Rapalink-1 uncovers TOR-related genes and a metabolic feedback system controlling TOR activity and lifespan in fission yeast

Our septation and cell size results indicate phenotypes related to TORC1 inhibition as previously shown and described^7–9,16. We, therefore, investigated whether drug treatments at these concentrations affect multiple aspects of TORC1 functions⁴. Negative cross-regulation between the AMPK and TORC1 pathways have been previously reported in fission yeast^23,24. We, therefore, examined the phosphorylation status of the fission yeast serine/threonine protein kinase AMPK catalytic (alpha) subunit Ssp2. Following two and five hours of treatment with 100 nM of rapamycin or rapalink-1, no observable changes in Ssp2 phosphorylation are found (Fig. 1E). TORC1 negatively regulates autophagy⁴. Examination of GFP-Atg8 patterns²⁵ following drug treatment in the aforementioned time points, do not show a significant degree of processing²⁵ indicating that effects are not pronounced, at least in these timeframes (Fig. 1E). However, global protein translation might be suppressed five hours post-treatment with both drugs, as seen through increase in phosphorylation of eIF2α while its total levels remain unchanged at both of the 2 h and 5h time points, examined (Fig. 1E, numbers indicate the ratios of total to phospho-eIF2α). We wondered whether drug treatment could affect the activity of Maf1, a Pol III repressor²⁶, through its TORC1-dependent phosphorylation status²³. Immunoblotting for Maf1-pK revealed that following treatment with both rapamycin and rapalink-1, Maf1-pK becomes dephosphorylated, more evidently in five hours of treatment (Fig. 1E, numbers indicate ratios of second band-a dephosphorylated form to top-phosphorylated form^23,24). While the above results indicate effects on TORC1-dependent processes, we examined whether TORC2-dependent effects are evident following rapalink-1 treatments. Towards this, we have examined the phosphorylation status of Gad8, a kinase dependent on TORC2⁸. The ratios of phosphorylated to dephosphorylated Gad8 forms do not change in the treatments and timelines examined (Fig. 1E, numbers indicate the ratio of phosphorylated-top and dephosphorylated-bottom forms). To confirm effects of rapalink-1 on protein translation (Fig. 1E) we have analysed the levels of Psk1-myc (Psk1 is an S6 kinase²⁷) following nitrogen starvation as well rapamycin and rapalink-1 treatments for two and five hours. In nitrogen starvation global translation and Psk1-myc drop drastically (Fig. 1F). Rapamycin (Rm) and rapalink-1 (Rl) treatments result in reduced amounts of Psk1-myc with the total protein content remaining in comparable levels with those of the control (Fig. 1F). This result strengthens the data, supporting that global protein translation is affected in rapalink-1 treatments as previously seen with rapamycin in fission yeast⁸.

We have previously shown that the GATA transcription factor Gaf1 mediates Pol II- and Pol III-mediated transcription downstream of TORC19. Gaf1 is localised mainly in the cytoplasm and upon TORC1 inhibition translocates to the nucleus and binds to target genes⁹. We examined localisation of Gaf1-GFP (controlled by the endogenous Gaf1 promoter as described in Rodríguez-López et al., 2020). Gaf1-GFP signal is found throughout the cell in untreated cells (Fig. 1G). Treatment with rapamycin and rapalink-1 results in nuclear localisation of Gaf1-GFP within 5 min (Fig. 1G) indicating that both drugs can inhibit TORC1 at sufficient levels to allow Gaf1 dephosphorylation and translocation. One of the major actions of Gaf1 is to repress all the Pol III transcribed tRNAs in fission yeast upon TORC1 inhibition. The result of Gaf1-GFP localisation agrees with Maf1 activation seen through Maf1-pK dephosphorylation (Fig. 1E). Overall, our results show that rapalink-1 primarily targets fission yeast TORC1 and, like rapamycin, can inhibit several but not all aspects of TORC1-dependent functions in the examined concentrations. We cannot exclude the possibility that in higher concentrations or prolonged exposures to the drug, TORC2 will not be affected (as previously has been shown in the case of rapamycin²⁸).

We then assessed the CLS of fission yeast cells when treated with rapamycin and rapalink-1 in YES media and at OD₆₀₀ 0.5 at 100 nM concentration. As expected, rapamycin induces CLS extension (Fig. 1H, log-rank p = 1.25*10⁻¹²). Similarly, rapalink-1 prolongs lifespan in similar fashion to rapamycin (Fig. 1I, p = 2.2*10⁻¹²) demonstrating the potential of the drug and its action in ageing through TORC1 inhibition. Rapalink-1 lifespan extension depends on fkh1 (Fig. 1J, p = 0.054) as has previously been observed in the case of rapamycin⁸ and observed here too (Fig. 1J, p = 0.067) demonstrating that the effects of rapalink-1 in these conditions are TORC1-dependent.

Beyond proteostasis, autophagy and global translation levels, TORC1 activity controls gene expression at the level of transcription through multiple transcription factors^5,35,36. We therefore performed RNA-seq analysis on fission yeast cells treated with 100 nM rapamycin and rapalink-1 for 5 h. This timepoint was chosen based on the observed changes in the relevant markers described at Fig. 1E, F.

High-throughput phenotyping approaches have majorly contributed to characterising the CLS of haploid fission yeast mutants^33,35. TORC1 is a major pro-ageing pathway, and TORC1-dependent genes are likely to be related in lifespan regulation. We, therefore, examined whether differentially expressed genes from rapalink-1-treated cells are linked with long or short reported lifespans of the corresponding mutants. While many rapalink-1 upregulated genes have reported lifespans, a significant number (331 genes) have no lifespan annotations (Fig. 3E). GO enrichment of these genes demonstrates that they are related to amino acid transport, amide transport, galactose, tyrosine, arginine and proline metabolism (Fig. S3A). Likewise, 150 genes that are found to be downregulated in rapalink-1 treated cells have not been annotated for lifespan (Fig. S3B) and these are related to L13a-mediated translational silencing, ribonucleoside biosynthesis, response to stimulus as well as to alanine, aspartate and glutamate metabolism (Fig. S3C). Examination of upregulated genes not linked yet to lifespan, revealed that arginine catabolism and specifically all the agmatinase enzymes involved in processing agmatine to putrescine are boosted following TORC1 inhibition (Fig. 3F). To validate the results from RNA-seq, we performed qPCRs for all three agmatinases from fast-growing cells (OD₆₀₀ = 0.5) treated with rapalink-1 for 5 h compared to untreated cells. The results from qPCRs showed a 2.5- to 3-fold increase in all agmatinases, confirming the RNAseq data (Fig. 3G). We therefore wondered whether agmatinases are required for normal CLS in fission yeast. CLS assays reveal that agm1Δ (log rank p = 9.7 × 10⁻⁵, Fig. 3H), agm2Δ (log rank p = 1.5 × 10⁻⁵, Fig. 3I) and agm3Δ (log rank, p = 3.9 × 10⁻⁵, Fig. 3J) mutant cells are short-lived compared to wt cells. In addition, cells mutant for both agm1 and agm3 are also short-lived (log rank p = 3.3 × 10⁻⁶, Fig. 3K). Rapalink-1 prolongs CLS in wt cells (Fig. 1I). But does it have effects on the agmatinase mutants? We therefore performed CLS on single agm mutants with and without rapalink-1 treatments. In all cases rapalink-1 can extend lifespan (Fig. S4A–C), probably due to compensation between the agmatinase enzymes. The lifespan extension for agm2Δ is less pronounced compared to the ones observed for agm1Δ and agm3Δ. Nevertheless, the change is significantly different in all cases (log rank p value < 0.01). We are not able to say whether the differences in rapalink-1-dependent lifespan extension signifie specialisation in agmatinase requirements, but it is plausible.

Close inspection of the interactors' list reveals that genes coding for stress and amino acid sensing that inhibit TORC1 such as gcn1⁴¹ and fil1⁴² positively interact with agm1 (Fig. 5E, green fonts). On the other hand, the TORC1 core component tco89³, negatively interacts with agm1 (Fig. 5E, red fonts). These results could mean that the absence of agm1 (and possibly the rest of the agmatinases) promotes growth that is enhanced through the absence of stress response players. The cellular growth enhancement is TORC1-dependent as absence of tco89 results in reduced cellular fitness/growth. Therefore, in this scenario, agmatinases may provide a metabolic regulatory feedback for TORC1 activity in fission yeast: when TOR is inhibited, cells process vacuole-stored amino acids such as arginine⁴³, thus ensuring that TOR inhibition signals are maintained, relayed within the cell and preventing anabolism while preparing for catabolic processes and autophagy. In this case, agm mutants would showcase phenotypes related to increased TORC1 activity. Indeed, we have found that all agmatinase deletion mutants have short chronological lifespans compared to wt cells (Fig. 3H–K). Agm1 mutants have already been reported to be fast-growing⁴⁴ and to have decreased mating efficiency⁴⁵, both cellular characteristics of increased TORC1 activity. We have then assessed cell size upon division for normal (wt) and agmatinase mutants. Our results demonstrate that agm mutant cells are larger (Fig. 5F, ***: p < 0.001, Wilcoxon testing) strengthening the case of increased TORC1 levels as the latter controls both temporal and spatial aspect of fission yeast growth. TORC1 activity growth through protein translation via phosphorylation of S6 kinases (such as Sck1, Sck2 and Psk1). Indeed, phospho-S6 activity is increased in agm mutants, almost 5-fold for agm1Δ, 3.4-fold for agm2Δ and 1.7-fold for agm3Δ (Fig. 5G). In addition, P-eIF2α, a negative marker of translation also used for assessing 'stress' levels of cells, is reduced in agm mutants (Fig. 5G, 0.5-fold in agm1Δ, 0.1-fold for agm2Δ and 0.3-fold for agm3Δ). Interestingly, while tco89 mutant cells are long-lived⁸, agm1Δ tco89Δ are short-lived as single agm1 mutants (Fig. 5H). This result shows that in the case of the CLS phenotype, agm1 is epistatic of tco89. Our data reveal that the agmatinergic branch of arginine catabolism is favoured during TOR inhibition and forms a regulatory loop required for maintaining a lower TOR level, according to what environmental pressures dictate (Fig. 5I).

972 h− was used as the wild-type. The strain containing C-terminally GFP-tagged Gaf1 (gaf1-GFP kanMX6) has been generated according to⁵⁸ with gaf1 being under the control of the endogenous promoter and as in refs. ^7,9,59. Strains Maf-1-pK and GFP-atg8 have been previously described^23,25. We have also used h+ agm1::kanMX6; h+ agm2::kanMX6; h+ agm3::kanMX6; h+ fkh1::kanMX6; agm1::natMX6 agm3::kanMX6; agm1::natMX6 tco89::kanMX6³⁰. Other strains have been obtained from the deletion collection and verified using PCR according to manufacturer's instructions. Microscopy was performed using an EVOS M5000 and Leica DMRA2 epi-fluorescent microscope fitted with a monochrome Orca-ER camera from Hamamatsu Photonics. YES (Yeast extract with supplements) and EMM2 media (Edinburgh Minimal Media 2) (Formedium) are used as indicated in the figures and main text. Drug treatments have been conducted in fast-growing cultures at OD₆₀₀ 0.5 (see main text for details). Liquid cultures were grown at 32 °C with shaking at 130 rpm.

Cells were treated with 100 nM rapamycin or 100 nM rapalink-1 for 2 or 5 h as indicated in figures and main text. Cells were disrupted using RIPA buffer and glass beads in a FastPrep 5 G MP lysis system. Antibodies directed against phospho-eIF2a (#9721, 1:1000), eIF2a (#9722, 1:1000), phospho-Ssp2 (#50081, 1:1000), Phospho-(Ser/Thr) Akt Substrate Antibody (#9611, 1:2000) were purchased from Cell Signalling Technologies. The antibody directed against myc (ab32, 1:1000) was purchased from Abcam while antibodies directed against V5 (sc-81,594, 1:1000) and HA (sc-7392, 1:1000) were purchased from Santa Cruz Biotechnology. Secondary antibodies Goat Anti-Mouse IgG H&L (HRP) (ab205719, 1:5000 in all cases except when V5 and GFP was the primary antibody where it was used at 1:10,000) and Goat Anti-Rabbit IgG H&L (HRP) (ab205718, 1:5000) were purchased from Abcam while the ECL Western Blotting Detection system was from Pierce™. ɑ-tubulin (T5168, 1:1000) and anti-GFP (11814460001, 1:1000) were purchased from Sigma-Aldrich. ɑ-tubulin and Ponceau S staining have been used as loading controls for western blots as and where indicated. Western blotting quantifications have been performed using ImageJ/Fiji^29,60.

To measure septation index, cells were resuspended in calcofluor solution (#18909, Sigma-Aldrich) and incubated at room temperature for five minutes. Cells were visualised using an EVOS M5000 microscope. Percentages of septation from 200 cells for each time point have been recorded. For cell length at division, measurements of 50–100 septated cells were conducted using Fiji/ImageJ software^7,60.

CLS was determined through counting colony-forming units every day and normalising numbers with day zero⁸. Fission yeast cells are grown to the stationary phase. However, in this study, cells are left for an additional 48 hours as seen in previous related studies¹⁵ and then the assay starts with this time point considered as the 100% survival for the time course. Two or three independent biological repeats with each repeat having three technical replicates have been used. Log rank tests are performed for all CLS assays with CFUs and percentages used in the statistical testing. Statistical significance has been also validated with AUC measurements⁶⁰ as also seen in ref. ⁹. Lifespan lengths vary largely in yeasts depending on the time point used following stationary phase entry and the media used. Different batches of YES media can generate different lifespan curve patterns depending on the Yeast Extract used in the particular batch. We have used a unique YES batch (Formedium) throughout the study.

Untreated as well as rapamycin and rapalink-1 treated cells (treatment was at OD₆₀₀ = 0.5) for 5 h were harvested and processed for RNA isolation⁸ using acidic phenol-chloroform extraction method followed by RNAeasy (Qiagen) cleanup with on column DNase treatments before double elution. Following appropriate quality control steps, RNA library formed by polyA capture and Illumina 150 bp paired-end sequencing followed by standard R-based bioinformatic analysis was performed. Differential gene expression analysis of two conditions/groups (two biological replicates per condition) was performed using DESeq2R package (1.20.0). Genes with an adjusted P value <=0.05 found by DESeq2 and at various fold cutoffs (see main text) were assigned as differentially expressed. Sequences are deposited at GEO (accession number: GSE272269).

As in the case for RNA sequencing untreated as well as rapalink-1 treated cells (at OD₆₀₀ = 0.5, treatment for 5 h) were harvested and processed for RNA isolation⁸ using acidic phenol-chloroform extraction method followed by RNAeasy (Qiagen) cleanup with on column DNase treatments before double elution. cDNA was prepared using LunaScript^® (NEB) while qPCR reactions were conducted on a QIAquant96 2plex cycler (Qiagen) using Luna^® (NEB) according to the manufacturer's instructions. Fold-changes of expression were calculated using the Ct method and normalising with alpha-tubulin 2 (atb2). Primers used are: agm1L1: 5'-TTTTGGAGGCGGCAAATCAA-3'; agm1R1: 5'-CCAACTCTGTCACGGATCCT-3'; agm2L1: 5'-TACTGTGCTTCCTCGAGTCC-3'; agm2R1: 5'-CCATCTGCTTCATCGCCATC-3'; agm3L1: 5'-TGATCAACAACGGCACATCC-3'; agm3R1: 5'-GCCAATCCAGGATCGACAAC-3'; atb2L1: 5'-TTCTGTGTATCCGGCTCCTC-3'; atb2R1: 5'-AGAGGCGGTGATGGAAGAAA-3.

The deletion library³⁰ was arrayed on solid YES media at 1536-spot density, with each strain represented by four spots with various concentrations of drugs as indicated for each assay. Plates were incubated at 32 °C and high-resolution images of the plates were acquired. Colony size quantitations were then performed using the R package Gitter⁶¹. Median colony sizes were calculated for each plate and replicate. Strain colony size data per condition were normalised to the corresponding growth on YES.

SGAs were performed using the Bioneer v5.0 haploid library³⁰ with each interaction examined in quadruplicate⁷. Query strains were the following: h− agm1::natMX6, and control query h− ade6::natMX6. Colony size was used as a proxy for double mutant fitness. Colony size measurements were obtained using the gitter package⁶¹. Library mutants that fell within 30 kb distance of query mutations were excluded from the respective datasets to avoid spurious interactions as well as mutants for which the ade6 double mutant colony size was less than 50 pixels as they were deemed to be present in too little amount for accurate interaction values to be calculated. Colony sizes were normalised to the plate median accounting for the effect of the query mutation and plate-specific effects and normalised for column- or row-specific effects. Medians of colony sizes of the SGAs were normalised with respect to the ade6 SGA (which represents fitness of library single mutants as this query does not affect the library) to calculate the value of genetic interactions between query and library mutants. Interactions with high within-replicate variability were excluded. Finally, the logarithms in base 10 of the interaction were used for interaction scores.

Enrichment analyses were performed using Metascape⁶². p-values were corrected for multiple tests according to FDR. Enrichment analysis was conducted by comparing lists of interacting genes to all genes in the dataset (Bioneer v5.0 collection strains³⁰).

CLS assays have been performed in two or three biological repeats with 3 technical repeats for each biological repeat. Statistical differences were determined using log-rank tests. Cell sizes have been determined through measurements of 200 cells, while statistical differences were determined using Wilcoxon testing. qPCRs were performed in three biological replicates while statistical differences are determined using student's t test. SGA data have been collected through quadruplicates for each mutant. Processing and normalisations are described in the corresponding paragraph of the Materials and Methods. p-values for RNAseq and Gene ontology enrichments have been determined through standard methods and were corrected for multiple tests according to FDR.

Further information on research design is available in the linked to this article. Nature Portfolio Reporting Summary