Alternative translation initiation by ribosomal leaky scanning produces multiple isoforms of the Pif1 helicase

Abstract

In budding yeast, the integrity of both the nuclear and mitochondrial genomes relies on dual-targeted isoforms of the conserved Pif1 helicase, generated by alternative translation initiation (ATI) of PIF1 mRNA from two consecutive AUG codons flanking a mitochondrial targeting signal. Here, we demonstrate that ribosomal leaky scanning is the specific ATI mechanism that produces not only these, but also novel, previously uncharacterized Pif1 isoforms. Both in-frame, downstream AUGs as well as near-cognate start codons contribute to the generation of these alternative isoforms. This has crucial implications for the rational design of genuine separation-of-function alleles and provides an explanation for the suboptimal behaviour of the widely employed mitochondrial- (pif1-m1) and nuclear-deficient (pif1-m2) alleles, with mutations in the first or second AUG codon, respectively. We have taken advantage of this refined model to develop improved versions of these alleles, which will serve as valuable tools to elucidate novel functions of this helicase and to disambiguate previously described genetic interactions of PIF1 in the context of nuclear and mitochondrial genome stability.

Graphical Abstract

Introduction

Pif1 belongs to the conserved superfamily 1B (SF1B) of 5′-3′ helicases. In addition to the characteristic SF1B helicase core region, Pif1 and its related proteins also contain a distinctive Pif1 signature motif that has led to the identification of Pif1-like helicases in practically all eukaryotes -including yeasts and human-, eubacteria, archaea and viruses (1). Although most eukaryotes encode a single Pif1 helicase, S. cerevisiae and closely related fungi possess a second paralog, Rrm3, with 40% identical residues to Pif1. In budding yeast, the Pif1 helicase is involved in DNA replication and recombination, safeguarding the integrity of both the mitochondrial and nuclear genomes. The proposed functions for nuclear Pif1 include (i) telomere maintenance, as a catalytic inhibitor of telomerase at chromosomal ends and double strand breaks, (ii) Okazaki fragment maturation, (iii) DNA synthesis during break-induced replication and recombination-dependent replication or (iv) support of the replication fork barrier at the rDNA, as reviewed elsewhere (2–4). To accomplish these functions, Pif1 activities are tightly regulated through protein–protein interactions and post-translational modifications, including lysine acetylation (5) and checkpoint-dependent phosphorylation (6,7). Comparatively, its essential function in the maintenance of mitochondrial DNA is poorly understood at the molecular level, but it has been reported that Pif1 travels with the replisome, facilitates repair of DSBs, protects mtDNA from oxidative damage (8), and is necessary to maintain iron and zinc homeostasis (9). To fulfil these roles PIF1 encodes two different isoforms produced by alternative translation initiation (ATI) from AUG¹ or AUG⁴⁰, which flank a mitochondrial targeting signal (MTS) (10). To avoid the pleotropic effects of PIF1 deletion in genetic studies, two separation-of-function alleles for the mitochondrial (pif1-m1) and nuclear (pif1-m2) functions were developed, in accordance with the described ATI rationale (10). In pif1-m1, the production of the mitochondrial Pif1 (mPif1) is abolished by mutation of AUG¹, while nuclear Pif1 (nPif1) can still be translated from the intact AUG⁴⁰ (Supplementary Figure S1a). Conversely, in pif1-m2, the mutation of AUG⁴⁰ abrogates the production of nPif1 without affecting the translation of mPif1 (Supplementary Figure S1b). pif1-m2 has been widely exploited to interrogate the involvement of Pif1 in nuclear DNA replication and repair. However, since its first description, pif1-m2 mutants have been acknowledged to retain a certain amount of nuclear protein, as inferred from their partial suppression of some of the pif1Δ phenotypes in different genetic set-ups (11–17).

Here, we demonstrate that ribosomal leaky scanning is the main ATI mechanism underlying start codon selection in PIF1 mRNA and characterize a novel alternative nuclear isoform of Pif1. Moreover, we uncover the ability of near-cognate AUG codons to produce a phenotypically relevant amount of functional Pif1. These previously unrecognized sources of Pif1 activity may explain not only the residual nuclear activity of pif1-m2, but also the less explored incomplete phenotype of pif1-m1 mutants. This refined model, together with the identification of Pif1 nuclear localization signal, has allowed us to develop improved nuclear- and mitochondrial-deficient alleles for Pif1, which can be employed to disambiguate misleading genetic interactions.

Materials and methods

Yeast manipulation and growing

All S. cerevisiae strains are derivatives of W303 background, as detailed in Supplementary Table S1. Yeast manipulations were performed according to standard procedures (18). Cells were typically grown in YP (1% yeast extract, 2% peptone) containing 2% glucose (YPD) or 2% raffinose (YPRaff), containing 40 μg/ml adenine. In some cases, adenine was omitted from media to allow development of the red pigment in W303 strains. For induction of genes under the control of the GAL1 promoter, 2% galactose was added to YPRaff (YPGal). YP containing 3% ethanol and 3% glycerol (YPEGly) was employed to test petite phenotypes. SC synthetic media were employed for selection of auxotrophic mutants and for live cell microscopy. Mutant strains were generated by PCR-based editing (19,20), followed by diploid formation and tetrad dissections when required. Point mutations at endogenous loci were introduced with delitto perfetto (21).

Tetrad dissection analysis

Strains of opposite mating types were mixed in patches and incubated on YPD plates for 4 h at 30°C. Diploids were selected by consecutive streaks in selective media, or by micromanipulation under a dissection microscope. Once selected, diploids were incubated on SPO plates (3 mg/ml CH₃COOK, 10 μg/ml Ade, 10 μg/ml His, 10 μg/ml Leu, 10 μg/ml Trp, 10 μg/ml Ura) for 3 days at 30°C before microdissection. After dissections, plates were incubated at 30°C and imaged after 48, 72 and 96 h in a Gel Doc XR+ (Bio-Rad).

Cloning and mutagenesis

All plasmids in this study are listed in Supplementary Table S2. The coding sequences of C-terminal 6xHis-6xFLAG-tagged PIF1, pif1-m1 and pif1-m2 were cloned in pRSII406 vectors (22). Alternatively, untagged versions of the same alleles were cloned in pDONR221 and shuttled to pAG306GAL-ccdB-eGFP vectors (23). From these vectors, derivative plasmids carrying different mutations or truncations were created by standard subcloning, inverse PCR and/or Gateway recombination. The sequences of all the oligonucleotides employed can be found in Supplementary Table S3. Details of each cloning can be provided upon request.

Protein analysis

Yeast cultures in mid log-phase were harvested and cell pellets disrupted using glass beads in 10% TCA. Precipitates were collected by centrifugation, resuspended in 2× NuPAGE sample buffer (Invitrogen, #NP0008) supplemented with 200 mM DTT, and neutralized with 1 M Tris-base. Samples were boiled at 95°C for 5 min, cleared by centrifugation, and separated in either NuPAGE 3–8% (Invitrogen, #EA0375BOX) or 7% Tris-Acetate polyacrylamide gradient gels (Invitrogen, #EA0355BOX).

For Western blotting, proteins were transferred onto Amersham™ Hybond® P PVDF membranes (GE Healthcare, #10600023) and detected with the following antibodies: FLAG-HRP (mouse, 1:10000, A8592-1MG, Sigma), eGFP (mouse 1:2000, 11814460001, Roche) and Pgk1-HRP (mouse, 1:5000, ab197960, Abcam). Alternatively, for fluorescence detection and quantitative western blotting, total proteins were detected with No-Stain Reagent (Invitrogen, #A44449) after transfer onto Immobilon - FL PVDF membranes (Millipore, #IPFL00010). FLAG-tagged proteins were detected using M2 anti-FLAG antibody (mouse, F1084, Merck) followed and fluorescent Anti-Mouse Alexa Fluor™ Plus 800, (Goat, 1:10000, A32730, Invitrogen) and imaged in a ChemiDoc™ MP (Bio-Rad).

Protein purification

A pET28b plasmid encoding C-terminal 6xHis-tagged nuclear Pif1 (Addgene plasmid #65047) (24) was employed to express and purify Pif1 from Escherichia coli BL21-CodonPlus (DE3)-RIL strain (Agilent) as described (25). The same plasmid was mutagenized to express and purify Pif1^nls, Pif1^107-859 and ATPase-dead Pif1^KA(K264A), employing the same protocol.

Splayed arm substrate preparation

Splayed arm substrates with 3′-6FAM labelling were prepared essentially as previously described (26), employing the following PAGE-purified ssDNA oligonucleotides (Sigma-Aldrich): A6-3′FAM (5′-ATTGGTTATTTACCGAGCTCGAATTCACTGG-3′-6FAM) and A9 (5′-CCAGTGAATTCGAGCTCGGTACCCGCTAGCGGGGATCCTCTA-3′) (27).

Briefly, labelled and unlabelled oligonucleotides were mixed in a 1:3 ratio, boiled in a water bath and cooled down to room temperature overnight. Substrates were subsequently purified from 10% polyacrylamide gels and eluted in TMgN buffer (10 mM Tris–HCl pH 8.0, 1 mM MgCl₂, 50 mM NaCl).

ATP hydrolysis assays

Pif1 (10 nM) was incubated with 0.165 μM [³²P-γ]-ATP at 30°C in 15 μl of ATPase buffer (35 mM Tris–HCl, pH 7.5, 1 mM DTT, 5 mM MgCl₂ and 100 μg/ml BSA, 42 mM KCl) containing 1 μM poli-T ssDNA. Reactions were terminated at the indicated times transferring 2 μl to pre-aliquoted tubes with 2X Stop solution (2% SDS, 40 mM EDTA). Thin-layer chromatography (Z122882-25EA, Sigma-Aldrich® TLC plates) and phosphorimaging analysis to quantify ATP hydrolysis were conducted as described elsewhere (28), using a Typhoon FLA 9500 scanner and the ImageQuant software (GE Healthcare).

DNA unwinding assays

Pif1 (4 nM) was incubated with a 10-molar excess of splayed arm substrate at 30°C for 3 min in 10 μl of buffer H (35 mM Tris–HCl, pH 7.5, 1 mM DTT, 5 mM MgCl₂, 100 ng/μl BSA, 5 mM ATP, 60 mM KCl). A 100-fold excess of ssDNA trap (unlabelled A6 oligo, see Splayed Arm substrate preparation) was added to avoid reannealing of the fluorescent-labelled oligo. The reactions were terminated at the indicated times by deproteinization with 0.5% SDS and proteinase K (2 mg/ml) at 37°C for 30 min. DNA species were resolved through native 12% polyacrylamide gels in TBE buffer. Gels were imaged on a Typhoon FLA 9500 and quantified with ImageQuant software (GE Healthcare)

DNA damage sensitivity assays

Cells grown to mid-log phase were normalized to OD₆₀₀= 0.5 and 10-fold serial dilutions were spotted onto YPGal or YPD with different concentrations of hydroxyurea (H9120, US Biological). Plates were incubated for 2–3 days at 30°C and imaged in a Gel Doc XR+ (Bio-Rad).

Telomeric Southern blot

PIF1/pif1Δ diploid strains carrying an additional copy of different PIF1 alleles, ectopically integrated in ura3-52, were sporulated and dissected at the same time. After replica-plating (1st passage), strains with genotypes of interest were selected (2nd passage) and re-streaked again (3rd passage) to allow telomere elongation in Pif1-deficient strains. Starter cultures were prepared by pooling three biological replicates for each genotype. Telomere length was assayed by Southern hybridization as described (13). Genomic DNA was first purified by phenol-chloroform and ethanol precipitation. Subsequently, DNA was further purified by silica-columns. 3 μg of genomic DNA was digested with XhoI overnight. 1 μg of each sample was loaded on 1% (w/v) agarose gels (20 cm length) and separated for 18 h at 1.5 V/cm in TBE 0.5X. After electrophoresis, DNA fragments were blotted onto Zeta-Probe nylon membranes (Bio-Rad, #1620159) by the alkaline transfer method, according to manufacturer's instructions. Synthesis of the labelled probes, membrane hybridization and detection were performed using DIG-High Prime DNA Labelling and Detection Starter Kit II (Roche, #11585614910) as per the manufacturer's instructions. Membranes were imaged on a Typhoon FLA 9500. The subtelomeric Y’ probe was amplified from pDL987 (kind gift from David Lydall) using oligos OMB363 and OMB364, respectively. The CDC15 probe was synthesized using oligos OMB349 and OMB350, as previously described (29). The GeneRuler 1 kb Plus DNA Ladder (ThermoFisher, #SM1331) was employed as a molecular-weight marker.

Fluorescence microscopy

Pif1 subcellular localization was addressed by fluorescence microscopy in live yeast cells as described (30). Briefly, mid-log phase cultures in SC-Raffinose + 100 μg/ml of adenine were induced for 3 h with 2% galactose. 30 min prior to the end of induction, DAPI was added to a final concentration of 10 μg/ml. After induction, 1 ml of cells were harvested (3 min, 1500 × g) and resuspended in 100 μl of SC-Raff-Gal + 100 μg/ml adenine. 3.5 μl of the cell suspension were transferred to microscope slides with a solidified patch of 1.2% agarose. The edges of the cover slip were sealed with melted VALAP before imaging. Cells were visualized using a THUNDER Imager Tissue (Leica Microsystems) with the following parameters: refractive index 1.34 and Thunder algorithm enabled for background subtraction.

In silico sequence analyses

Potential regions encoding NLSs in Pif1 sequence were identified using the cNLS mapper software (31) with the discovery threshold set to 4.0, together with NucPred (32) and NLStradamus (33) with their default parameters. Potential mitochondrial pre-sequences were analysed with Mitofates (34), using fungal or metazoan parameters where applicable. For the prediction of secondary structures and disordered regions in Pif1, the PSIPRED package (35) -including PSIPRED 4.0 and DISOPRED3- was employed with the default parameters.

Results

The pif1-m2 allele does not fully recapitulate the phenotypes due to loss of nuclear Pif1 functions in pif1Δ mutants

The study of Pif1 roles in both nuclear and mitochondrial genome stability has taken advantage of the two classical separation-of-function mutants pif1-m1 and pif1-m2 (10). However, results from several groups (11–17) and our own suggest that pif1-m2 retains variable levels of nuclear activity. To assess the relevance of this residual nuclear activity, we compared directly the phenotypes derived from pif1-m2 and pif1Δ for some of the described genetic interactions. Importantly, in our experiments we also included the pif1-m1 mutant to clarify whether any phenotypic differences between pif1-m2 and pif1Δ arise from the residual nuclear activity in pif1-m2 or the combined effect of Pif1 loss at both nucleus and mitochondria (Figure 1). First, we tested the synthetic sickness between pif1Δ and the absence of DIA2, a member of the E3 ligase SCF complex (36) by tetrad dissection analyses of DIA2/dia2Δ heterozygous diploids in combination with PIF1, pif1Δ, pif1-m1 or pif1-m2 (Figure 1A). Neither pif1-m1 nor pif1-m2 conferred an overt synthetic sickness phenotype, as observed in pif1Δ dia2Δ double mutants, and consistently with a previous report (37). Similarly, the hypersensitivity to HU in pif1Δ rad3-102 (a hypomorphic allele of the essential Rad3 helicase involved in nucleotide excision repair) double mutants (38), could not be recapitulated with the pif1-m1 or pif1-m2 mutants (Figure 1B). We next compared how these separation-of-function alleles could suppress the lethality caused by the absence of the essential Dna2 nuclease-helicase (Figure 1C). Consistent with the literature (39), the complete absence of Pif1 provides a robust suppression of dna2Δ lethality that is not observed with the pif1-m1 allele. However, pif1-m2 dna2Δ mutants grow poorly compared to the pif1Δ dna2Δ mutants, indicative of reduced, but functionally relevant amounts of nuclear Pif1 in those cells. In this sense, replacement of the endogenous pif1-m2 promoter with the constitutive ADH1 promoter is sufficient to restore dna2Δ lethality (Figure 1C, right panel), suggesting that functional nuclear Pif1 can still be produced from the pif1-m2 mRNA. Interestingly, the synthetic lethality observed between pif1Δ and top3Δ (40) is fully recapitulated by pif1-m2 (but not pif1-m1), suggesting that nuclear levels of Pif1 in pif1-m2 cells are too low to prevent lethality (Figure 1D). In this case, the replacement of the endogenous promoter of pif1-m2 with P_ADH1 is sufficient to suppress this synthetic lethality (Figure 1D, right panel), in accordance with increased Pif1 levels being able to ameliorate the slow growth of top3Δ mutants (40). Finally, given Pif1 role in preventing aberrant telomeric expansions (10,41), we analysed telomeric length in these mutants. In agreement with the presence of functional nuclear Pif1 in pif1-m2 mutants, we recapitulated the previously observed (41) intermediate telomere length of these mutants compared to wild-type and pif1Δ strains (Figure 1E).

Figure 1.

pif1-m1 and pif1-m2 cannot disambiguate if the mitochondrial or nuclear loss of Pif1 function underlies various pif1Δ genetic interactions. (A) Neither pif1-m1 nor pif1-m2 recapitulate the synthetic sickness between pif1Δ and dia2Δ. Tetrad microdissection of diploid strains carrying heterozygous mutations for the indicated wild-type and mutant alleles of DIA2 and PIF1. Images were taken after 2 days of incubation at 30ºC. Dashed lines indicate the stitching of two images from the same plate. (B) Neither pif1-m1 nor pif1-m2 recapitulate the hypersensitivity to HU of pif1Δ and rad3-102. Ten-fold serial dilutions of strains with the indicated genotypes were plated on YPD containing different concentrations of hydroxyurea (HU) and imaged after 2 days of incubation at 30ºC. (C) pif1-m2 behaves as a hypomorphic allele for the loss of Pif1 nuclear function in the context of the suppression of dna2Δ lethality. As in (A), but employing heterozygous strains for the indicated wild-type and mutant alleles of DNA2 and PIF1. Images were taken after 3 days of incubation at 30 ºC. Right panel: overexpression of pif1-m2 is lethal in dna2Δ cells. As in the left panels but employing the constitutively overexpressed allele P_ADH1 -pif1-m2. (D) pif1-m2 recapitulates the synthetic lethality between pif1Δ and top3Δ. As in (C), but employing heterozygous strains for the indicated wild-type and mutant alleles of TOP3 and PIF1. Right panel: overexpression of pif1-m2 sustains growth of top3Δ cells. As in the left panels but employing the constitutively overexpressed allele P_ADH1 -pif1-m2. (E) Neither pif1-m1 nor pif1-m2 drive telomere extension to the extent of pif1Δ. Southern blot of XhoI-digested DNA from strains carrying the indicated PIF1 alleles using the subtelomeric Y′ probe (see Materials and Methods). Telomeric DNA can be observed as a smeared band at the bottom of the blot. A second probe targeted to CDC15 was employed as a loading control.

Altogether, these results demonstrate that the separation-of-function allele pif1-m2 still retains a certain level of nuclear activity that may be sufficient to mask some of the phenotypes observed in the complete absence of PIF1. Moreover, this also implies that in those pif1Δ interactions not recapitulated by pif1-m1 or pif1-m2, it cannot be formally disambiguated whether this is a consequence of the incomplete separation-of-function of these mutant alleles or whether such interactions require the simultaneous loss of nuclear and mitochondrial Pif1 functions.

An additional nuclear isoform of Pif1 produced by ATI is increased in pif1-m2 mutants

The mitochondrial and nuclear isoforms of Pif1 are produced through the alternative translation initiation of PIF1 mRNA at two different, in-frame AUG codons, corresponding to methionines at positions 1 and 40 of Pif1 (Figure 2A). The full-length isoform contains an N-terminal mitochondrial targeting signal (MTS) that drives its import into mitochondria (10). Once in the mitochondrial matrix, the full-length protein is proteolytically processed to yield the mature mitochondrial Pif1 (mPif1) starting at arginine 46 (42). Conversely, when translation starts at AUG⁴⁰, the MTS is excluded and the protein is imported into the nucleus (nPif1), presumably through a yet-uncharacterized nuclear localization signal (Figure 2A). This renders the mature nPif1 (∼93 kDa) just fractionally bigger than the mature mPif1 (∼92 kDa) and, therefore, their electrophoretic mobility is very similar. In PIF1 cells both isoforms form a tight doublet when analysed by western blotting, although their different migrations are more apparent in the pif1-m1 and pif1-m2 mutants (Figure 2B). We noticed the presence of a mild, but specific, third band in all PIF1 alleles and in two different genetic backgrounds, displaying faster electrophoretic mobility and being particularly enriched in pif1-m2 mutants (Figure 2B, red arrow). In these cells, this band is over 3 times more abundant than in PIF1 or pif1-m1 cells, representing around 30% of total Pif1 (Figure 2C and Supplementary Figure S2a).

Figure 2.

Alternative translation initiation of PIF1 mRNA by ribosomal leaky scanning drives the production of a novel Pif1 isoform. (A) Schematic representation of the current model for Pif1 isoforms productions by ATI. (B) Detection of a novel, fast-migrating Pif1 isoform. Protein extracts from strains expressing untagged PIF1 (-), PIF1-6flag (WT), pif1-m1-6flag (m1) and pif1-m2-6flag (m2) strains were probed with an anti-FLAG antibody to assess Pif1 expression (two different exposures shown). Black arrows indicate the tight doublet formed by nuclear (n) and mitochondrial Pif1 isoforms (m). Red arrows mark the fast-migrating isoform. Pgk1 was employed as a loading control. (C) The fast-migrating isoform of Pif1 is significantly enriched in pif1-m2 mutants. Quantification of total Pif1 or the fast-migrating isoform in PIF1-6flag, pif1-m1-6flag and pif1-m2-6flag strains from fluorescent western blots (see Supplementary Figure S2). Histograms represent mean values ± standard deviation (n = 6). Statistical analysis was performed using ANOVA Kruskal-Wallis test followed by Dunn's post hoc multiple comparison analysis. P-value represented by ns (not significant), *P< 0.05, ** P< 0.005 and *** P< 0.005. (D) Schematic representation of the 5′ region of PIF1, indicating the in-frame ATG codons before the helicase domain. (E) Simultaneous mutation of Met codons to Ile at positions 107, 113 and 127 abrogates the production of the fast-migrating isoform. Western blot analysis of Pif1 in wild-type of pif1-m2 strains harbouring the indicated single and triple Met to Ile substitutions. Arrows as in (B). (F) Optimization of the Kozak context surrounding AUG⁴⁰ in PIF1 and pif1-m1 abrogates the production of Pif1^107-859. Western blot analysis of Pif1 was carried out as in (E). The specific mutations introduced to optimize the Kozak context of start codons (green) AUG¹ and AUG⁴⁰ in each PIF1 allele are shown in orange.

We speculated that this band could arise from an additional alternative translation initiation event downstream AUG⁴⁰. There are five in-frame AUG codons between this position and the beginning of the Pif1 helicase domain, clustered in positions corresponding to the methionines 107, 113 and 127, and methionines 187 and 196 (Figure 2D). Initiation of translation from any of these positions would result in a Pif1 isoform that would retain an intact helicase domain but would lack the N-terminal region between M40 and this alternative start site. To test this hypothesis, we ectopically expressed PIF1 alleles (Supplementary Figure S2b) with single conservative methionine-to-isoleucine mutations at each of these positions and determined whether the expression of the fast-migrating isoform was affected, but no obvious changes were observed (Supplementary Figure S2c). However, since an ectopically expressed Pif1^107-859 truncation co-migrates electrophoretically with the fast-migrating isoform (Supplementary Figure S2d), we decided to focus on the M107–M113–M127 cluster for further analysis (Figure 2d). Interestingly, double Met-to-Ile substitutions in this cluster revealed that translation initiation may occur from any of these methionines, albeit M107 appears as the most prominent one (Supplementary Figure S2e). In this sense, a fainter band with further increased mobility can be observed when only M127 is maintained (Supplementary Figure S2e). Importantly, the fast-migrating band became undetectable when these three methionines were simultaneously mutated (Figure 2E). These results demonstrate the existence of, at least, a third cellular Pif1 isoform that is translated from M107 (hereinafter, Pif1^107–859) and that M113 and M127 can also function as alternative initiation points.

The nuclear Pif1 isoforms are produced through ribosomal leaky scanning

Several mechanisms of alternative translation initiation have been proposed to explain the generation of different protein isoforms from a single mRNA, including internal ribosomal entry sites, reinitiation, ribosomal shunting and ribosomal leaky scanning (43,44). Leaky scanning involves the ribosomal skipping of the first AUG codon and initiation of translation at a downstream, in-frame AUG codon (45). In this context, the efficiency of any given AUG for translation initiation depends on the sequence context around this codon, known as the Kozak sequence (46,47). Interestingly, the Kozak sequence for AUG¹ in PIF1 mRNA is a relatively poor context for initiation (Supplementary Figure S2f), consistent with a model in which ribosomes must have a high probability of skipping AUG¹ (that produces mPif1) in order to reach AUG⁴⁰ and produce nPif1. Since the Kozak context around AUG⁴⁰ is also sub-optimal, we could further extend this model and hypothesize that a fraction of ribosomes skipping AUG⁴⁰ reach AUG¹⁰⁷ and promote translation of Pif1^107-859. If true, we would predict that reducing the leakiness of ribosomal scanning over AUG⁴⁰ should decrease the translation of Pif1^107-859. Therefore, we designed optimized, but conservative, Kozak contexts around AUG¹ or AUG⁴⁰ in the PIF1 and pif1-m1 alleles and verified the expression of Pif1^107-859 by western blotting (Figure 2F). Confirming our hypothesis, no Pif1^107-859 could be detected in those cells harbouring constructs with optimized Kozak contexts at AUG⁴⁰. Also, optimization of AUG¹ in PIF1 cells produced an apparent increase in mobility of the Pif1 doublet that would be consistent with increased proportion of mPif1 over nPif1. Altogether, our data demonstrate that the main mechanism underlying the alternative translation of Pif1 isoforms is ribosomal leaky scanning and explain the specific increase of Pif1^107-859 in pif1-m2 mutants, since the mutation of AUG⁴⁰ in this allele would allow a higher proportion of scanning ribosomes to reach AUG¹⁰⁷. However, the fact that mutation of AUG¹ in pif1-m2 mutants (i.e. a pif1-m1-m2 mutant) does not overtly increase the final amounts of Pif1^107-859 (Supplementary Figure S2g) suggests that other translational regulatory mechanisms may also influence the levels of different Pif1 isoforms.

Pif1^107-859 is catalytically active and retains partial functionality in the nucleus

Pif1^107-859 lacks the MTS region and, consequently, cells expressing this allele display a petite phenotype and are unable to grow on non-fermentable carbon sources (Supplementary Figure S3a). Hence, to investigate if the cause of the incomplete separation-of-function of pif1-m2 alleles is due to the increased levels of nuclear Pif1^107-859, we first decided to assess its subcellular localization by fluorescence microscopy. For this purpose, we integrated constructs carrying PIF1-eGFP alleles under the control of the inducible GAL1 promoter in a strain with endogenous, untagged pif1-m2(to avoid the confounding effect of the petite phenotype). Upon induction, Pif1-eGFP signal could be observed both in the nucleus and mitochondria in cells expressing PIF1, while in those expressing pif1-m1 or pif1^107–859 the Pif1 staining was highly enriched in the nucleus (Figure 3A). Cells expressing pif1-m2-eGFP, in addition to the predicted mitochondrial localization, also displayed a mild Pif1 nuclear enrichment (Figure 3A), in accordance with the genetic data in this paper and the literature. We next determined whether Pif1^107-859 retained catalytic activity with respect to the wild-type protein. For this purpose, we expressed Pif1, Pif1^107-859 and the catalytically inactive Pif1^KA (41) as C-terminal His-tagged fusion proteins in E.coli (Supplementary Figure S3b) and compared their ATPase (Supplementary Figure S3c and d) and helicase (Figure 3B and Supplementary Figure S3e) activities. Our results indicate that the truncation of the N-terminal region does not impair either Pif1^107-859 ATPase or helicase activity with respect to wild-type Pif1.

Figure 3.

Pif1^107-859 is a source of catalytically active Pif1 in the nucleus. (A) Pif1^107-859 is a nuclear Pif1 isoform. Subcellular localization of Pif1 isoforms in strains overexpressing different PIF1-eGFP alleles was assessed in live cells by fluorescence microscopy. DNA was visualized with DAPI staining and cell contour by DIC microscopy. (B) Helicase activity of Pif1^107-859in vitro is comparable to wild-type Pif1. DNA unwinding activity of purified Pif1 and Pif1^107-859 was estimated as percentage of substrate dissociation in time-course assays from three replicates (see Supplementary Figure 3e for representative image) and represented as mean values ± standard deviation. (C) Overexpression of pif1^107-859 is not lethal. Ten-fold dilutions of strains with the indicated genotypes were plated on YPD or YPGal plates and imaged after 2 days of incubation at 30ºC. (D) pif1^107-859 suppresses synthetic sickness of dia2Δ pif1Δ double mutants. Tetrads from diploid strains carrying heterozygous mutations for the indicated wild-type and mutant alleles of PIF1 and DIA2 were microdissected on YPD plates and incubated at 30ºC for 2 days before imaging. (E) pif1^107-859 is not as efficient as pif1Δ in suppressing the lethality of dna2Δ. Tetrads from diploid strains carrying heterozygous mutations for the indicated wild-type and mutant alleles of PIF1 and DNA2 were microdissected on YPD plates and incubated at 30ºC for 3 days before imaging. (F) pif1^107-859 is synthetic lethal with top3Δ. Tetrads from diploid strains carrying heterozygous mutations for the indicated wild-type and mutant alleles of PIF1 and TOP3 were microdissected on YPD plates and incubated at 30 ºC for 3 days before imaging. (G) pif1^107-859 recapitulates the hypersensitivity to HU of pif1Δ in rad3-102 mutants. Ten-fold serial dilutions of strains with the indicated genotypes were plated on YPD in the presence or absence of hydroxyurea (HU) and imaged after 2 days of incubation at 30ºC.

To understand if this nuclear and catalytically proficient isoform could substitute for full-length nPif1, we first addressed the biological functionality of Pif1^107–859 using as a proxy the lethality caused by the overexpression of nuclear Pif1 (5,48,49). While the overexpression of PIF1 or pif1-m1 derived in extensive cell death, overexpression of pif1^107–859 caused no detectable reduction in cell viability (Figure 3C), despite their comparable protein levels (Supplementary Figure S3f), suggesting that the N-terminal region in Pif1 may be relevant for some Pif1 functions in vivo.

Given these results, we directly addressed whether Pif1^107–859 could be responsible for incomplete loss of Pif1 nuclear function in pif1-m2 mutants. First, tetrad dissection analyses showed that dia2Δ pif1^107-859 mutants displayed reduced synthetic sickness compared to dia2Δ pif1Δ mutants (Figure 3D), indicating that Pif1^107-859 can support cell growth similarly to Pif1 in dia2Δ mutants. Second, unlike the lethality caused by the absence of DNA2 in PIF1 and pif1-m1 strains, dna2Δ pif1^107–859mutants are able to grow, but form smaller colonies than dna2Δ pif1Δ mutants, indicating that in this context Pif1^107–859 is partially functional (Figure 3E). Third, top3Δ pif1^107-859 strains failed to grow after tetrad dissection, similarly to top3Δ pif1Δ strains, suggesting that Pif1^107-859 cannot substitute for Pif1 in these conditions (Figure 3F). Fourth, pif1^107–859 could not reduce the hypersensitivity to HU observed in rad3-102 pif1Δ mutants, unlike pif1-m1 (Figure 3G), an unexpected result that could arise due to various reasons and will be further discussed below. Taken together, our data indicate that in pif1-m2 mutants there is an increase in the translation of the Pif1^107–859 isoform, which is able to enter the nucleus and fulfil some of the functions of nuclear Pif1.

Identification of the nuclear localization signal in Pif1

The presence of alternative nuclear Pif1 isoforms may lead to the misinterpretation of genetic results obtained with the classical pif1-m2 allele. Therefore, we argued that a more robust separation-of-function allele could be attained by blocking the nuclear import of all isoforms. While it has been proposed that Pif1 contains a nuclear localization signal (NLS) (42), this had not been formally shown at the time this project was started. Therefore, we employed an in silico analysis of the Pif1 sequence to search for potential NLSs and selected four candidate regions for further analysis (Supplementary Figure S4a, see also Materials and Methods). As a first approach to test their functionality, we took advantage of the P_GAL1-PIF1 overexpression-dependent lethality to search for those regions whose deletion restored viability upon GAL1 promoter induction (Supplementary Figure S4b). We observed that deletion of two different regions (650–659 and 775–803) resulted in an increased survival, suggesting that both regions might be involved in Pif1 nuclear import. However, fluorescence microscopy of C-terminal eGFP-fusions of these PIF1 variants revealed that only deletion of the 775–803 region resulted in loss of nuclear accumulation of Pif1, concomitantly with an increase in cytoplasmic signal (Supplementary Figure S4c). These results indicate that the 775–803 region is required for the nuclear import of Pif1, while the 650–659 deletion probably results in a general loss-of function of the protein. To pinpoint the key residues in this NLS, we made substitutions to alanine in two patches of basic residues present in this region, (Figure 4A) and determined whether these mutations affected functionality of this NLS by fluorescence microscopy (Figure 4B) and suppression of overexpression-dependent lethality of P_GAL1-PIF1 (Figure 4C, D). Substitution of all nine basic aminoacids (Pif1^9A) in this region or exclusively in the first basic patch (⁷⁸¹KKRK⁷⁸⁴; Pif1^4A) impaired the function of the Pif1 NLS, as judged by the reduction of nuclear accumulation and tolerance to overexpression of these mutants. To confirm that this region contains a fully functional NLS, we fused the Pif1 775–803 region to eGFP and confirmed that it was sufficient to drive its nuclear accumulation (Figure 4E). Consistently, this was impaired by mutation to alanines in the first basic patch. These results confirm the existence of a C-terminally located NLS in Pif1 and that the ⁷⁸¹KKRK⁷⁸⁴ basic patch is essential for the nuclear import of Pif1, in line with a recent report (50).

Figure 4.

The basic patch ⁷⁸¹KKRK⁷⁸⁴ is required for Pif1 nuclear import. (A) Schematic representation of the 775–803 region of Pif1 (purple), highlighting the amino acids substituted to alanine (green/red) in the indicated mutants. (B) Disruption of the ⁷⁸¹KKRK⁷⁸⁴ basic patch abolishes Pif1 nuclear import. Subcellular localization of the indicated Pif1 variants in live yeast cells was assessed by fluorescence microscopy as in Figure 3A. (C) Disruption of the ⁷⁸¹KKRK⁷⁸⁴ basic patch prevents the lethality of Pif1 overexpression. Viability of the indicated strains upon overexpression of the indicated Pif1 variants was assessed as in Figure 3C. (D) Expression levels of various Pif1-eGFP NLS mutants is comparable. Abundance of Pif1 variants in the strains from (C) was compared by western blotting using anti-GFP antibodies. Protein extracts were prepared from liquid cultures grown in YPRaff and induced for 3 h by addition of 2% galactose. Pgk1 was used as a loading control. (E) Fusion of a wild-type 775–803 Pif1 fragment to eGFP is sufficient to drive its nuclear accumulation. The subcellular localization of the indicated eGFP-fusions was assessed as in Figure 3A.

A nuclear-null allele of Pif1 disambiguates unclear genetic interactions

Following the identification of Pif1 NLS, we decided the assess the consequences of impaired nuclear import for Pif1. For this, we employed the pif1^4A allele (Figure 4A), which hereinafter we will rename as pif1^nls (this is to avoid nomenclature overlapping with the previously described pif1^{T763A,S765A,S766A,S769A} phosphomutant allele (6), also referred to as pif1^4A). First, to exclude if the substitutions in Pif1^nls could affect its catalytical properties, we purified Pif1^nls as described for Pif1 (Supplementary Figure S3c) and compared its ATPase and helicase activities to the wild-type protein (Supplementary Figure S5a–d). In both cases, Pif1^nls displayed similar reaction kinetics to Pif1, indicating that the four mutations in the NLS do not interfere with its enzymatic activity.

We then tested the effect of NLS mutations on Pif1 genetic interactions by combining pif1^nls with the dia2Δ, dna2Δ and top3Δ mutants followed by tetrad dissection analyses. Disruption of Pif1 NLS resulted in weak, but detectable, synthetic sickness with the dia2Δ mutants (Supplementary Figure S5e). However, pif1^nls was able to both suppress dna2Δ lethality (Supplementary Figure S5f) and induce lethality in the top3Δ mutants (Supplementary Figure S5g). These results indicate that disruption of the NLS in an otherwise wild-type PIF1 allele severely compromises its nuclear functions.

Despite the NLS inactivation in pif1^nls, we were still surprised that this allele did not fully recapitulate the pif1Δ effects on the growth of dia2Δ and dna2Δ mutants (Supplementary Figure S5b and c). We hypothesized that even in pif1^nls strains, some fraction of functional Pif1 can reach the nucleus. To minimize even further this residual nuclear Pif1 activity, we combined the mutations in both the pif1^nls and pif1-m2 alleles into a new allele which we have termed pif1^mit (Figure 5A) due to its improved separation-of-function features (Figure 5B-D). Tetrad dissection analyses revealed that when combined with dia2Δ mutants, pif1^mit phenocopies pif1Δ synthetic sickness more closely than pif1-m2 (Figure 5B). Likewise, the suppression of dna2Δ lethality by pif1^mit is similar to that of pif1Δ, and more robust than for pif1-m2 mutants (Figure 5C). Moreover, pif1^mitstrains also display slightly increased telomeric length compared to pif1-m2 mutants (Figure 5D). These results indicate that pif1^mit represents an improved separation-of-function allele that translates almost exclusively into mPif1.

Figure 5.

pif1^mit constitutes an improved separation-of-function allele with respect to pif1-m2. (A) Schematic comparison of the pif1-m2 and pif1^mit alleles. (B) pif1^mit recapitulates the synthetic sickness of pif1Δ in dia2Δ strains. Tetrad microdissections of the indicated diploid strains were carried out as in Figure 1A. (C) pif1^mit suppresses the lethality of dna2Δ strains similarly to pif1Δ. Tetrad microdissections of the indicated diploid strains were carried out as in Figure 1C. (D) Strains expressing pif1^mit display extended telomeres comparable to those in pif1Δ strains. Telomeric Southern blots of the indicated strains were carried out as in Figure 1G. (E) pif1^mit does not increase the hypersensitivity to HU of rad3-102 mutants. Sensitivity assays were carried out as in Figure 1B.

Interestingly, pif1^mit expression in rad3-102 cells did not increase their sensitivity to HU with respect to PIF1 or pif1-m2 (Figure 5E). Since these PIF1 alleles retain mitochondrial Pif1, this result underscores the possibility that the hypersensitivity of pif1Δ rad3-102 mutants to HU arises from the loss of mitochondrial Pif1 activity. However, we did not detect any increased sensitivity to HU when we analysed the pif1-m1 rad3-102 mutants (Figure 1B), which should be devoid of mitochondrial Pif1. Hence, it cannot be ruled out that, akin to pif1-m2, the separation of function provided by the pif1-m1 allele may also be incomplete.

Residual mitochondrial Pif1 is present in pif1-m1 mutants

In addition to the aforementioned result (Figure 5E), during the course of our studies, we repeatedly observed an anomalous behaviour of the pif1-m1 strains (Figure 6A) with respect to their expected loss of mitochondrial function. While haploid pif1Δ strains freshly generated by tetrad dissection of PIF1/pif1Δ diploids had acquired an overt petite phenotype by the time the colony is visible, pif1-m1 colonies arising from PIF1/pif1-m1 diploids had not. To formally confirm this observation, we created diploid PIF1/pif1Δ strains with one additional ectopic copy of the pif1-m1 and analysed their meiotic progeny by tetrad dissection. pif1Δ colonies displayed a clear petite phenotype as judged by their lack of pigmentation and inability to grow in a medium lacking a fermentable carbon source (Figure 6B and C). Contrarily, when pif1Δ co-segregated with the ectopic pif1-m1 allele, normal pigmentation and the ability to grow on YPEGly were restored (Figure 6B and C), suggesting that this theoretically mitochondrial-null allele may also be subject to some sort of genetic bypass.

Figure 6.

ATI from near-cognate start codons underlies incomplete separation-of-function of the pif1-m1 allele. (A) Schematic representation of the N-terminal MTS region of pif1-m1. Start codons are depicted in green. (B) pif1-m1 strains do not display a petite phenotype after tetrad dissection. Tetrads from a diploid strain carrying a heterozygous deletion of PIF1 and an ectopic pif1-m1 allele integrated at ura3-52 were microdissected on YPD plates and incubated at 30ºC for 3 days. Adenine was omitted from YPD to allow for colour discrimination of petite colonies. (C) Non-fermentable carbon sources can sustain growth of pif1-m1 strains. Freshly generated strains harbouring the indicated PIF1 alleles were streaked on rich medium with a fermentable (YPD, 2% glucose) or non-fermentable (YPEGly, 3% ethanol, 3% glycerol) carbon source. Plates were incubated at 30ºC for 2 days and photographed. (D) Schematic representation of the N-terminal MTS region of the pif1-m1^4L, depicting start codons (green), as well as the mutations in downstream near-cognate ATGs (red). (E) pif1-m1^4L strains display a petite phenotype after tetrad dissection. As in (B), but the diploid strain harbours an ectopic copy of pif1-m1^4L. (F) Non-fermentable carbon sources cannot sustain growth of pif1-m1^4L strains. Strains carrying the indicated alleles were streaked as in (C). (G) pif1-m1^4L, pif1-m1^L9L and pif1Δ confer similar HU hypersensitivity to rad3-102 mutants. Ten-fold serial dilutions of strains with the indicated genotypes were plated on YPD containing different concentrations of hydroxyurea (HU) and imaged after 2 days of incubation at 30ºC.

While this situation could seem reminiscent of the ATI by ribosomal leaky scanning observed for pif1-m2, there are no additional in-frame start codons in the vicinity of AUG¹ that could support the translation initiation of a protein with a functional MTS. However, it is well established that ribosomes can initiate translation at near-cognate start codons when they are embedded in a favourable Kozak context or mutations in eIF2 are introduced (51–53). We identified five in-frame near-cognate codons in the region of AUG¹ (one upstream: AUC⁻⁵(Ile); four downstream: AUA⁵(Ile), UUG⁹(Leu), AUU¹²(Ile), AUA¹³(Ile)) that could give rise to Pif1 isoforms with functional MTSs (Supplementary Figure S6a and b). Therefore, we introduced a series of mutations in these near-cognate codons in a pif1-m1 construct, being as conservative as possible with respect to amino acid changes in those downstream AUG¹(Supplementary Figure S6c). While mutation of the upstream near-cognate codon (pif1-m1^I(-5)A) had a minor effect on the mitochondrial proficiency of pif1-m1 strains (Supplementary Figure S7a–c), substituting the four downstream codons to different leucine codons (pif1-m1^4L) resulted in a complete loss of Pif1 mitochondrial function (Figure 6D–F), with most of this effect arising from the single substitution of codon UUG⁹(pif1-m1^L9L) (Supplementary Figure S7d–f). The mitochondrial defects in pif1-m1^4L are not due to reduced protein levels (Supplementary Figure S6d) or disruption of Pif1 MTS, as the same substitutions in a wild-type allele (pif1^4L) do not result in a petite phenotype (Supplementary Figure S7g–i). Given its potential as a novel separation-of-function allele, the pif1-m1^4L phenotype was also confirmed introducing the corresponding mutations in the endogenous PIF1 locus (Supplementary Figure S8a–f). Altogether, these results indicate that translation initiation from near-cognate start codons in the vicinity of AUG¹ can give rise to biologically relevant amounts of mitochondrial Pif1, which, in our hands, can suppress the mitochondrial defects predicted for pif1-m1.

With the pif1-m1^4L allele in hand, we decided to test whether the hypersensitivity of pif1Δ rad3-102 mutants to HU derives from the loss of mitochondrial Pif1. In agreement with this hypothesis, the presence of the pif1-m1^4L or pif1-m1^L9L in rad3-102 mutants resulted in similar HU hypersensitivity to the pif1Δ rad3-102 strain (Figure 6G and Supplementary Figure S8g), revealing a counterintuitive mitochondrial connection between Pif1 and Rad3 in their response to replication stress. These results underscore the importance of the development of improved separation-of-function alleles to disambiguate the biological roles of this helicase in the nucleus and mitochondria.

Discussion

In this work, we have determined that ribosomal leaky scanning is the specific ATI mechanism responsible for the generation of mPif1 and nPif1. Under this model, the stringency or weakness of the Kozak context influences the probability for the scanning 43S pre-initiation complex (PICs) to engage or skip the first AUG, which could enable translation initiation from downstream AUG codons. The sub-optimal Kozak context surrounding AUG¹ in PIF1 mRNA (Supplementary Figure S2f) allows a significant fraction of the scanning PICs to bypass AUG¹, responsible for the translation of mPif1, and start from AUG⁴⁰, resulting in the translation of nPif1. However, our results demonstrate that the weak Kozak motif around AUG⁴⁰ (Supplementary Figure S2f) is permissive enough to let a minor fraction of scanning PICs traverse until the next in-frame start codon, AUG¹⁰⁷, and produce detectable amounts of the shorter Pif1^107-859 isoform even in wild-type cells (Figure 2B). Consistently with this leaky scanning model, mutations in the preceding start codon (AUG⁴⁰), as in pif1-m2 mutants, enable a higher proportion of ribosomes to reach AUG¹⁰⁷ and increase the abundance of this isoform (Figure 2B and C), while an enhanced engagement of ribosomes by AUG⁴⁰ renders this isoform undetectable (Figure 2F). Moreover, the sole mutation of AUG¹⁰⁷ is not sufficient to abrogate the generation of a fast-migrating isoform. Instead, the complete disappearance of this band requires the concomitant mutation of the AUG¹⁰⁷/AUG¹¹³/AUG¹²⁷ cluster, further substantiating the intrinsic tendency of PIF1 mRNA to undergo ATI by leaky scanning (Figure 2E). This provides a compelling explanation for the incomplete separation of function observed for the pif1-m1 and pif1-m2 alleles (Figure 7). However, in the case of pif1-m1 mutants, residual mitochondrial Pif1 exists even in the absence of any in-frame AUG codons that could enable the translation of MTS-containing isoforms. Here we demonstrate that near-cognate start codons in the vicinity of AUG¹ may serve as alternative translation initiation points for isoforms that still contain a functional MTS and render mature mPif1 (Figures 6, 7 and Supplementary Figure S6b). Indeed, such residual mPif1 can be unveiled in western blots of pif1-m1-m2 double mutants (Supplementary Figure S2g) and suffices to support mitochondrial function and growth on non-fermentable carbon sources (Figure 6C and Supplementary Figure S8c). In pif1-m2 mutants, the increased amounts of the biochemically proficient and nuclear Pif1^107-859 isoform could underlie the oft-invoked residual nuclear function of these mutants (11–17), as suggested by the similar restoration of cell growth produced by the expression of pif1^107-859 or PIF1 in dia2Δ mutants (Figure 3D). However, our results are more compatible with a scenario where the residual nuclear activity in pif1-m2 mutants derives from a combined effect between such increased levels of Pif1^107–859 and, akin to pif1-m1 mutants, the existence of other undetectable isoforms produced by translation initiation from non-AUG codons (Figure 7). This is supported by the existence of several near-cognate codons with a favourable context near AUG⁴⁰ that display an enrichment in scanning PICs by TCP-seq (Supplementary Figure S9) and by the reduced functionality of pif1^107–859 compared to PIF1 in two different scenarios: (i) pif1^107-859 overexpression does not impair cell growth (Figure 3C) and (ii) dna2Δ cells maintain viability in the presence of pif1^107–859(Figure 3E).

Figure 7.

A refined model for the translation of PIF1 mRNA. PIF1 mRNA (upper panel) undergoes ATI from at least three AUG codons (green). AUG¹ and AUG⁴⁰ are responsible for the production of the main isoforms of mPif1 and nPif1, with a small fraction of PICs reaching AUG¹⁰⁷ to generate the nuclear Pif1^107–859. Minor contributions from near-cognate start codons (^ncAUGs, blue) are not depicted here. In pif1-m1 mutants (middle panel), mutation of AUG¹ increases translation from proximal, downstream ^ncAUGs. Despite their small N-terminal truncations, these mPif1 precursors retain an operative MTS and therefore may sustain mitochondrial function. In pif1-m2 mutants (bottom panel), the absence of AUG⁴⁰ allows a higher proportion of scanning PICs to reach AUG¹⁰⁷ and increase the production of nPif1^107–859, which may support some of nPif1 functions. The potential contribution of ^ncAUGs between AUG⁴⁰ and AUG¹⁰⁷ to produce additional isoforms of nPif1 with intermediate truncations is also indicated.

Noteworthy, the Pif1 N-terminal domain (40–237) is responsible for the toxicity associated to PIF1 overexpression (54) and it is the target of several PTMs, including checkpoint- and cell-cycle dependent phosphorylation (6,7,55), as well as acetylation (5). Then, it is arguable that Pif1^107–859 might lack regulatory domains located between amino acids 40–107 that could be included in isoforms arising from non-AUG codons within that region, thus allowing them to fulfil those functions that Pif1^107–859 cannot satisfy. We have hypothesized that the 40–107 region in Pif1 may contain the motif responsible for the interaction between Pif1 and RPA (56) which could contribute to Pif1 recruitment to chromatin and, indeed, the existence of such RPA-binding motif has been recently confirmed (57). Under this new scenario, it is tempting to speculate that the ability to produce a catalytically active version of Pif1 that is devoid of this RPA-binding motif might help in fine-tuning the cellular response to specific types of genotoxic stress. Analogous explanations have been invoked to rationalise the existence of a DNA damaged-induced splicing isoform of human PIF1 (vPIF1) (58) or the truncated form of human MUS81 nuclease (MUS81_short) that cannot interact with SLX4 scaffold (59).

It is important to underscore that our results cannot fully rule out that other regulatory mechanisms may be contributing to the balance between Pif1 isoforms. When comparing PIF1 versus pif1-m1, or pif1-m2 versus pif1-m1-m2 strains, no obvious increase in the expected nuclear isoform (nPif1 or Pif1^107-859, respectively) is observed (Supplementary Figure S2g). This would imply that the number of scanning 43S PICs that reach the region of codon 40 is relatively similar regardless of the presence or absence of AUG¹. One possibility could be that other out-of-frame, non-AUG codons between AUG¹ and AUG⁴⁰ are engaging a considerable fraction of scanning PICs and leading to translation initiation of short, missense peptides. Other mechanisms involving termination and reinitiation (60) after AUG⁴⁰ or ribosomal shunting (61) could further contribute to the regulation of start codon selection. Additionally, transcriptional effects could also influence the balance of Pif1 isoforms, as cryptic transcription start sites in the region of PIF1 codon 90, which could potentially give rise to Pif1^107–859, have been reported in the YeasTSS database (http://www.yeastss.org; (62)) under various stress conditions, including osmotic, thermal and genotoxic stress. However, it is unlikely that under normal growth conditions this alternative TSS may influences the production of mPif1, nPif1 or Pif1^107–859.

Considering all this plasticity for translation initiation of PIF1 mRNA, we have built on the pif1-m1 and pif1-m2 alleles to minimize their residual mitochondrial and nuclear functions, respectively. In this sense, a recent report has employed a similar rationale to ours in order to minimize the nuclear levels of Pif1 by independently identifying and disrupting the so-far uncharacterized Pif1 NLS (50). Both works identify the ⁷⁸¹KKRK⁷⁸⁴ basic patch as required for Pif1 nuclear functions, while dispensable for its mitochondrial roles, by either deleting these residues (50) or mutating them to alanines (Figure 5 and Supplementary Figure S5). The involvement of this sequence in nuclear import was also independently confirmed either by suppression of the pif1-NLSΔ mutant phenotypes by fusing it to an SV40 NLS (50) or directly by fluorescence microscopy, showing: (i) loss of Pif1-eGFP nuclear accumulation when this patch is mutated (Figure 4B) and (ii) nuclear accumulation of eGFP when fused to the Pif1 fragment containing this basic patch (Figure 4E). Moreover, our results also confirm that nuclear phenotypes in an NLS-deficient PIF1 mutant do not arise from reduced catalytic activity of the helicase (Supplementary Figure S5). The impairment of Pif1 NLS does not suffice to completely abolish Pif1 nuclear functions (Supplementary Figure S5). Potential reasons behind this observation, including diffusion through the nuclear pore, a secondary NLS in Pif1 or association to a carrier protein, have been elegantly laid out by Lee et al. (50) and will not be further discussed here. Given these results, both groups have resorted to the impairment of the NLS in pif1-m2 to further diminish nuclear Pif1 levels. Therefore, both pif1-m2-NLSΔ (50) and pif1^mit (Figure 5) represent the best alleles to date to abrogate nuclear Pif1 functions without affecting its mitochondrial activities. Additionally, we have combined mutations in near-cognate start codons with pif1-m1 to produce the improved mitochondrial-deficient allele pif1-m1^4L(Figure 6D–F and Supplementary Figure S8). We envisage that the combination of pif1^mit and pif1-m1^4L(= pif1^nuc) will enable the discovery of subtle genetic interactions that could have been previously missed due to the residual activities of pif1-m1 and pif1-m2 or the confounding effect of pif1Δ pleiotropy. As a proof of principle, we have disambiguated the origin of the hypersensitivity of pif1Δ rad3-102 mutants to HU (38), which, unexpectedly, arises from the loss of Pif1 mitochondrial activity, rather than nuclear (Figure 6G). Therefore, these alleles may contribute to establish more controlled genetic set-ups for the study of Pif1 in the replication and repair of the mitochondrial and nuclear genomes.

This plasticity in the translation initiation step has important implications for the design of separation-of-function alleles that are likely applicable to other dual-targeted DNA repair factors or, more generally, to any other genes where ATI is responsible for the generation of functionally different isoforms. Akin to the problematic described for S. cerevisiae, in S. pombe, the PIF1 ortholog pfh1 also generates essential nuclear and mitochondrial isoforms through ATI, but nuclear-deficient alleles of pfh1 (mutated for ATG²¹, the start codon for nuclear Pfh1) can sustain cell division indefinitely (Pinter et al., 2008), suggesting an additional source of nuclear Pfh1. In fact, the combined mutation of ATG²¹ and the next three in-frame, downstream ATGs, does not suffice to recapitulate the pfh1⁻ phenotype (63). This strongly suggests that near-cognate start codons, like in PIF1, could also enable ATI of nuclear Pfh1 isoforms that become phenotypically relevant in the absence of the main start codons (64).

It is intuitive to expand the potential impact of ATI for the study of other dual-targeted genes in budding and fission yeasts. For instance, the conserved ligase I (CDC9 in S. cerevisiae / cdc17 in S. pombe) also presents nuclear and mitochondrial isoforms derived from ATI at two in-frame start codons flanking its MTS (65). The phenotypes exhibited by their separation-of-function alleles again suggest that individual mutations of the first and second start codons do not completely abolish the production of mitochondrial or nuclear isoforms. In budding yeast, mutation of the CDC9 second ATG, which should abrogate the expression of the nuclear isoform, still yields viable cells and, like in pif1-m2, correlates with the appearance of a fast-migrating isoform (65). In S. pombe, the equivalent mutation of the second ATG (ATG²⁰) in cdc17 failed to completely abolish nuclear function unless the NLS was also mutated (66). Given that no in-frame start codons exist before this NLS, the most likely explanation for the residual nuclear protein upon mutation of ATG²⁰ would be ATI events from two near-cognates at positions 27 and 32 (66).

In human cells, ATI by ribosomal leaky scanning and near-cognate start codon usage also contribute to the generation of protein variants, frequently in combination with alternative splicing, thus leading to similar scenarios to that of Pif1 in budding yeast. For instance, the mRNAs of disease-linked DNA repair factors TOP3α (67) and RNAseH1 (68) undergo ribosomal leaky scanning for their dual targeting to nucleus and mitochondria. Both genes present downstream ATGs and near-cognates with strong Kozak contexts that could sustain a relevant amount of translation initiation in mutants of the primary start codons. Consistently, in order to achieve the exclusive expression of the mitochondrial isoform of human TOP3α, the mutation of both its second ATG (which drives the translation of the nuclear isoform) and its NLS is required (69). Incomplete separation-of-function mutants for dual-targeted DNA repair factors have been also described for human PIF1 (70) and ADAR1 (71,72). Recently, the striking difference between the complete inactivation of the APC/C complex coactivator CDC20 and its first ATG codon has also been highlighted. Unlike the full-length protein, the N-terminally truncated isoforms produced by ribosomal leaky scanning that specifically increase after mutation of the first CDC20 ATG codon cannot be inhibited by the spindle-assembly checkpoint and therefore allow progression into mitosis despite its activation (73).

Cumulative evidence from genome-wide analyses and bioinformatic predictions indicate that thousands of genes may undergo ATI events from both in-frame AUG and near-cognate start codons. This enriches the diversity of the cellular proteome by producing extended or truncated protein isoforms, with the potential gain or loss of functional domains in physiological and pathological situations (74–82). Given this complexity, genetic approaches mutating the main start codons to understand the function of specific protein isoforms must consider the potential contribution of ATI events to the protein pool.

Data availability

The data underlying this article are available in the article and in its online supplementary data.

Supplementary data

Supplementary Data are available at NAR Online.

Acknowledgements

The authors would like to thank Virginia Zakian and Rodrigo Bermejo for various yeast strains, Joao Matos and David Lydall for plasmids, and Maria Crugeiras for the critical reading of the manuscript and technical support.

Author contributions: Tomas Lama-Diaz: Conceptualization, Formal analysis, Methodology, Investigation, Validation, Visualization, Writing—original draft, Writing—review & editing. Miguel G. Blanco: Conceptualization, Funding acquisition, Writing—original draft, Writing—review & editing.

Funding

The Blanco lab was supported from Ministerio de Ciencia e Innovación/Agencia Estatal de Investigación/10.13039/501100011033 [PID2020-115472GB-I00 to M.G.B.]; Xunta de Galicia/FEDER ‘Una manera de hacer Europa’ [ED431C 2019/013 to M.G.B.]; CIMUS receives financial support from the Xunta de Galicia/FEDER [ED431G 2019/02, Centro Singular de Investigación de Galicia, accreditation 2019–2022]; T.L.-D. was a recipient of a pre-doctoral fellowship from Xunta de Galicia [ED481A-2018/042]. Funding for open access charge: Ministerio de Ciencia e Innovación / Agencia Estatal de Investigación [PID2020-115472GB-I00].

Conflict of interest statement. None declared.