RNA-Based 5-Fluorouracil Toxicity Requires the Pseudouridylation Activity of Cbf5p
Jason Hoskins, J. Scott Butler

Abstract

The chemotherapeutic drug 5-fluorouracil (5FU) disrupts DNA synthesis by inhibiting the enzymatic conversion of dUMP to dTMP. However, mounting evidence indicates that 5FU has important effects on RNA metabolism that contribute significantly to the toxicity of the drug. Strains with mutations in nuclear RNA-processing exosome components, including Rrp6p, exhibit strong 5FU hypersensitivity. Studies also suggest that 5FU-containing RNA can inhibit pseudouridylation, the most abundant post-transcriptional modification of noncoding RNA. We examined the effect of modulating the expression and activity of the essential yeast rRNA pseudouridylase Cbf5p on the 5FU hypersensitivity of an rrp6-Δ mutant strain. Depletion of Cbf5p suppressed the 5FU hypersensitivity of an rrp6-Δ strain, while high-copy expression enhanced sensitivity to the drug. A mutation in the catalytic site of Cbf5p also suppressed the 5FU hypersensitivity in the rrp6-Δ mutant, suggesting that RNA-based 5FU toxicity requires the pseudouridylation activity of Cbf5p. High-copy expression of box H/ACA snoRNAs also suppressed the 5FU hypersensitivity of an rrp6-Δ strain, suggesting that sequestration of Cbf5p to a particular guide RNA reduces Cbf5p-dependent 5FU toxicity. On the basis of these results and previous reports that certain pseudouridylases form stable adducts with 5FU-containing RNA, we suggest that Cbf5p binds tightly to substrates containing 5FU, causing their degradation by the TRAMP/exosome-mediated RNA surveillance pathway.

5-FLUOROURACIL (5FU) is a commonly used and potent chemotherapeutic drug developed as an inhibitor of the enzyme thymidylate synthetase (TS), which converts dUMP to dTMP, thus providing the sole de novo source of thymidylate (Parker and Cheng 1990; Longley et al. 2003). Inhibition of TS leads to “thymineless death,” a depletion of dTTP pools accompanied by an accumulation of dUTP, resulting in the mis-incorporation of deoxyuridine into newly synthesized DNA (Ladner 2001). Uracil DNA glycosylase and the base excision repair pathway repair the resulting abasic lesion (Boiteux and Guillet 2004). However, due to the continued imbalance in the ratio of dUTP to dTTP, repair of the lesion will likely lead to further deoxyuridine mis-incorporation. This futile cycling causes irreversible DNA damage, replication fork collapse, and cell cycle arrest (Ladner 2001; Dornfeld and Johnson 2005; Kouzminova and Kuzminov 2006; Seiple et al. 2006). These DNA-based effects of 5FU are fairly well understood, and the majority of co-therapeutics target the effect of 5FU on DNA (Longley et al. 2003). However, 5FU is also incorporated into RNA, which inhibits rRNA processing, post-transcriptional modification of tRNA, rRNA, and snRNA and mRNA splicing (Kaiser 1971; Parker and Cheng 1990; Ghoshal and Jacob 1994; Lenz et al. 1994; Yu et al. 1998; Longley et al. 2003; Fang et al. 2004; Zhao and Yu 2004, 2007).

Mounting evidence suggests that the RNA-based effects contribute significantly to the efficacy of 5FU. Several experiments have shown that uridine, not thymidine, relieves the cytotoxic and apoptotic effects of 5FU (Engelbrecht et al. 1984; Linke et al. 1996; Pritchard et al. 1997; Bunz et al. 1999; Hoskins and Butler 2007). Furthermore, a good correlation exists between the incorporation of 5FU into RNA and the cytotoxicity of the drug, and the co-therapeutic methotrexate enhances incorporation of 5FU into RNA (Greenhalgh and Parish 1990; Parker and Cheng 1990; Longley et al. 2003). Despite the apparent significance of the RNA-based effects of 5FU, the mechanism of this RNA-based toxicity remains elusive.

Recent genomewide 5FU-induced haplo-insufficiency screens in Saccharomyces cerevisiae provided clues to the metabolic pathways involved in the toxic effects of 5FU (Giaever et al. 2004; Lum et al. 2004). Strikingly, strains heterozygous for components of the nuclear RNA exosome showed a significant degree of 5FU hypersensitivity, with the RRP6+/− strain exhibiting the greatest sensitivity (Giaever et al. 2004; Lum et al. 2004). The RNA exosome, an exoribonucleolytic complex, plays a critical role in RNA processing, quality control, and degradation in the nucleus and cytoplasm (Butler 2002; Houseley et al. 2006). The nuclear 3′–5′ exoribonuclease Rrp6p associates and functions with the exosome in the nucleus, but has several roles distinct from the rest of the exosome (Butler 2002; Houseley et al. 2006). Microarray and Northern analyses revealed the accumulation of polyadenylated noncoding RNA, especially rRNA, upon 5FU treatment of wild-type yeast, and deletion of RRP6 enhanced this effect (Fang et al. 2004). Polyadenylation of noncoding RNA reflects the activity of a quality control system in which the TRAMP complex recognizes aberrantly processed noncoding RNA and then polyadenylates and enhances the degradation of the aberrant RNA by Rrp6p and the nuclear exosome (Kadaba et al. 2004; LaCava et al. 2005; Wyers et al. 2005; Houseley and Tollervey 2006). Accordingly, one view suggests that 5FU may cause toxicity in RNA by disrupting processing of noncoding RNA (Fang et al. 2004).

Post-transcriptional modifications commonly occur during the maturation of noncoding RNA, and pseudouridylation is the most widespread modification (Charette and Gray 2000; Spedaliere and Mueller 2004; Reichow et al. 2007). Pseudouridylation is an isomerization of a uracil base in which the glycosidic bond to the N1 of uracil is broken and reformed to the C5 of uracil, effectively flipping the base over and providing an additional hydrogen bond donor at N1 (Gu et al. 1999; Hamilton et al. 2006). 5FU-containing RNA inhibits certain pseudouridylases due to tight binding between the substrate 5FU and the pseudouridylase after the isomerization reaction, although this inhibition appears idiosyncratic among pseudouridylases (Samuelsson 1991; Gu et al. 1999; Spedaliere and Mueller 2004; Hamilton et al. 2006). All pseudouridylation in rRNA, and some in snRNA, requires the activity of H/ACA snoRNP complexes composed of four proteins and a box H/ACA guide snoRNA complementary to at least one RNA site containing a substrate uracil (Reichow et al. 2007). In yeast, CBF5 (DKC1 in humans, NOP60B in Drosophila, and NAP57 in rats) encodes the essential pseudouridylase component of these H/ACA snoRNPs (Lafontaine et al. 1998; Zebarjadian et al. 1999; Reichow et al. 2007). In addition to pseudouridylation, Cbf5p also stabilizes box H/ACA snoRNPs and enhances 35S pre-rRNA transcription initiation by RNA polymerase I (Cadwell et al. 1997; Lafontaine et al. 1998).

Since 5FU-containing RNA inhibits some pseudouridylases and strongly affects rRNA processing in particular, we examined the role of Cbf5p in the RNA-based toxicity of 5FU. Our results show that partial Cbf5p depletion suppresses the 5FU hypersensitivity of an rrp6-Δ mutant. Conversely, high-copy expression of Cbf5p increases the 5FU sensitivity of both wild-type and rrp6-Δ strains. This suggests that Cbf5p is actually responsible for the RNA-based toxicity of 5FU. To determine which role of Cbf5p causes the RNA-based toxicity, we examined the 5FU sensitivity of CBF5 mutants defective in pseudouridylation or 35S pre-rRNA transcription initiation. We found that impaired pseudouridylation activity suppresses the 5FU hypersensitivity of an rrp6-Δ mutant, while a mutation (cbf5-1) causing a defect in 35S pre-rRNA transcription initiation results in 5FU hypersensitivity. Furthermore, high-copy expression of box H/ACA snoRNAs, but not a box C/D snoRNA, suppresses the 5FU hypersensitivity of an rrp6-Δ mutant, suggesting that sequestering Cbf5p to complexes containing the overexpressed guide snoRNA reduces the Cbf5p-dependent toxic effects of 5FU's RNA incorporation into RNA. We propose a model in which pseudouridylation of 5FU in substrate rRNA causes Cbf5p to bind tightly to rRNA during processing, which leads to polyadenylation of the rRNA by the TRAMP complex and degradation by Rrp6p and the nuclear exosome. However, defects in Rrp6p or the exosome lead to the accumulation of these polyadenylated rRNAs with toxic downstream effects.

MATERIALS AND METHODS

Yeast strains, plasmids, and reagents:

The yeast strains used are described in Table 1. Unless noted otherwise, yeast strains were grown in yeast extract–peptone–dextrose or synthetic complete dextrose (SCD) medium (Sherman 1991).

View this table:
TABLE 1

Yeast strains used in this study

Northern blot analysis

Total RNA was extracted from the indicated strains grown to an OD600 of 0.7–1.1 as previously described (Patel and Butler 1992), and Northern blot analysis was performed as described by Briggs et al. (1998).

Western blot analysis

Western analysis was performed on whole-cell extract from an rrp6-Δ strain with the pCBF5-BFG plasmid grown in SCD–LEU medium with 0, 20, or 200 μm 5FU for 4 hr to an OD600 of 0.3–0.4 as described previously (Burkard and Butler 2000). Cbf5p-HA was detected with a monoclonal anti-HA antibody (12CA5; 1:1500 dilution), and Pgk1p was detected with a monoclonal anti-Pgk1p antibody [22C5, Molecular Probes (Eugene, OR); 1:1000 dilution].

RESULTS

Modulation of Cbf5p levels directly affects the 5FU hypersensitivity of an rrp6-Δ strain:

5FU treatment of yeast strains causes the accumulation of a wide range of polyadenylated noncoding RNA species, an effect enhanced by recessive RRP6 mutations (Fang et al. 2004). In response to aberrant processing, improper protein associations, or a lack of post-transcriptional modification, the TRAMP complex polyadenylates noncoding RNA as a prelude to their degradation by Rrp6p and the nuclear exosome (Kadaba et al. 2004; LaCava et al. 2005; Wyers et al. 2005; Houseley and Tollervey 2006). Pseudouridylation, the most abundant modification in rRNA, tRNA, and snRNA requires excision and reattachment of uracil by pseudouridylases, and previous studies showed that 5FU strongly inhibits pseudouridylation by some pseudouridylases (Samuelsson 1991; Gu et al. 1999; Charette and Gray 2000; Spedaliere and Mueller 2004; Hamilton et al. 2006; Reichow et al. 2007). Cbf5p catalyzes all pseudouridylation in rRNA and therefore seemed a likely candidate for inhibition by 5FU.

Accordingly, we asked if depletion of Cbf5p affects the RNA-based 5FU hypersensitivity of an rrp6-Δ mutant (Hoskins and Butler 2007). We used strains containing CBF5 under the control of the tetO7 repressible promoter, which allows repression of CBF5 transcription by the addition of doxycycline to the media (Gari et al. 1997). We placed dilutions of wild-type, rrp6-Δ, tet-CBF5, and tet-CBF5 rrp6-Δ strains on media containing no drug, 5FU, doxycycline, or both 5FU and doxycycline. Strikingly, depletion of Cbf5p on the 5FU and doxycycline plate caused a strong suppression of the 5FU hypersensitivity in the tet-CBF5 rrp6-Δ mutant (Figure 1A). Weaker suppression of the 5FU hypersensitivity occurs in the tet-CBF5 rrp6-Δ strain compared to the rrp6-Δ mutant on the plate containing only 5FU, a result likely due to lower-than-normal expression of CBF5 from the tetO7 promoter (Figure 1A). If the RNA-based toxicity of 5FU results simply from the inhibition of pseudouridylation, we would expect that further inhibition of pseudouridylation by depleting Cbf5p should increase the 5FU sensitivity of the cell. However, the opposite effect occurs, indicating that some activity of Cbf5p causes the RNA-based toxicity of 5FU. This experiment also revealed improved growth of the tet-CBF5 and tet-CBF5 rrp6-Δ strains on the 5FU and doxycycline plate compared to the plate with only doxycycline (Figure 1A). Western and Northern analyses do not reveal any increase in steady-state Cbf5p or CBF5 transcript levels upon 5FU treatment (Figure 1, C and D), suggesting that this effect represents a reduced need for Cbf5p in cells treated with 5FU. We did not observe this effect upon increasing doxycycline concentrations two- to fourfold. Moreover, suppression of the 5FU hypersensitivity of the rrp6-Δ mutant does not occur upon more extensive depletion of Cbf5p (Figure 2, empty vector strains), suggesting that this suppression occurs only upon partial depletion of Cbf5p.

Figure 1.—

Modulating Cbf5p levels modulate 5FU toxicity. Cultures of YSB3001 (WT), YSB3013 (rrp6-Δ), YSB3007 (tet-CBF5), and YSB3020 (tet-CBF5 rrp6-Δ) (A) or YSB1002 (WT) and YSB1005 (rrp6-Δ) with pESC-LEU or pCBF5-BFG (B) were diluted to OD600 = 0.5, serially diluted 10-fold, and spotted on (A) SCD plates containing no drug, 200 μm 5FU, 5 μg/ml doxycycline, or both 200 μm 5FU and 5 μg/ml doxycycline. The no-drug plate was incubated at 30° for 3 days, and the other three plates were incubated for 4 days. (B) SCD–LEU plates containing no drug or 200 μm 5FU. The 0 and 200 μm 5FU plates were incubated at 30° for 3 and 4 days, respectively. (C) Northern analysis of total RNA from YSB1005 (rrp6-Δ) treated with 0 or 200 μm 5FU for 1 hr. CBF5 and ACT1 mRNAs were detected with complementary 3′-end radiolabeled oligonucleotides, oSB687 and oSB400, respectively. (D) Western analysis of whole-cell extract from YSB1002 (WT) and YSB1005 (rrp6-Δ) containing the pCBF5-BFG plasmid and treated with 0 or 200 μm 5FU for 4 hr. Cbf5-HA fusion protein from the pCBF5-BFG plasmid was detected with an anti-HA antibody, and Pgk1p was detected with an anti-Pgk1p antibody.

Figure 2.—

Pseudouridylation activity is required for Cbf5p-dependent 5FU toxicity. Cultures of YSB3001 (WT), YSB3027 (rrp6-Δ), YSB3007 (tet-CBF5), and YSB3028 (tet-CBF5 rrp6-Δ) containing pESC-LEU, pCBF5-BFG, or pcbf5D95A-BFG plasmids were diluted to OD600 = 0.5, serially diluted 10-fold, and spotted on SCD–LEU plates containing no drug, 200 μm 5FU, 20 μg/ml doxycycline, or both 200 μm 5FU and 20 μg/ml doxycycline. The no-drug plates were incubated at 30° for 3 days, and the others were incubated for 4 days.

If Cbf5p causes the toxicity of 5FU in RNA, then overexpression of the CBF5 gene would presumably increase the 5FU sensitivity of the cell. To test this, we transformed wild-type and rrp6-Δ strains with an empty vector and a high-copy plasmid containing a CBF5-triple HA-tag gene fusion under the constitutive PGK1 promoter (pCBF5-BFG; Zebarjadian et al. 1999). Expression of CBF5 from this plasmid increased the 5FU sensitivity of both the wild-type and rrp6-Δ strains, supporting the view that some activity of Cbf5p causes toxicity of 5FU (Figure 1B).

The pseudouridylation activity of Cbf5p causes the RNA-based toxicity of 5FU:

To investigate the involvement of the pseudouridylation activity of Cbf5p in the toxicity of 5FU, we used a catalytically inactive Cbf5p point mutant (Cbf5-D95A) (Zebarjadian et al. 1999). Wild-type, rrp6-Δ, tet-CBF5, and tet-CBF5 rrp6-Δ strains carrying empty vector, pCBF5-BFG, or the mutant-containing plasmid pcbf5D95A-BFG were diluted and plated on SCD–LEU media containing no drug, 5FU, doxycycline, or 5FU and doxycycline (Figure 2). The suppression of the 5FU hypersensitivity of the rrp6-Δ mutant by doxycycline-induced depletion of Cbf5p does not occur in this experiment because the fourfold increase in doxycycline concentration compared to previous experiments presumably depletes the essential Cbf5p beyond an acceptable limit for viability. This level of depletion of endogenous wild-type Cbf5p enhanced the phenotype of the catalytically inactive mutant Cbf5 protein in the presence of 5FU. While both CBF5 plasmids complement the depletion of Cbf5p in the tet-CBF5 strains, only the D95A mutant plasmid allows for suppression of 5FU hypersensitivity of the rrp6-Δ strains depleted of endogenous wild-type Cbf5p (Figure 2). Moreover, the Cbf5p-depleted rrp6-Δ strain containing the D95A mutant plasmid grows much better on the 5FU and doxycycline plates than the rrp6-Δ mutant with endogenous levels of wild-type Cbf5p. These results indicate that the toxicity of 5FU observed in rrp6-Δ mutants requires the pseudouridylase activity of Cbf5p.

High-copy expression of H/ACA snoRNA suppresses the 5FU hypersensitivity of the rrp6-Δ mutant:

In addition to pseudouridylation, Cbf5p enhances the stability of H/ACA snoRNA (LaFontaine et al. 1998). Accordingly, we examined whether the loss of this activity might account for the ability of cbf5 mutants to suppress 5FU sensitivity. We constructed high-copy plasmids containing SNR30, SNR37, or SNR47 under their endogenous promoters. SNR30 and SNR37 encode box H/ACA snoRNAs, while SNR47 encodes a box C/D snoRNA that does not interact with Cbf5p and functions as a control for any general effects of snoRNA overexpression. We transformed the snoRNA plasmids or empty vector into wild-type and rrp6-Δ strains and plated dilutions of the resulting strains on media with or without 5FU (Figure 3A). Northern blot analysis of total RNA extracted from wild-type and rrp6-Δ strains containing the empty vector, pRS-SNR30, pRS-SNR37, or pRS-SNR47 plasmids showed that each plasmid increases the level of its respective snoRNA about fourfold (Figure 3B). The high-copy expression of the H/ACA snoRNAs snR30 and snR37 clearly suppressed the 5FU hypersensitivity of rrp6-Δ mutants, while the empty vector or the snR47 plasmid did not (Figure 3A). Since overexpression of H/ACA snoRNAs suppresses 5FU hypersensitivity rather than enhances it, the stabilization of H/ACA snoRNA by Cbf5p likely does not play a role in the RNA-based toxicity of 5FU.

Figure 3.—

High-copy expression of H/ACA snoRNAs suppresses the 5FU hypersensitivity of an rrp6-Δ strain. (A) Cultures of YSB1002 (WT) and YSB1005 (rrp6-Δ) containing pRS423, pRS-SNR30, pRS-SNR37, or pRS-SNR47 were diluted to OD600 = 0.5, serially diluted 10-fold, and spotted on SCD–HIS plates containing 0 or 200 μm 5FU. Both plates were incubated at 30° for 3 days. (B) Total RNA from YSB1002 (WT) containing pRS423, pRS-SNR30, pRS-SNR37, or pRS-SNR47 was separated by denaturing 2% agarose gel electrophoresis and transferred to a GeneScreen membrane. Transcripts were visualized through hybridization of complementary radiolabeled oligonucleotide probes for detection of snR30, snR37, snR47, and ACT1 mRNA.

Impaired 35S pre-rRNA transcription initiation increases toxicity of 5FU:

Previous experiments indicated a role for Cbf5p in 35S rRNA transcription initiation (Cadwell et al. 1997). The cbf5-1 temperature-sensitive mutation causes a deficiency only in its ability to enhance initiation of 35S pre-rRNA transcription by RNA polymerase I. To assess the role of the 35S pre-rRNA transcription initiation activity of Cbf5p on 5FU toxicity, we tested the growth of the cbf5-1 mutant on 5FU-containing media by placing dilutions of the cbf5-1 strain and its isogenic wild type on media with or without 5FU (Figure 4A). (Because all strains exhibit much greater sensitivity to 5FU in the absence of uracil, less 5FU was added to SCD–URA plates). The cbf5-1 mutant displayed significant 5FU hypersensitivity, suggesting that impaired 35S pre-rRNA transcription causes increased sensitivity to 5FU. Next, we checked whether or not this hypersensitivity results from reduced 35S pre-rRNA transcription initiation by examining the effect of a high-copy plasmid expression of RRN3 in the cbf5-1 strain grown on 5FU-containing media (Figure 4A). RRN3 encodes a rate-limiting RNA polymerase I transcription initiation factor previously shown to cause high-copy suppression of cbf5-1 phenotypes (Cadwell et al. 1997). Indeed, the high-copy plasmid expressing RRN3 suppresses the 5FU hypersensitivity of the cbf5-1 mutant (Figure 4A), suggesting that defective 35S pre-rRNA transcription initiation results in increased 5FU toxicity.

Figure 4.—

5FU hypersensitivity of the cbf5-1 mutant is due to reduced 35S pre-rRNA transcription initiation efficiency. Cultures of YSB2030 (WT) and YSB2031 (cbf5-1) containing YEplac181 or YEpRRN3 (A) or containing pRS423 or pRS-SNR30 (B) were diluted to OD600 = 0.5, serially diluted 10-fold, and spotted on (A) SCD–URA plates containing 0 or 2 μm 5FU. Both plates were incubated at 30° for 3 days. (B) SCD–HIS plates containing 0 or 200 μm 5FU. Both plates were incubated at 30° for 3 days.

In light of this observation, we asked whether high-copy expression of H/ACA snoRNAs might reduce 5FU sensitivity by affecting 35S pre-rRNA transcription. We transformed the empty vector and pRS-SNR30 plasmids into the wild-type and cbf5-1 strains and examined their effect on the 5FU hypersensitivity by plating dilutions on media with or without 5FU (Figure 4B). The results show that SNR30 overexpression has no effect on the 5FU hypersensitivity of the cbf5-1 mutant, suggesting that H/ACA snoRNA does not suppress the 5FU hypersensitivity of the rrp6-Δ strain by affecting 35S pre-rRNA transcription.

DISCUSSION

Our findings show that depletion of Cbf5p, the pseudouridylase component of H/ACA snoRNPs, suppresses the 5FU hypersensitivity of an rrp6-Δ mutant. Moreover, expression of Cbf5p from a high-copy plasmid increases the sensitivity of both wild-type and rrp6-Δ strains to 5FU. These observations implicate Cbf5p in the toxic effects of 5FU incorporation into RNA. Further analyses aimed at discerning which of the known functions of Cbf5p causes the 5FU toxicity implicated the pseudouridylase activity, rather than the rRNA transcription or snoRNA stabilization activities. Specifically, a mutation in the catalytic site of Cbf5p suppressed the 5FU hypersensitivity of an rrp6-Δ strain. High-copy expression of box H/ACA snoRNAs also suppressed the 5FU hypersensitivity of an rrp6-Δ mutant, while overexpression of a box C/D snoRNA had no effect. Finally, the cbf5-1 mutant, defective only for the 35S pre-rRNA transcription initiation activity of Cbf5p, caused 5FU hypersensitivity, indicating that 5FU toxicity does not result from enhanced rRNA transcription. The enhanced 5FU toxicity caused by lowering the rRNA transcription rate probably reflects the facts that the drug already causes a defect in ribosome biogenesis and mutations that inhibit rRNA processing cause 5FU hypersensitivity (Fang et al. 2004). These findings lead us to conclude that pseudouridylation by Cbf5p plays a significant role in RNA-based 5FU toxicity in yeast.

Previous reports suggested that 5FU inhibits all pseudouridylases, but more recent results showed that 5FU-containing RNA inhibits TruA and RluA, but had no effect on the activity of TruB (Spedaliere and Mueller 2004). Biochemical analyses of the 5FU-inhibited TruA and RluA pseudouridylases showed that the enzymes isomerize 5FU, but form heat-stable adducts with the rearranged products (Gu et al. 1999; Hamilton et al. 2006). It remains unclear whether these intermediates contain covalent adducts between enzyme and substrate. Although TruA, RluA, and TruB represent different families of pseudouridylases, all of them utilize the same general active-site structure, catalytic aspartate residue, and reaction mechanism (Hamilton et al. 2006). The variability in 5FU inhibition may represent minor differences in active-site geometry that favor either release or adduct formation between the enzyme and the isomerized 5FU (Hamilton et al. 2006).

Cbf5p resides in the TruB pseudouridylase family, but while 5FU does not inhibit TruB, it strongly inhibits Pus4p, another member of the family found in yeast (Samuelsson 1991). Therefore, 5FU may inhibit Cbf5p by causing the formation of adducts with some 5FU-containing substrate RNA. We suggest that formation of such adducts between Cbf5p and 5FU-containing RNA substrates might lead to RNA-based 5FU toxicity. In this view, Cbf5p-RNA adducts would then become substrates for the RNA surveillance pathway in which TRAMP complexes polyadenylate the aberrantly processed RNA adducts, marking them for degradation by Rrp6p and the nuclear exosome. In the absence of Rrp6p, these polyadenylated Cbf5p-RNA adducts would accumulate as indicated by the fact that deletion of RRP6 further exacerbates the 5FU-dependent accumulation of polyadenylated rRNAs in yeast (Fang et al. 2004). Since Cbf5p catalyzes >40 different pseudouridylation events in multiple substrate RNAs, we suspect that the effects of 5FU on Cbf5p would have a significant impact on the RNA-based toxicity of 5FU.

This model provides an explanation for the suppression of the 5FU hypersensitivity in an rrp6-Δ mutant caused by overexpressing a box H/ACA snoRNA. Yeast contain 28 box H/ACA snoRNAs that guide pseudouridylation of 44 different uracils throughout 18S and 25S rRNA in yeast (Torchet et al. 2005). Since Cbf5p must form a snoRNP complex with each of these guide RNAs, increased expression of one would bind a significant portion of the available Cbf5p, leading to a shift in the equilibrium of snoRNPs and a preference for pseudouridylation of the RNA targeted by the overexpressed snoRNA. Recent experiments in yeast estimated the level of 5FU substitution in RNA at ∼3–4% for yeast treated with 150 μm 5FU, a concentration similar to that used in most of the experiments described here (Seiple et al. 2006). Assuming an even distribution of 5FU throughout all transcripts, there will typically be only one or two 5FU residues among the 44 positions targeted for pseudouridylation per 35S transcript. High-copy expression of one guide RNA could shift the equilibrium of snoRNPs to target specific uracils, allowing an increase in the proportion of 5FU-containing 35S pre-rRNA that eludes quality control, and continues with subsequent processing into mature rRNA. This dilution of snoRNP activity would be especially effective for high-copy expression of snR30 since it binds Cbf5p and plays a role in 18S pre-rRNA cleavage, but not pseudouridylation. Therefore, high-copy expression of a box H/ACA snoRNA may suppress the 5FU hypersensitivity of an rrp6-Δ mutant by preferentially redirecting Cbf5p to a limited number of pseudouridylation sites, thereby reducing the number of transcripts that form an adduct with Cbf5p due to pseudouridylation of a substituted 5FU.

The toxicity caused by 5FU may result, at least in part, from the polyadenylated rRNAs that accumulate in an rrp6-Δ strain treated with 5FU, which could inhibit the processing of other rRNA transcripts by sequestering components of the rRNA-processing machinery. This could cause the defects in rRNA processing and ribosome biogenesis observed upon 5FU treatment and would suggest a mechanism by which the accumulation of polyadenylated rRNAs causes toxicity (Parker and Cheng 1990; Ghoshal and Jacob 1994; Fang et al. 2004; Lum et al. 2004). Alternatively, sequestering of Cbf5p to complexes with rRNAs could cause 5FU toxicity by reducing the pool of Cbf5p available for its essential functions. In this view, degradation of the RNA in the Cbf5p-RNA adducts by Rrp6p would release bound Cbf5p. However, this explanation seems unlikely since depletion of Cbf5p actually suppresses the growth defect of the rrp6-Δ mutant on 5FU-containing media.

One unexpected observation was the improved growth of the tet-CBF5 and tet-CBF5 rrp6-Δ strains on 5FU and doxycycline plates vs. plates containing only doxycycline. This effect occurs only at lower doxycycline concentrations, suggesting that 5FU does not completely negate the need for Cbf5p in the cell. 5FU does not appear to affect CBF5 expression since both CBF5 mRNA and Cbf5p levels remain unaltered by 5FU treatment. Therefore, 5FU treatment must compensate for some deficiency in cells partially depleted of Cbf5p. Since the pcbf5D95A-BFG plasmid complements a Cbf5p-depleted strain without expressing a pseudouridylation-competent enzyme, it seems unlikely that 5FU improves the growth of a strain partially depleted of Cbf5p by affecting pseudouridylation. It seems equally unlikely that 5FU enhances 35S pre-rRNA transcription, since 5FU would then improve the growth of a cbf5-1 mutant rather than inhibit it. Depletion of Cbf5p also leads to a depletion of box H/ACA snoRNAs, and at least one box H/ACA snoRNA, snR30, plays an essential role in pre-rRNA cleavage. However, Northern analyses failed to show a significant increase in snR30 transcript levels in strains treated with 5FU (data not shown). This suggests an uncharacterized role for Cbf5p that is less important upon 5FU treatment or that 5FU treatment is able to mimic.

Although 5FU clearly affects rRNA processing and rRNA represents an abundant target of Cbf5p pseudouridylation, it remains possible that Cbf5p causes 5FU toxicity by pseudouridylating 5FU in other types of RNA. Recent experiments showed that 5FU incorporation into the branch-site recognition region (BSRR) of U2 snRNA blocks pseudouridylation and prevents packaging of U2 into splicing complexes (Zhao and Yu 2004, 2007). Moreover, 5FU-containing U2 fails to reconstitute splicing in Xenopus oocytes depleted of endogenous U2, and 5FU treatment causes inhibition of some pre-mRNA splicing in HeLa cells (Zhao and Yu 2007). Cbf5p catalyzes the formation of at least one of the pseudouracils in the BSRR of U2 snRNA, suggesting that the Cbf5p-dependent 5FU toxicity may result from the pseudouridylation of U2 snRNA (Ma et al. 2003; Zhao and Yu 2007). Further study is required to determine the degree to which each 5FU-containing Cbf5p target contributes to the overall RNA-based toxicity of 5FU.

In conclusion, we have shown that the pseudouridylation activity of Cbf5p causes 5FU toxicity in RNA. In light of our results and previous studies on the effect of 5FU on other pseudouridylases, we suggest that pseudouridylation of substituted 5FU in an RNA substrate might lead to the formation of Cbf5p-RNA adducts that undergo polyadenylation by the TRAMP complex and subsequent degradation by Rrp6p and the nuclear exosome. Although the downstream toxic effects of these proposed Cbf5p-RNA adducts remain unclear, it seems likely that accumulation of polyadenylated Cbf5p-RNA adducts causes toxicity, since rrp6-Δ mutants incapable of clearing these aberrant RNAs exhibit enhanced sensitivity to 5FU. Alternatively, Cbf5p might catalyze the formation of RNA products containing 5-fluoropseudouracil, and the metabolism of these molecules could play a significant role in 5FU toxicity in yeast. Further experimentation will clarify these issues.

Acknowledgments

We thank the members of our lab for fruitful discussions and critical reading of the manuscript. We are grateful to John Carbon for providing strains and plasmids.

Footnotes

  • Communicating editor: B. J. Andrews

  • Received October 2, 2007.
  • Accepted March 8, 2008.

References

View Abstract