA and 10 mM GTP (Flag-Ttf1a mix). The mixture was then incubated with purified Bms1l-WT, Bms1l-163, or Bmsl1-X with or without 50 mM GTP for 60 min. Only Bms1l-WT-HA obviously dissociates the Ttf1a-DNA complex. Lower panels: the Co-IP products (Ttf1a, Bms1l, or derivatives) were examined by western blotting with respective antibodies. ***P < 0.001.

Zebrafish lines and maintenance
Zebrafish AB line was used as the WT control in this work. Fish were raised and maintained according to standard procedures. bms1l sq163 and p53 M214K mutant lines and transgenic line Tg(lfabp: RFP; elaA: EGFP) were obtained as previously described (Berghmans et al., 2005;Wan et al., 2006;Wang et al., 2012). The bms1l zju1 mutant line and rcl1 À/À mutant line were generated by CRISPR-Cas9 technology as detailed in Supplementary Materials and methods. Homozygous mutant embryos (e.g. bms1l sq163/sq163 and bms1l zju1/zju1 ) were obtained by PCR-based genotyping of the progenies derived from the crosses between heterozygous (e.g. bms1l sq163/þ or bms1l zju1/þ ) male and female fish. Considering the progenies  RPA2 and Fibrillarin (E) showing that the nucleolar enrichment of RPA2 after siBMS1#6 treatment is alleviated by Flag-BMS1 R but not Flag-BMS1-163 R (F). Scale bar, 5 lm. (G) Western blotting showing the upregulation of CHK2, phosph-CHK2 (T68), and TTF1 in Hela cells after BMS1 knockdown. In C and F, **P < 0.01, ***P < 0.001. from one cross were raised in the same container, the pool of the bms1l þ/þ (i.e. WT) and heterozygous (i.e. bms1l sq163/þ or bms1l zju1/þ ) progenies, being called 'the sibling' in this work, was sometimes used as a control.

Cell lines, plasmids, and siRNA transfection
Human Hela and 293 T cells were used to express different target proteins by plasmid transfection as detailed in Supplementary Materials and methods. Corresponding siRNA sequences for gene knockdown and primer sequences for constructing plasmids are listed in Supplementary Table S7.
Northern blotting, 18S/28S ratio analysis, and qPCR Total RNA was extracted from WT, mutant siblings, and mutant embryos using Trizol Reagent (Invitrogen, 15596-026). The digoxigenin (DIG)-labeled 5 0 -ETS and ITS1 DNA probes were prepared and RNA gel blot hybridization and qPCR was performed as described previously (Chen et al., 2009;Wang et al., 2012). For 18S/28S ratio analysis, total RNA was analyzed by Agilent Bioanalysis 2100 (Agilent). Primer pairs used for qPCR analysis are listed in Supplementary Table S8.
Whole-mount in situ hybridization fabp10a, ifabp, and trypsin gene fragments were respectively cloned into the pGEM-T vector to produce DIG-labeled RNA probes (Roche DIG RNA Labeling mix 11277073901) for whole-mount in situ hybridization as previously described (Chen et al., 2005).

Protein analysis and preparation of antibodies
Total protein extraction from zebrafish embryos or cultured cells, western blotting analysis, and Co-IP assays were performed as previously described (Chen et al., 2009;Guan et al., 2016). All antibodies used in this study are listed in Supplementary Materials and methods. Detailed protocols are described in Supplementary Materials and methods.

Cryo-sectioning and immunostaining
For cryo-sectioning, embryos were fixed, embedded, and sectioned as detailed in Supplementary Materials and methods. Immunostaining was performed as described (Wang et al., 2016) and also detailed in Supplementary Materials and methods. All immunofluorescence staining images were taken under an Olympus BX61WI confocal microscope.

EdU and BrdU incorporation assay
EdU single-or EdU and BrdU double-incorporation assay (1 nl and 10 mM, respectively) was performed as described in Supplementary Materials and methods. Injected embryos were incubated at 28.5 C till the desired time point for fixation in 4% paraformaldehyde for 2 h prior to cryo-sectioning. BrdU was detected by immunostaining and EdU incorporation by Alexa Fluor 488 Azide (Life Technologies, A10266).

Flow cytometry analysis
Approximately 100 Tg(lfabp: RFP) zebrafish embryos were collected and fixed. Liver was then dissected under fluorescence microscope, followed by treatment with trypsin as previously described (Guan et al., 2016). For DNA content detection, cells were washed by phosphate-buffered saline (PBS) and resuspended in PBS, then incubated with PI (50 lg/ml), and subjected to flow cytometry analysis of the cell cycle by using a BD FACS Calibur flow cytometer. Detailed protocols are described in Supplementary Materials and methods.

RNA-seq and data analysis
Total RNA was extracted from WT and bms1l zju1 mutant embryos at 3dpf, respectively. RNA library construction and highthroughput sequencing were performed by Beijing Annoroad Gene Technology Company. Briefly, multiplexed libraries were sequenced for 150 bp at both ends using an Illumina HiSeq4000 platform. Clean reads were mapped to the zebrafish genome (Danio_rerio.GRCz10.84 from ENSEMBL). The threshold parameters for DEGs were an absolute fold change !2 and P <0.05. A GO enrichment analysis was performed using DAVID (version 6.8) (Huang et al., 2009).

Immuno-TEM detection of RPA
Embryos were first fixed in 2.5% glutaraldehyde followed by three times washes in PBS. Liver was then dissected and fixed at 4 C overnight. Detailed protocols are described in Supplementary Materials and methods.

ttf1a/ttf1b expression analysis
Total RNA was extracted from WT embryos at desired stages for generating cDNA by using M-MLV Reverse Transcriptase kit (Invitrogen, 28025-021). RT-PCR products were obtained by using a pair of primers (F: 5 0 -CGACTCATTAAAGCGATGTATGA-3 0 ; R: 5 0 -CTATTGATTAAAGCTGTTGTTCT-3 0 ) perfectly matching both ttf1a and ttf1b sequences and cloned into the pGEM-T vector. Then, 96 individual Escherichia coli colonies were randomly picked for DNA sequencing to identify clones corresponding to ttf1a or ttf1b based on single-nucleotide polymorphisms between ttf1a and ttf1b.

MO efficiency assay
MO was purchased from Gene Tools. The ttf1-MO (5 0 -AATC TGACAGCATCTCATCCATCGT-3 0 ) was designed to target the TSC ATG region in both ttf1a and ttf1b. ttf1a/ttf1b TSC sequence was cloned upstream to the 5 0 -end of the EGFP gene to construct the pCS2 þ -TS-EGFP plasmid for checking the efficiency of ttf1-MO (Supplementary Table S7).

ChIP-seq and ChIP-qPCR
The ChIP-seq experiment was performed based on previous zebrafish studies with modifications. ChIP-qPCR primers were designed based on CHIP DNA peaks and listed in Supplementary Table S8. Detailed protocols are described in Supplementary Materials and methods. ChIP-seq data GEO accession number is GSE176455.

Assay for Ttf1a binding to peak-3
For DNA-binding assay, purified Flag-Ttf1a protein was mixed first with 5 fmol peak-3 DNA fragment with or without GTP (Thermol Scientific), and then with purified Bms1l-WT-HA, Bms1l-163-HA, or Bms1l-X-HA protein at 28 C for 1 h. The supernatant was collected for qPCR analysis, and the pellet was used as the protein sample for western blotting analysis. Detailed protocols are described in Supplementary Materials and methods.

GTPase activity assay
The GTPase activity assay is based on checking the inorganic phosphate (Pi) level released from a phosphorylated substrate. This assay is performed under standard procedure of PiColorLock Phosphate Detection System (303-0030, Expedeon). Using purified Bms1l-WT-HA, Bms1l-163-HA, and Bms1l-X-HA with GTP (Thermol Scientific) to start the enzyme reaction for 5 and 60 min. The absorbance values at a wavelength between 590 and 650 nm were recorded. EDTA was used to stop the enzyme activity.

Statistical analysis
For statistical analysis, comparisons were made using the Student's t-test assuming a two-tailed distribution, with significance being defined as *P < 0.05, **P < 0.01, and ***P < 0.001. Details for each category of statistical analysis are provided in Supplementary Materials and methods.

Supplementary material
Supplementary material is available at Journal of Molecular Cell Biology online.