The yeast phosphofructokinase β-subunit has RNA unwinding activity and modulates cell cycle progression
Waleed S Albihlal, Ana M Matia-González, Tobias Schmidt, Wieland Mayer, Joe Bryant, Martin Bushell, Dierk Niessing, Matteo Barberis, Alexander Schmidt, Jürgen J Heinisch, André P Gerber
TL;DR
A subunit of a glycolytic enzyme in yeast can unwind RNA and influence cell cycle progression, linking energy metabolism to cell proliferation.
Contribution
Discovery of RNA unwinding activity in the Pfk2 subunit of yeast phosphofructokinase and its role in cell cycle regulation.
Findings
Pfk2p unwinds RNA in a 5′–3′ direction and binds specific RNA motifs in cell cycle-related mRNAs.
Pfk2p promotes translation of cell cycle genes and associates with ribosomes.
Loss of Pfk2p causes increased cell size and delayed G1/S phase transition independent of glycolytic activity.
Abstract
Phosphofructokinase (PFK) is a rate-limiting glycolytic enzyme that also possesses an unexplored RNA binding activity. Here, we show that the α- and β-subunits of yeast PFK, encoded by PFK1 and PFK2, respectively, bind hundreds of functionally related messenger RNAs (mRNAs) in cells, including one’s coding for proteins involved in the regulation of mitotic cell cycle. Both Pfk1p and Pkf2p directly bind to short GA-, UC-, AU-, and U-rich motifs overrepresented in their mRNA targets. Strikingly, Pfk2p displays directional 5′–3′ double-stranded RNA unwinding activity not seen with Pfk1p. Furthermore, Pfk2p dynamically associates with ribosomes and promotes translation of cell cycle genes. Consequently, pfk2∆, but not pfk1∆, mutants display increased cell sizes and severely delayed G1/S phase transition, independent of the enzyme’s glycolytic activity. Our results uncovered a hidden…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —Biotechnology and Biological Sciences Research Council10.13039/501100000268
- —Royal Society Wolfson Research Merit Award
- —Cancer Research UK10.13039/501100000289
- —BBSRC10.13039/501100000268
- —EPSRC10.13039/501100000266
- —UK Research and Innovation10.13039/100014013
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCancer, Hypoxia, and Metabolism · RNA modifications and cancer · Bacterial Genetics and Biotechnology
Introduction
RNA-binding proteins (RBPs) play vital roles in the post-transcriptional control of gene expression [1, 2]. The development of various proteomics approaches during the last decade has greatly expanded the repertoire of eukaryotic RBPs. Besides “conventional” RBPs, which possess one or more well-characterized or predicted RNA-binding domains (RBDs), hundreds of “unconventional” RBPs lacking defined RBDs but bearing other well-established cellular functions such as metabolic enzymes were identified [1–3]. For instance, several early studies suggested that a number of glycolytic enzymes act as RBPs [3, 4]. More recently, crosslinking-assisted RNA interactome capture (RIC) studies revealed that most, if not all, glycolytic enzymes can interact with polyadenylated [poly(A)] RNA in the yeast Saccharomyces cerevisiae and in other eukaryotes, suggesting conserved functions with consequences for the bound RNAs, the interacting enzyme and/or for both together [3, 5–8].
Glycolysis is an ancient and evolutionarily conserved energy generating pathway breaking down glucose (glc) into pyruvate with net production of two ATP molecules and two molecules of the essential cofactor NADH. Phosphofructokinase (PFK) is the ‘gate keeper’ of glycolysis and catalyses the irreversible conversion of fructose-6-phosphate (F6P) and ATP to fructose-1,6-bisphosphate (F1,6BP) and ADP [9]. Importantly, PFK is highly allosterically regulated allowing for fine-tuning of the glycolytic flux and adjusting cellular energy demands. It is activated by AMP/ADP and fructose 2,6-biphosphate (F2,6BP), and inhibited by high levels of ATP, citrate, and other metabolites [10, 11]. This triggers different conformational states: molecular signals of low cellular energy (ADP/AMP) promote an enzymatically active R-state, whereas high energy signals (ATP) promote the enzymatically inactive T-state [10–12]. Each subunit of eukaryotic Pfks consists of two homologous halves which probably evolved by gene duplication of a potential bacterial ancestor. The N-terminal half retained substrate and ATP binding sites with catalytic functions, while the C-terminal half bears effector-binding sites for allosteric regulation [12–14].
In the yeast S. cerevisiae, PFK is comprised of two paralogous subunits, termed α (Pfk1p) and β (Pfk2p) that can form a heterooctameric complex [13–18]. While both yeast Pfk subunits cooperate in glycolysis, early mutational studies suggested additional, and possibly distinct, functions for both paralogs. For instance, deletion of PFK2 leads to significantly slower cell growth, while deleting PFK1 shows no noticeable effect on growth, despite the commonly observed lack of in vitro catalytic activity in the respective deletions or catalytic mutants [19–21]. Interestingly, a recent report suggested that the yeast Pfk subunits besides eight other glycolytic enzymes, assemble into cytoplasmic granules called glycolytic bodies (G-bodies) under hypoxic stress conditions. There they associated with numerous RNAs, including messenger RNAs (mRNAs) coding for glycolytic enzymes [7]. Moreover, photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP) followed by deep sequencing suggested that Pfk2p binds to numerous mRNAs that contain AU/U-rich elements at binding sites preferentially located in 3′UTRs [7]. This finding further confirmed the RNA binding activity of Pfk2p although its functional relevance remains unclear [7].
Here we describe a novel function of the yeast Pfk2p in RNA metabolism under normal physiological growth conditions which expands its role in glycolysis. We found that Pfk proteins bind to hundreds of mRNAs coding for functionally related proteins including mitotic cell cycle regulators, in a glucose dependent manner in vivo. The bound mRNAs contain distinct repetitive RNA triplet motifs preferentially located either in CDS or 3′UTRs, to which purified recombinant Pfks binds. Intriguingly, Pfk2p exhibits RNA unwinding activity with high 5′–3′ polarity, and it supports translation of cell cycle gene transcripts such as cyclin 3 (CLN3) and budding uninhibited benzimidazole 3 (BUB3). Consistently, pfk2∆, but not pfk1∆, mutant cells are larger and display delayed progression into S phase of the cell cycle which is independent of the enzyme’s catalytic activity. In conclusion, our results suggest that Pfk2p has the potential to act as a molecular relay/switch that controls cell cycle progression by promoting the energy-dependent translation of cell cycle regulators.
Materials and methods
Reagents
Oligonucleotides: DNA and RNA nucleotides are displayed in Supplementary Table S5.
Enzymes: AcTEV™ protease (Invitrogen, 12575-023), Proteinase K (Invitrogen, AM2546), Trypsin (Promega, V5113), T7 RNA polymerase (NEB, M0251), RNase A (Thermo Fisher, EN0531), RNasin (Promega, N2511), SUPERase•In™ RNase Inhibitor (Invitrogen, AM2696), aldolase (Sigma, A8811), glycerophosphate dehydrogenase-triosephosphate isomerase (Sigma, G1881).
Antibodies: peroxidase anti-peroxidase complex (1:5000; Sigma, P1291) for detection of the TAP-tag; mouse anti-Pab1 (1:5000; Antibodies online, ABIN1580454); mouse anti-Act1 (1:2500; MP Biomedicals, 08691002), rabbit anti-PFK1 (1:20 000 [22], rabbit anti-Rpl35 (1:5000 [23]), rabbit anti-Scp160 (1:10 000 [23]), mouse anti-V5 Tag (SV5-Pk1) (1:2000; Thermo Fisher, R960-25). Secondary antibodies: horseradish peroxidase (HRP)-conjugated donkey anti-rabbit IgG (1:5000; Cytiva, NA9340V), HRP-conjugated sheep anti-mouse IgG (1:5000; Cytiva, NXA931).
Kits (in order of appearance): Q5^®^ Hot Start High-Fidelity 2× Master Mix (New England Biolabs, M0494S), QIAquick polymerase chain reaction (PCR) Purification Kit (Qiagen, 28104), Kinase-Ligase-DpnI (KLD) enzyme mix (New England Biolabs, M0554S), LR Clonase II plus kit (Thermo Fisher, 12538120), Q5 site-directed Mutagenesis kit (NEB, E0554S), Dynabeads™ mRNA DIRECT™ Purification kit (Thermo Fisher, 61011), Pierce™ bicinchoninic acid (BCA) Protein Assay Kit (Thermo Fisher, 23227), TURBO DNA-free kit (Thermo Fisher, AM1907), Ribo-Zero Gold ribosomal RNA (rRNA) Removal Kit (Illumina, MRZY1306), SensiFast complementary DNA (cDNA) synthesis kit (Meridian Bioscience, BIO-65053), SensiFast SYBR lo-ROX mix (Meridian Bioscience, BIO-94005), NEBNext^®^ Ultra II Non-Directional RNA Second Strand Synthesis Module (New England Biolabs, E6111S), Bioanalyzer DNA high sensitivity kit (Agilent, 5067-4626); JetSeq DNA library preparation kit (Meridian Bioscience, BIO-68025), MyTaq Red Mix (Meridian Bioscience, BIO-25044), TMTpro 16-plex Label Reagent Set (Thermo Fisher, A44520), Biotin RNA Labelling Mix (Roche, 11685597910), Quick RNA Microprep kit (Zymo Research, R1050).
Nonstandard chemicals: AMPPNP (Merck, 10102547001), 1 × cOmplete™ ULTRA Tablets, Mini, ethylenediaminetetraacetic acid (EDTA) free (Roche, 05892791001); polyadenylic acid [poly(A)] (Sigma, 9403), formaldehyde (Thermo Fisher, 28908), Pan mouse IgG Dynabeads (Invitrogen, 11041), linear polyacrylamide (Invitrogen, AM9520); cycloheximide (CHX; Sigma, C1988), Trizol (Thermo Fisher, AM9738), streptavidin M280 Dynabeads (Invitrogen, 11205D), α-factor (Zymo Research, Y1001).
Biological resources
Saccharomyces cerevisiae strains and media: Haploid BY4741, BY4742, and diploid BY4743 wild-type (wt) strains, the BY4741 derived PFK1 (YGR240c), PFK2 (YMR205c) and MAP1 (YLR244c) gene deletion strains [24], as well as tandem-affinity purification (TAP)-tagged strains [25] were obtained from Open Biosystems/EUROSCARF. The BY4739 derived CLN3 (YAL040c) gene deletion strain was obtained from Horizon/Dharmacon (YSC6272-201917649). All gene deletions/modifications were verified with PCR. Heterozygous PFK1:TAP, PFK2:TAP diploid strains were generated by mating the BY4741 (MATa) derived PFK1:TAP and PFK2:TAP haploid cells with wt MATα (BY4742) cells in yeast–peptone–dextrose (YPD; 1% yeast extract, 2% peptone, 2% D-glucose) media [26]. Heterozygous diploids were selected on synthetic complete media lacking histidine (SC-His) and confirmed by PCR [27]. Unless otherwise stated, cells were grown in YPD media at 30°C with constant orbital shaking at 220 r.p.m to mid-log phase [optical density (OD) at 600 nm ∼ 0.6].
Nonstandard bacterial strain: ArticExpress (DE3) RIL competent Escherichia coli cells (Agilent, 230193).
Web sites/data base referencing
AgriGO v2.0 database (https://systemsbiology.cau.edu.cn/agriGOv2) [28]; BioMart (https://www.ensembl.org/info/data/biomart/index.html) [29]; MEME suite version 5.5.1 (https://meme-suite.org/meme/) [30]; Saccharomyces Genome Database (https://www.yeastgenome.org/) [31].
Plasmids and cloning
CEN/ARS based plasmids for expression of PFK wild-type genes [YCplac111:PFK1 (pJJH2101); YCplac111:PFK2 (pJJH2105)] and those carrying mutations in the Pfk2p catalytic Mg-ATP binding site [YCplac111:PFK2^D301T^ (pJJH2504)] and at a proton acceptor for F6P [YCplac111:PFK2^D348S^ (pJJH2892)] are based on previously described clones [20]. YCplac111 was further used as empty plasmid control (Pfk2E).The PFK2 double mutant (DM) at both active sites was constructed by in vivo recombination, yielding YCplac111:PFK2^D301TD348S^ (pJJH3063). Therefore, the EcoRI site in the polylinker of pJJH2105 was first removed, leading to plasmid pJJH2912. The EcoRI digested and dephosphorylated plasmid (pJJH2912) was then co-transformed with a 786 bps DNA fragment (Thermo Fisher’s “Gene Art” sting synthesis) covering the two mutation sites and overlapping the two internal EcoRI sites. Positive clones were selected on SC-Leu, and plasmids recovered from yeast and transformed into E. coli for further propagation and validation by sequencing.
For construction and expression of C-terminally tagged 3 × V5 PFK2 fusion proteins, the coding sequence (CDS) of wild-type PFK2 (pJJH2105), PFK2^D301T^ (pJJH2504), PFK2^D348S^ (pJJH2892), and PFK2^DM^ (pJJH3063) plasmids were amplified by PCR with primers PFK2_V5_F and PFK2_V5_R that anneal upstream and downstream of the PFK2 stop codon and contain parts of the 3 × V5-tag sequence, respectively (Supplementary Table S5). Whole plasmid sequences were PCR amplified with Q5^®^ Hot Start High-Fidelity 2 × Master Mix (New England Biolabs, M0494S) for 25 cycles, and products cleaned up with the QIAquick PCR Purification Kit (Qiagen, 28104). PCR products were phosphorylated, self-ligated and treated with DpnI with the Kinase-Ligase-DpnI (KLD) enzyme mix (New England Biolabs, M0554S) and transformed into E. coli for propagation. All plasmids were sequenced and finally transformed into the pfk2Δ strain. The expression of Pfk2-3 × V5 fusion proteins was confirmed by immunoblot analysis with anti-V5 antibodies.
To generate E. coli expression constructs, the CDS of PFK2 and PFK1 were subcloned from the respective pBG1805-PFK1/2 plasmids [32] into pDONR221 (Thermo Fisher, 12536017) by Gateway cloning. The CDSs were then subcloned into the pDEST17 gateway vector (Thermo Fisher, 11803012) using LR clonase II kit (Thermo Fisher, 12538120) to generate pDEST17-Pfk1 and pDEST17-Pfk2 expression vectors. A corrected stop-codon at the end of the CDS of PFK1 and PFK2 was inserted using the Q5^®^ site-directed Mutagenesis Kit (NEB, E0554S) using primers designed with NEBaseChanger according to manufacturer’s instructions. All constructs were validated by sequencing.
RNA interactome capture
Wt (BY4741), PFK1:TAP, PFK2:TAP, PGK1:TAP, pfk1Δ, and pfk2Δ cells were grown in YPD at 30°C to mid-log phase (OD_600 _∼ 0.6) and collected by centrifugation. Crosslinking, cell collection and lysis were essentially performed as described with our formaldehyde-assisted crosslinking RNA-protein immunoprecipitation (fRIP) method, except for initial Pfk RIC experiments (Fig. 1A) that applied ultraviolet (UV)-irradiation of cells for crosslinking as described [33]. Essentially, 300 µl of oligo[dT]25 coupled beads (Dynabeads™ mRNA DIRECT™ Purification Kit, Thermo Fisher, 61011) were equilibrated four times with 1 ml lysis buffer (20 mM Tris–HCl, pH 7.5, 150 mM NaCl, 500 mM LiCl, 2 mM EDTA, 5 mM dithiothreitol (DTT), 0.2% sodium dodecyl sulphate (SDS), 1% Triton X-100, 1 × cOmplete™ ULTRA Tablets, Mini, EDTA-free, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 0.2 U ml^−1^ SUPERase•In™). Three milligrams of total protein extracts were mixed with beads and incubated at 25°C for 10 min in a Thermomixer (peqlab) at 1000 r.p.m. For competition experiments, 20 µg of poly(A) (Sigma, P9403) was added to extracts (Sigma, P9403) Beads were washed once with 1 ml of wash buffer A (20 mM Tris–HCl, pH 7.5, 600 mM LiCl, 0.5 mM EDTA, 0.1% Triton X-100, 1 × cOmplete™ ULTRA Tablets, Mini, EDTA-free,1 mM PMSF and 0.2 U ml^−1^ SUPERase•In™), and twice with 1 ml of wash buffer B (20 mM Tris–HCl, pH 7.5, 600 mM LiCl, 0.5 mM EDTA, 1 × cOmplete™ ULTRA Tablets, Mini, EDTA-free, 1 mM PMSF, and 0.2 U ml^−1^ SUPERase•In™) each for 15 s. Beads were resuspended in 40 µl of 10 mM Tris–HCl (pH 7.5) and incubated at 85°C for 5 min. Eluates were collected and stored at –80°C until use.
Pfk1p and Pfk2p bind to functionally related RNAs in vivo. (A) Immunoblot analysis of RIC eluates. Top: Pfk1:TAP and Pfk2:TAP detected with Peroxidase-Anti-Peroxidase Soluble Complex (PAP) reagent. Middle: endogenous Pfk1p and Pfk2p proteins detected with Pfk antibodies; Tal1:TAP is as unconventional RBP used as positive control [5]. Bottom: Endogenous Pfk1p and Pfk2p detected in pfk2Δ and pfk1∆ cells, respectively. Scp160p is an RBP control; Act1p is non-RBP negative control. Input refers to cell extracts; poly(A) refers to the addition of excess competitor polyadenylic acids. (B) Heatmap representation of the abundance of 1249 fRIP selected Pfk1p and Pfk2p RNA targets. Columns refer to independent Pfk1:TAP, Pfk2:TAP, and untagged mock control fRIPs; rows denote individual transcripts. The white-blue colour bar represents normalized read counts for respective transcripts. (C) Venn diagram showing overlap of Pfk1p and Pfk2p mRNA targets. The P-value (hypergeometric test) relates to the significance of overlap. (D) Overrepresented GO terms among 671 common Pfk1p and Pfk2p RNA targets. The circle diameter is proportional to the number of RNA targets, the colour refers to the −log10 FDR. X-axis specifies the enrichment score. GO categories: BP, biological process; MF, molecular function. (E) Agarose gel showing products from reverse transcriptase-polymerase chain reaction (RT-PCR) reactions for detection of cell cycle related mRNA targets in fRIP eluates as marked in panel (B). Actin (ACT1) is a negative control, PFK2 mRNA was previously shown to associate with Pfk2p [5].
Formaldehyde-assisted crosslinking RNA-binding protein immunoprecipitation (fRIP)
Heterozygous PFK1:TAP, PFK2:TAP, and wild-type (BY4743; mock control) diploid cells were grown in 400 ml of YPD at 30°C to mid-log phase (OD_600_∼0.6). RNAs and proteins were crosslinked by the addition of 0.3% formaldehyde (v/v) (Thermo Fisher, 28908) to the culture for 10 min at room temperature with vigorous shaking of the culture every 2 min. The crosslinking reaction was subsequently quenched by the addition of 125 mM glycine (pH 7.0) for 5 min at room temperature and constant agitation of cells. Cells were harvested by centrifugation at 2000 × g for 5 min at room temperature, washed twice with ice-cold phosphate buffered saline (PBS) and snap-frozen in liquid nitrogen. Frozen cell pellets were cryogenically lysed by grinding in a mortar filled with liquid nitrogen and resuspended in 3.5 ml of ice-cold fRIP lysis buffer [20 mM Tris–HCl, pH 7.5, 150 mM NaCl, 500 mM LiCl, 2 mM EDTA, pH 8.0, 5 mM DTT, 0.2% SDS, 1% Triton X-100, 1 × cOmplete™ ULTRA Tablets, Mini, EDTA-free; 1 mM PMSF, and 0.2 Uml^−1^ SUPERase•In™]. Cell lysates were clarified by four subsequent centrifugations at 12 000 × g for 10 min at 4°C. Total protein concentrations of extracts were determined with the Pierce™ BCA protein assay kit using bovine serum albumin (BSA) as a reference standard (Thermo Fisher, 23227). Total RNA was isolated by phenol/chloroform extraction and precipitated with ethanol.
The crosslinked TAP-tagged proteins were captured from the extract as follows: 350 µl of Pan mouse IgG Dynabeads (Invitrogen, 11041) were equilibrated in 1 ml fRIP lysis buffer at 4°C. Then, 2.5 mg of cell extract was then added to the beads and mixed on rotator for 3 h at 4°C. Beads were collected with a magnet and supernatants kept for monitoring the IP efficiency by western blot. Beads were thoroughly washed five times for 1 min with 1 ml of high salt buffer [20 mM Tris–HCl (pH 7.5), 500 mM NaCl, 500 mM LiCl, 2 mM EDTA, 0.2% SDS, 1% IGEPAL, 1 × cOmplete™ ULTRA Tablets, Mini, EDTA-free, 1 mM PMSF, 0.2 U/ml SUPERase•In™]; and a further five times with 1 ml of 1 × AcTEV reaction buffer [50 mM Tris–HCl (pH 8.0), 0.5 mM EDTA (pH 8.0), 5 mM DTT, and 0.2 U/ml SUPERase•In™]. Ten microlitres (∼3%) of the washed beads were collected for monitoring IP efficiency. To elute Pfk–RNA complexes, the beads were resuspended in 400 µl of 1 × AcTEV reaction buffer containing 50 U of AcTEV protease (Invitrogen, 12575–023) and 0.2 U/ml SUPERase•In™ and incubated for 2 h at room temperature. The beads were collected with a magnet and the supernatant (= eluates) transferred to a new microtube.
To reverse crosslink and digest proteins, 400 µg of Proteinase K (Invitrogen, AM2546) was added to the eluate supplemented with 250 mM NaCl and 0.5% SDS, and incubated for 2 h at 65°C. RNA was isolated by acidic phenol/chloroform extraction and precipitated with ethanol and with addition of 3 µg of linear polyacrylamide (Invitrogen, AM9520) for at least 1 h at –80°C. The RNA pellet was washed twice with 70% ethanol, resuspended in 20 µl 1 × DNase reaction mix containing 1 µl TURBO DNase (Invitrogen, AM1907), and incubated at 37°C for 30 min to digest residual genomic DNA. RNA was finally ethanol-precipitated and resuspended in 20 µl of nuclease-free water.
RT-PCR and sequencing library preparation
For fRIP experiments, 15 μl (75%) of fRIP-RNA and 250 ng of rRNA depleted total RNA (Ribo-Zero Gold rRNA Removal Kit; Illumina, MRZY1306) were used for reverse-transcription with the SensiFast cDNA synthesis kit containing a mixture of oligo-dT and random hexamers primers (Meridian Bioscience, BIO-65053). The cDNA was further converted to double-stranded cDNA (dscDNA) using NEBNext^®^ Ultra II Non-Directional RNA Second Strand Synthesis Module as described in the manufacturer’s guide (New England Biolabs, E6111S). DscDNA was sheared to 150 bp fragments by sonication (Covaris S220) with the following settings: Peak incident power: 175 W, Duty Factor: 10%, Cycle per burst: 200 and treatment time: 600 s (10 min). Size distribution of dscDNA was validated with the Bioanalyzer DNA high sensitivity kit (Agilent, 5067-4626). Sequencing libraries were constructed using JetSeq DNA library preparation kit (Meridian Bioscience, BIO-68025). DNA was sequenced at the Bauer Core Facility (Harvard University, USA) using one lane of HiSeq2000 flow cell with 50 bp single-end reads with a minimum depth of 5 mio reads per library.
To validate selected mRNA targets from independent fRIP experiments, reverse transcription (RT) was performed as described above with 75% of the IPed RNA and corresponding total RNA. Twenty microlitre PCR reactions were performed with 1 μl (5% vol.) of cDNA, 0.5 µM of each primer (Supplementary Table S5) and 1 × MyTaq Red Mix (Meridian Bioscience, BIO-25044). The following temperature profile was applied: 5 min at 95°C; 35 cycles of the sequence [95°C for 30 s, 58°C for 30 s, 72°C for 30 s], and 7 min at 72°C. The products were resolved and visualized on a 2% agarose gel.
fRIP-seq data analysis
The quality of reads of the fRIP-seq and RNA-seq data were checked with fastqc v0.12.1 (Babraham institute; https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Low quality reads were trimmed using FastX toolkit v0.0.14 (Hannon’s lab; http://hannonlab.cshl.edu/fastx_toolkit/) with minimum phred33 score of 20 and minimum length threshold of 20 bp. Total RNA, Pfk1 and Pfk2 fRIP and mock IP reads were mapped to the yeast reference genome R64-2-1 with GMAP/GSNAP [34] allowing a maximum of 3 mismatches and output in BAM format. BAM files were sorted and indexed using SAMtools [35]. Mapping statistics were checked with SAMtools subcommand “flagstat”. Unmapped and low mapping quality reads were removed from BAM files using SAMtools view on the basis of MAPQ score. Mapped reads were counted on all genomic features using BEDtools subcommand “multicov” [36]. Read counts transformed using the variance stabilizing transformation method (VST) [37]. VST-transformed total RNA, fRIP, and Mock IP were normalized using trimmed means of M-values (TMM) with limma algorithm [38]. The Mock control normalized reads were subtracted from fRIP-seq normalized reads of the log_2_ transformed data [log_2_ (fRIP) – log_2_ (mock IP)]. P-values were calculated using moderated t-statistic test and adjusted with Benjamini–Hochberg to obtain false discovery rates (FDR). Significantly enriched RNA targets were at least four-fold enriched over mock IP [log_2_ (fRIP) – log_2_ (mock IP) ≥ 2] with FDR ≤ 5%. Processed data is given in Supplementary Table S1.
GO analysis and sequence motif searches
Gene ontology (GO) analysis was performed with the Singular Enrichment Analysis tool in AgriGO v2.0 database [28] using the S. cerevisiae reference dataset. The CDS of the top 881 Pfk1 and all 867 Pfk2 mRNA targets were retrieved from ENSEMBL with BioMart. The associated 3′UTR and 5′UTR sequences were extracted from Saccharomyces Genome Database (SGD) [31] and duplicates and sequences shorter than 10 nts were removed, resulting in 875 and 756 3′-UTRs, and 869 and 752 5′-UTR sequences for Pfk1 and Pfk2 mRNA targets, respectively. De novo motif searches were performed with Multiple Expectation Maximization for Motif Elicitation (MEME) version 5.5.1 [30] with the following settings: searching the sense strand, RNA sequence, motif width between 6 and 12 nts, minimum and maximum number of sites 20–500 per sequence, any occurrence mode and default P-value cut-off < 0.0001. To estimate the occurrence of motifs in the background/population; the MEME output motifs were transferred into FIMO implemented in the MEME suite and searched against all CDS (6255 sequences), and the available 5211 5′UTR and 3′UTR sequences by searching the given strand only. Motif enrichment was calculated to the respective background/population of sequences, and P-values were determined (hypergeometric test). MEME search outputs and statistics are given in Supplementary Table S2.
Glucose starvation and polysomal profiling
For standard polysomal profiling, yeast cells were grown in 200 ml of YPD at 30°C to mid-log phase and treated with 100 µg/ml of CHX (Sigma, C1988) for 1 min. Cells were harvested by centrifugation at 1500 × g at room temperature for 5 min and immediately snap-frozen in liquid nitrogen. For glucose starvation and recovery experiments, 880 ml cultures were grown to mid-log phase. Then 250 ml of cells were isolated, CHX treated, collected and snap-frozen in liquid nitrogen. The remaining 550 ml of culture was collected and washed twice with pre-warmed media lacking glucose (YP) and finally resuspended in 550 ml of YP media and grown for 20 min at 30°C in an orbital shaker. Then, 250 ml of cells were treated with CHX for 1 min and collected. The remaining 300 ml of cells was pelleted and resuspended in 300 ml of pre-warmed YPD to activate recovery from glucose starvation and grown for 20 min at 30°C. Finally, cells were treated with CHX, and 250 ml collected and snap-frozen as described above.
Cell pellets were ground in liquid nitrogen using pre-chilled mortar and pestle and resuspended in 2 ml polysome lysis buffer (PLB; 20 mM Tris–HCl, pH 8.0, 140 mM KCl, 5 mM MgCl_2_, 1% Triton X-100, 0.5 mM DTT, 100 µg/ml CHX, 1 × cOmplete™ ULTRA EDTA-free, 1 mM PMSF, and 0.2 U/ml SUPERase•In™). For ribosome disassembly experiments, 30 mM EDTA (pH 8.0) was added to extracts. Cell lysates were centrifuged three times at 16 000 × g at 4°C for 10 min. Five milligrams of the resulting extract (total protein) were loaded on top of a 10%–50% linear sucrose density gradient prepared in PLB without Triton X-100. Samples were centrifuged at 125 000 × g for 2.5 h at 4°C in a Beckman Coulter Optima XPN-100 ultracentrifuge with SW41Ti swing rotor. Fractions (900 µl) were collected while continuously recording the absorbance at 254 nm (A_254_) with a flow cell UV detector (Teledyne ISCO). The ratio (P/S) of polysomal (fraction 7–13) to subpolysomal peak areas (fractions 1–6) were quantified with ImageJ (v1.53) [39].
Total RNA was isolated with Trizol (Thermo Fisher, AM9738) and ethanol precipitation. Essentially, 250 µl from each fraction was diluted with 250 µl (1 vol) of cold nuclease-free water, and 10 ng of a control RNA (LysA) was added prior organic extraction as described [40]. RNA was precipitated by addition of 3 µg of linear polyacrylamide (Invitrogen, AM9520), 0.25 vol. of 10 M ammonium acetate, and 2.5 vol. of ethanol; and RNA pellets washed twice with 70% ethanol, air-dried and resuspended in 20 µl of nuclease-free water. RNA was treated with 1 µl of DNase (TURBO DNA-free) prior RT.
For SYBR Green quantitative real-time PCR (qPCR) analysis, RT was performed with 3 µl of RNA from each fraction with the SensiFast cDNA synthesis kit. qPCR was then performed with 1 µl of the cDNA (accounting for between 30 and 400 ng of RNA extracted from polysome fractions) with the SensiFast SYBR lo-ROX mix (Meridian Bioscience, BIO-94005) according to manufacturer’s instructions. The temperature profile used for qPCR was 95°C for 2 min, followed by 40 cycles of (95°C for 5 s, 60°C for 10 s, 72°C for 5 s). Ct values were normalized to the LysA spike-in control as previously described [40].
Immunoblot analysis
A total of 0.1% (∼0.7–5 μg of protein) of the input extract, 10% of fRIP eluates, 10% of RIC eluates, and 5 µl of each polysomal fraction were resolved by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE). Specifically, protein samples were resolved on 4%–15% Mini-PROTEAN^®^ TGX™ Precast Protein Gels (Bio-Rad, 4561084) and transferred to nitrocellulose membranes (Cytiva, 10600003) with a semi-dry electrophoretic transfer cell (Trans-Blot Turbo, Bio-Rad). Membranes were blocked in PBS containing 0.1% Tween-20 (PBST) and 5% skimmed milk, probed with designated antibodies and HRP-coupled secondary antibodies. Membranes were developed with the Immobilon Western Chemiluminescent HRP Substrate (MerckMillipore, WBKLS0500) and luminescence recorded with a Fusion FX gel documentation system (Vilber Lourmat Sté).
Proteome analysis
The strains pfk1Δ, pfk2Δ, map1Δ, and isogenic BY4741 wild-type yeast cells were grown in 100 ml of YPD to mid-log phase and collected by centrifugation. Cells were lysed in 50 μl of lysis buffer [100 mM Tris–HCl (pH 8.5), 10 mM Tris(2-carboxyethyl)phosphine (TCEP), 1% sodium deoxycholate] by strong ultra-sonication (10 cycles, Bioruptor, Diagnode). Protein concentrations were determined by Pierce BCA assay kit using BSA as a reference standard (Thermo Fisher, 23227). Sample aliquots containing 50 μg of total proteins were reduced for 10 min at 95°C and alkylated with 15 mM chloroacetamide for 30 min at 60°C. Proteins were digested by incubation with sequencing-grade modified trypsin [2% (w/w); Promega, V5113] overnight at 37°C. Peptides were cleaned up using iST cartridges (PreOmics, Munich, Germany), according to the manufacturer’s instructions. Samples were dried under vacuum and stored at –80°C.
Sample aliquots comprising 10 μg of peptides were labelled with isobaric tandem mass tags (TMTpro 16-plex; Thermo Fisher, A44520) as described previously [41]. TMT-labelled peptides were fractionated and pooled into six fractions by high-pH reversed phase separation using a XBridge Peptide BEH C18 column (3.5 µm, 130 Å, 1 mm × 150 mm, Waters) on an Agilent 1260 Infinity high performance liquid chromatography (HPLC) system as previously described in [42]. Peptides were subjected to liquid chromatography tandem mass spectrometry (LC–MS/MS) analysis using a Q Exactive HF Orbitrap Mass Spectrometer fitted with an EASY-nLC 1000 (both Thermo Fisher Scientific) and a custom-made column heater set to 60°C using the same LC and MS settings as previously described [43]. The acquired raw data files were searched against a protein database containing sequences of the predicted SwissProt entries of S. cerevisiae (www.ebi.ac.uk, release date 21 April 2020), the six calibration mix proteins [41] and commonly observed contaminants (in total 12 882 sequences) using the SpectroMine software (Biognosys, version 1.0.20235.13.16424). Standard Pulsar search settings for TMTpro (“TMTpro_Quantification”) were used and resulting identifications and corresponding quantitative values were exported on the PSM level using the “Export Report” function. Acquired reporter ion intensities in the experiments were employed for automated quantification and statistical analysis using an in-house developed SafeQuant R script (v2.3) [41]. Processed MS data is available in Supplementary Table S3; associated GO searches are given in Supplementary Table S4.
Expression and purification of histidine-tagged fusion proteins in E. coli
Expression vectors (pDEST17-PFK1; pDEST17-PFK2*)* were transformed into ArticExpress (DE3) RIL competent E. coli cells (Agilent, 230 193). Cells were grown in 2YT media (16 g/l tryptone, 10 g/l yeast extract, 5 g/l NaCl) supplemented with 100 µg/ml ampicillin (Sigma) at 14°C to mid-log phase (OD_600_ ∼ 0.4). Expression of His-tagged fusion proteins was induced by addition of 0.25 mM of isopropyl-β-D-thiogalactopyranoside (IPTG) for 18 h, respectively. Cells were collected by centrifugation at 5000 × g for 15 mins at 4°C, washed with ice-cold PBS and snap-frozen in liquid nitrogen then stored at –80°C. Cells from 2 l cultures were then resuspended in lysis buffer [LS; 50 mM HEPES, pH 7.5, 1 M NaCl, 10% glycerol, 2 mM MgCl_2_, 0.05% Tween-20, 10 mM imidazole, 1 mM β-mercaptoethanol (β-Me)] and disrupted by four to six passages at 1200 bar through a Microfluidizer (Microfluidics, LM10). Cell lysates were cleared by centrifugation at 30 000 × g for 45 min at 4°C and loaded on a 5 ml HisTrap FF column (Cytiva, 17531901). The column was washed with 10 column volumes (CV) of high-salt buffer (HS, NB2 buffer with 2 M NaCl, 20 mM imidazole) and 5 CV of LS buffer; and His-tagged proteins were eluted with a linear gradient to 100% elution buffer (EB; 20 mM HEPES, pH 7.5, 0.3 M NaCl, 10% glycerol, 2 mM MgCl_2_, 500 mM Imidazole, 1 mM β-Me). Fractions containing recombinant proteins were pooled and dialysed in buffer A (20 mM HEPES, pH 7.5, 150 mM NaCl, 5% glycerol, 2 mM MgCl_2_, 1 mM β-Me) and loaded on a HiTrap Heparin HP column (Cytiva, 17040601). The column was first washed with 10 CV of buffer A. Bound protein was eluted by applying a linear gradient from buffer A to buffer B (containing 2 M NaCl) over 100 ml. Fractions with recombinant protein were pooled and loaded on a HiLoad 16/600 Superdex 200 (Cytiva, 28989335) size-exclusion column and separated with buffer A at 1 ml/min. Protein concentrations were determined using a Nanophotometer (Implen, NP80). Required theoretical molecular weights and extinction coefficients were calculated using ProtParam (Expasy).
UV-crosslinking assays
Fluorescently labelled RNA oligos with IRDye^®^ 800CW (LI-COR, Inc, USA) were obtained from Integrated DNA technologies (IDT) (Supplementary Table S5). UV crosslinking was performed by combining 500 fmol of fluorescently labelled RNA and 0.5 µg of His_6_-Pfk1 or Pfk2 protein in 10 µl reaction buffer (20 mM HEPES–KOH, pH 7.5, 50 mM NaCl, 2 mM MgCl_2_, 5% glycerol, 1 mM DTT). To test nucleotides, 50 mM of the respective nucleotide (Sigma) was incubated in 50 mM MgCl_2_ for 15 min at room temperature prior addition of 1 µl to the reaction mix (final conc. 5 mM). Mixed reactions were incubated for 30 min at room temperature and crosslinked at 254 nm on ice with 400 mJ/cm^2^ in a UV crosslinker (Boekel Scientific, 234100). Samples were resolved on 4%–15% gradient SDS–PAGE (Bio-Rad) and scanned with an Odyssey CLx scanner (LI-COR, Inc, USA).
Synthesis of biotinylated RNAs and RNA pull-down experiments
cDNA oligonucleotides with T7 promotor sequences were designed that cover consensus motifs in CLN3 and BUB3 mRNAs and additional 4–20 nts of flanking sequences (Supplementary Tables S2 and S5). Specifically, the CLN3-CDS covers an UCRE at starting nts 1373 in the CDS, the CLN3-5′UTR covers two proceeding U-rich element at position 103 and 113 of the 5′UTR, and the CLN3-3′UTR covers an U-rich element at position 81 of the 3′UTR. The BUB3-3′UTR fragment covers and A-rich element at position 91 of the 3′UTR. The DNA oligonucleotides were annealed with T7 RNA polymerase promotor sequences for transcription. Essentially, 25 µM of the T7 promotor and DNA target oligos were mixed and incubated at 95⁰C for 2 min in 10 µl of water and then cooled down slowly to room temperature. Biotin labelled RNAs was produced with the Biotin RNA Labelling Mix (Roche, 11685597910) and T7 RNA polymerase (NEB, M0251), as instructed by the manufacturer. The 32–37 nts long RNA products were purified with the RNA Micro kit (Zymo, R1050) following the manufacturers protocol to isolate small RNAs. RNA was quantified by UV spectrometry with a Nanodrop device, and the integrity of the RNA validated by agarose gel electrophoresis.
Biotin RNA pull-down experiments were performed essentially as described [44]. In brief, 0.9 µg of recombinant His_6_-Pfk1 and His_6_-Pfk2 was combined with 20 pmol of biotinylated RNAs in 50 µl reaction buffer II [20 mM HEPES–KOH (pH 7.5), 100 mM NaCl, 5% glycerol, 2 mM MgCl_2_, 0.1% Triton X-100, 1 mM DTT, 0.1% BSA, 1 mM PMSF, 0.5 µg/ml leupeptin, 0.8 µg/ml pepstatin, 40 U/ml of RNAsin]; and incubated for 30 min at room temperature by rotation on a daisy wheel. RNA protein complexes were further captured with 25 μl of streptavidin M280 Dynabeads^®^ (Invitrogen, 11205D) for 30 min, and the beads recovered with a magnet and washed five times with reaction buffer II. RNA–protein complexes were resolved by 4%–15% gradient SDS-PAGE and analysed by immunoblotting with Pfk antibodies.
RNA electrophoretic mobility shift assays
RNA electrophoretic mobility shift assays (REMSA) was essentially performed as described [45] with minor modifications. The reaction was set in 20 µl by adding 2 µl of 10 × REMSA buffer [400 mM Tris–HCl (pH 7.5), 300 mM KCl, 10 mM MgCl_2_, 50% glycerol, 0.1% IGEPAL, 10 mM DTT], 0.4 µl of 100 nM fluorescently labelled RNA oligo (= 2 nM), 10 µl of 1:2 serially diluted recombinant His_6_-Pfk2p. Reactions were mixed by pipetting and incubated at room temperature for 1 h. Samples were loaded on a 5% native Tris-acetate-EDTA (TAE) polyacrylamide gel and run in TA buffer at 90 V for 50 min. The gel was scanned with Odyssey CLx scanner (LI-COR, Inc, USA).
RNA unwinding assays
Annealing of the RNA substrates and real-time fluorescence-based RNA unwinding assays were performed as described [46]. Shortly, 50 nM of annealed RNAs substrates were incubated with 1 µM of the indicated proteins in 10 μl assay buffer (20 mM HEPES–KOH, pH 7.5, 100 mM KCl, 1 mM TCEP, 1% glycerol) in 384-well plates and incubated for 30 min at room temperature. Reactions were started by the addition of Mg^2+^-ATP, or the indicated nucleotides to a final concentration of 5 mM and fluorescent readings taken in a Spark (Tecan) with excitation at 535 nM and emission at 575 nM for 70 min at room temperature. Data was analysed according to [46]. For baseline correction, the buffer control trace (negative control) was subtracted from all other reactions, and the data was normalized to the reference, which is the reporter strand only. The initial linear part of the traces was fitted to yield the initial rate of unwinding and converted to 10^6^ min^−1^. Experiments were performed in triplicates and repeated with at least two different batches of recombinant Pfk1 and Pfk2 proteins.
Enzymatic activity assays
BY4741 and pfk2Δ cells harbouring CEN plasmids expressing wild-type and mutant PFK2 were grown in 5 ml SC-Leu containing 2% glucose (except BY4741, which was grown in SC). Cells were collected by centrifugation, washed once with 3 ml water and once with 3 ml 50 mM potassium phosphate, pH 7.0 (PP) buffer, cells were collected and the cell pellet was snap frozen. Cells were resuspended in 500 µl of PP buffer and broken mechanically with 500 mg of glass beads (0.3–0.5 mm in diameter) upon vigorous shaking for 7 min at 4°C. Five hundred microlitres of cold PP buffer was added and the crude extract was cleared by centrifugation at 14 000 × g for 10 min at 4°C; the supernatant was then used in enzymatic activity assays. PFK activity was measured in a coupled enzyme assay resulting in the oxidation of NADH as described [20]. Essentially, 2–25 µg of the extract was combined with 700 µl of the reaction mixture prepared in PP buffer containing 10 mM MgCl_2_, 5 mM fructose 6-phosphate (F6P), 5 µM fructose 2,6-biphosphate (F2,6BP), 1 mM ATP, 0.2 mM NADH, 0.3 Units (U) aldolase (Sigma, A8811), and at least 1 U of glycerophosphate dehydrogenase-triosephosphate isomerase from rabbit muscle (Sigma, G1881). NADH oxidation was measured by following the decrease in absorbance at 340 nM [20].
Yeast cell growth and size measurements
Yeast cells were diluted from overnight pre-cultures to a starting OD_600_ ∼ 0.1 in 2 ml of YPD. 0.5 ml of culture (triplicates) were transferred into a 48-well Multidish (Thermo Fisher, 150787) placed in a plate reader (CLARIOstar, BMB Labtech), and cells were continuously grown at 800 r.p.m and 30°C. OD_600_ was recorded every 15 min for 48 h. Data was exported to GraphPad Prism and doubling time was calculated with the nonlinear regression curve fit analysis and selecting the least squares fit method. To measure cell size, cells were grown in 2 ml of YPD at 30°C to mid-log phase (OD_600_ = 0.6–0.8) and collected by centrifugation at 5000 × g for 3 min, washed twice with PBS, and finally resuspended in 250 µl PBS. Five microlitres of resuspended cells were placed on a glass slide and imaged in phase contrast with a Life Technologies EVOS FL microscope with a 40×/0.65 EVOS LPlanFL objective and images were captured with a Sony™ ICX285AL CCD camera. The Feret diameter (µm) of at least 100 individual cells per sample were quantified with ImageJ (v1.53) [39] with the following settings: Images were set to 8 bit, threshold set at ∼2% to select cells without background, scale = 4.6 pixels/µm, size from 4-infinite µm, and circularity was set to 0.8–1.00 to compensate for cells that were not perfectly circular. The Mann–Whitney U-test implemented in GraphPad Prism (v10.5) was used for statistics (GraphPad Software, San Diego, CA, USA).
Cell synchronization and flow cytometry analysis
Yeast pre-cultures were diluted to OD_600_ ∼ 0.1 and grown to mid-log phase in 25 ml YPD at (30°C). Cells were diluted again in 25 ml YPD to OD_600 ∼ 0.1 and allowed to grow to early-log phase OD_600 ∼ 0.3. One millilitre of culture was collected and fixed with 70% ethanol to be used as asynchronous control for the Flow Cytometry (FC) experiment. Cell cycle was blocked by adding 20 µg/ml of the α-factor (Zymo Research, Y1001) mating pheromone for 180 min at 30°C upon shaking. To release the α-factor arrest, cells were collected by centrifugation (2000 × g, 3 min, 30°C), gently washed with 25 ml of pre-warmed (30°C) YPD, and finally resuspended in 25 ml of pre-warmed YPD. One millilitre of cells was immediately collected and fixed with 1 ml of 70% ice-cold ethanol, corresponding to start of cell cycle measurements (time = 0). The remaining cells were further incubated on the shaker at 30°C with 200 r.p.m. Likewise, 1 ml of cells were collected every 10 min and immediately fixed with 70% ice-cold ethanol over a time course of 200 min (20 samples). The collected samples were kept in 70% ethanol at 4°C until use. Fixed cells were centrifuged to remove ethanol, washed twice with 1 ml PBS, resuspended in 1 ml PBS and left in a stand for 1 h at room temperature for rehydration. Cells were then collected by centrifugation and resuspended in 200 μl of RNase A reaction buffer [2 mM Tris–HCl (pH 7.5), 25 µg of RNase A (Thermo Fisher, EN0531)] and incubated at 42°C for 2 h. Cells were collected by centrifugation (3000 × g, 5 min, room temperature) and the supernatant was discarded. The cell pellets were further washed twice with 1 ml of PBS, resuspended in 200 μl PBS supplemented with 100 µg of proteinase K and incubated at 42°C for 40 min. Finally, cells were pelleted at 2000 × g for 3 min, washed twice with PBS and resuspended in 1 ml of PBS. DNA was stained with 16 µg/ml propidium iodide (Thermo Fisher, J66584.AB) for 1 h at room temperature in the dark. Cells were briefly sonicated and DNA content at each time point was determined on 20 000 cells by flow cytometry analysis with Attune NxT flow cytometer (Life technologies) and the following settings: flow rate of 200 µl/s and YL-2 detector with long-pass filter and maximum absorbance of 620 nm. Data was saved in FCS 3.0 file format and analysed and visualized using FlowJo (BD Biosciences).
Results
Pfk1p and Pfk2p bind RNA in vivo
To reconcile the RNA-binding activities of Pfk1p and Pfk2p seen in previous RIC-MS studies [5, 6], we first performed RIC with yeast strains expressing either PFK1 or PFK2 with a C-terminal tandem-affinity purification (TAP)-tag under the control of their endogenous promoters. As expected, Pfk1:TAP and Pfk2:TAP were detected in RIC eluates but not in corresponding samples supplemented with a poly(A) competitor. No signals were detected for Act1p, a non-RBP control (Fig. 1A). Likewise experiments with wild-type cells confirmed interaction of both endogenous (untagged) Pfk1 and Pfk2 proteins, using an antibody that detects both Pfk subunits. RIC performed with pfk1Δ and pfk2Δ strains further revealed that each subunit can interact with poly(A)-RNA in the absence of the other, indicating that formation of a heterooctameric Pfk protein complex is not required for interaction with poly(A) RNAs (Fig. 1A).
Pfk1p and Pfk2p bind hundreds of functionally related mRNAs in cells
Next, we profiled the RNA targets for Pfk1p and Pfk2p by carrying out fRIP followed by high-throughput sequencing of bound RNAs (fRIP-seq). fRIP-seq was performed on yeast heterozygous diploid cells expressing PFK1:TAP/PFK1, PFK2:TAP/PFK2, and the untagged parental wild-type was used as an IP background control strain. To estimate relative enrichment scores, we also performed RNA-seq on rRNA depleted total RNA isolated from extracts. The IP efficiency was tested and confirmed that our chemical crosslinking does not interfere with the capture of the tagged proteins (Supplementary Fig. S1A).
Overall, 1025 and 895 transcripts were significantly associated with Pfk1 and Pfk2, respectively (selection based on threshold: FDR ≤ 5%, ≥4-fold enriched over mock-IP control; data given in the Supplementary Table S1A). Among those, 671 transcripts were commonly selected with both Pfk proteins (p= 1.83 × 10^−463^) (Fig. 1B and C). Interestingly, RNA biotype analysis showed that most of the associated transcripts were mRNAs (1017 and 867 of the Pfk1p and Pfk2p RNA targets, respectively; Supplementary Fig. S1B), which contrasts other yeast glycolytic enzymes that preferentially interacted with short ncRNAs such as transfer RNAs (tRNAs) [47]. Nevertheless, 13 tRNAs were associated with Pfk2p including the glutamine–tRNA [tQ(CUG)M, tRNA-gln] CDC65, which has a role in the regulation of entry of cells into the cell cycle, while Pfk1p was only associated with one tRNA [tI(UAU)L, tRNA-Ile] also bound by Pfk2p.
We then searched for significantly enriched GO terms among Pfk1p and Pfk2p RNA targets. No GO terms were enriched among RNA targets exclusive to either Pfk subunit when applying our stringent selection criteria. However, the 671 common RNA targets were significantly enriched (FDR < 0.05) for mRNAs coding for proteins acting in transcription (e.g. ‘transcription from RNA polymerase II promoter’, p = 1.7 × 10^−5^), DNA replication (p = 0.007), chromatin organization (p = 0.0022), and associated with the mitotic cell cycle (p = 7.4 × 10^−4^) (Fig. 1D; Supplementary Table S1B). Intrigued by the link to the cell cycle, we validated the interaction of Pfk1p and Pfk2p with cell cycle factor mRNAs with independent fRIP followed by RT-PCR (Fig. 1E). Notably, we observed stronger PCR signals with Pfk2 than with Pfk1 fRIPs, which could indicate preferential binding to Pfk2p in vivo. Overall, these findings indicate that Pfk1 and Pfk2 proteins interact with functionally related mRNAs which could facilitate the coordination of post-transcriptional gene regulation for specific biological functions.
Pfk proteins bind RNA sequence repeats in vitro
To identify common sequence/structural elements among mRNA targets that could mediate selective interactions with Pfk proteins, we retrieved the CDSs as well as the 3′- and 5′-UTR sequences for Pfk1 and Pfk2 mRNA targets from the SGD and performed de novo motif discovery analysis with MEME. UC-rich (UCRE) and GA-rich sequence elements (GARE) were overrepresented within the CDSs of mRNA targets associated with either Pfk1p and Pfk2p (Fig. 2A; Supplementary Table S2). Both motifs are arranged as trinucleotide repeats, spanning at least 4 repeats covering 11–12 nucleotides (nts). Interestingly, these triplet repeats are preferentially positioned in frame (i.e., 96% of the GARE and 87% of the UCRE repeats in Pfk2p mRNA targets), with GA(C/U) coding for the amino acid aspartate (D), GA(A/G) coding for glutamate (E), and UCN coding for serine (S). Hence, proteins encoded by those mRNAs bearing stretches of D/E codons were also overrepresented among Pfks mRNA targets (e.g., MEME identified a stretch of eight D/E amino acids among Pfk2p mRNA target encoded proteins; 156 sites in 101 proteins, *p *= 3 × 10^−18^) (Supplementary Table S2). Moreover, searching for the occurrence of UCRE and GARE motifs in all CDSs of the yeast genome confirmed the significant enrichment of these motifs among experimentally determined Pfk targets. However, the recorded in-frame periodicity of the motif triplets is also apparent across all CDSs of the yeast genome [i.e., 7973 (94%) of the 8400 GARE motifs, and 3504 (81.7%) of the 4317 UCRE’s encoded in all genomic CDSs are in frame]. Thus, the periodicity observed does not seem selective for Pfk RNA associations (Supplementary Table S2).
Likewise, motif searches with UTR sequences revealed enrichment of AU-rich elements (AURE) in 3′UTRs, as well as less-well defined A-rich and U-rich sequences in 5′UTRs and 3′UTRs. The latter is reminiscent to a U-rich element previously identified in Pfk2p PAR-CLIP studies [7] (Fig. 2A and Supplementary Table S2). Importantly, the ‘U-rich’ elements contained interspersed ‘C’ nucleotides; and the ‘A-rich’ elements are intermingled with ‘G’ nucleotides which seem distantly related to the UCRE and GARE’s identified in CDS, respectively. A search with the individual consensus motifs, across all genome-encoded UTR sequences, revealed that the U-rich element located in 3′UTR, and the A-rich elements in the 5′UTR were both significantly enriched, while the AURE was not. However, this statistical analysis may be substantially obtruded by the limited information contained in AU-rich sequences that are prevalent in the 3′UTR. For example, there are 2925 UAUAUAUAUAUA sites in 1144 different 3′UTRs, as compared to 354 sites in 185 5′UTRs, and 319 sites in 137 CDS (Supplementary Table S2).
*Pfk2p interacts with short RNA motifs and has directional RNA-unwinding activity in vitro. (A) Overrepresented sequence motifs within the ORFs and 3′UTRs of all Pfk2p mRNA targets identified with MEME. E-values are indicated below each motif. (B) UV-crosslinking experiments with recombinant His6-Pfk1/2 proteins and fluorescently labelled RNA oligos, including UCRE (UC), GARE (GA), AURE (AU), and poly(U) (pU) RNA oligos matching the consensus sequences as well as double-stranded RNA (dsRNA) motif (dsR) and poly(A) RNA (pA). The top gel illustrates crosslinking of Pfk1 and Pfk2 to RNA substrates indicated on top. The gel at the bottom shows interaction of His6-Pfk2p with UCRE RNA in the presence of 5 mM of indicated Mg-complexed nucleoside phosphates; or upon addition of a 50× excess of unlabelled RNAs. A marker (M) with molecular weights in kDa is indicated to the left. (C) Biotinylated RNA pull-downs with His6-Pfk2p. Lane 1, marker; lane 2, 10% of the input (Ex); lanes 3 and 4, 10% of unbound supernatant (Sn); lanes 5–10, 50% of eluates after incubation with indicated RNAs fragments or in the absence of RNA (lane 11). (D) REMSAs with UCRE and dsR RNA substrates and increasing concentration of His6-Pfk2p. Free RNA and RNA–protein complex are marked to the right. (E) RNA unwinding by Pfk1p, Pfk2p, and eIF4A1 using 24 bp substrates with the indicated 20 nts repeat overhang sequences at the 5′ or 3′ ends. Data are mean values with standard deviations (± stdev) from repeat experiments (n = 3). The UCUC 5–3 RNA tested with indicated adenosine phosphates or MgCl2 only is shown to the right (Student’s t-test, *P <0.01). A reaction scheme for 5′ overhang RNA substrates is boxed, indicating the position of the quencher (Q) and the Cy3 fluorescent dye (green dot).
To assess whether Pfk proteins could directly interact with those RNA elements, we performed RNA–protein interaction assays using recombinant His_6_-tagged Pfk1p and Pfk2p expressed in E. coli and purified to homogeneity (Supplementary Fig. S2A). We first applied UV-crosslinking experiments using fluorescently labelled short RNA oligonucleotides containing UCRE, GARE, AURE, poly(U) motifs, and poly(A) serving as a control. Given that UCRE and GARE are complementary and could form dsRNA, we also tested an RNA oligo with both elements (dsMotif/dsR). Both His_6_-Pfk1p and His_6_-Pfk2p bound specifically to GARE, UCRE, AURE, and poly(U) motifs, with binding attenuated in the presence of excess unlabelled competitor RNA; whereas the observed weak crosslinks to dsRNA may reflect UV-induced perturbation of short dsRNA duplexes that permit capture of transient Pfk–RNA interactions (Fig. 2B). Addition of Mg²⁺-ADP decreased Pfk2p binding—but not that of Pfk1p—whereas Mg²⁺-AMP or ribonucleoside triphosphates (CTP, GTP, UTP) caused a more pronounced reduction with both Pfk proteins (Fig. 2B and Supplementary Fig. S2B). These results indicate that Pfk proteins engage specific RNA elements in a nucleotide-sensitive manner, with affinity diminished in the allosterically activated R-state, likely stabilized by elevated intracellular AMP or ADP.
To validate these interactions in the context of native RNA sequences, we performed RNA pull-down assays using 33–37 nucleotide biotinylated RNA fragments derived from motif-containing regions. We focused on two transcripts with central and also distinct roles in cell cycle regulation that were validated with RT-PCR (Fig. 2C): CLN3, encoding a G1 cyclin that promotes Start and entry into S phase and is subject to translational control [48], and BUB3, encoding a spindle assembly checkpoint protein that ensures proper chromosome attachment and delays anaphase onset during mitosis (M-phase) [49]. Specifically, we tested a UCRE within the CLN3 CDS (CLN3-CDS), U-rich elements in the 5′- and 3′-UTRs of CLN3 (CLN3-5U and CLN3-3U), and an A-rich element in the BUB3 3′-UTR (BUB3-3U). Biotinylated UCRE oligonucleotides were used as positive controls, and poly(A) as a negative control. Both Pfk1p and Pfk2p bound strongly to all CLN3 and BUB3 RNA fragments, whereas minimal interaction was observed with the poly(A) control (Fig. 2C and Supplementary Fig. S2C).
Finally, we employed REMSA with both Pfk proteins to further evaluate binding specificity and complex formation (Fig. 2D, and Supplementary Fig. S2D and E). We observed distinct mobility shifts with UCRE, GARE, and AURE oligonucleotides in the nanomolar concentration range, indicative of specific binding. No shifts were observed with the dsMotif/dsR or poly(A) controls in the presence of His6-Pfk2p, although a weak shift was seen with His6-Pfk1p at highest protein concentrations (Supplementary Fig. S2D and E). However, interpretation was complicated by the apparent formation of high-molecular-weight complexes or aggregates that did not resolve well during electrophoresis, preventing a reliable quantification of binding affinities at this point.
Taken together, these RNA–protein interaction assays consistently demonstrated that Pfk proteins selectively associate with unstructured GA-, UC-, and AU/U-rich RNA motifs. Moreover, RNA binding appears to be modulated by nucleotide-induced conformational changes, likely reflecting the enzyme’s transition to the R-state upon AMP/ADP binding. While the conformational status of the purified recombinant proteins remains unresolved, these findings indicated that metabolic cues could influence RNA-binding of Pfk proteins in vivo.
Pfk2p has RNA unwinding activity
Curious about the observed nucleotide sensitivity and potential structural preferences in Pfk–RNA interactions, we hypothesized that Pfk proteins may not only bind but also remodel RNA. Specifically, we reckoned that Pfkps might facilitate structural rearrangements or unwinding of RNA, particularly in motif-rich or partially structured elements. To test this, real-time fluorescence-based RNA unwinding assays were employed using RNA substrates previously tested with eIF1A1, a human DEAD-box family RNA helicase involved in translation initiation [46]. The RNA substrates contained an identical 24 bp duplex with 20 nts overhang sequences related to our identified GA and UC-rich sequence elements placed either at the 5′ or the 3′ end of the RNA [46]. Surprisingly, we found that His_6_-Pfk2p displayed RNA unwinding activity with RNA substrates bearing the overhang sequence at the 5′ end, while no unwinding activity was detected with overhang sequences placed at the 3′ end (Fig. 2E and Supplementary Fig. S3). As expected, eIF1A1 displayed unwinding activity irrespective of the position of the overhangs. No activity in either direction was detectable with likewise expressed and purified His_6_-Pfk1p. These results demonstrate that while both PFK subunits are capable of RNA binding, only Pfk2p exhibits 5′ to 3′ directional unwinding activity on short RNA duplexes. Furthermore, unlike eIF4A1, whose RNA unwinding activity was abrogated by 5 mM adenylyl-imidodiphosphate (AMPPNP; a nonhydrolysable ATP analogue), His_6_–Pfk2p retained comparable activity under the same conditions. Similarly, the addition of Mg^2+^ ions caused only a modest reduction of Pfk2-mediated unwinding (Fig. 2E). These findings suggest that ATP hydrolysis is not strictly required for Pfk2p’s unwinding activity. Rather, residual ATP co-purified with His_6_-Pfk2p [11] or the addition of excess AMPPNP may stabilize Pfk2p in the T-state configuration required for RNA interaction. Consistent with this model, addition of Mg-ADP and Mg-AMP significantly reduced unwinding activity by 40%–60% (Fig. 2E), mirroring the previously observed nucleotide-dependent decrease in Pfk2p-RNA crosslinking in the presence of these nucleotides, and consistent with the stabilization of the R-state. The remaining unwinding activity in the presence of ADP and AMP could reflect their modest activating properties, possibly in combination with competition from residual ATP that co-purified with Pfk2p.
Pfk–RNA and ribosome interactions are dependent on the energy-status of cells
Given that nucleotide addition (e.g., AMP) impaired Pfkp-RNA interactions and specifically reduced Pfk2ps unwinding activity—potentially stabilizing the enzymatically active R-state—we hypothesized that RNA binding in vivo could be influenced by the cellular energy status linked to the the enzyme’s conformational state. To test this, we performed RIC with cells subjected to glucose depletion and after re-addition of glucose (Fig. 3A). The removal of glucose from the media immediately affects the levels of various metabolites and drastically depletes cellular ATP levels by ∼70% within minutes while increasing the cellular AMP/ATP ratio in budding yeast [50–52]. Indeed, we found that glucose starvation abolished Pfk1:TAP and Pfk2:TAP binding to poly(A) RNA and re-addition of glucose to media largely restored their interactions with RNA (Fig. 3B). This change in RNA association was not observed with Pgk1:TAP, which is another RNA-binding glycolytic enzyme that remained bound to poly(A) RNA under all conditions. These findings support the notion that glucose starvation, inducing a conformational change of the enzyme from the inactive T-state to the active R-state, reduces Pfk–RNA interactions. Hence, RNA binding by Pfk is likely modulated by the enzyme’s allosteric state and is responsive to energy availability.
Yeast Pfk2p and human orthologs, such as the liver (PFKL) and platelet (PFKP) isoforms have previously been detected in polysomes or in association with cytoplasmic ribosomes [53, 54]. This prompted us to investigate whether the yeast Pfk paralogs play a role in translation. Hence, we performed sucrose density fractionations using wild-type, pfk1Δ and pfk2Δ strains to assess the association of Pfk subunits with polysomes. While pfk1Δ cells showed similar polysomal profiles like wild-type cells (ratio of polysomes to subpolysomes (P/S) = 1.85 – 1.95), pfk2Δ cells showed a ~50% decrease in polysomes (P/S = 0.86) indicating diminished global translation (Fig. 3C, Supplementary Figure S4A). This is in line with the reported slow-growth of pfk2Δ cells, which manifests in reduced protein synthesis. We further found that ~20% of Pfk1p and ~11% Pfk2p co-sedimented with polysomes in heavy sucrose fractions. However, only Pfk2p dissociated from these polysomal fractions upon treatment of extracts with 30 mM EDTA, a condition that disassembles polysomes and ribosomes. In contrast Pfk1p did not disassemble, suggesting that Pfk1p could be part of another high order complex that co-sediments with polysomes, such as previously reported highly filamentous structures that can form in the absence of PFK2 [55]. We next tested whether the association of Pfk2p with ribosomes is affected by glucose levels in the growth media. We found that Pfk2p dissociated from polysomes in pfk1Δ cells upon glucose removal and fully reassociated with polysomes after re-addition of glucose to media (Supplementary Figure S4B). Conversely, Pfk1p did not dissociate from high-density gradient fractions upon glucose removal and even after disassembly of polysomes with EDTA, corroborating the notion that Pfk1p is not directly associated with translating ribosomes in the absence of Pfk2p (Supplementary Figure S4B). Taken together, these results suggest a specific dynamic and reversible association of Pfk2p - but not of Pfk1p - with translating ribosomes.
*Pfk2p dynamically associates with mRNAs and polysomes and impacts translation. (A) Experimental design for glucose (glc) recovery experiments. Cells were grown in YPD (+G), then starved in media lacking glc for 20 min (−G) and recovered upon re-addition of glc for 20 min (R). (B) Immunoblot of input extracts (left) and RIC eluates (right) from glc recovery experiments at indicated stages (+G, −G, R). Monitored proteins are labelled to the right; a molecular weight marker is indicated to the left. Poly(A) binding protein (Pab1p) is a RBP control. (C) Polysomal absorbance profiles of pfk1∆ and pfk2Δ cells grown in YPD medium are shown at the top. Fractions numbers are indicated and those containing polysomes are highlighted in purple. Immunoblot analysis of fractions monitoring the distribution of Pfk1p and Pfk2p; and upon treatment of extracts with 30 mM EDTA to dissociate ribosomes are given below. Rpl35p is a ribosomal protein of the large subunit, Act1p is a nonribosomal associated control protein. (D) Distribution of CLN3, BUB3, and ACT1 mRNA levels across sub-polysomal and polysomal fractions (purple) obtained from wild-type (wt; grey), pfk1∆ (blue) and pfk2∆ (red) cells grown in YPD medium. RNA was isolated from each fraction and quantified by RT-qPCR. The y-axis denotes mRNA levels in each fraction calculated as the percentage of the total (mean values ± stdev, n = 3). (E) BUB3 and CLN3 mRNA levels relative to ACT1 levels determined by RT-qPCR in total RNA isolated from wt, pfk1∆ and pfk2∆ cells (mean values ± stdev, n = 3). P <0.05 (Student’s t-test).
Pfk2 promotes translation of CLN3 and BUB3 mRNAs
As many ribosome-associated RBPs act as translational regulators of selected mRNAs, we further monitored the distribution of Pfk mRNA targets across the sucrose density fractions isolated from wild-type, pfk1Δ and pfk2Δ cells with RT-qPCR. Specifically, we investigated the mRNAs for CLN3 and BUB3 coding for proteins involved in cell cycle entry and mitotic progression, respectively—and confirmed their direct interaction with both Pfk subunits (Fig. 2C). Indeed, we found a dramatic shift of both BUB3 and CLN3 mRNAs from heavy polysome fractions to ‘subpolysomal’ fractions [comprised of ribonucleoprotein complexes (RNPs), free ribosomal subunits and monosomes] in pfk2Δ but not in pfk1Δ cells or wild-type cells. Specifically, ∼10% of BUB3 was detected in polysomal fractions of pfk2∆ cells, compared to 40% and 32% in pfk1∆ and wild-type cells, respectively. Similarly, only 31% of CLN3 was observed in polysomal fraction of pfk2∆ mutants, but 63% and 60% in pfk1∆ and wild-type cells, respectively. No change was seen in the distribution of ACT1 mRNA as a nontarget control in all samples (Fig. 3D). The corresponding total RNA levels isolated from extracts were not altered for CLN3 in pfk1Δ and pfk2Δ compared to wild-type cells; however, BUB3 mRNA levels were reduced in pfk2∆ cells, which could indicate reduced mRNA stability associated with the observed drastic shift to ribosome devoid fractions (Fig. 3E). The dynamic association of Pfk2p with polysomes together with the severely reduced co-sedimentation of selected Pfk target mRNAs with polysomes in pfk2∆ cells, suggests that Pfk2p could act as a translational activator for selected mRNAs.
Deletion of PFK2 affects abundance of a large fraction of the proteome
To assess the proteome-wide impact of PFK expression and the possible implications for cell-physiology, we next profiled changes of the proteome of pfk1Δ and pfk2Δ deletion mutants compared to wild-type cells using isobaric tandem mass tag (TMT)-based quantitative proteomics. We also profiled slow-growing map1Δ cells [31, 56] to evaluate whether altered protein levels could simply be associated with slow growth as reported for pfk2Δ strains. Data were obtained for 3856 proteins (n = 3; 1% FDR). Irrespective of the chosen cut-off, the largest fraction of protein level changes was seen in pfk2Δ cells (e.g. 1740 proteins at FDR ≤ 10%). Substantial fractions also changed in map1Δ cells (1100 proteins, FDR ≤ 10%), whereas considerably less changes were observed in pfk1Δ cells (672 proteins, FDR ≤ 10%; Fig. 4A; MS data given in Supplementary Table S3). As may be expected, proteins changing in pfk1∆ and pfk2∆ cells greatly overlapped, most of them showing reduced relative abundances (Fig. 4A). GO analysis revealed that the commonly ‘down-regulated’ proteins, i.e., less abundant proteins in pfk1∆, pfk2∆, and map1∆ compared to wild-type cells, refer to ‘cytoplasmic translation’, ‘ribosome biogenesis’, and/or are part of ‘ribonucleoprotein complexes’ (Fig. 4B, complete GO analysis is given in Supplementary Table S4). Along the commonly observed slightly increased protein levels associated with the oxidative stress response, these changes are obviously not selective for the lack of PFK genes as they are also seen in the slow-growing map1∆ cells. A second subgroup of proteins showing increased abundances specifically in pfk1∆ and pfk2∆, but not in map1∆ cells, was assigned to the ‘carbon metabolic process’ and the ‘generation of precursor metabolites and energy’, including glycolytic enzymes. The increased levels of carbon metabolism components in pfk mutants could relate to a compensation mechanism to cope with the reduced glycolytic activity in pfk mutants. Finally, a third group of proteins acting in DNA replication, tRNA processing and transcription were particularly lower abundant in pfk2Δ but not in pfk1Δ cells; i.e., representing pfk2∆ specific implications in the proteome (Fig. 4B and Supplementary Table S4). These functional themes were reminiscent to the relationships of experimentally determined Pfk mRNA targets and hence, indicated functional links between mRNA targets and aberrant expression in pfk2∆ cells.
*Proteome changes in pfk1∆, pfk2∆ and map1∆ cells. (A) Venn Diagram displaying the overlap of numbers of proteins with altered levels (FDR ≤ 10%) in the mutant compared to wild-type (wt) cells. (B) Heatmap displaying a subset of GO terms (rows) enriched among the proteins with increased (upright arrow) or reduced (downward arrow) levels in indicated mutants (columns). The colour intensity corresponds to Benjamini-Hochberg FDRs. (C) Overlap of Pfk1 and Pfk2 target mRNA encoded proteins for which MS data was obtained (outer circles) with those having altered levels (<10% FDR) in pfk1∆ and pfk2∆ cells, respectively (inner circles). (D) Boxplots depicting relative fold-changes of protein levels in pfk1∆ (blue), pfk2∆ (yellow) and map1∆ (grey) cells compared to wt cells (log2 scale). Whiskers extend from the 10th to the 90th percentile. The distribution of all proteins is shown at the top. GO terms are indicated to the left and the number of plotted Pfk2p mRNA target encoded proteins within the respective GO group is indicated in brackets. Asterisks refer to P-values determined in a Student’s t-test with Welch’s correction comparing the distribution of Pfk2p mRNA target encoded proteins assigned to specified GO term with the distribution of all measured features: ***P <0.001; **P <0.01; P <0.05.
To further establish proteome-wide connections to mRNA targets, we mapped experimentally determined Pfk1p and Pfk2p targets to changes in proteomes (MS data was available for 730 and 620 of Pfk1p and Pfk2p mRNA targets, respectively) (Fig. 4C). We found only a slight overrepresentation of proteins encoded by Pfk1p and Pfk2p mRNA targets among the proteins selected with 10% FDR in pfk1∆ and pfk2∆ cells, respectively (Pfk1p: 111 proteins out of 730 detected Pfk1 target proteins, 15%, p = 0.043; Pfk2p: 274 of 620 detected Pfk2 targets, 44.1%, p = 0.3). Of note, choosing more stringent FDRs, significant associations were also revealed for Pfk2p mRNA target encoded proteins (FDR < 1%, 88 proteins, p = 0.051; among those 41 less abundant in mutants, p = 0.002). However, combinatorial control through other RBPs and secondary effects induced through metabolic changes in pfk mutants may confuse ‘simple’ associations between changes in protein levels and the encoded mRNA targets. Hence, we searched for specific instances on subsets of mRNAs attributed to particular GO terms. Indeed, we found significant directional responses at the protein levels for functionally related mRNA subsets (Fig. 4D). For instance, reduced protein levels for Pfk2p mRNA targets coding for the ‘mitotic cell cycle’, ‘DNA replication’, and Pol I related subjects like ‘tRNA processing’ were found, including Bub3p. Importantly, the generally decreased abundance of those functionally related target groups, which was slightly more prevalent in pfk2∆ than in pfk1∆ cells, agrees with Pfk2’s proposed role as a translational activator.
PFK2 modulates cell size and cell cycle progression independent of its catalytic activity
We wondered whether the observed reduced expression of cell cycle related mRNAs is also mirrored by the cell’s phenotype. pfk2∆ cells elicit a slow-growth phenotype not seen with pfk1∆ cells but the source of these phenotypic differences remained unclear [21]. Furthermore, it was noted that in vitro measured catalytic activity is depleted in cell extracts derived from either pfk gene knock-out strains, indicating that catalytic activity is unlikely a major source driving the pfk2∆ phenotype [21]. To confirm these findings in our yeast strains, we first compared the growth of wild-type, pfk1Δ and pfk2Δ cells, and reintroduced wild-type genes as well as mutants with compromised catalytic activity for functional complementation (Fig. 5A). As expected, pfk2Δ cells showed a slow-growth phenotype while pfk1Δ cells grew like wild-type cells, confirming that the presence of heteroctameric Pfk1p-Pfk2p complexes is not required for cell fitness. Additionally, in agreement with previous observations [20], the slow-growth of pfk2∆ cells was rescued by reintroduction of the wild-type PFK2 gene as well as catalytically dead mutants expressed from a single-copy plasmid. Specifically, catalytic mutants that contained either an exchange of aspartate 348 for serine (D348S) that eliminates the proton acceptor in the substrate binding site reduced activity up to 50%, or a mutation of aspartate 301 for threonine (D301T), which affects the catalytic Mg^2+^-ATP binding site reduced catalytic activity by almost 40% compared to PFK2 rescued cells (Fig. 5B). To see whether both mutations could act synergistically, we further created a DM at both sites (D301T, D348S). The DM showed reduced enzymatic activity and rescued the growth defect of pfk2∆ cells like single catalytic mutants and the wild-type PFK2 gene. This suggests that the remaining catalytic activity is contributed by the intact Pfk1p subunits in complex with the catalytically dead Pfk2p (since neither extract from pfk1∆ nor pfk2∆ cells have measurable enzymatic activity in vitro) as observed previously [20]. Conversely, the slow-growth phenotype of pfk2∆ cells can obviously not be explained by the reduced catalytic activity of Pfk2p.
*Slow growth and cell cycle progression defects of pfk2∆ cells can be rescued with catalytic mutants. (A) Doubling times of indicated yeast strains grown in YPD media at 30°C. (B) PFK activity measurements. Bar depicts PFK specific activity (mU/mg) of 5–6 paired replicate measurements in extracts derived from indicated cells. Standard deviations are indicated in brackets; Student’s t-test, *P <0.05, **P <0.01. (C) Immunoblot of RIC eluates for indicated Pfk2-V5 tag expressing strains. The input extract is shown to the left. (D) Cell size analysis of indicated strains grown in YPD media to mid-log phase. The cell diameter (µm) relating to the cell size is depicted for > 100 cells. The median is marked with a red line. P-values indicate the significance of increased cell diameters relative to wild-type cells (Mann–Whitney test; **P <0.01, ***P <0.0001). (E) Analysis of cell cycle progression. Yeast strains indicated at the top were synchronized in late G1 phase using α-factor treatment. After release from α-factor, samples were collected every 10 min (y-axis), fixed, and analysed for DNA content (1C/ 2C; x-axis) at specified time points with flow cytometry. Red arrows depict aberrant ‘1C’ and ‘2C’ peaks in pfk2∆ cells.
To evaluate potential impact on the association of the Pfk complex with RNA, we performed RIC with the different mutants. To improve the visualization of Pfk2p-RNA interactions, we introduced a triple V5 tag at the C-terminal end of Pfk2 into plasmids, enabling specific detection of tagged Pfk2 proteins with V5 antibodies (Fig. 5C). We also assessed the recovery of untagged Pfk complexes in separate experiments with Pfk antibodies (Supplementary Fig. S5). We found that the complementation of pfk2∆ cells with single (D348S) and double (DM) catalytically-dead pfk2 mutants did not greatly affect poly(A) RNA association in vivo. We though observed some diminished RNA associations with the D301T mutant (∼35% reduction compared to D348S or DM mutants). However, whether this effect relates to altered stability of the protein in the extract or compromised RNA binding will need further investigation, especially since no impact on RNA binding was seen with the DM. Overall, unlike for the enzymatic activity, we conclude that formation of an RNA-binding complex is not significantly compromised in those mutants, which can all rescue the slow-growth phenotype (Fig. 5A).
Since deletion of cell cycle regulators can influence cell size, including cln3∆ that shows increased cell size and promotes timely progression from G1 to S phase [57], we assessed the size (diameter) of wild-type, pfk1∆ and pfk2∆ mutants, using cln3∆ cells as a reference control (Fig. 5D). Indeed, pfk2∆ but not pfk1∆ cells showed significantly increased cell sizes as compared to wild-type cells. The increased size of pfk2∆ cells was rescued by reintroduction of plasmid-born PFK2 wild-type (Pfk2R) as well as the catalytically dead DM mutant, while no complementation was seen with the empty plasmid (Pfk2E). Hence, the increased size of pfk2∆ cells seems independent of Pfk2ps enzymatic activity. Nonetheless, we wish to note the remaining slightly increased cell size upon rescue with the D301T mutant, which is reminiscent to the potentially altered stability and impaired RNA-binding seen with RIC experiments.
Finally, we assessed potential cell cycle defects within the single deletion mutants. Cells were synchronized by α-factor treatment, which arrests cells in G1 by preventing passage through START and therefore captures cells that have completed much of early G1. Whilst pfk1Δ mutants exhibited cell cycle progression like wild-type cells, we observed that a significant portion of pfk2Δ cells showed delayed progression from G1 into S phase and further progression into G2/M phase (Fig. 5E). The cell cycle defect was fully rescued by reintroducing either wild-type PFK2 or the catalytically inactive DM. Notably, cells carrying PFK2 rescue plasmids progressed through the cell cycle more rapidly than wild-type or pfk1Δ/pfk2Δ strains, likely due to LEU2 overexpression from the selectable marker in the plasmid. Hence, as described above, the cell cycle defect in pfk2Δ cells can neither be explained by the absence of heteromeric complexes nor the compromised enzymatic activity of Pfk2p. Accordingly, we postulate that RNA binding functions, and the observed effect on translation in cell cycle regulators, constitute a novel “moonlighting” role specific for Pfk2p.
Discussion
Although PFK has been extensively characterized for its central role in glycolysis over the past seven decades, only recent findings have implicated PFK in RNA binding, with the functional relevance of this interaction remaining largely unresolved. In this study, we provide insight into yeast PFK-RNA binding and uncover a functional divergence between its two paralogous subunits, Pfk1p and Pfk2p. Our data reveal that while both subunits exhibit RNA-binding capacity, Pfk2p uniquely functions as an mRNA-binding protein bearing dsRNA unwinding activity with 5′ to 3′ directional polarity. Importantly, Pfk2p promotes the post-transcriptional ‘upregulation’ of mRNAs involved in cell cycle progression and facilitates G1-to-S phase transition independently of its canonical enzymatic role. These findings support a model in which Pfk2p serves as a regulatory hub or ‘molecular switch’, linking cellular metabolic status to the translational control of cell cycle regulators - thereby enabling adaptive proliferative responses to fluctuating nutrient conditions in budding yeast.
Pfks are selective mRNA binding proteins
We used a chemical crosslinking approach to identify cellular RNA targets for Pfk1p and Pfk2p. This analysis revealed that both Pfk proteins bind to a substantially overlapping set of hundreds of different mRNAs and a few noncoding RNAs. Given that Pfk1 and Pfk2 are paralogous proteins which form a stable hetero-octameric protein complex, this significant overlap in RNA targets may be expected. However, alternate multi- and monomeric complexes may also exist, as Pfk proteins can associate with cellular RNA independently of their paralog (Fig. 1A). The commonly bound mRNAs encode functionally related protein groups, a characteristic of canonical RBPs that form post-transcriptional operons or RNA regulons for coordination of post-transcriptional events [58, 59]. These functional associations—particularly those linked to the mitotic cell cycle—suggested roles for Pfk proteins in cell cycle control reminiscent to previous observations [60], which we finally confirmed (Fig. 5E). Furthermore, bioinformatic analysis identified short and unstructured single-stranded RNA sequences (UC-, GA-, AU-, or U-rich) that interact with Pfk proteins (Fig. 2B–D). Hence, Pfks selectively bind to RNA regions, albeit not being as specific as canonical RBPs but still more selective than most RBP without canonical RBDs [61], such as GAPDH [62]. In this context, the binding selectivity of Pfks could align with RNA helicases that do not display well-defined sequence or structural preferences, besides the polarity requirements for unwinding RNA duplexes of some enzymes [63]. As RNA-binding was reported for PFK orthologues across diverse organism and human cells [1, 8, 64], it will be interesting to explore insofar RNA-binding selectivity and functions are evolutionary conserved.
Pfk2p has nonconventional RNA unwinding activity
RNA helicases are a prominent class of enzymes that use ATP to bind or remodel RNA or RNPs with diverse functions from splicing to translation and degradation [63, 65]. Surprisingly, we discovered that purified recombinant Pfk2p, but not Pfk1p, demonstrated unwinding activity with RNA substrates bearing single-stranded regions at the 5′, but not at 3′ end, defining polarity or directionality with respect to the duplex. This contrasts with the DEAD-box RNA helicase eIF4A, a member of the SF2 subfamily, which exhibits only limited strand displacement directionality (Fig. 2E). It may rather resemble the eukaryotic RNA helicases of the SF1b family, such as MOV10 with 5′ to 3′ directionality for strand displacement [66]. At this point, it is though unclear whether RNA displacement by Pfk2p occurs via a processive unidirectional mechanism or repetitive local strand separation, and further experiments would be required to test these instances. Most eukaryotic RNA helicases bear two RecA subdomains that close around the RNA upon NTP binding and promote strand-displacement. Analogous to this, a conformational shift of Pfk2p induced and stabilized by ATP could prompt the low-activity T-state that involves re-orientation of the N-terminal half (containing the active site) and C-terminal half (housing allosteric regulatory sites) in enzyme subunits [10, 67]. The low activity T-state could facilitate RNA interaction and enable directional unwinding of the RNA. It is though unclear how this occurs, but it may involve the assembly of several Pfk2ps on the RNA leading to larger complexes as possibly seen in REMSA experiments (Fig. 2D and Supplementary Fig. S2D). Finally, a conformational shift to the high-activity R-state with bound ADP/AMP may be required for Pfk2p’s final release from RNA substrates. In this regard, the release of Pfk2p from the RNA may not be required to measure activity in our assay since proteins were in excess (1 µM) over RNA substrates (50 nM) and hence, ATP that co-purifies with Pfk2p may be sufficient for activity without the need for repeated cycles of interactions. Although speculative, this model could also account for the persistence of unwinding activity in the presence of AMPPNP, a nonhydrolysable ATP analogue that potently inhibits catalytic kinase activity while mimicking ATP-induced allosteric regulation, as well as for the reduced activity upon competition with ADP or AMP, which may shift Pfk2p toward the high-activity R-state (Fig. 2D).
Pfk2p can promote the translation of cell cycle mRNAs
As observed for many RNA helicases, we found that translation is disrupted in pfk2Δ cells, further suggesting that Pfk2p specifically enhances the translation of mRNAs coding for pivotal regulators of cell cycle progression, such as CLN3 and BUB3 (Fig. 3D). Notably, these translational defects occur exclusively in pfk2Δ cells. This correlates well with the observation that only Pfk2p, and not Pfk1p, associates with ribosomes in a manner dependent on the cell’s energy state, as found in the glucose-depletion assays (Supplementary Fig. S4). Since Pfk2p demonstrates RNA unwinding activity and associates with ribosomes, it seems likely that Pfk2p facilitates translation by resolving mRNA secondary structures or displacing mRNP complexes, thereby improving ribosome processivity and enhancing translation efficiency.
Proteomic analysis of pfk1Δ, pfk2Δ and map1Δ yeast mutants revealed that pfk2Δ exhibited the most significant changes on protein levels, with a substantial overlap of all three mutants among ‘downregulated’ proteins, particularly those involved in cytoplasmic translation and ribosome biogenesis (Fig. 4A and B). Likewise, proteins related to oxidative stress showed increased levels in all mutants, indicating a generalized stress response. In contrast, proteins with roles in carbon metabolism were exclusively ‘upregulated’ in both pfk1Δ and pfk2Δ, likely compensating for reduced glycolytic activity. Moreover, proteins related to DNA replication and transcription were preferentially downregulated in pfk2Δ, suggesting a subunit-specific impact on the proteome (Fig. 4B). Further analysis connected Pfk1p and Pfk2p mRNA targets to these proteomic changes, particularly in pfk2Δ. Specifically, the analysis of functional GO clusters revealed a significantly lower abundance of proteins encoded by Pfk2p mRNA targets associated with the mitotic cell cycle, DNA replication, and tRNA processing, supporting the suggested role of Pfk2p in translational activation (Fig. 4D). Overall, these results affirm Pfk2p’s specific impact on mRNA target expression and align with its role in cell cycle progression.
Physiological relevance of Pfk2p mRNA binding
Phenotypically, pfk2∆ but not pfk1∆ cells, exhibit a pronounced slow-growth phenotype that can be effectively rescued by reintroducing exogenously expressed PFK2 or its catalytically dead mutant alleles on a centromeric vector (Fig. 5A). Consistent with earlier reports showing no detectable catalytic activity in extracts from either single mutant [19, 21, 22], our findings strongly suggest that the growth phenotype in pfk2∆ cells is not dependent on the enzyme’s catalytic functions (Fig. 5B). Likewise, we observed that pfk2∆, but not pfk1∆, cells showed significantly increased cell sizes and delayed progression from the G1 to the S phase of the cell cycle independent of the enzyme’s catalytic activity (Fig. 5D and E). Combined with evidence that Pfk2p binds to mRNAs coding for cell cycle-related proteins—and RNA binding remaining intact in catalytically inactive mutants (Fig. 5B)—alongside Pfk2p’s role in translation and reduced levels of cell cycle proteins in pfk2∆ cells, we propose that the cell cycle defect arises from Pfk2p’s function as a post-transcriptional regulator of cell cycle genes; emerging as a further RBP involved in the post-transcriptional control of cell cycle related mRNAs [48]. For instance, previous studies have shown that CLN3 is post-transcriptionally regulated by Whi3p, an RNA-binding protein that associates with CLN3 mRNA, promotes its destabilization and affects translation [68], and localizes it to discrete cytoplasmic foci [69].
Is Pfk2p a molecular relay switch?
We further propose that Pfk2p acts as a molecular relay for transduction of metabolic signals to the post-transcriptional level to control cell cycle progression and possibly other processes (Fig. 6). When cellular energy levels are low with limited protein synthesis activity, Pfk proteins preferentially adopt the high activity state (R-state) through binding of AMP/ADP at allosteric activator sites and engage in glycolysis for ATP production and weak binding to RNA. Conversely, when energy levels become sufficiently high, Pfk2p adopts the low-activity conformation (T-state) with low glycolytic activity but increased binding to mRNAs with unwinding activity. Through Pfk2p’s association with ribosomes, the translation of cell cycle and other transcripts is facilitated which ultimately promotes cell cycle progression. In this way, Pfk2p could constitute a ‘molecular relay switch’ to balance cellular needs: cell proliferation is repressed in energy deprived cells where resources must be allocated to energy production to secure cell survival, while favourable conditions mounted by high energy levels reallocate resources towards production of proteins for cell proliferation and enabling colony expansion. Such a molecular switch at the post-transcriptional level can be highly efficient and rapid, allowing for immediate response to changes in available nutrients and circumvents the need for reshaping the transcriptome. As such, it creates an ‘economic’ solution for short-term adaptation to changing energy levels imposed by the availability of nutrients coordinating glycolytic flux and cell proliferation. Given that numerous metabolic enzymes have been shown to interact with RNA, potentially affecting bound RNAs or the enzyme’s activity, it is plausible that many more relay switches between intermediary metabolites and RNA could exist. In particular, through specialization of paralogous enzymes [70], forming a vast network that we are only beginning to discover.
A relay switch model for yeast Pfk2p. The enzymatic Pfk1-Pfk2 protein complex is shown to the left as a tetramer for simplicity; allosteric adenosine phosphate binding is indicated only (ATP required for catalysis is not shown). The R-state refers to high enzyme activity; the T-state to low activity. Cellular energy levels are indicated with a colour bar (AMP/ADP – ATP levels) to the left. Mg²⁺-ATP favours the T-state, thereby enhancing Pfk2p association with RNA and ribosomes, which could in turn stimulate mRNA translation and promote cell-cycle progression (schematic, right). Schematic components were created with BioRender.com (https://BioRender.com/yycdn34).
Limitations of the study
Our study concentrated on the yeast S. cerevisiae Pfk proteins, defining RNA-related functions for the Pfk2p paralog, while possible other functions for the Pfk1p paralog remain to be determined. Furthermore, since PFK proteins from diverse species have been shown to bind RNA, it will be of interest to investigate whether the described yeast Pfk2p RNA-binding preferences and RNA unwinding activity is evolutionary conserved, restricted to specific isoforms, and possibly associated with metabolic diseases and/or cancer.
Despite our intense efforts, we have not been able to generate convincing RNA-binding mutants of Pfk2p, which could allow for distinction between RNA binding and enzymatic activity. Hence, a systematic mutational approach based on structural data of Pfk2p in complex with RNA will be of prime interest to deepen our understanding of underlying RNA–protein interactions and to pinpoint relevant domains for RNA unwinding activity; as well as to solidify the proposed involvement for translation of mRNA targets and cell-cycle progression. Regarding the latter, α-factor arrest captures cells that have completed much of early G1 and therefore provides limited resolution for interrogating events occurring prior to START. Synchronization at alternative cell-cycle stages, such as S phase or G2/M, may provide additional insight into the role of Pfk2p as a post-transcriptional regulator of the cell cycle. Furthermore, most of our experiments were performed in asynchronous cell populations, while it is unclear whether RNA-binding activity could depend on a particular phase in the cell cycle. This seems important because oscillations between reductive and oxidative metabolism during mitosis have been observed, opening the possibility that Pfk2p might alternate its roles between mRNA binding and glycolysis as cells transition through various phases of the cell cycle.
Supplementary Material
gkag184_Supplemental_Files
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Hentze MW, Castello A, Schwarzl T et al. A brave new world of RNA-binding proteins. Nat Rev Mol Cell Biol. 2018;19:327–41., 10.1038/nrm.2017.130.29339797 · doi ↗ · pubmed ↗
- 2Gebauer F, Schwarzl T, Valcarcel J et al. RNA-binding proteins in human genetic disease. Nat Rev Genet. 2021;22:185–98. 10.1038/s 41576-020-00302-y.33235359 · doi ↗ · pubmed ↗
- 3Albihlal WS, Gerber AP. Unconventional RNA-binding proteins: an uncharted zone in RNA biology. FEBS Lett. 2018;592:2917–31. 10.1002/1873-3468.13161.29908064 · doi ↗ · pubmed ↗
- 4Scherrer T, Mittal N, Janga SC et al. A screen for RNA-binding proteins in yeast indicates dual functions for many enzymes. P Lo S One. 2010;5:e 15499. 10.1371/journal.pone.0015499.21124907 PMC 2988813 · doi ↗ · pubmed ↗
- 5Matia-Gonzalez AM, Laing EE, Gerber AP. Conserved m RNA-binding proteomes in eukaryotic organisms. Nat Struct Mol Biol. 2015;22:1027–33. 10.1038/nsmb.3128.26595419 PMC 5759928 · doi ↗ · pubmed ↗
- 6Beckmann BM, Castello A, Medenbach J. The expanding universe of ribonucleoproteins: of novel RNA-binding proteins and unconventional interactions. Pflugers Arch - Eur J Physiol. 2016;468:1029–40. 10.1007/s 00424-016-1819-4.27165283 PMC 4893068 · doi ↗ · pubmed ↗
- 7Fuller GG, Han T, Freeberg MA et al. RNA promotes phase separation of glycolysis enzymes into yeast G bodies in hypoxia. Elife. 2020;9:e 48480. 10.7554/e Life.48480.32298230 PMC 7162659 · doi ↗ · pubmed ↗
- 8Wegener M, Dietz KJ. The mutual interaction of glycolytic enzymes and RNA in post-transcriptional regulation. RNA. 2022;28:1446–68. 10.1261/rna.079210.122.35973722 PMC 9745834 · doi ↗ · pubmed ↗