HMA mutants exhibit resistance to chronic HU treatment

We previously found that an S. pombe mutant lacking Rtf1/Prf1 (prf1∆) was hypersensitive to the microtubule-destabilizing agent thiabendazole and the DNA-damaging agent methyl methanesulfonate (MMS) (Chen et al, 2020). However, the prf1∆ mutant exhibited an unexpected resistance to hydroxyurea (HU) (Figs 1A and S1). RNR inhibition by HU depletes cellular dNTP pools, leading to S-phase arrest and stalled replication forks. Collapse of stalled forks leads to DNA damage (Giannattasio & Branzei, 2017). HU can also cause oxidative stress independently of its effects on RNR (Singh & Xu, 2016). A small-scale, qualitative screen of mutants affecting transcription and chromatin (Fig S1) revealed HU resistance in mutants defective in H2B monoubiquitylation (H2Bub1) and H3K4 methylation (H3K4me). These included htb1-K119R (harboring a K119R substitution in H2B), brl1∆ and brl2∆ (harboring deletions of either of two E3 ubiquitin ligases that target H2B K119), set1∆ (harboring a deletion of the H3K4 methyltransferase), and hht2-K4R (a strain expressing a single copy of histone H3 with a K4R substitution) (Tanny et al, 2007; Xhemalce & Kouzarides, 2010). In fact, all of the strongly HU-resistant mutants we found were components of the HMA (Fig S1). These mutants grew similar to wild type or worse than wild type in the presence of the DNA-damaging agent phleomycin, in accord with our previous results and demonstrating that HU resistance does not reflect general resistance to DNA damage (Fig S2A). Therefore, HU resistance is a unique and specific feature of S. pombe mutants with defects in the HMA. The fact that histone point mutations shared the resistance observed in the strains lacking the modifying enzymes points to involvement of the modifications themselves in the phenotype.

Not all tested mutants affecting this regulatory axis displayed HU resistance. For example, a knockout of rhp6⁺, encoding a Rad6 ortholog that is the E2 ubiquitin–conjugating enzyme for H2Bub1, was strongly sensitive to HU (Fig 1A). This was expected based on the known role of Rad6/Rhp6 in post-replication DNA repair pathways that involve substrates other than H2B (Game & Chernikova, 2009). Deletion of tpr1⁺, encoding the largest PAF complex subunit (Ctr9 in humans), did not lead to HU resistance, despite the fact that H2Bub1 and H3K4me are strongly reduced (Fig 1A) (Mbogning et al, 2013). This may be due to pleiotropy caused by the loss of the entire PAF complex, which masks the resistance phenotype. However, this finding also aligns with the phenotypic divergence observed between mutants affecting PAF and Rtf1/Prf1, further supporting their functional independence (Mbogning et al, 2013).

We quantified the resistance by monitoring growth in the presence of HU over time in liquid culture (Fig 1B and C). A strain harboring a knockout of cds1⁺, which encodes the S. pombe ortholog of the S-phase checkpoint kinase Rad53, was included as an HU-sensitive control (Giannattasio & Branzei, 2017). All untreated strains grew to saturation by 24 h. When treated with HU, wild-type cells exhibited decreased cell growth throughout the experimental period, even at the lowest dose of HU tested (10 mM). Conversely, the growth curves for htb-K119R and prf1∆ strains in the presence of HU more closely resembled the untreated condition (Fig 1B). We quantified these differences by fitting the curves to a Gompertz growth curve model to derive the saturation values (Ymax). Ymax decreased with increasing HU concentrations in the wild-type strain, as indicated by the increasing ΔYmax value (defined as the difference in Ymax between the relevant HU dose and no treatment) (Fig 1C). These differences were significantly attenuated in htb-K119R and prf1∆ strains (Fig 1C). We observed more modest attenuation in brl1∆, set1∆, and hht2-K4R strains that was limited to lower doses of HU. As expected, cds1∆ was strongly sensitive to HU, whereas set2∆ was similar to wild type. These data implicate the upstream components of the HMA, particularly Rtf1/Prf1 and H2Bub1, in mediating the sensitivity of S. pombe to HU.

We considered the possibility that the resistance phenotype was driven by emergence of resistance in a small fraction of the mutant cells. To assess penetrance of the resistance phenotype, we plated cells on either control medium (YES) or HU medium plates (Fig S2B and C). Whereas ∼10% of wild-type cells formed colonies on HU media, ∼40–75% of the resistant cells formed colonies, and the colonies that did form grew faster (Fig S2B and C). The Colony size was generally a more reliable indicator of resistance than colony counts in these assays, particularly in the strains expressing histone H3 from a single hht2 locus. The strain with wild-type H3 expressed in this background was able to form colonies on 12 mM HU, indicating a higher basal level of resistance, but colonies formed by the hht2-K4R mutant were notably larger (Fig S2B). Thus, resistance to HU is a highly penetrant phenotype in the mutant strains.

HMA mutants exhibit increased dNTPs compared with wild type upon acute HU treatment

We previously used acute treatment with 12 mM HU to synchronize htb1-K119R cells in the S phase, suggesting that resistance to this dose of HU does not preclude the acute effect of HU on cellular dNTP pools (Page et al, 2019). To confirm this observation in other HU-resistant mutants, we monitored the DNA content by FACS in HU-treated cells. Our treatment was with 12 mM HU for 4 h, conditions we used previously and that are well established for cell synchronization (Page et al, 2019). Untreated S. pombe of all genotypes tested displayed a single, prominent peak corresponding to the 2N DNA content that is characteristic of asynchronous growth (because G2 is the longest cell cycle phase) (Fig S3) (Hoffman et al, 2015). HU-treated cells of all genotypes tested still displayed a single peak that was shifted to lower PI-A values, indicating a lower DNA content and near-uniform arrest in the S phase. However, the resistant htb-K119R, prf1∆, brl1∆, and set1∆ mutants displayed a broad 1N peak after HU treatment that partially overlapped with the peak observed in the absence of HU (Fig S3). This was not observed in the wild-type or in the set2∆ strain, indicating a correlation with HU resistance and suggesting that S-phase arrest was not complete under these conditions in the resistant mutants. These findings pointed to an alteration in the acute HU response in the resistant mutants that protects the cells from the toxic effects of chronic HU treatment.

Because HU is an inhibitor of RNR, we tested whether genetic ablation of the HMA affected cellular dNTP levels in response to HU. A liquid chromatography–coupled mass spectrometry method was used to quantify dNTP levels both before and after HU treatment. All dNTP levels were normalized to the corresponding NTP levels within each sample. We observed that HU resistance correlated with increased levels of dATP and/or dGTP in HU-treated mutant cells compared with wild type (Fig 2A–C). This effect was especially pronounced in the htb-K119R, brl1∆, and prf1∆ mutants, in which we observed an increase in dGTP even in the absence of HU (Fig 2A). This suggests that these mutations may alter nucleotide metabolism independently of their response to HU. In the set1∆ and hht2-K4R mutants, the increase was more subtle, affecting only dATP in the HU-treated condition (Fig 2B and C). No effects on purine dNTP levels were observed in set2∆, although dCTP levels were significantly elevated compared with wild type in the absence of HU (Fig 2B). The preferential effects of the mutations on dNTP levels in the presence of HU are consistent with reduced efficacy of HU in the resistant mutants.

In wild-type cells, HU treatment depleted dATP levels by 70–80% and dGTP levels by ∼50%, but had minimal effects on levels of dCTP or dTTP (Fig 2D). This is consistent with previous results demonstrating that HU affects cellular levels of purine dNTPs more strongly than pyrimidine dNTPs (Hendricks & Mathews, 1998). The mean of the HU/untreated ratios of dATP across experiments was increased in resistant mutants compared with wild type, further supporting reduced HU efficacy. These differences reached statistical significance in the htb-K119R, prf1∆, and hht2-K4R mutants (Fig 2E and F). The HU/untreated ratios for set1∆ and brl1∆ were not significantly different from wild type (Fig 2G). We conclude that increased levels of purine dNTPs, particularly in HU-treated conditions, confer protection from the deleterious effects of HU treatment in the resistant mutants, likely because the efficacy of HU treatment is diminished.

HMA mutations suppress the HU sensitivity of S-phase checkpoint mutants

HU treatment results in activation of the S-phase checkpoint, which blocks entry into mitosis in response to perturbed DNA replication (Giannattasio & Branzei, 2017). In S. pombe, the checkpoint is initiated by activation of the kinase Rad3 (ortholog of human ATR) at stalled replication forks, which transduces a signal to effector kinases Cds1 and Chk1 (ortholog of human CHK1) (Giannattasio & Branzei, 2017). We expected that the HU resistance mutations we identified would ameliorate HU hypersensitivity of checkpoint kinase mutants by avoiding replication fork stalling and subsequent DNA damage. We thus tested HU sensitivity of double mutant strains in which either htb-K119R or set1∆ was combined with a checkpoint kinase deletion. As expected, single checkpoint mutants were sensitive to HU at doses that did not affect wild-type growth: 2 mM for rad3∆ and cds1∆, and 6 mM for chk1∆ (Fig 3A). Both htb-K119R and set1∆ were able to suppress the sensitivity of all three checkpoint mutants to low doses of HU, consistent with the effects of these mutants on HU efficacy (Fig 3A).

HU-induced DNA damage resulting from stalled replication forks can be repaired by error-prone translesion bypass involving alternate DNA polymerase recruitment by monoubiquitylation of PCNA (Hedglin & Benkovic, 2015). Knockout of rhp18⁺, encoding the E3 ligase specific for PCNA, led to moderate HU hypersensitivity that was suppressed by either htb-K119R or set1∆ (Fig S4). The ability of HMA mutations to suppress multiple defects in DNA damage signaling and repair in response to HU further supports the idea that ablation of the HMA reduces HU efficacy.

Consistent with our previous results demonstrating that HU-resistant mutants were not resistant to other DNA-damaging agents (Fig S2A), neither htb-K119R nor set1∆ suppressed the sensitivity of checkpoint mutants to MMS (Fig 3B). In fact, MMS sensitivity of the double mutants was enhanced compared with the single checkpoint mutants (Fig 3B). These results support a specific effect of the resistance mutations on HU toxicity.

Mutations in the Rpb1 CTD, but not the Spt5 CTD, also cause HU resistance

The Spt5 CTD is a well-established interacting partner of Rtf1/Prf1 and is required for Prf1 recruitment to chromatin (Mayekar et al, 2013; Mbogning et al, 2013; Wier et al, 2013; Chen et al, 2020). Therefore, we tested whether Spt5 CTD mutants were HU-resistant. We used strains harboring mutations in Thr1 of the CTD amino acid repeat motif that is phosphorylated by Cdk9 (either to alanine or to the phosphomimetic glutamate) in spot tests. The mutations were introduced in the context of the full-length Spt5 CTD (18 repeats) or a truncated CTD (7 repeats) (Schneider et al, 2010; MacKinnon et al, 2023). We also used a mutant strain lacking the entire CTD (spt5∆C). Surprisingly, mutations in the Spt5 CTD did not cause HU resistance; the spt5∆C mutant was, in fact, HU-sensitive (Figs 4A and S1). Therefore, the ability of Rtf1/Prf1 to regulate the response to HU does not depend on its interaction with the Spt5 CTD.

In parallel, we tested HU sensitivity of mutants of the Rpb1 CTD. The Rpb1 CTD is dynamically phosphorylated throughout the transcription cycle at multiple residues. We used rpb1 mutants in which the amino acids Tyr1, Ser2, Thr4, or Ser7 were individually mutated in all 29 CTD heptad repeats (Sanchez et al, 2018). Phosphorylation of Tyr1, Ser2, and Thr4 all peak near the 3′ ends of genes and have been linked with transcription termination. Ser7 phosphorylation peaks at the 5′ ends of genes, and its function is poorly understood (Harlen & Churchman, 2017). We found that the rpb1-T4A, rpb1-Y1F, and rpb1-S2A mutants were all HU-resistant, whereas the rpb1-S7A mutant was sensitive to HU (Figs 4A and S1). Liquid growth assays in the presence of varying HU concentrations were consistent with the spot tests, as were single-colony plating assays (Figs 4B and S5A and B), and revealed more robust resistance in rpb1-Y1F and rpb1-T4A mutants than in rpb1-S2A. Y1 and T4 are not known targets of Cdk9 in S. pombe, suggesting that they function in a pathway parallel to the HMA to promote HU resistance.

We quantified dNTP levels in the most resistant Rpb1 CTD mutant, rpb1-T4A, and the most sensitive mutant, rpb1-S7A (Fig 4C). The rpb1-T4A mutant displayed significantly higher levels of dATP after treatment with HU compared with the wild-type strain, but had levels similar to wild type in the absence of HU. Conversely, dGTP levels in the rpb1-S7A strain were slightly elevated in untreated cells, but comparable to wild type in HU (Fig 4C). The HU/untreated ratio trended higher in rpb1-T4A than in wild type or rpb1-S7A (Fig 4D). These data mirror our findings on the HMA mutants.

To further corroborate these findings, we subjected the Rpb1 CTD mutants to the same battery of phenotypic and genetic interaction tests we performed on htb-K119R and set1∆. In summary, we found that rpb1-Y1F, rpb1-S2A, and rpb1-T4A were not resistant to phleomycin (Fig S6A); that rpb1-Y1F and rpb1-T4A suppressed HU hypersensitivity of rhp18∆, whereas rpb1-S7A enhanced it (Fig S6B); and that rpb1-Y1F and rpb1-T4A suppressed HU hypersensitivity of checkpoint kinase mutants, whereas rpb1-S7A did not (Fig S7). Collectively, these results suggest that the removal of Rpb1 CTD repeat residues Y1 or T4 leads to HU resistance by reducing HU efficacy, similar to what we found for HMA mutants.

We previously showed that mutations in multiple Rpb1 CTD phosphosites modestly affect global H3K4me3 levels (Mbogning et al, 2015). These experiments employed S. pombe strains in which phosphosite mutations were introduced in the context of a truncated CTD repeat array (Schwer & Shuman, 2011). We thus tested whether effects on H3K4me3 levels could account for the HU resistance phenotypes of the full-length CTD mutants used in this study. Immunoblotting did not detect significant differences in global H3K4me3 levels between the CTD mutants and wild type (either in the absence or in the presence of HU) (Fig S8). H3K4me3 levels trended higher in the rpb1-S2A mutant, in agreement with our previous findings (Mbogning et al, 2015). Thus, HU resistance of Rpb1 CTD mutants is not likely a consequence of diminished H3K4me3.

HU-resistant mutants exhibit unique transcriptome features in the absence of HU treatment

We hypothesized that HU resistance in HMA and Rpb1 CTD mutants most likely arises from effects on RNAPII transcription. To investigate potential transcriptional defects underlying these phenotypes, we performed spike-in normalized RNA-seq on poly(A)+ RNA isolated from two sets of strains: wild type, prf1∆, htb-K119R, set1∆, set2∆, brl1∆, and cds1∆; and rpb1-WT, rpb1-Y1F, rpb1-S2A, rpb1-T4A, and rpb1-S7A. Each set included its own isogenic wild-type control. Three biological replicates were included for each strain, grown under control or HU-treated conditions.

The sets of differentially expressed transcripts determined in comparisons between untreated wild type and untreated prf1∆, brl1∆, htb-K119R, or set1∆ reflected the hierarchy of the HMA: 506 coding transcripts were dysregulated in prf1∆, 317 in brl1∆, 380 in htb1-K119R, and 135 in set1∆ (Fig S9A). This recapitulates previous observations demonstrating broader gene regulatory roles of H2Bub1 than of H3K4me, and extends this trend further upstream to imply still broader roles of Rtf1 (Tanny et al, 2007; Zofall & Grewal, 2007; Fleming et al, 2008; Batta et al, 2011). We assessed the overlap between differentially expressed transcripts identified by RNA-seq in the HMA mutants and in set2∆ by performing pairwise comparisons (Fig S9B). Fewer than 100 transcripts were differentially regulated in cds1∆, so these data were not included in this analysis. As expected, we found highly significant overlaps between all classes of differentially expressed transcripts in the HMA mutants. The coding and non-coding transcripts that were increased compared with wild type in the HMA mutants overlapped significantly with those in set2∆. The coordinate regulation of non-coding transcripts by the HMA and Set2 is consistent with the known roles of these factors in repressing antisense transcription within protein-coding genes, and argues that many of the targets of antisense suppression are common to both factors (Sansó et al, 2020). In contrast, the decreased coding transcripts in the HMA mutants were distinct from those identified in set2∆, suggesting that they reflect unique properties of the HMA mutants (Fig S9B).

RNA-seq identified similar numbers of differentially expressed transcripts in all of the Rpb1 CTD mutants. All classes of transcripts showed broadly significant overlaps in pairwise comparisons between the mutants, consistent with previous results (Fig S10A and B) (Garg et al, 2021). This suggests that gene regulatory roles of these CTD residues are shared at least in part. Overlaps between groups of increased coding or increased non-coding transcripts were consistently less significant in comparisons involving rpb1-S7A, indicating that the steady-state transcriptome of this mutant under normal growth conditions diverges from that of rpb1-Y1F, rpb1-S2A, and rpb1-T4A (Fig S10B). This is in accord with the distinct phenotypic behavior of this mutant (Schwer et al, 2014).

We also determined the overlap between groups of differentially expressed genes identified in HMA mutants, set2∆, and Rpb1 CTD mutants to ascertain whether there were expression signatures associated with the HU resistance phenotype. Pairwise comparisons did not reveal any consistent patterns that distinguished resistant and non-resistant mutants, although they did uncover a particularly strong association between non-coding transcripts increased in the HMA mutants, set2∆, and rpb1-S2A (Fig S11). This aligns with previous data implicating S2 phosphorylation in non-coding RNA repression (Schwer et al, 2014; Garg et al, 2021). We identified only four genes that were differentially expressed in all the HU-resistant mutants, but not in set2∆ or rpb1-S7A: SPCC18B5.02c⁺(a cinnamoyl-CoA reductase pseudogene), mug182⁺ (encoding a NADHX epimerase), SPNCRNA.472, and SPNCRNA.980. None of the known functions of these genes suggest a role in HU resistance.

HU-resistant mutants exhibit a blunted transcriptional response to acute HU treatment

We next compared the transcriptome profiles of wild-type and mutant strains in the presence or absence of HU treatment. In wild-type cells treated with HU, 893 total transcripts were differentially expressed compared with control conditions, 85% of which (763/893) were up-regulated and 15% of which (130/893) were down-regulated. In total, HU treatment affected 6.8% (347/5118) of fission yeast protein-coding genes (Fig 5A). Strikingly, HU treatment of the resistant mutants prf1∆, htb-K119R, brl1∆, and set1∆ exhibited 54, 134, 280, and 534 differentially expressed transcripts, respectively, compared with control conditions. Thus, HMA mutations blunt the transcriptome changes induced by HU. The extent of the blunting effect roughly correlated with the strength of the resistance phenotype we measured in quantitative growth assays (Fig 1C). In contrast, the set2∆ exhibited 971 differentially expressed transcripts in response to HU and cds1∆ exhibited 925. Transcripts induced by HU were highly overlapping between wild type and both resistant and non-resistant mutants, indicating that the character of the gene expression response was maintained (Fig S12A). To visualize the transcriptome changes induced by HU in the different strains, we performed hierarchical clustering by the expression level for the set of HU-regulated transcripts in wild-type cells. For most of the strains, the HU-treated samples clustered separately from the untreated samples (Fig S12B). In the case of the strongly resistant prf1∆ and htb-K119R mutants, the distance between HU-treated and untreated expression patterns in the cluster diagram was diminished, such that control and HU-treated conditions could not be distinguished. This result further demonstrates that HU-induced transcriptome changes are blunted in HMA mutants that are resistant to HU.

The blunting effect was evident, albeit less pronounced, in HU-resistant Rpb1 CTD mutants as well (Fig 5B). There were 865 differentially expressed transcripts for the rpb1-WT strain comparing HU treatment with control, whereas there were 342 for rpb1-Y1F, 577 for rpb1-T4A, and 639 for rpb1-S2A. The pattern in the HU-sensitive rpb1-S7A mutant was similar to wild type, as 813 differentially expressed transcripts were identified. We also verified that the HU-induced transcriptome alterations in rpb1-WT were comparable to those observed for the wild-type control in Figs 6A and S13A. The blunting effect was less apparent in the hierarchical clustering analysis performed on the Rpb1 CTD mutants than what we observed for the HMA mutants. Nonetheless, gene expression patterns for resistant rpb1-Y1F and rpb1-T4A mutants clustered more closely to one another across treatment conditions than those for non-resistant strains (Fig S13B).

We used gene set enrichment analysis to further characterize the groups of HU-regulated coding transcripts we observed. In the wild-type strain, HU-repressed coding transcripts were enriched for cell cycle genes, specifically targets of the transcription factors Sep1 and Ace2 that are activated in mitosis (Fig 5C) (Rustici et al, 2004; Marguerat et al, 2006). This is consistent with HU-mediated cell cycle arrest in the S phase, leading to the decreased expression of mitotic genes (Chu et al, 2007). A similar pattern of enrichment was observed for HU-repressed coding transcripts in all of the resistant mutants, and in the non-resistant rpb1-S7A, consistent with the fact that 4-h HU treatment at 12 mM led to efficient S-phase arrest in all of these strains (Fig S3 and data not shown). Enrichment of mitotic genes was less prominent in prf1∆ and htb1-K119R, reflecting the stronger blunting effect in these mutants. As expected, the HU-induced coding transcripts in the wild-type strain showed enrichment for targets of the Cdc10 transcription factor that activates G1/S genes (Fig 5D) (Rustici et al, 2004; Marguerat et al, 2006). In addition, we observed enrichment for meiotic genes, commonly induced in mitotically growing S. pombe under stress conditions, as well as oxidative stress response genes (Tanay et al, 2005; Chen et al, 2008). Targets of the oxidative stress–responsive transcription factor Pap1 were notably enriched (Chen et al, 2008). This observation was expected given that HU is known to cause oxidative stress (Singh & Xu, 2016). These gene set enrichments were also a feature of the HU-induced genes in all of the mutant strains examined, including the HU-sensitive rpb1-S7A mutant. However, meiosis and stress response genes were less enriched in the strongly resistant prf1∆ and htb1-K119R strains.

The blunting of the HU-induced changes in gene expression may be related to gene expression changes in untreated cells that resemble the effect of HU. Indeed, there was significant enrichment of oxidative stress genes and Pap1 targets among transcripts induced in the prf1∆ and htb1-K119R mutants in the absence of HU (Fig S14A). Moreover, cell cycle–regulated genes were enriched among the genes repressed by the htb1-K119R mutation in the absence of HU. These included Ace2 targets and, surprisingly, Cdc10 targets (Fig S14B). We assessed the proportion of transcripts induced by prf1∆ in untreated cells that were also induced by HU in the wild type. We found over 50% overlap for both coding (122/233) and non-coding (260/392) transcripts. The overlaps were comparable in the set2∆ strain (89/169 coding; 217/304 non-coding), despite the fact that there is no blunting effect in this strain (Fig S14C). Thus, blunting of the HU-induced transcriptome changes in resistant mutants is not simply due to pre-existing differences in transcript levels.

To determine whether the blunting effect resulted from changes in transcription, we assessed genome-wide RNAPII occupancy by performing ChIP-seq in wild-type and mutant strains expressing HA-tagged Rpb3 (Coudreuse et al, 2010). Each strain was profiled under control conditions or after a 4-h treatment with 12 mM HU. Metagene plots of normalized RNAPII occupancy across all protein-coding genes revealed that, for all strains analyzed, the profiles were similar in HU-treated and untreated samples (Fig S15). The same analysis focused on the set of 262 protein-coding genes induced by HU in wild-type cells revealed increased average RNAPII occupancy in HU-treated cells compared with untreated cells, as expected (Fig 6A). This increase was blunted in HU-resistant mutants, but was similar to wild type in the HU-sensitive rpb1-S7A mutant. We quantified the difference in RNAPII occupancy in untreated and HU-treated cells determined from the metagene analyses for each strain (Fig 6B). This difference was decreased in all the resistant mutants compared with wild type; variability in the magnitude of the decrease in the two ChIP-seq replicates precluded the difference from reaching significance in the prf1∆ and brl1∆ strains. In contrast, the occupancy difference in rpb1-S7A was similar to wild type. Thus, the transcriptional response to HU was impaired in HU-resistant mutants.

We did not observe increased RNAPII occupancy in response to HU at HU-induced non-coding RNA genes (Fig S16A). Most of these RNAs are antisense to protein-coding genes, and thus, HU-induced changes in RNAPII occupancy may be obscured by ongoing transcription in the sense direction. However, we made the same observation when we analyzed HU-induced antisense and intergenic non-coding RNA genes separately (using high-confidence non-coding RNA annotations described previously [Atkinson et al, 2018]) (Fig S16A). This suggests that post-transcriptional stabilization is important for the observed increase in abundance of these transcripts in response to HU.

Given that Rpb1 CTD Y1/S2/T4 phosphorylation is involved in transcription termination, we interrogated termination efficiency at HU-induced protein-coding genes. We computed an average termination index for this group of genes in each strain in the absence or presence of HU by determining the ratio of Rpb3-HA occupancy in a 0.5-kb interval downstream of the transcription end site to that in a 0.5-kb interval upstream (Fig S16B). We found a significant increase in termination index in the rpb1-S2A mutant compared with wild type, but no significant effects in any of the other strains. This suggests that blunting of the HU-induced transcriptional response in resistant mutants was not related to defective termination.

Loss of the HMA or Rpb1 CTD Thr4 confers a multi-drug resistance phenotype

In the absence of specific transcriptional changes that could be obviously implicated as causal to HU resistance, we tested several hypotheses for how resistance arises. First, we asked whether resistance could be due to increased RNR activity. We did not observe significant increases in cdc22⁺ or suc22⁺ mRNAs (encoding RNR subunits) in the resistant mutants by RNA-seq, nor did the resistant htb1-K119R mutant show elevated levels of the corresponding proteins in the absence or presence of HU (Fig S17A). RNR activity is also regulated by subcellular localization of Suc22. Outside of the S phase, Suc22 is sequestered in the nucleus by the RNR inhibitor protein Spd1. Proteasomal degradation of Spd1 in G1/S liberates Suc22 to move to the cytoplasm and form the active RNR complex with Cdc22 (Nestoras et al, 2010; Guarino et al, 2014). We did not observe increased cytoplasmic Suc22 localization in htb1-K119R or set1∆ cells compared with wild-type cells by fluorescence microscopy (Fig S17B). Moreover, htb1-K119R, set1∆, and rpb1-T4A caused HU resistance in a genetic background lacking Spd1, indicating that resistance operates independently of this RNR regulatory mechanism (Fig S17C).

Because HU is known to cause oxidative stress, and the strongly HU-resistant mutants partially activated the transcriptional response to oxidative stress (Fig 5D), we tested growth in the presence of hydrogen peroxide. HU-resistant mutants were in fact sensitive to doses of H₂O₂ that were permissive to wild-type growth (Fig S18A and B). We further tested the possibility that chronic, low-level oxidative stress in the resistant mutants activated a drug efflux response that allowed HU tolerance. Indeed, we detected significant accumulation of reactive oxygen species in the resistant mutants prf1∆ and brl1∆, and a similar trend in htb1-K119R, using the fluorescent reactive oxygen species sensor dichlorodihydrofluorescein diacetate (DCFDA), suggesting chronic oxidative stress (Fig S19A and B). The DCFDA labeling in the mutants in the absence of HU was more robust than that induced by HU treatment, and was not augmented by HU. DCFDA labeling in set1∆, rpb1-Y1F, and rpb1-T4A did not significantly differ from wild-type controls. Furthermore, deletion of pap1⁺, encoding the transcription factor that activates antioxidant and drug efflux responses in response to low levels of oxidative stress, did not prevent HU resistance in htb1-K119R or set1∆ backgrounds (Fig S19C) (Vivancos et al, 2006; Calvo et al, 2012). These data show that HU resistance does not depend on the oxidative stress response.

We further tested whether activation of other general drug tolerance mechanisms could account for HU resistance. If this were the case, one would expect that resistance to HU would correlate with tolerance to other drugs. We found that all HU-resistant mutants tested exhibited cross-resistance to the clinical antifungal agent fluconazole, whereas the set2∆ and rpb1-S7A mutants grew similar to wild type (Fig 7A). Most HU-resistant mutants also showed increased tolerance to clotrimazole and the pesticide tebuconazole, although these phenotypes were shared with rpb1-S7A (Fig 7B). We conclude that HU resistance is part of a multi-drug–resistant phenotype induced by the loss of the HMA or Rpb1 CTD Thr4.

Yeast strains

S. pombe strains used in this study are listed in Table S1. For strains obtained from the National BioResource Project (NBRP; Japan), the strain number is provided. All genetic manipulations, including crosses and tetrad dissection, used standard techniques as previously described (Chen et al, 2020). Standard YES media (5 g/l yeast extract, 30 g/l D-glucose, and 250 mg/l of each histidine, leucine, adenine, and uracil) were used for the growth of all cultures (Moreno et al, 1991).

Spot tests

S. pombe strains were grown in 2 ml YES to an OD₆₀₀ of 0.3–0.6. 1 ml of culture was removed and adjusted to an OD₆₀₀ of 2 with YES media. Fivefold serial dilutions were performed in sterile water, and 3 μl of each dilution was spotted onto the corresponding plates. The plates were then incubated at 30°C for 2–5 d and photographed. Hydroxyurea (HU), MMS, phleomycin, fluconazole, clotrimazole, and tebuconazole were purchased from MilliporeSigma and added to YES plates at the indicated concentrations. Twofold serial dilutions were used for spot tests on phleomycin plates.

Liquid growth assays

S. pombe strains were grown in 2 ml YES to an OD₆₀₀ of 0.3–0.6. 1 ml of culture was removed and diluted to an OD₆₀₀ of 0.05 in YES. 200 μl of culture from each strain was added to each well of a 96-well plate (triplicate wells for each strain/condition). HU was added to the wells at the indicated concentrations. The plate was then incubated at 30°C with shaking, and OD₆₀₀ readings were taken with a plate reader for 36 h. The triplicate values for each timepoint and strain were plotted in GraphPad Prism 8 to generate individual growth curves for each experiment. All growth curves were fitted by the Gompertz growth modeling to extrapolate the Y_max and k values (Tjørve & Tjørve, 2017). The growth curve experiments were conducted at least three times for each strain. To generate average growth curve figures for each strain, the triplicate values of OD₆₀₀ readings from each timepoint were averaged for each experiment, and then, these averaged OD₆₀₀ values from each experiment were plotted.

Colony plating assay

S. pombe strains were grown in 2 ml YES to an OD₆₀₀ of 0.3–0.6. 15 μl of each culture was removed, and cells were counted using a hemocytometer. 100–1,000 cells were then plated onto YES or YES+12 mM HU plates and incubated at 30°C. YES plates were incubated for 2 d, whereas YES+HU plates were incubated for 6 d. Plates were then photographed, and colonies were counted.

FACS analysis

S. pombe cells were grown to an OD₆₀₀ of 0.2 in 50 ml YES at 30°C and then split into two 25 ml cultures. HU was added to one of the cultures to a final concentration of 12 mM, and incubation was continued for 4 h. Cultures were then harvested by centrifugation at 4°C and fixed with ice-cold 70% ethanol. Fixed cells were centrifuged, and cell pellets were washed once with 1 ml 50 mM sodium citrate, followed by resuspension in 0.5 ml of 50 mM sodium citrate containing 0.1 μg/ml RNase A. After a 4-h incubation at 37°C, 0.5 ml of 50 mM sodium citrate containing 4 μg/ml propidium iodide was added. 500,000 events were analyzed per sample on a FACSCalibur-I instrument (BD Biosciences). Data were processed using FlowJo software as previously described (Knutsen et al, 2011; Page et al, 2019).

For staining of reactive oxygen species, cells were cultured in the presence or absence of HU as described above. For the last hour in culture, 50 μM DCFDA (MilliporeSigma) was included in the media. The cells were pelleted and washed twice with 50 mM sodium citrate. The cells were then stained with propidium iodide (to allow the removal of dead cells) and analyzed as described above.

RNA-seq

S. pombe cells were grown to an OD₆₀₀ of 0.2 in 50 ml YES at 30°C and then split into two 25 ml cultures. HU was added to one of the cultures to a final concentration of 12 mM, and incubation was continued for 4 h. Cells were then harvested by centrifugation at 4°C, washed once with distilled water, and frozen at −80°C. Total RNA was extracted using a hot phenol method (Tanny et al, 2007), followed by purification using RNeasy Mini Kit (QIAGEN). The quality of the RNA samples (RIN score >7) was verified on Agilent 2100 Bioanalyzer using the RNA 6000 kit (cat# 5067-1513; Agilent). 1 μg of each RNA sample was then used for poly(A) selection and library preparation. Before poly(A) selection, 2 μl of a 1:100 dilution of ERCC RNA Spike-In Mix 1 (cat# 4456740; Ambion by Life Technologies) was added to the total RNA. Poly(A) selection was conducted using NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs). Libraries were prepared using NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs). The quality of the libraries was assessed using the High Sensitivity DNA kit (cat# 5067-4626; Agilent) on Agilent 2100 Bioanalyzer.

RNA-seq libraries (three replicates for each strain/condition) were sequenced at the Michael Smith Genome Sciences Centre (BC Cancer, University of British Columbia, Vancouver, BC). Multiplexed libraries were normalized to 15 nM using JANUS G3 Varispan Automated Workstation (PerkinElmer) and pooled (26 samples per pool). Each of the three pools was sequenced in a single lane of a HiSeq X instrument with paired-end 150 base reads. The quality of the reads was determined using FastQC before and after trimming with TrimGalore (0.6.6) (Babraham Institute Bioinformatics; https://www.bioinformatics.babraham.ac.uk/). The trimmed reads were then aligned to a reference genome consisting of the S. pombe ASM294v2 genome combined with the ERCC RNA Spike-In transcripts. Alignment was with STAR (2.7.8a) (Dobin et al, 2013). S. pombe transcripts were assembled using StringTie (2.1.5) and normalized to the spike-in RNA using RUVSeq (1.30.0) (Risso et al, 2014; Pertea et al, 2015). Pearson’s correlations between normalized datasets within each triplicate were between 0.97 and 1, with the exception of rpb1-Y1F n1/rpb1-Y1F n3 (0.94), and set1∆ n1/set1∆ n3 (0.94). Differential expression was assessed using DESeq2 (1.36.0) using all three replicates (Love et al, 2014). The adjusted P-value threshold of 0.05 and log₂FoldChange of 2 were used for the comparisons. Heat maps were generated in R using pheatmap (1.0.12). Gene ontology enrichment analysis, comparison with other S. pombe gene expression datasets, and pairwise comparison between sets of differentially expressed genes were performed using AnGeLi (Bitton et al, 2015).

Chromatin immunoprecipitation sequencing (ChIP-seq)

S. pombe strains expressing rpb3-HA were grown to an OD₆₀₀ of 0.2 in 100 ml YES at 30°C and then split into two 50 ml cultures. HU was added to one of the cultures to a final concentration of 12 mM, and incubation was continued for 4 h 1.5 × 10⁷ cells from each sample were then crosslinked with formaldehyde, harvested by centrifugation, and frozen as described previously (Chen et al, 2020). Chromatin extracts were prepared in a volume of 1 ml as described previously, and 100 μl of extract was removed and processed separately (“input”). Immunoprecipitation was performed by adding 3 μg of HA antibody (monoclonal 12CA5; MilliporeSigma) to the remaining extract and incubating at 4°C for 4 h, followed by the addition of 15 μl magnetic protein G beads (Dynabeads; Thermo Fisher Scientific) and a further 1-h incubation at 4°C. The beads were recovered and washed, followed by elution and purification of the immunoprecipitated DNA as described (Chen et al, 2020).

ChIP-seq libraries (two replicates for each strain/condition) were prepared with 2 ng of input or IP DNA using the NEBNext Ultra II DNA Library kit for Illumina (New England Biolabs). The quality of the libraries was assessed using the High Sensitivity DNA kit (cat# 5067-4626; Agilent) on Agilent 2100 Bioanalyzer. Multiplexed samples were pooled and sequenced at the Michael Smith Genome Sciences Centre as described above. The quality of the reads was determined using FastQC before and after trimming with TrimGalore (0.6.6) (Babraham Institute Bioinformatics; https://www.bioinformatics.babraham.ac.uk/). The trimmed reads were aligned to the reference S. pombe ASM294v2 genome using bowtie2 (Langmead & Salzberg, 2012). Duplicate reads were marked with Picard (Genome Analysis Toolkit; https://gatk.broadinstitute.org). The aligned, deduplicated bam files were sorted and indexed in SAMtools (Li et al, 2009). The deeptools package (version 2.0 [Ramírez et al, 2016]) was then used to derive bigWig files corresponding to IP/input ratios for each sample, and to scale based on read depth using the SES method. A median genic Rpb3-HA occupancy across all protein-coding genes was computed for each bigWig file; this value was set to 1 for the wild-type control and used to scale the other bigWig files using WiggleTools (Zerbino et al, 2014). Metagene plots were generated from the scaled bigWig files using deeptools.

Quantification of dNTP levels

S. pombe cells were grown to an OD₆₀₀ of 0.2 in 100 ml YES at 30°C and then split into two 50 ml cultures. HU was added to one of the cultures to a final concentration of 12 mM, and incubation was continued for 4 h. Equal numbers of cells from each sample (1 × 10⁷) were harvested by centrifugation. Samples were kept on ice, and all solutions were pre-chilled in subsequent steps. The cell pellet was washed with ice-cold water and then resuspended in 380 μl 50% MeOH/water and 220 μl acetonitrile (ACN). Chilled glass beads were added to the meniscus of each sample, and cells were lysed by bead beating (4 × 60 s pulses in a Biospec bead beater). A hole was poked into the bottom of the tube to recover the lysates by centrifugation. Lysates were extracted with 600 μl dichloromethane and 300 μl ice-cold water. Samples were vortexed for 1 min and then allowed to partition on ice for 10 min. After centrifugation for 10 min at 1,600g at 1°C, the upper aqueous phase was transferred to a pre-chilled 1.5-ml tube on dry ice and dried in SpeedVac. Samples were resuspended in 50 μl of water and subjected to LC-MS analysis. For nucleotide analysis, a 10x dilution was prepared by adding 3 μl of sample to 27 μl of water.

The relative concentrations of the targeted nucleotides and deoxynucleotides were measured using a triple quadrupole mass spectrometer (QQQ 6470) equipped with a 1290 ultra–high-pressure liquid chromatography system (Agilent Technologies). Chromatographic separation was achieved using a Scherzo SM-C18 column 3 μm, 3.0 × 150 mm (Imtakt Corp). The chromatographic gradient started at 100% mobile phase A (5 mM ammonium acetate in water) with a 5-min gradient to 100% B (200 mM ammonium acetate in 20% ACN/80% water) at a flow rate of 0.4 ml/min. This was followed by a 5-min hold time at 100% mobile phase B and a subsequent re-equilibration time (6 min) before the next injection. To ensure proper instrumental duty cycle, samples were injected twice: nucleotide analysis followed by deoxynucleotide analysis. A sample volume of 5 μl was injected for each run.

Multiple reaction monitoring transitions were optimized on standards for each metabolite measured. Multiple reaction monitoring transitions and retention time windows are summarized in Table S2. An Agilent JetStream electrospray ionization source was used in a positive ionization mode in which a gas temperature and flow were set at 300°C and 5 l/min, respectively, nebulizer pressure was set at 45 psi, and capillary voltage was set at 3500 V. Relative concentrations were determined from external calibration curves prepared in water. Ion suppression artifacts were not corrected; thus, the presented metabolite levels are relative to the external calibration curves and should not be considered as absolute concentrations. Data were analyzed using MassHunter Quant (Agilent Technologies). Values for dNTPs were normalized to the corresponding NTPs (Nestoras et al, 2010).

Fluorescence microscopy

S. pombe strains expressing CFP-Suc22 were grown to an OD₆₀₀ of 0.2 in 10 ml YES at 30°C and then split into two 5 ml cultures. HU was added to one of the cultures to a final concentration of 12 mM, and incubation was continued for 4 h. Live cells were imaged on an Olympus IX83 fluorescence microscope with a 100X objective. Focal CFP-Suc22 fluorescence was scored as nuclear localization, whereas diffuse fluorescence was scored as cytoplasmic. At least 200 cells were analyzed in each experiment.

Immunoblotting

S. pombe strains expressing Cdc22-YFP or CFP-Suc22 were grown to an OD₆₀₀ of 0.2 in 40 ml YES at 30°C and then split into two 20 ml cultures. HU was added to one of the cultures to a final concentration of 12 mM, and incubation was continued for 4 h. Cells were harvested and washed in 5% trichloroacetic acid, and extracts were prepared by bead beating in 20% trichloroacetic acid as described (Chen et al, 2020). SDS–PAGE and immunoblotting were as described using a GFP monoclonal antibody (Clontech). Histone immunoblots were performed the same way using antibodies against H3K4me3 (ab8580; Abcam) or total H3 (ab1791; Abcam).

Hydrogen peroxide (H₂O₂) sensitivity

S. pombe strains were grown in 2 ml YES to an OD₆₀₀ of 0.3–0.6. 1 ml of culture was removed and adjusted to an OD₆₀₀ of 0.05 with sterile water. 200 μl of each culture was added to one well of a 96-well plate (each strain was analyzed in triplicate). H₂O₂ was added at the indicated concentrations. The plate was then incubated at 30°C with shaking, and OD₆₀₀ was monitored at 24 and 48 h.