Targeting of Hmo1 to subcompartments of the budding yeast nucleolus

Abstract

The nucleolus is a multilayered, membraneless organelle made up of liquidlike biogenesis compartments surrounding an array of ribosomal RNA genes (rDNA). Biogenesis factors accumulate in the outer compartments through RNA binding and phase separation promoted by intrinsically disordered protein regions. In contrast, the nucleolar localization of rDNA-binding proteins, which reside in the central chromatin compartment, is less well characterized. To gain mechanistic insight, we analyzed the localization, mitotic segregation, nucleic acid binding, and nuclear dynamics of the budding yeast rDNA-binding protein Hmo1. Deletion of the main DNA-binding domain, the HMG boxB, compromised Hmo1 transfer to daughter cells in mitosis and transcription-independent rDNA association but still allowed nucleolar localization. The C-terminal lysine-rich region turned out to be a combined nuclear and nucleolar localization sequence (NLS-NoLS). Its integrity was required for maximal enrichment and efficient retention of Hmo1 in the nucleolus and nucleolar localization of the ΔboxB construct. Moreover, the NLS-NoLS region was sufficient to promote nucleolar accumulation and bound nucleic acids in vitro with some preference for RNA. Bleaching experiments indicated mobility of Hmo1 inside the nucleolus but little exchange with the nucleoplasm. Thus, a bilayered targeting mechanism secures proper localization of Hmo1 to the nucleolus.

INTRODUCTION

The nucleolus is the most prominent membraneless compartment of the eukaryotic nucleus and provides the cell with ribosomes (Hernandez-Verdun et al., 2010; Pederson, 2011). The basic organizational principle is conserved among eukaryotes (Hernandez-Verdun et al., 2010). Repeats of ribosomal DNA (rDNA), so-called nucleolar organizer regions (NORs), are the core structures, and further compartments assemble around rDNA arrays depending on transcriptional activity. The central element of the budding yeast nucleolus is the chromatin compartment consisting of the rDNA array on chromosome XII and associated proteins that function in transcription, recombination, and maintenance (Egidi et al., 2020; Tartakoff et al., 2021). The chromatin compartment is polymer-polymer phase-separated from the rest of the genome with the help of self-interacting rDNA-binding proteins and anchored to the nuclear envelope (Mekhail et al., 2008; Hult et al., 2017; Lawrimore et al., 2021). As rDNA is transcribed by RNA polymerase I (pol I), ribosomal rRNA (rRNA) emerges and promotes the self-organizational assembly of different layers of liquid-liquid phase-separated compartments by the recruitment of a multitude of different ribosome biogenesis factors (Brangwynne et al., 2011; Woolford and Baserga, 2013; Grob and McStay, 2014; Falahati et al., 2016; Feric et al., 2016; Yao et al., 2019; Lawrimore et al., 2021; Tartakoff et al., 2021). In these liquidlike biogenesis compartments cleavage, folding, and assembly of rRNA with ribosomal proteins take place. In mammals, the central chromatin compartment is termed fibrillar center, and the liquidlike rDNA surrounding biogenesis compartments are traditionally termed dense fibrillar component and granular component, respectively, according to their appearance in electron microscopy (Hernandez-Verdun et al., 2010). Electron microscopy initially suggested the presence of three different nucleolar subdomains also in budding yeast (Léger-Silvestre et al., 1999). A few years later, reinspection of the literature leads to the view of a bipartite nucleolus in budding yeast (Thiry and Lafontaine, 2005). Recent live-cell imaging with metaphase-arrested budding yeast suggests the existence of at least three compartments (Tartakoff et al., 2021): the central chromatin compartment and two layers of liquidlike biogenesis compartments surrounding the central rDNA. An inner layer, in which the small ribosomal subunit rRNA is processed, and an outer layer, in which the large ribosomal subunit rRNAs are processed.

It is still not absolutely clear by which mechanisms proteins are targeted to and accumulated in different compartments of the nucleolus. No consensus targeting sequences have been identified, suggesting individual mechanisms for nucleolar localization. Functional interactions with other nucleolar components are thought to be a major mechanism for nucleolar targeting (Grob and McStay, 2014). In the case of the rDNA-binding high mobility group box (HMGB) protein UBF in vertebrates, the HMG boxes, which bind rDNA, and its acidic C-terminal tail, which interacts with SL-1 at the rDNA promoter, participate in nucleolar localization (Maeda et al., 1992; Ueshima et al., 2017; Kihm et al., 1998). In the case of the budding yeast nucleolin Nsr1, which acts in rRNA processing, RNA recognition motifs participate in nucleolar localization (Yan and Mélèse, 1993).

Although no consensus sequences for nucleolar localization have been identified, there is a group of short targeting sequences that show a high degree of similarity (Scott et al., 2010). The so-called nucleolar localization sequences (NoLSs) are often rich in positive charge and are preferentially located near the termini of proteins. NoLSs are further subdivided in NoLS-only and joint NLS-NoLS (nuclear localization sequence–nucleolar localization sequence), which also function in nuclear localization. For such sequences, binding of rRNA or negatively charged nucleolar proteins have been suggested as mechanisms for nucleolar targeting (Siomi et al., 1988; Lechertier et al., 2007; Houmani and Ruf, 2009; Musinova et al., 2011). Although a more precise systematic classification of NoLSs with regard to subnucleolar targeting is currently missing, such signals in general seem to accumulate specifically in the granular component of human nucleoli (Lechertier et al., 2007; Scott et al., 2010; Musinova et al., 2011). In contrast, no sequences of high similarity targeting proteins to the chromatin compartment have been described so far.

The budding yeast UBF orthologue Hmo1 is part of a special form of rDNA chromatin and colocalizes with rDNA-binding proteins (Merz et al., 2008; Gadal et al., 2002; Tartakoff et al., 2021). Genetic experiments suggested that Hmo1 is a part of the pol I machinery (Gadal et al., 2002). Deletion of HMO1 shows synthetic lethality with deletions of genes coding for pol I holoenzyme subunits and overexpression of Hmo1 rescues the growth defect caused by deletion of the gene coding for the pol I holoenzyme subunit Rpa49. Chromatin immunoprecipitation analysis showed that Hmo1 is associated with the promoter and the entire 35S gene transcribed by pol I (Hall et al., 2006). More detailed analyses by chromatin endogenous cleavage (ChEC) revealed that Hmo1 is part of a special chromatin form of rRNA genes devoid of nucleosomes (Merz et al., 2008). Transcription by pol I is needed to convert the “closed” nucleosomal form into the “open” Hmo1-associated form (Wittner et al., 2011).

In addition to its nucleolar function in pol I transcription, Hmo1 has a multitude of additional functions throughout the nucleus (Alekseev et al., 2002; Hall et al., 2006; Gonzalez-Huici et al., 2014; Achar et al., 2020). Hmo1 binds to a subset of ribosomal protein gene (RPG) promoters and contributes to their transcriptional regulation (Hall et al., 2006; Zencir et al., 2020; Kasahara et al., 2007). It was suggested that Hmo1 might play a role in coordinating expression of rRNA and RPGs (Hall et al., 2006). Furthermore, Hmo1 preserves genome integrity by contributing to DNA repair and buffering torsional stress between pol II transcribed genes (Alekseev et al., 2002; Gonzalez-Huici et al., 2014; Achar et al., 2020).

Sequence alignments in the initial study that described Hmo1 suggested the presence of two HMG boxes, which in general are nucleic acid binding domains composed of three alpha helices adopting an L-shaped fold (Lu et al., 1996; Bustin, 1999; Weir et al., 1993; Read et al., 1993). The HMG boxA of Hmo1 spans amino acids 12-86, and the HMG boxB spans amino acids 106-178. Biochemical characterization revealed that the HMG boxA of Hmo1 binds DNA with only low affinity, and that the HMG boxB confers most DNA-binding activity (Kamau et al., 2004; Bauerle et al., 2006). HMG boxes have been reported to bind special DNA structures like stem-loops, four-way junctions, or kinked DNA (Bustin, 1999). As demonstrated by extensive structure function analyses, both HMG boxes of Hmo1 are necessary for stimulation of cellular growth and correct binding to RPG promoters (Kasahara et al., 2016). In addition to the two HMG boxes, Hmo1 contains a conserved region implicated in stimulation of pol I transcription and a C-terminal lysine-rich region implicated in nuclear transport and DNA bending (Bauerle et al., 2006; Albert et al., 2013).

rRNA binding and weak multivalent interactions with other nucleolar proteins conferred by intrinsically disordered regions are proposed to drive nucleolar targeting and accumulation via liquid-liquid phase separation in the case of biogenesis components (Emmott and Hiscox, 2009; Brangwynne et al., 2011; Falahati et al., 2016; Feric et al., 2016). In contrast, knowledge about the mechanisms driving nucleolar localization of rDNA-binding proteins is scarce. In this study, we used the budding yeast HMGB protein Hmo1 as an example to study the nucleolar localization of an rDNA-binding protein.

RESULTS

To investigate the localization of Hmo1 derivatives in living cells, constructs were fused with GFP, stably integrated at the ura3 locus, and expressed from the constitutive pTEF2 promoter in an hmo1 background. Protein levels of the derivatives were 4.6- to 9.6-fold elevated compared with endogenous Hmo1-GFP (Supplemental Figure S1A), and endogenous Hmo1-GFP was similar to untagged endogenous Hmo1 (1.3-fold; Supplemental Figure S1B). All strains were spotted on solid agar plates to test for complementation of the growth defect conferred by deletion of endogenous HMO1 (Supplemental Figure S2). Genotypes are given in Supplemental Table S1. We quantified the colocalization of Hmo1 derivatives with different comarkers using a center of mass-based approach (Bolte and Cordelières, 2006). Colocalization applies when the center of mass of a GFP signal coincides with the center of mass of the respective comarker. A constitutively nuclear localizing derivative of the SWI5 NLS fused with mCherry was used as a marker for the nucleoplasm (Arnold et al., 2015). Fob1 fused to tdTomato and Nop56 fused to 3mCherry were used as markers for the nucleolus. Fob1 binds the replication fork barrier site of rDNA and was reported to colocalize with Hmo1 in fluorescence microscopy (Gadal et al., 2002; Kobayashi, 2003). Nop56 is a component of the U3 snoRNP (small nucleolar ribonucleoprotein particle) required for processing of the 18S rRNA (Lafontaine and Tollervey, 2000). To include total nuclear protein in the colocalization analyses, a low threshold was used.

Nucleolar localization independent of the HMG boxB

Hmo1 is known to bind the entire pol I transcribed region of rDNA, and the HMG boxB of Hmo1 was shown to confer most DNA affinity in vitro (Kamau et al., 2004; Bauerle et al., 2006; Hall et al., 2006; Merz et al., 2008). The HMG boxB is, therefore, the prime candidate to investigate regarding nucleolar targeting of Hmo1. We created an internal deletion substituting the HMG boxB with GFP, and we created a full-length version carrying an internal GFP tag as a control. The constructs are termed ΔboxB and Hmo1-intGFP, respectively. An endogenous C-terminal fusion termed Hmo1-GFP served as an additional reference. The subnuclear localization of the GFP fusions was determined in exponentially growing cells by colocalization with the three different comarkers described above (Figure 1).

FIGURE 1:

Nucleolar enrichment in the absence of the HMG boxB. Cartoons illustrate the GFP-fused constructs used for colocalization analysis. boxA: HMG boxA (aa 12-86, according to Lu et al., 1996); boxB: HMG boxB (aa 106-178, according to Lu et al., 1996); NLS: Nuclear localization sequence (aa 212-236, predicted by cNLS Mapper; Kosugi et al., 2009). Note that GFP is not drawn to scale. (A–C) Colocalization of nucleoplasmic Swi5-NLS fused to mCherry, rDNA-binding protein Fob1 fused to tdTomato, and biogenesis component Nop56 fused to 3mCherry with (A) endogenous Hmo1 C-terminally fused to GFP (Hmo1-GFP), (B) overexpressed Hmo1 with internal GFP fusion (Hmo1-intGFP), (C) overexpressed Hmo1 with internal GFP fusion lacking the HMG boxB (ΔboxB). (B–C) Endogenous HMO1 deleted. Bar graphs depict the percentage of GFP mass centers colocalizing with the mass center of the respective comarker. From top to bottom: n = 205, 208, 207, 249, 226, 210, 212, 200, 205. Scale bar, 5 µm.

Hmo1-GFP was highly enriched in a peripheral subregion of the nucleoplasm as judged by comparison with Swi5-NLS and matched well with Fob1-tdTomato and Nop56-3mCherry (Figure 1A): colocalization of Hmo1-GFP with Swi5-NLS was low (2.4%, n = 205), intermediate with Fob1-tdTomato (57.2%, n = 208), and highest with Nop56-3mCherry (82.1%, n = 207). The better match with Nop56-3mCherry might be due to low nucleoplasmic pools of Hmo1-GFP and Nop56-3mCherry. No nucleoplasmic pool could be detected in the case of Fob1-tdTomato. Hmo1-intGFP was also highly enriched in a subregion of the nucleoplasm and overlapped well with Fob1-tdTomato and Nop56-3mCherry (Figure 1B). Colocalization with Swi5-NLS was low (1.2%, n = 249), moderate with Fob1-tdTomato (49.6%, n = 226), and high with Nop56-3mCherry (84.8%, n = 210). Hence, the overexpressed full-length control carrying an internal GFP shows efficient nucleolar enrichment, which is quantitatively well comparable to that of endogenous Hmo1-GFP. Hmo1-intGFP fully complemented deletion of endogenous HMO1 at least at 25 and 30°C (Supplemental Figure S2). Interestingly, deletion of the HMG boxB caused a rather subtle change in subnuclear localization (Figure 1C). ΔboxB accumulated in a subregion of the nucleoplasm and overlapped well with Fob1-tdTomato and Nop56-3mCherry. In addition, 50% of cells of an asynchronous culture showed a single dotlike condensate for ΔboxB within the nucleolus (n = 100). This might be a consequence of overexpression but nonetheless was specific for the ΔboxB construct. Supplemental Figure S3 shows the formation of such nucleolar condensates in a time series. Colocalization of ΔboxB with Swi5-NLS was comparable to that of the full-length controls (1.4%, n = 212), indicating that Hmo1 is not markedly delocalized throughout the nucleoplasm when the HMG boxB is missing. Colocalization with Fob1-tdTomato was moderate and therefore also comparable to the controls (49.0%, n = 200), and colocalization with Nop56-3mCherry was substantial but reduced when compared with the full-length controls (45.4%, n = 205). In agreement with earlier work, ΔboxB did not complement the growth defect of the hmo1 background (Kasahara et al., 2016; Supplemental Figure S2). These data indicate that the HMG boxB, the main DNA-binding domain of Hmo1, is dispensable for efficient nucleolar enrichment.

To analyze the in vivo association with the rDNA array in more detail, we investigated the mitotic segregation of Hmo1 derivatives. The geometry of chromosome segregation is asymmetric as one sister chromatid is being pulled from the larger mother cell into the smaller daughter cell through the bud neck (Renshaw et al., 2010). Total nuclear amounts of rDNA-binding proteins segregate equally in mitosis, like chromosomal rDNA itself, whereas pol I subunits and rRNA biogenesis factors segregate asymmetrically (Girke and Seufert, 2019). Mitotic segregation most probably reflects the in vivo association of a nucleolar protein with the rDNA array. We quantified the proportion of Hmo1-intGFP and ΔboxB inherited by daughter cells in mitosis and used the median as a measure of center and the interquartile range (IQR) as a measure of spread, respectively.

Hmo1-intGFP essentially equally segregated in mitosis and therefore behaved similarly to endogenous Hmo1-GFP (Figure 2, A and C; Supplemental Figure S4, A and B). Daughter cells received 46.9% (IQR: 5.3) of total nuclear signal in the case of Hmo1-intGFP. The slight discrepancy to endogenous Hmo1-GFP, of which daughter cells received 54.9% (IQR: 7.5), might be explained by overexpression or internal tagging. Moreover, the Hmo1-intGFP and Hmo1-GFP signals transported to daughter cells appeared more compact and concentrated than the signal in mother cells, most probably reflecting the higher degree of compaction of chromatin in daughter cells (Figure 2A; Supplemental Figure S4A; Neurohr et al., 2011; Girke and Seufert, 2019). In the case of ΔboxB, we wanted to specifically investigate the mitotic segregation of the Fob1 and Nop56 overlapping nucleolar population. We therefore excluded condensate-containing cells in our first round of analyses. In contrast to the full-length controls, ΔboxB segregated highly asymmetrically (Figure 2, B and C). Daughter cells received 28.9% of the total nuclear signal (IQR: 10.8), and ΔboxB did not seem to be more concentrated in daughter cells. In a second round of analyses, we included cells containing condensates (Supplemental Figure S5, A and B). In this case, the mitotic segregation of ΔboxB was even more asymmetric. Daughter cells received 10.2% (IQR: 16.1) of the total nuclear signal. In more detail, 17 of the 20 investigated cells contained a condensate at the onset of anaphase. In 14 cases thereof, the condensate remained in the mother cell (daughter cells receiving 5.6 to 15.2% of total nuclear signal), and in three cases, the condensate was passed to the daughter cell (daughter cells receiving 42.0, 51.8, and 63.4% of total nuclear signal, respectively). In the case of the three cells that did not contain condensates at anaphase onset, daughter cells received 22.0, 24.0, and 24.8% of total nuclear signal, respectively. Thus, besides the HMG boxB, other unknown domains establish contacts with the rDNA array and suffice for efficient nucleolar enrichment. However, rDNA association mediated by those domains seems to be rather loose as suggested by the highly asymmetric mitotic segregation of ΔboxB.

FIGURE 2:

Equal segregation of Hmo1 during mitosis in budding yeast depends on the HMG boxB. (A, B) Time series showing the mitotic segregation of (A) overexpressed Hmo1 internally fused with GFP (Hmo1-intGFP) and (B) overexpressed Hmo1 with internal GFP fusion lacking the HMG boxB (ΔboxB) at 2-min intervals with Nop56-3mCherry as comarker. Endogenous HMO1 was deleted; t = 0 is the first point in time with completely segregated GFP signals. Scale bar, 5 µm. (C) Box plot showing the percentage of total nuclear signal inherited by daughter cells for Hmo1-intGFP and ΔboxB, respectively (n = 19). Cells containing ΔboxB condensates at the onset of anaphase were excluded.

A lysine-rich NLS-NoLS targets Hmo1 to the nucleolus

Trying to identify a domain that targets Hmo1 to the nucleolus in the absence of the HMG boxB, we directed our attention to the C-terminal region (Figure 3A). This region is rich in lysine and necessary for nuclear localization of Hmo1 (Albert et al., 2013). However, protein sequences rich in basic amino acids are implicated not only in nuclear transport but also in nucleolar targeting (Siomi et al., 1988; Houmani and Ruf, 2009; Scott et al., 2010; Musinova et al., 2011). More precisely, regions rich in basic amino acids are involved in nuclear transport and targeting to the liquidlike granular component of human nucleoli. The program cNLS Mapper (Kosugi et al., 2009) predicts an NLS with the highest score ranging from amino acids 212 to 236. To investigate the role of the C-terminal region in nuclear transport and possibly nucleolar targeting in more detail, we cloned two C-terminal fragments N-terminally fused to GFP (Figure 3B, and C). In one case, we used a fragment spanning only the predicted NLS with the highest score from amino acids 212 to 236 not including the 10 C-terminal amino acids. In the other case, we used a fragment ranging from amino acids 212 to 246, extending the predicted NLS by the 10 C-terminal amino acids (four of which are lysine). We termed the constructs Hmo1-NLS and Hmo1-NLS-C10, respectively.

FIGURE 3:

A joint NLS-NoLS in the C-terminal region targets Hmo1 to the nucleolus. (A) Sequence of the C-terminal region of Hmo1. Hmo1-NLS: Nuclear localization sequence of Hmo1 (aa 212-236) predicted by cNLS Mapper (Kosugi et al., 2009). C10: The 10 C-terminal amino acids following the predicted Hmo1-NLS. (B, C) Colocalization of nucleoplasmic Swi5-NLS fused to mCherry, rDNA-binding protein Fob1 fused to tdTomato, and biogenesis component Nop56 fused to 3mCherry with (B) overexpressed Hmo1-NLS N-terminally fused with GFP, (C) overexpressed Hmo1-NLS-C10 N-terminally fused with GFP. (B, C) Endogenous HMO1 deleted. Line-plots depict the normalized intensity profiles along the yellow lines from left to right. Scale bar, 5 µm.

As predicted, Hmo1-NLS localized in the nucleus (Figure 3B). Using the same comarkers as before, we did not observe nucleolar enrichment. In contrast, Hmo1-NLS-C10 localized in the nucleus and in addition was enriched in subregions of the nucleus that were also enriched with Fob1-tdTomato and Nop56-3mCherry (Figure 3C). As both constructs were highly dispersed throughout the nucleus, we applied line-plots instead of the center of mass-based approach for comparison of those two specific constructs. The data suggest that the C-terminal lysine-rich region of Hmo1 ranging from amino acids 212 to 246 constitutes a joint NLS-NoLS that partially accumulates in the nucleolus. Interestingly, the NoLS activity can be uncoupled from the NLS activity by removing the 10 C-terminal amino acids.

As ΔboxB was highly enriched in the nucleolus, we wanted to test if the 10 C-terminal amino acids would be relevant for nucleolar recruitment in this context. We created a double deletion construct lacking the HMG boxB as well as the 10 C-terminal amino acids that are crucial for NoLS activity of the C-terminal region. The double deletion construct ΔboxBΔC10 was again quantitatively analyzed regarding colocalization with the three comarkers.

In contrast to ΔboxB, which hardly coincided with Swi5-NLS and well coincided with both nucleolar markers (Figure 1C), the double deletion construct ΔboxBΔC10 was relatively homogeneously distributed throughout the entire nucleoplasm (Figure 4). Its center of mass highly colocalized with that of Swi5-NLS (80.7%, n = 223), whereas colocalization with Fob1-tdTomato and Nop56-3mCherry was low (Fob1-tdTomato: 2.0%, n = 203; Nop56-3mCherry: 16.3%, n = 203). The data show that the C-terminal lysine-rich region ranging from amino acids 212 to 246 contains a combined NLS-NoLS, which is critical for nucleolar targeting when the HMG boxB is missing.

FIGURE 4:

C-terminal NoLS activity is crucial for nucleolar enrichment in the absence of the HMG boxB. Colocalization of nucleoplasmic Swi5-NLS fused to mCherry, rDNA-binding protein Fob1 fused to tdTomato, and biogenesis component Nop56 fused to 3mCherry with overexpressed Hmo1 with internal GFP fusion lacking the HMG boxB and the 10 C-terminal amino acids (ΔboxBΔC10). Endogenous HMO1 deleted. Bar graphs depict the percentage of GFP mass centers colocalizing with the mass center of the respective comarker. From top to bottom: n = 223, 203, 203. Scale bar, 5 µm.

NoLS activity is necessary for maximum nucleolar enrichment

Due to the drastic effects observed in the ΔboxB context, we wondered how the C-terminal NoLS activity contributes to nucleolar localization in context of the full-length protein. We expressed an internally tagged construct lacking the 10 C-terminal amino acids to ablate NoLS function (ΔC10). Protein levels were comparable to the other internally tagged derivatives (Supplemental Figure S1).

In addition to a highly concentrated population overlapping with Fob1-tdTomato and Nop56-3mCherry, a nucleoplasmic pool was well apparent in the case of the overexpressed ΔC10 construct (Figure 5). Colocalization with Swi5-NLS was low (8.1%, n = 234) but slightly higher than in the case of the full-length controls and ΔboxB. Colocalization with Fob1-tdTomato was low (12.4%, n = 233), and colocalization with Nop56-3mCherry was moderate (49.6%, n = 234). Thus, ablation of NoLS function in context of the full-length protein results in partial delocalization from the nucleolus. The partial delocalization does not shift the mass center of ΔC10 far enough to result in high coincidence with Swi5-NLS but far enough to markedly reduce coincidence with the nucleolar comarkers. Note that raising the threshold for GFP detection to exclude the nucleoplasmic pool of ΔC10 resulted in decreasing colocalization with Swi5-NLS and a massive increase in colocalization with Fob1-tdTomato and Nop56-3mCherry (Figure 5). This demonstrates that the highly concentrated nucleolar population of ΔC10 is well colocalizing with Fob1-tdTomato and Nop56-3mCherry.

FIGURE 5:

C-terminal NoLS activity contributes to nucleolar recruitment in context of the full-length protein. Colocalization of nucleoplasmic Swi5-NLS fused with mCherry, rDNA-binding protein Fob1 fused to tdTomato, and biogenesis component Nop56 fused to 3mCherry with overexpressed Hmo1 internally fused with GFP and lacking the 10 C-terminal amino acids (ΔC10). Bar graphs depict the percentage of GFP mass centers colocalizing with the mass center of the respective comarker at different GFP thresholds (threshold value of 15 was used in Figures 1 and 4). From top to bottom: n = 234, 233, 234. Scale bar, 5 µm.

In addition to the overexpressed construct, we also generated an endogenous truncation lacking the C-terminal NoLS activity (endΔC10) and an appropriate full-length control (Hmo1-GFP). To this end, we used a PCR-based approach for C-terminal GFP tagging (Sheff and Thorn, 2004).

Both constructs were concentrated in a subregion of the nucleoplasm where the nucleolar comarkers were enriched (Figure 6, A and B). Just like the overexpressed construct, endΔC10 was slightly delocalized throughout the nucleoplasm. We quantified the signals and compared the maximum signal intensities as a measure for maximum enrichment in the nucleolus (Figure 6C). Indeed, the maximum values were significantly reduced for the truncated construct, indicating impaired nucleolar enrichment in the absence of NoLS activity. Importantly, the maximum values for neither Fob1-tdTomato nor Nop56-3mCherry were significantly altered. This excludes the possibility that global changes, for example in rDNA compaction, could be the cause for reduced nucleolar concentration of the truncated construct. Moreover, protein levels of both Hmo1 constructs were similar according to Western blot analysis (Figure 6D). This excludes reduced protein levels as a possible cause for reduced nucleolar enrichment. Growth of the strains carrying the C-terminally truncated versions was well comparable to that of wild type (Supplemental Figure S2). This agrees with the literature (Lu et al., 1996; Albert et al., 2013). Thus, the C-terminal NoLS activity is not essential for nucleolar recruitment in context of the full-length protein but contributes to the authentic nucleolar localization of Hmo1. Maximum enrichment in the nucleolus is only reached in the presence of the 10 C-terminal amino acids.

FIGURE 6:

C-terminal NoLS activity is necessary for maximum nucleolar enrichment of endogenous Hmo1. (A, B) Colocalization of nucleoplasmic Swi5-NLS fused to mCherry, rDNA-binding protein Fob1 fused to tdTomato, and biogenesis component Nop56 fused to 3mCherry with (A) endogenous Hmo1 C-terminally fused with GFP (Hmo1-GFP), (B) endogenous Hmo1 truncated for the 10 C-terminal amino acids and C-terminally fused with GFP (endΔC10). Scale bar, 5 µm. (C) Box plots showing the maximum signal intensities (maximum pixel gray value per nucleolus) of Hmo1-GFP, endΔC10, Fob1-tdTomato, as well as Nop56-3mCherry in cells expressing either Hmo1-GFP or endΔC10, respectively; n = 50. (D) To the left: Western blot with monoclonal anti-GFP antibody from mouse (Roche) showing the protein levels of Hmo1-GFP and endΔC10, respectively. A secondary goat anti-mouse IRDye800CW antibody (Licor Biosciences) was used for detection with a Licor Odyssey scanner. Heterozygous diploid precursor strains to those used for microscopy were used. con: precursor strain without GFP fusion. To the right: Ponceau S staining.

The HMG boxB and the NLS-NoLS contribute to both rDNA and rRNA binding in vitro

In the case of the HMG boxB, it is assumed that nucleolar targeting results from binding of rDNA or special rDNA chromatin structures. In contrast, NoLSs rich in basic amino acids were suggested to target proteins to human nucleoli by binding of rRNA (Siomi et al., 1988; Houmani and Ruf, 2009). We conducted electrophoretic mobility shift assays (EMSAs) to test whether nucleolar targeting by the NLS-NoLS of Hmo1 might be caused by nucleic acid binding. We cloned C-terminal fragments analogous to those used for in vivo localization as well as a full-length control carrying an N-terminal 6His fusion in addition to GFP and purified the recombinant proteins from Escherichia coli. Coomassie staining of the purified proteins is shown in Supplemental Figure S6. We used a 400-bp fragment of the 18S rDNA, which is known to be bound by Hmo1 in vivo (Hall et al., 2006), and the respective sequence as single-stranded rRNA, as it is part of the primary transcript. The precise sequences are shown in Supplemental Table S2. 6His-GFP as negative control shifted neither rDNA nor rRNA at protein concentrations up to 10 µM, which was the maximum concentration tested for all protein constructs (Supplemental Figure S7).

GFP-Hmo1 bound rDNA and rRNA with similar affinity in vitro (Figure 7A; EC50 rDNA: 261 ± 33 nM; EC50 rRNA: EC50: 286 ± 96 nM). In contrast, Hmo1-NLS showed no binding of rDNA at concentrations up to 10 µM and only low affinity for rRNA (Figure 7B; EC50 rRNA: 7047 ± 484 nM). We calculated the relative affinity as the ratio EC50(full-length)/EC50(Hmo1-NLS) to evaluate rRNA binding of Hmo1-NLS in relation to the full-length control. The relative affinity for rRNA was 0.041 ± 0.0029 (Figure 7C). Hence, Hmo1-NLS constitutes a weak rRNA-binding domain. We next tested Hmo1-NLS-C10, which included the 10 C-terminal amino acids essential for NoLS function. Hmo1-NLS-C10 shifted both rDNA and rRNA and showed a bias as it bound rRNA with higher affinity than rDNA (Figure 7D; EC50 rDNA: 1552 ± 222 nM; EC50 rRNA: 756 ± 45 nM). The relative affinity for rDNA (0.17 ± 0.026) was significantly lower than that for rRNA (0.38 ± 0.022), suggesting that the combined NLS-NoLS also has a preference for rRNA (Figure 7E). Our EMSAs show that Hmo1 binds single-stranded rRNA in addition to double-stranded rDNA in vitro. Nucleolar targeting by the NLS-NoLS might indeed be caused by binding of nucleolar nucleic acids.

FIGURE 7:

The C-terminal NLS-NoLS binds rDNA and rRNA in vitro. Electrophoretic mobility shift assays of N-terminal 6His-GFP fusions for (A) GFP-Hmo1, (B) Hmo1-NLS, and (D) Hmo1-NLS-C10 with rDNA (to the left) and rRNA (to the right), respectively. EC50 values and standard deviations were calculated from three technical replicates. (C, E) rDNA and rRNA affinities of Hmo1-NLS and Hmo1-NLS-C10, respectively, relative to the full-length control shown in A. Relative affinities were calculated as the ratio EC50(full-length control)/EC50(construct). Mean values with standard deviations are shown.

To get a more comprehensive view about nucleic acid binding by Hmo1, additional derivatives were analyzed. We purified Hmo1-intGFP, ΔboxB, ΔC10, and ΔboxBΔC10 containing an additional 6His tag at the N-terminus from E. coli and tested for rDNA and rRNA binding.

Hmo1-intGFP efficiently shifted rDNA with an EC50 value of 196 ± 19 nM and rRNA with an EC50 value of 303 ± 29 nM (Figure 8A). The values were thus comparable to the GFP-Hmo1 construct (Figure 7A). ΔboxB, which efficiently accumulated in the nucleolus (Figures 1 and 2; Supplemental Figure S3), showed only moderate reduction of nucleic acid binding in vitro (Figure 8B; EC50 rDNA: 549 ± 54 nM; EC50 rRNA: 552 ± 111 nM). There was no significant difference in relative affinities (Figure 8C; relative affinity rDNA: 0.36 ± 0.037; relative affinity rRNA: 0.57 ± 0.13). ΔC10, which showed a mild delocalization throughout the nucleoplasm (Figure 5), also showed mild reduction of both rDNA and rRNA binding (Figure 8D; EC50 rDNA: 279 ± 20 nM; EC50 rRNA: 477 ± 82 nM). The difference in relative affinities was not significant (Figure 8E; relative affinity rDNA: 0.70 ± 0.050; relative affinity rRNA: 0.65 ± 0.12). The double deletion version ΔboxBΔC10, which was massively delocalized throughout the nucleoplasm (Figure 4), also showed the most pronounced reduction in both rDNA and rRNA binding (Figure 8F; EC50 rDNA: 1523 ± 171 nM; EC50 rRNA: 1474 ± 135 nM) with significantly higher relative affinity for rRNA (Figure 8G; relative affinity rDNA: 0.13 ± 0.014; relative affinity rRNA: 0.21 ± 0.018). These in vitro studies suggest that nucleolar targeting by the HMG boxB and the NLS-NoLS might be mediated by binding of rDNA as well as rRNA.

FIGURE 8:

The HMG boxB and the C-terminal NLS-NoLS cooperate for maximum nucleic acid binding in vitro. Electrophoretic mobility shift assays of N-terminal 6His fusions for (A) Hmo1-intGFP, (B) ΔboxB, (D) ΔC10, and (F) ΔboxBΔC10 with rDNA (to the left) and rRNA (to the right), respectively. All constructs carried internal GFP fusions analogous to the constructs used for localization studies. EC50 values and standard deviations were calculated from three technical replicates. (C, E, G) rDNA and rRNA affinities of ΔboxB, ΔC10, and ΔboxBΔC10, respectively, relative to the full-length control shown in A. Relative affinities were calculated as the ratio EC50(full-length control)/EC50(construct). Mean values with standard deviations are shown.

Transcription-independent rDNA association by the HMG boxB and transcription-dependent rDNA association by the NLS-NoLS

To determine if in vivo targeting to the rDNA array is dependent on rDNA or rRNA binding, we investigated the localization of Hmo1 derivatives after treatment with rapamycin. Exposure to rapamycin inhibits the TOR pathway and leads to a cessation of rRNA transcription by down-regulation of the Rrn3 dependent promoter recruitment of pol I (Zaragoza et al., 1998; Powers and Walter, 1999; Claypool et al., 2004). Fob1 stays bound to rDNA and separates from rRNA biogenesis factors in cells that are exposed to rapamycin (Mostofa et al., 2018). As first construct, we investigated the effect of rapamycin on the colocalization of Hmo1-NLS-C10 with Fob1-tdTomato and Nop56-3mCherry, respectively.

Line-plots indicate that the nucleolar population of Hmo1-NLS-C10 and Fob1-tdTomato overlap in DMSO-treated control cells, while the nucleolar signals separate following rapamycin treatment (Figure 9A). In contrast, Hmo1-NLS-C10 overlaps well with Nop56-3mCherry under both conditions (Figure 9B). This indicates that recruitment of the C-terminal NLS-NoLS to rDNA is dependent on transcriptional activity.

FIGURE 9:

The combined NLS-NoLS detaches from rDNA after transcriptional inhibition. Colocalization of (A) rDNA-binding protein Fob1 fused to tdTomato and (B) biogenesis component Nop56 fused to 3mCherry with overexpressed Hmo1-NLS-C10 in DMSO-treated control cells (to the left) and rapamycin-treated cells (to the right). Cells were exposed to DMSO or DMSO+rapamycin for 4 h. Endogenous HMO1 deleted. Line-plots depict the normalized intensity profiles along the yellow lines from left to right. Scale bar, 5 µm.

To validate if the center of mass-based colocalization approach readily detects rapamycin-induced separation of rDNA-bound Fob1 from a biogenesis factor, we analyzed the colocalization of Fob1 and Nop56 using different combinations of fluorescent protein fusions (Supplemental Figure S8, A and B). Colocalization of Fob1-tdTomato with Nop56-GFP and Fob1-GFP with Nop56-3mCherry revealed a high center of mass coincidence in DMSO-treated control cells (Fob1-tdTomato + Nop56-GFP: 80.2%, n = 283; Fob1-GFP + Nop56-3mCherry: 90.5%, n = 275) and reduced coincidence when cells were treated with rapamycin (Fob1-tdTomato + Nop56-GFP: 41.9%, n = 308; Fob1-GFP + Nop56-3mCherry: 11.1%, n = 280). Although the specific fluorescent fusions seem to influence the assay, rapamycin-induced separation of rDNA-bound Fob1 from the rRNA biogenesis factor Nop56 can readily be measured using the center of mass-based approach. Interestingly, colocalization of endogenous Hmo1-GFP with Fob1-tdTomato was not affected by rapamycin (Supplemental Figure S8C). Colocalization was high in DMSO control cells (72.8%, n = 268) as well as in rapamycin-treated cells (73.0%, n = 270). In contrast, rapamycin massively affected colocalization of Hmo1-GFP and Nop56-3mCherry (Supplemental Figure S8D). Colocalization was high in DMSO control cells (82.0%, n = 261) and low in rapamycin-treated cells (14.2%, n = 275). Therefore, endogenous Hmo1-GFP stays associated with rDNA when pol I transcription is shut off and separates from the rRNA biogenesis factor Nop56. We next analyzed the effect of rapamycin on the colocalization of Hmo1-intGFP, ΔboxB, ΔC10, and ΔboxBΔC10 with Fob1-tdTomato and Nop56-3mCherry, respectively.

Hmo1-intGFP behaved well comparable to endogenous Hmo1-GFP (Figure 10, A and B). Coincidence with Fob1-tdTomato was high in DMSO-treated cells (78.5%, n = 246) and in cells exposed to rapamycin (73.0%, n = 244). Like the endogenous control, Hmo1-intGFP separated from Nop56-3mCherry following rapamycin treatment as colocalization was markedly reduced (DMSO: 82.2%, n = 242; rapamycin: 21.8%, n = 239). Interestingly, ΔboxB differed considerably from the full-length controls (Figure 10, C and D). Colocalization of ΔboxB with Fob1-tdTomato markedly dropped after rapamycin treatment (DMSO: 64.7%, n = 235; rapamycin: 7.5%, n = 239). This was also the case for colocalization with Nop56 (DMSO: 35.3%, n = 238; rapamycin: 7.1%, n = 241). Importantly, the percentage of cells containing a nucleolar condensate in addition to the Fob1 and Nop56 overlapping population was clearly increased when cells were exposed to rapamycin (DMSO: 43%, n = 200; rapamycin: 88%, n = 200). This indicates that ΔboxB leaves the rDNA array and accumulates in condensates adjacent to rDNA when transcription is shut off. This behavior would be expected for an rRNA biogenesis factor. Nonetheless, ΔboxB and Nop56-3mCherry also separate from each other when pol I transcription is inhibited. In the case of ΔC10, the colocalization with Fob1-tdTomato increased when cells were treated with rapamycin (Figure 10, E and F). Colocalization in DMSO control cells was 55.9% (n = 263) and in rapamycin-treated cells was 79.0% (n = 238). It is noteworthy that a low nucleoplasmic pool was visible in DMSO-treated cells but not in rapamycin-treated cells. Thus, the slight delocalization of ΔC10 throughout the nucleoplasm might be caused by transcriptional activity and transcriptional shutoff may enable ΔC10 to better associate with rDNA. Colocalization of ΔC10 and Nop56-3mCherry was moderate in control cells and dropped due to rapamycin exposure (DMSO: 58.6%, n = 239; rapamycin: 29.4%, n = 238). The colocalization of ΔboxBΔC10 with Fob1-tdTomato was very low under both conditions (Figure 10, G and H; DMSO: 0.4%, n = 269; rapamycin: 0%, n = 275) as was the case with Nop56-3mCherry (DMSO: 11.6%, n = 277; rapamycin: 7.4%, n = 283). In addition to the main findings described above, it is worth mentioning that the center of mass coincidences of DMSO-treated cells were in general different from those of untreated cells. This might be due to the fact that DMSO reduces nucleolar incorporation of 5-ethynyl uridine into rRNA even at low concentrations (Bryant et al., 2022). The data suggest that the HMG boxB establishes profound direct contact with rDNA independent of transcription, whereas the NLS-NoLS establishes indirect contact with rDNA, possibly by binding of rRNA.

FIGURE 10:

Transcription-dependent and -independent association of Hmo1 with rDNA. Colocalization of (A, C, E, G) rDNA-binding protein Fob1 fused to tdTomato and (B, D, F, H) biogenesis component Nop56 fused to 3mCherry with overexpressed (A, B) Hmo1-intGFP, (C, D) ΔboxB, (E, F) ΔC10, and (G, H) ΔboxBΔC10 in DMSO-treated control cells (to the left) and rapamycin-treated cells (to the right). Cells were exposed to DMSO or DMSO+rapamycin for 4 h. Endogenous HMO1 deleted. Bar graphs depict the percentage of GFP mass centers colocalizing with the mass center of the respective comarker. From left to right and top to bottom: n = 246, 244, 242, 239, 235, 239, 238, 241, 263, 238, 239, 238, 269, 275, 277, 283. Scale bar, 5 µm.

In addition to rapamycin treatment, we wanted to analyze a physiological condition where pol I activity is low. To this end, we grew cultures to stationary phase (5 d at 30°C) and investigated the colocalization of Hmo1-intGFP and ΔboxB with Fob1-tdTomato and Nop56-3mCherry, respectively. Note that higher thresholds had to be used in the case of stationary cells to faithfully exclude background fluorescence.

In general, effects were very similar to rapamycin treatment (Supplemental Figure S9). Hmo1-intGFP highly colocalized with Fob1-tdTomato in stationary cells (Supplemental Figure S9A; 77.0%, n = 148), and ΔboxB hardly colocalized with Fob1-tdTomato (Supplemental Figure S9A; 1.5%, n = 134). Both constructs were markedly separated from Nop56-3mCherry in the stationary phase (Supplemental Figure S9B). Colocalization of Hmo1-intGFP with Nop56-3mCherry was 6.2% (n = 177), and colocalization of ΔboxB with Nop56-3mCherry was 17.2% (n = 134). However, although the center of mass colocalization was low in both cases, ΔboxB seemed to be more intimately connected to Nop56-3mCherry (Supplemental Figure S9B). While Hmo1-intGFP spatially separated from Nop56-3mCherry, spherical structures of ΔboxB were surrounded by a layer of Nop56-3mCherry. This is reminiscent of the multiphase nucleolar droplets formed by fibrillarin and nucleophosmin in Xenopus laevis (Feric et al., 2016). Increasing the threshold for Nop56-3mCherry to exclude the lowly concentrated nucleoplasmic fraction resulted in a massive increase in colocalization of ΔboxB and Nop56-3mCherry but not in the case of Hmo1-intGFP and Nop56-3mCherry (Supplemental Figure S9C). Thus, Hmo1 is highly associated with rDNA in the stationary phase when the HMG boxB is present. Association of Hmo1 with rDNA in postlog cells was reported previously (Johnson et al., 2013). In contrast, Hmo1 detaches from rDNA when nucleolar targeting is mediated by the NLS-NoLS. This is in line with the rapamycin experiments and suggests an indirect association of the NLS-NoLS with rDNA. Although both Hmo1-intGFP and ΔboxB separate from Nop56-3mCherry in stationary cells, application of higher thresholds suggests that separation of ΔboxB and Nop56-3mCherry is less pronounced. Altogether, Hmo1 localizes like an rRNA biogenesis factor when the HMG boxB is missing.

Hmo1 is highly mobile within the nucleolus but only moderately exchanges between nucleolus and nucleoplasm

Bleaching studies on various nucleolar proteins suggested rapid exchange between the nucleolus and the nucleoplasm in mammals (Phair and Misteli, 2000; Chen and Huang, 2001; Dundr et al., 2002). Subunits of the pol I holoenzyme, processing factors, and the Hmo1 orthologue UBF were investigated. However, from the equal segregation of Hmo1 during asymmetric mitosis of budding yeast we observed here, one might expect stable association of Hmo1 with the rDNA array. We performed fluorescence loss in photobleaching (FLIP) experiments to investigate the in vivo dynamics of endogenous Hmo1 fused to GFP. First, we wanted to compare the mobility of Hmo1 with that of the nucleoplasmic marker Swi5-NLS. To this end, we cobleached the mother cell-located portions of both signals in midanaphase cells with completely separated nucleolar structures. We applied 30 consecutive bleaching events and quantified the mother- and daughter-located portions of both signals immediately after bleaching and again after 4 min.

Bleaching of mother cells led to an immediate and drastic decrease of the Swi5-NLS signal also in daughter cells (Figure 11, A and B). The mother-located signal decreased to 0.08 ± 0.04 relative to the prebleach value and the daughter-located signal to 0.18 ± 0.13 immediately after bleaching. On the contrary, bleaching of the mother cell nucleolus led to complete extinction of the nucleolar Hmo1-GFP signal in mother cells but did not markedly affect the daughter-located signal (0.98 ± 0.05 relative to prebleach). Even after 4 min, no obvious equilibration could be observed. Both the mother- and the daughter-located signals were almost unchanged (mother: 0.08 ± 0.09; daughter: 0.94 ± 0.16). This shows that the exchange between nucleolus and nucleoplasm for Hmo1 is rather low.

FIGURE 11:

Nucleolar retention and intranucleolar mobility of Hmo1. (A) Cobleaching of Swi5-NLS fused with mCherry and endogenous Hmo1 fused with GFP (Hmo1-GFP) in the mother cell body of midanaphase cells. The indicated area (orange outline) was bleached for 26.49 s (30 bleaching events, 883 ms each). Images before bleaching (pre), immediately after bleaching (0 min), and 4 min after bleaching (4 min) are shown. (B) The bar graphs depict integrated signal intensities in mother and daughter cell bodies normalized to the prebleach image. Unbleached cells were used as controls. Mean values and standard deviations are shown (n = 5). (C) Cells were repeatedly bleached for 8.83 s (10 cycles consisting of 10 bleaching events, 883 ms each, and one imaging step) in the nucleoplasm or the nucleolus, respectively, with areas of identical size (orange circles). Scale bars, 2 µm. (D) The line graphs depict integrated signal intensities of Hmo1-GFP in nonbleached nucleolar areas of identical size from (C) (orange squares). Signals were normalized to the prebleach image. Unbleached cells were used as controls. Mean values and standard deviations are shown (n = 8). p(bleaching nucleoplasm vs. bleaching nucleolus, t = 224 s) = 6.05 × 10^–5.

To further characterize the dynamics of Hmo1, we compared its intranuclear and intranucleolar mobilities. We therefore used areas of identical size to repeatedly bleach the nucleoplasm or a portion of the nucleolus, respectively. A prebleach imaging step, followed by 10 cycles consisting of 10 bleaching events and one imaging event each, was executed. A nonbleached nucleolar region of identical size was quantified in both experiments to enable direct comparison of the data.

Repeated bleaching of the nucleoplasm resulted in a moderate decrease of the Hmo1-GFP signal in the nucleolus (Figure 11, C and D). The nucleolar signal intensity relative to prebleach dropped to 0.47 ± 0.15 after 224 s, that is, after 100 bleaching events. In contrast, the nucleolar signal intensity dropped much more rapidly when we bleached a region of the nucleolus instead of the nucleoplasm (Figure 11, C and D). A value of 0.48 ± 0.19 relative to prebleach was reached already after 50 s, that is, after 30 bleaching events. Unbleached control cells showed a relative nucleolar signal intensity of 0.99 ± 0.09 by the end of the experiment (Figure 11D). Note that Hmo1-GFP is also mobile within an elongated anaphase nucleolus (Supplemental Figure S10). The data suggest that Hmo1 is highly mobile within the nucleolus, but passage from the nucleolus to the nucleoplasm is limited.

NoLS activity promotes nucleolar retention

To delineate the contribution of Hmo1 domains to nucleolar retention, we performed bleaching experiments with Hmo1-intGFP, ΔboxB, ΔC10, and ΔboxBΔC10. We continuously bleached nucleoplasmic regions of identical size and measured the decrease of total nuclear signal over time.

In the case of Hmo1-intGFP, the decrease of total nuclear signal was modest (Figure 12, A and B). The signal decreased to 0.33 ± 0.15 relative to the prebleach value by the end of the experiment, that is, after 100 bleaching events. ΔboxB was only slightly more mobile as the signal decreased to 0.22 ± 0.15 (Figure 12, A and B). Interestingly, deletion of the 10 C-terminal amino acids resulted in a marked increase in mobility (Figure 12, A and B). The relative nuclear signal of ΔC10 decreased to 0.08 ± 0.08 by the end of the time series. Deletion of the 10 C-terminal amino acids also increased mobility in context of the endogenous endΔC10 construct (Supplemental Figure S11). The double deletion construct ΔboxBΔC10 showed the highest mobility of all constructs (Figure 12, A and B). The signal was reduced to 0.02 ± 0.04 compared with the prebleach value at the end of the experiment. Note that only constructs lacking the 10 C-terminal amino acids showed significant differences compared with Hmo1-intGFP at the last point in time of the experiment. Deletion of the HMG boxB did not significantly alter mobility. This suggests that predominantly interactions mediated by the NLS-NoLS promote nucleolar retention of Hmo1.

FIGURE 12:

NoLS activity contributes to nucleolar retention. (A) Cells overexpressing Hmo1-intGFP, ΔboxB, ΔC10, or ΔboxBΔC10 were repeatedly bleached for 8.83 s (10 cycles consisting of 10 bleaching events, 883 ms each, and one imaging step) in the nucleoplasm with areas of identical size (orange circles). Scale bar, 2 µm. (B) Total nuclear signals were quantified and normalized to the prebleach image. Unbleached cells were used as controls. Mean values and standard deviations are shown (n = 10). p(Hmo1-intGFP bleach vs. ΔboxB bleach, t = 224 s) = 0.804, p(Hmo1-intGFP bleach vs. ΔC10 bleach, t = 224 s) = 6.57 × 10^–4, p(Hmo1-intGFP bleach vs. ΔboxBΔC10 bleach, t = 224 s) = 8.58 × 10^–5.

DISCUSSION

In this study, we provide detailed analyses for the nucleolar localization of the budding yeast rDNA-binding HMGB protein Hmo1. We show that two different targeting mechanisms cooperate to establish the authentic nucleolar localization. Based on our data, we propose a model for the nucleolar localization of Hmo1 (Figure 13). The model suggests that the HMG boxB mainly recruits Hmo1 to the nucleolus by direct rDNA binding. In addition, a C-terminal NLS-NoLS (amino acids 212 to 246) mediates recruitment to rDNA mainly indirectly through binding of rRNA. Hmo1 association with rDNA is dynamic and NoLS activity of the C-terminal region helps to retain Hmo1 in the nucleolus. Altogether, our data suggest that the authentic nucleolar localization of Hmo1 results from an equilibrium between rDNA- and rRNA-based compartments of the nucleolus rather than simply from binding of rDNA.

FIGURE 13:

Model for the nucleolar localization of Hmo1. The HMG boxB is mainly responsible for direct association with the rDNA array and therefore targets Hmo1 to the polymer-polymer phase-separated chromatin compartment. rDNA association is dynamic in exponentially growing cells and interaction with the rDNA surrounding liquid-liquid phase-separated biogenesis compartment suppresses high exchange of Hmo1 between nucleolus and nucleoplasm. Interaction with the biogenesis compartment is mainly mediated through rRNA-binding by the C-terminal region of Hmo1 (amino acids 212-246), which constitutes a combined NLS-NoLS.

Although different Hmo1 domains directly bind rDNA in vitro, experiments on mitotic segregation and inhibition of transcription suggest that only the HMG boxB establishes profound direct interactions with rDNA in vivo. In a previous study, we investigated the mitotic segregation of a subset of nucleolar proteins and found that rDNA-binding proteins segregate equally, whereas biogenesis components segregate asymmetrically (Girke and Seufert, 2019). We argued that stable rDNA binding confers equal segregation in the course of asymmetric anaphase in budding yeast when one of the sister chromatids is asymmetrically pulled from the nucleolus of the mother cell. Hmo1, like other rDNA-binding proteins and rDNA itself, segregates equally in mitosis and shows a higher degree of compaction and concentration in daughter nucleoli after complete segregation (Figure 2; Supplemental Figure S4; Neurohr et al., 2011; Girke and Seufert, 2019). Hmo1 reflects the behavior of rDNA during mitotic segregation but only when the HMG boxB is present (Figure 2). In addition, the rDNA-binding protein Fob1 was reported to stay associated with rDNA when transcription is shut down, and colocalization of Hmo1 with Fob1 under such conditions is highly dependent on the HMG boxB (Mostofa et al., 2018; Figures 9, 10). HMG boxes have been reported to bind special DNA structures with higher affinity than B-form DNA and Hmo1 was shown to preserve negative supercoils at gene boundaries (Bustin, 1999; Achar et al., 2020). The rDNA locus might harbor a high density of special DNA topologies, which might be high affinity binding sites specifically for the HMG boxB.

The HMG boxB tightly links the nucleolar Hmo1 population to rDNA in vivo according to our segregation analyses, but the association is nonetheless dynamic as indicated by our FLIP experiments (Figure 11). It is indeed unlikely that Hmo1 is statically bound to rDNA in exponentially growing cells. Hmo1 mainly associates with the pol I transcribed region of rDNA, and pol I activity was shown to enable Hmo1 binding to rDNA by removing nucleosomes, which are assembled during each round of replication (Wittner et al., 2011). In more detail, the lobe-binding subunits confer an intrinsic capacity on pol I to read through nucleosomes (Merkl et al., 2020). The high processivity and density of pol I might cause transient displacement of Hmo1 from rDNA in the transcribed region.

Albeit rDNA binding of Hmo1 is dynamic in growing cells, high exchange with the nucleoplasm is suppressed mainly by interactions of a lysine-rich NLS-NoLS with rRNA in the rDNA surrounding biogenesis compartment. The C-terminal lysine-rich region was shown to be necessary for nuclear transport in an earlier study, and we show here that it is a joint NLS-NoLS that recruits Hmo1 to the nucleolus (Albert et al., 2013; Figures 1, 3, and 4). Deletion of the 10 C-terminal amino acids results in the loss of NoLS activity, reduced nucleolar enrichment in context of the full-length protein, reduced nucleic acid binding, and enhanced exchange with the nucleoplasm (Figures 3–8 and 12; Supplemental Figure S11). Nonamers of arginine or lysine are sufficient for nucleolar targeting in HeLa cells, suggesting that unspecific electrostatic interactions with both rDNA and rRNA could drive nucleolar accumulation (Musinova et al., 2011). Interestingly, the Hmo1-NLS contains one arginine and nine lysine, albeit only eight lysine in a row. The Hmo1-NLS might be right on the edge to nucleolar localization but of course, nucleolar targeting by NoLSs in metazoans and yeast might differ. Although nucleolar targeting by NoLSs rich in positive charge might be due to unspecific interactions with nucleic acids in general, such sequences accumulate specifically in granular components of human nucleoli, and electrostatic interactions with rRNA were suggested as drivers for nucleolar accumulation (Scott et al., 2010; Musinova et al., 2011). Although the Hmo1 NLS-NoLS binds both rDNA and rRNA in vitro, its association with rDNA in vivo is highly dependent on transcriptional activity (Figures 7, 9, and 10). This is in line with a view that the Hmo1 NLS-NoLS confers nucleolar recruitment mainly by binding of rRNA. This might, however, not be due to high specificity for certain RNA structures in vivo but simply due to the high concentration of rRNA relative to rDNA in nucleoli of living cells.

The HMG boxB and the NLS-NoLS are two critical domains that establish contacts with other nucleolar components and thereby promote nucleolar recruitment (Figures 1, 4, 5). In addition to the targeting domains, self-interaction mediated by the HMG boxA is a prerequisite for highly efficient nucleolar accumulation (Albert et al., 2013). It was shown that deletion of the boxA leads to delocalization throughout the nucleoplasm with only slight enrichment of the derivative in the nucleolus (Albert et al., 2013). This is reminiscent of the localization of the NLS-NoLS alone shown here (Figure 3C). Moreover, the boxA fused to GFP localizes throughout the cytoplasm in the absence of endogenous Hmo1 but is efficiently recruited to the nucleolus in the presence of endogenous Hmo1 (Albert et al., 2013). In our study, we observed high colocalization of the ΔboxBΔC10 construct with nucleoplasmic Swi5-NLS and very low colocalization with the nucleolar comarkers (Figure 4). Overall, the data suggest that the main function of the boxA for nucleolar enrichment is self-interaction rather than interaction with other nucleolar components. However, the precise mode of self-interaction is still controversial. It has been proposed that the boxA might function as dimerization or tetramerization module, but it has also been suggested that Hmo1 forms oligomers and that multiple domains of Hmo1 contribute to self-interaction (Bauerle et al., 2006; Xiao et al., 2010; Albert et al., 2013; Kasahara et al., 2016).

It might indeed be possible that Hmo1 forms a highly coherent liquidlike phase surrounding the rDNA array by the binding of rDNA, rRNA, and self-interaction. In contrast to rRNA biogenesis factors, however, this liquidlike Hmo1-containing phase would be tightly yet dynamically cross-linked to rDNA via the HMG boxB and therefore shifts toward the chromatin compartment. Although we overexpressed the full-length control construct carrying an internal GFP, it still efficiently accumulated in the nucleolus and almost equally segregated in mitosis (Figures 1 and 2). This indicates that binding sites for Hmo1 in the nucleolus cannot be easily saturated. This would at least be consistent with a liquidlike phase, as the relative amount of interaction partners inside liquid droplets can vary over a broader range (Feric et al., 2016). Furthermore, specifically the ΔboxB construct showed a massive spherical nucleolar condensate in addition to the Nop56-like pool in 50% of cells in a cycling culture (Supplemental Figures S3 and S5). This spherical condensate and the asymmetric mitotic segregation of ΔboxB might reflect reduced association of a liquidlike Hmo1-rRNA phase with the rDNA array when the HMG boxB is missing. A highly coherent, yet dynamic liquidlike phase would also agree with our bleaching experiments (Figure 11). Such a view could at least explain equal mitotic segregation, efficient nucleolar retention, and intranucleolar mobility at the same time. The mitochondrial HMGB protein TFAM, which is structurally highly similar to Hmo1, has lately been shown to phase-separate in vitro (Feric et al., 2021).

Moderate overexpression of full-length Hmo1 from the pTEF2 promoter used here well recapitulates the properties of endogenous Hmo1 regarding colocalization with other nucleolar proteins, mitotic segregation, behavior after rapamycin treatment, and nucleolar retention and fully supports cellular growth at least at 25 and 30°C (Figures 1, 2, 10–12; Supplemental Figures S2, S4, and S8). However, strong overexpression from the inducible pGAL1 promoter is toxic (Xiao et al., 2011). It would be interesting to investigate the localization at higher expression levels. Would Hmo1 be even more concentrated in the nucleolus or might nucleolar binding sites eventually get saturated and a highly concentrated nucleoplasmic pool of Hmo1 would emerge? It might be the case that strong overexpression of Hmo1 interferes with nucleolar processes but also processes at loci outside the nucleolus might be affected. For example, expression of RPGs or remodeling of nucleosomes throughout the entire genome might be aberrant (Hall et al., 2006; Malinina et al., 2022). As Hmo1 cross-links and compacts rDNA (Murugesapillai et al., 2014; Hult et al., 2017), extensive expression might theoretically also hamper mitotic segregation of rDNA or chromosomes in general. Noteworthy, Hmo1 negatively regulates its own expression (Xiao et al., 2011). This might help to prevent toxic effects by down-regulating expression when specific binding sites get saturated.

The classical view of nucleolar organization is that distinct nucleolar proteins exclusively localize to distinct compartments. However, recent work suggests that not only Hmo1 but also other rDNA-binding proteins might be targeted to different compartments of the nucleolus. The rDNA-associated budding yeast protein Net1 shows physical characteristics of both the chromatin compartment and the biogenesis compartment (Lawrimore et al., 2021). The full-length protein inhomogeneously localizes to the center of nucleoli, coincides well with rDNA markers in localization studies, and fits less well with Nop1 or Nop56 (Straight et al., 1999; Hannig et al., 2019; Lawrimore et al., 2021). In contrast, the nucleolar fraction of a C-terminal Net1 fragment coincides very well with Nop56 (Hannig et al., 2019). Net1 might be targeted to the chromatin as well as the biogenesis compartment via different protein domains. Moreover, different derivatives of the Hmo1 orthologue UBF expressed in U2OS cells localize to different compartments of nucleoli (Ueshima et al., 2017). The full-length protein colocalizes with pol I in fibrillar centers but an N-terminal fragment comprising the dimerization domain and the first two HMG boxes accumulates in granular components. Authentic nucleolar localization of certain rDNA-binding proteins might be the consequence of a dynamic cooperation of domains establishing contacts with different compartments rather than rDNA binding only. A possible functional relevance, for example, physical linkage of nucleolar compartments or coordination of biochemical processes of different compartments, will have to be tested in the future.

MATERIALS AND METHODS

Request a protocol through Bio-protocol.

Yeast methods and strains

Standard protocols were used for growth, transformation, sporulation, and mating of budding yeast cells (Ausubel et al., 2005). Strains used in this study are listed in Supplemental Table S1 and are isogenic derivatives of W303. Deletion of HMO1 in the W303 background was done by transformation of a PCR-based deletion construct. The deletion construct was amplified from genomic DNA of an hmo1 deletion strain purchased from EUROSCARF (Frankfurt). The standard growth medium was YPD+ (10 g/l yeast extract, 20 g/l peptone, 2% glucose, 0.2 g/l tryptophan, 0.1 g/l adenine, 10 mM KH₂PO₄).

DNA constructs and genetic manipulation

For endogenous tagging with fluorescent protein genes, either a PCR-based approach (Sheff and Thorn, 2004) or a plasmid based approach was used. For the PCR-based tagging of HMO1 and FOB1 with GFP or tdTomato, respectively, PCR products were transformed. For the plasmid-based endogenous tagging of HMO1 and NOP56, C-terminal fragments were PCR-amplified, sequenced, and finally cloned into pRS-based (Sikorski and Hieter, 1989) GFP or 3mCherry plasmids, respectively. Plasmids were linearized within the C-terminal fragment using a single cutter enzyme for integration at the respective endogenous locus. The Swi5-NLS construct (pTEF2-mCherry-SWI5(codons 569–709)S646A/S664A-MYC13-tCYC1) has been described (Arnold et al., 2015). It consists of the constitutive TEF2 promoter, mCherry, a SWI5-derived NLS fragment, and a myc13 epitope. Serine residues S646 and S664 within the Swi5-NLS were mutated to alanine to confer permanent nuclear localization (Moll et al., 1991; Arnold et al., 2015). The Swi5-NLS construct was integrated at the trp1 marker locus. HMO1 derivatives were PCR-amplified, sequenced, and cloned into pRS-based GFP plasmids. Constructs were integrated at the ura3 marker locus of a heterozygous diploid HMO1/hmo1 NOP56/NOP56-3mCherry strain and expressed from the constitutive TEF2 promoter. Similar expression levels of different GFP fused Hmo1 derivatives were ensured by Western blot (Supplemental Figure S1). Haploid strains bearing a specific GFP fusion, NOP56-3mCherry, and the deletion of endogenous HMO1 were then generated by sporulation. Resulting from sporulation of the same diploid strains, haploid progeny carrying a specific GFP construct and the deletion of endogenous HMO1, but not NOP56-3mCherry, were mated with haploid strains carrying either FOB1-tdTomato or pTEF2-mCherry-SWI5(codons 569–709)S646A/S664A-MYC13-tCYC1, respectively. The resulting diploid strains were then sporulated to generate haploid strains bearing the respective GFP fusion, FOB1-tdTomato or pTEF2-mCherry-SWI5(codons 569–709)S646A/S664A-MYC13-tCYC1, respectively, and the deletion of endogenous HMO1. This way we ensured similar expression of different Hmo1 derivatives and similar expression of each specific Hmo1 derivative when investigating the localization together with different comarkers. Cartoons for the different derivatives were created with the program Corel Draw (GFP is not drawn to scale). For expression in E. coli, HMO1 constructs were PCR-amplified, sequenced, and cloned into derivatives of the expression vector pJOE4056.2 containing an L-rhamnose-inducible rhaP_BAD promoter and an N-terminal 6His fusion (Wegerer et al., 2008). Identical HMO1 codons were used as in the case of live-cell imaging (Supplemental Table S1).

Protein expression and purification

Expression vectors were transformed in the E. coli expression strain BL21 Codon Plus (Stratagene). Cells were inoculated in LB medium (10 g/l tryptone, 5 g/l yeast extract, 10 g/l NaCl, 300 µM NaOH) containing 100 µg/ml ampicillin and 50 µg/ml chloramphenicol and induced with 0.2% rhamnose for 24 h. Harvested cells were resuspended in wash buffer (20 mM NaPi 500 mM NaCl 30 mM imidazole pH7) and cell disruption was done by three cycles in a French press. Lysates were centrifuged, filtrated, and applied to an ÄKTA Start chromatography system (Cytiva). The proteins of interest bound to HisTrap FF Crude columns (Cytiva), columns were washed with wash buffer; proteins were eluted in 20 mM NaPi 500 mM NaCl, pH 7, with an imidazole gradient (20 1-ml fractions, imidazole concentration ranging from 30 to 359 mM), concentrated with Biomax—10 K centrifugal filter devices (Millipore) if necessary; and dialyzed against buffer without imidazole (20 mM NaPi 500 mM NaCl, pH 7). Proteins were stored at 4°C. The concentrations were determined with a bovine serum albumin calibration curve from polyacrylamide gels using a quantitative Coomassie-staining protocol (Luo et al., 2006). All purified proteins were adjusted to 6His-GFP allowing a mean deviation of no greater than 10% for three technical replicates. Image processing and quantification of Coomassie-stained gels was done using the software ImageJ (National Institutes of Health, Bethesda, MD; Schneider et al., 2012). For quantification, the background was measured from a region of the gel without signal and subtracted from all pixels. The raw integrated density was then determined using threshold setting.

EMSAs

Binding reactions contained a final concentration of 10 mM MgCl₂, 8 mM NaPi, 200 mM NaCl, 5% glycerol, 100 µM EDTA, 1 mM Tris-HCl, pH 6.7, 0.025% bromphenol blue, 0.025% xylene cyanol ff, 20 nM of the indicated nucleic acid, and the indicated concentration of purified protein in a total volume of 10 µl. After incubation for 10 min at 22°C, reactions were applied to 1.5% agarose gels and electrophoresis was run for 1 h at 80 V and 22°C in TAE buffer (40 mM Tris, 1.3 mM EDTA, pH 8). Gels were stained with 1 µg/ml EtBr in TAE buffer under shaking for 1 h in the case of rDNA and 45 min in the case of rRNA, respectively. Gel images were recorded on an UVP Gelstudio Plus imager (Analytic Jena) using identical settings. Images were processed and quantified in ImageJ. The mean background values of the gels were determined via regions of interest not including signal and subtracted from each pixel in the image. Signals were quantified for each lane (raw integrated density) using a rectangular region of interest fitted to the band of the negative control. Identical intensity and size threshold settings were used in all cases. Normalization to the signal of the negative control then yielded the fraction of free nucleic acid for each lane. The resulting bound fractions were then used to calculate the EC50 values by curve fitting using the Hill equation and the Excel Solver Add-In (Microsoft). The mean EC50 values for three technical replicates with standard deviations are shown. For comparison with the control constructs, the relative affinities for rDNA and rRNA were calculated for each Hmo1 derivative. To this end, the mean EC50 value for the full-length control was divided by the EC50 value for each technical replicate of the respective derivative. Mean values for the relative affinities and standard deviations are shown. EMSAs for GFP, full-length Hmo1, Hmo1-NLS-C10, Hmo1-NLS, and ΔboxB are reproductions. Although originally different fusions were used in some cases (full-length Hmo1 without GFP fusion, Hmo1-NLS-C10 and Hmo1-NLS with GST instead of 6His-GFP fusion), the results were reproducible.

Microscopy of live yeast cells

Cultures were grown in YPD+ containing 2% glucose overnight at 25°C. Cells in exponential growth phase were harvested the next day, resuspended in a small volume of supernatant of the medium they were cultivated in, pipetted on cover slides, and covered with an agarose block containing synthetic complete medium (2% glucose, yeast nitrogen base, amino acids, nucleobases) for analysis with an inverted microscope. For the rapamycin experiments, exponential overnight cultures grown in YPD+ at 25°C were split, rapamycin (dissolved in DMSO) was added to a final concentration of 200 ng/ml (respective volume of DMSO was added to control culture), and cells were incubated for 4 h at 25°C, harvested, and covered with an agarose block containing synthetic complete medium and 200 ng/ml rapamycin (or respective volume of DMSO in the case of control experiments). For experiments in the stationary phase, cells were grown for 5 d in YPD+ at 30°C, harvested, and covered with an agarose block containing yeast nitrogen base, amino acids, and nucleobases but devoid of glucose. Spinning disk microscopy was done with an Axio Observer.Z1 (Carl Zeiss) in combination with a CSU-X1 spinning disk unit (Yokogawa), an AxioCam MRm (Carl Zeiss) for detection, and ZEN software (Carl Zeiss). Fourteen or 16 z-slices with a slice-to-slice distance of 500 nm were acquired. FLIP experiments were done using a TCS-SP8 confocal laser scanning microscope (Leica) with photomultiplier tubes (Leica) and Leica Application Suite software (Leica). Eighteen z-slices with a slice-to-slice distance of 460 nm were acquired (except in the case of Supplemental Figure S10: single z-plane). Plan Apochromat 63×/1.40 oil objectives were used for all microscopy. mCherry and tdTomato were excited using a 561-nm laser line. GFP was excited using a 488-nm laser line. All experiments were performed at room temperature (21–23°C). We originally investigated the localization of all Hmo1 derivatives in the S288c background, where they showed the same localization behavior as in the W303 background presented here.

Image processing and quantification was done using the software ImageJ. Colocalization analysis was carried out from unprojected z-stacks with the plugin JACoP v2.1.4 (Bolte and Cordelières, 2006). The mean background value calculated from regions outside cells was subtracted, and low thresholding was generally used to include the entire nuclear population of proteins. In the case of stationary cells, higher thresholds had to be used to exclude background fluorescence from the colocalization analyses. Line-plots were done from maximum projections after the background was calculated from regions outside cells and subtracted from the entire image. For quantification of the mitotic segregation of Hmo1 derivatives, z-stacks were summed up, the mean background value calculated from regions outside cells was subtracted, and the raw integrated density of the nuclear population was measured. To quantify the mitotic segregation of fluorescence signals, total signals in mother- and daughter-located parts of anaphase nuclei were averaged for three time points after complete segregation as defined by threshold setting. The daughter signal, as a percentage of the total signal, was then determined by normalization of the daughter portion to the added mother and daughter signal. Maximum values were quantified from unprojected z-stacks using the 3D object counter plugin for ImageJ (Bolte and Cordelières, 2006; http://imagejdocu.tudor.lu/doku.php?id = plugin:analysis:3d_object_counter:start; Schneider et al., 2012). The mean background value was calculated from regions outside cells for single z-planes and subtracted. For quantification of FLIP experiments, z-stacks were summed up and the mean background value calculated from regions outside cells was subtracted from all pixels. Threshold settings were used for all quantifications except the quantification from ROIs in FLIP experiments (Figure 11, C and D). All micrographs shown are maximum projections of z-stacks, except Supplemental Figure S10, which shows a single z-plane. In color merges, GFP fusions are depicted in cyan and tdTomato as well as mCherry fusions are depicted in red.

Protein analysis

Growing yeast cultures were harvested and cells were resuspended in 1 ml ice-cold H₂O. Lysis buffer (150 μl: 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, 50 mM NaF, 5 mM EDTA, 0.1% IGEPAL CA-630) was added, and cells were lysed by shaking with an equal volume of glass beads in a mixer mill (Retsch, Haan, Germany) at 4°C. Cell debris was pelleted by centrifugation for 3 min at 13,200 rpm and 4°C. Equal volumes of protein lysate and 2× Laemmli sample buffer were mixed and incubated for 10 min at 100°C. SDS–PAGE and Western blot analysis were performed as described in Schwab et al. (1997, 2001). Mouse monoclonal anti-GFP antibody (11814460001, Roche, Basel, Switzerland) was used for comparison of GFP-fused derivatives, and a polyclonal anti-Hmo1 serum from rabbit (Lu et al., 1996) was used for comparison of endogenous Hmo1-GFP and endogenous untagged Hmo1. Goat anti-mouse or anti-rabbit IRDye800CW secondary antibodies (LI-COR Biosciences) were used for detection with an Odyssey Infrared Imaging System (LI-COR Biosciences, Bad Homburg, Germany). Image processing and quantification of Western blots were done using the software ImageJ. For quantification, the background was measured from a region of the membrane without signal and subtracted from all pixels. The raw integrated density was then determined using threshold setting.

Graphs and statistical analyses

Box plots and statistical analyses were done with the software Origin (OriginLab Corporation). Box plots are defined as follows: the central box spans the first and third quartile (Q1 and Q3, respectively); the line inside the box represents the median, which equals the second quartile (Q2); and a square represents the mean. Whiskers extending from the first and third quartile have a maximum length of 1.5 × IQR. Values <Q1 – 1.5 × IQR or >Q3 + 1.5 × IQR are considered outliers and were excluded from statistical tests (no values were removed in the case of EMSAs). For statistical analyses, a two-sample, two-sided, unpaired Student’s t test and the unequal-variance t procedure (Welch correction) was used. Differences are regarded significant with a p value < 0.05. In the case of all tested datasets, normal distribution was not rejected according to a Kolmogorov–Smirnov test with a significance level set to 0.05. Bonferroni correction was applied to the statistical analyses in Figure 12B because of multiple testing (p values were multiplied by three). Graphs other than box plots were done in Excel (Microsoft).

Abbreviations used:

EC50

half maximal effective concentration

EMSA

electrophoretic mobility shift assay

FLIP

fluorescence loss in photobleaching

GFP

green fluorescent protein

HMGB

High Mobility Group box

IQR

interquartile range

NLS

nuclear localization sequence

NoLS

nucleolar localization sequence

pol I

RNA polymerase I

rDNA

ribosomal DNA

RPG

ribosomal protein gene

rRNA

ribosomal RNA.