Introduction

Cellular aging results in several conserved phenotypes, including chromosome mis-segregation, fragmentation of mitochondria, accumulation of damaged or misfolded proteins, and alteration of nuclear morphology^1–3. Strikingly, in organisms as distinct as fungi, worms and mammals^4–6 aging also correlates with remodelling of the Nuclear Pore Complex (NPC), a ∼50 MDa gateway that controls mRNA export and the nucleo/cytoplasmic distribution of proteins ⁷. At least in dividing cells and upon expression of the progeriatric variant of Lamin A, progerin, the most prominent change observed on NPC is the loss of their nuclear basket^8,9. This structure is formed by the protein TPR on the nucleoplasmic side of the NPC and acts in processes as diverse as mRNA export, the retention of pre-mRNAs in the nucleus, protein quality control and gene regulation ^10–15. The central position of the NPC in eukaryotic information flow and the broad range of function assumed by its basket suggests that altering their function could have very diverse effects on cell function. However, despite this and the fact that the reorganization of NPC architecture with age is prominent in a wide range of organisms, we know little about whether and how NPC remodelling contributes to aging.

The unicellular fungus Saccharomyces cerevisiae, a budding yeast, is a remarkably useful system for studying the role of NPCs in cellular aging ^1,3. Indeed, budding yeast undergoes replicative aging, causing the mother cell to bud off a limited number of daughter cells until it starts slowing down its division rate, enter senescence and die¹⁶. Importantly, several studies have emphasized that the yeast NPC is remodelled during replicative aging^4,17,18, but the exact consequences of its reorganization are not fully understood.

A prominent driver of replicative aging in budding yeast is the accumulation of extrachromosomal DNA circles (also called eccDNAs) in the mother cell nucleus ^19–21. These DNA circles form during the cell’s lifetime through excision of genomic segments, generally by recombination between repeated sequences ²². The vast majority of these circles (>99%) stems from the rDNA locus and are referred to as extrachromosomal rDNA circles (ERCs)¹⁹. As the cell age, these circles accumulate exponentially through their replication in S-phase and their retention in the mother cell at mitosis^19,23. By the end of their life, yeast cells contain about 1000 ERCs, which represents roughly the same amount of DNA as the rest of the genome²³. Importantly, their retention in the mother cell relies on their anchorage to NPCs via the megadalton protein complex called the SAGA complex ^4,24. These NPC-ERC units are then confined to the mother part of the dividing nucleus by a lateral diffusion barrier forming in the nuclear envelope in the future plane of cleavage ^24–26. Indeed, yeast cells undergo a closed mitosis, i.e., they do not undergo nuclear envelope breakdown (NEBD). Remarkably, anchorage of ERCs to NPCs results in the dissociation of their nuclear basket due to SAGA acetylating nucleoporins such as Nup60 on the nuclear side of the NPC ²⁴. Consequently, a majority of the NPCs of old yeast cells lack a basket⁴. However, it is unknown whether this has consequences for the physiology and survival of the cell and contributes to the aging process.

One way to approach the question of whether displacement of the basket from the NPCs contributes to the aging process is to determine whether other aging phenotypes depend on basket displacement and if so, how. Here we focused on chromosome instability in old yeast cells. In many metazoans, including humans, aging results in increasing rates of chromosome mis-segregation and aneuploidy, correlating with increasing prevalence of cancer ^27,28. Our understanding of the molecular mechanisms of chromosome segregation errors in old cells remains rudimentary. Here, we report evidence that NPC remodelling during yeast aging leads to chromosome loss from old mother cells and investigate the underlying mechanisms.

Results

Aging yeast cells lose chromosomes

To examine whether aging affects the karyotype of yeast cells, we visualized and followed the fate of individual chromosomes in replicatively aging mother cells, using either TetO-labelled chromosome II or chromosome IV as reporters. Chromosome IV is the second longest chromosome, while chromosome II is of average size. Both chromosomes are labelled through inserting an array of 256 TetO repeats in the vicinity of their centromere^29,30. These chromosomes were visualized by expressing fluorescently tagged versions of the TetR protein (TetR-GFP or TetR-mCherry; Figures 1A-B). We then monitored them throughout the entire replicative lifespan of yeast mother cells trapped under the microscope using a microfluidics platform ^4,31. The replicative age of each mother cell was measured by counting the number of daughters that it budded off over time, defining how many budding events they have already completed at a given time point (completed budding events, CBE). Bud size and the number and position of the spindle pole bodies (SPBs/Centrosome equivalent, labelled with mCherry) were used to identify the cell cycle stages at which the cells were imaged (Figure 1A).

Figure 1

Analysis of these images showed that all young mother cells (∼0-3 CBE; t=0h) and their daughters contained a fluorescent dot, documenting the presence of the labelled chromosomes in all of them. However, when reaching the age of 18 to 22 CBE (t=24h), 10-15% of these now old mother cells had lost their dot (Figures 1B-C). This was observed irrespectively of which chromosome was labelled and which fluorescent tag was used. Fluorescence of the labelled SPB was not affected, indicating that dot loss was not caused by some instability of fluorophores in old cells. To test if the loss of these fluorescent chromosomal foci was due to the loss of the TetO array specifically or of the entire chromosome, we included a second label (an array of 256 repeats of the LacO sequence) in the middle of the long arm of chromosome IV (Figure 1B) and asked whether the cells that lost the TetO array lost the LacO repeats as well. Analysis of these images showed that 85% of the cells lacking the TetO array had indeed lost the LacO array (50/59). Thus, these cells had probably lost the entire chromosome. Supporting this notion, loss of the TetO array was fatal to the cells. Indeed, more than 99% (145/146 chromosome II loss events) of the dot-less cells failed to divide further, as expected for a haploid cell losing an entire chromosome. We concluded that much like many other organisms, yeast undergoes a burst of chromosome instability when reaching an old age (Figures 1A, C).

In contrast to chromosomes, extra-chromosomal DNA circles accumulate in the yeast mother cell with age rather than being lost ^30,32. Similarly, deleting the centromere from a chromosome causes the retention of both sister chromatids in the mother cell ³⁰. Remarkably, inserting a centromeric sequence (CEN) into a DNA circle, was sufficient to cause this mini-chromosome to be lost from aged mother cells at a similar rate as that observed for full-size chromosomes (∼20% of old cells; Figures 1B-C). To exclude the possibility that the TetO array used to label them is the cause of chromosome and mini-chromosome loss in aging, we also characterized the stability of a circular mini-chromosome expressing short-lived GFP as a marker instead of TetO repeats. Scoring of the GFP signal demonstrated that old mother cells lost these mini-chromosomes at essentially the same rate as they lose the TetO labelled one (Figures 1B-C). Therefore, the presence of a centromere was necessary and sufficient for chromosomes and mini chromosomes to be lost from aging mother cells. We concluded that chromosome loss in old cells is driven by their centromere.

Aged cells asymmetrically partition both sister chromatids to the bud

Given the prominent role of the centromere in chromosome loss, we rationalized that the loss of chromosomes in old mother cells could be due to mitotic mis-segregation events. To determine whether it was the case, we next focused our analysis on the cells that were imaged during their mitosis in our time-lapse recordings of aging cells (Figure 2A). Analysis of anaphases in old mother cells revealed a striking increase in the rate of chromosome II, IV and minichromosome missegregation (Figure 2B). In 600 total anaphase events captured in our movies we found that 66 (11%) of them co-segregated the labelled sister-chromatids to a single pole with, 54/66(82%) of the missegregation events resulting in asymmetric partition of the chromatids to the bud of these old mother cells (Figure 2C). This was virtually never observed in young cells (Figure 2A-C). Remarkably, the frequency of chromosome loss in the old mother cells and of sister-chromatids segregating asymmetrically to the bud were in the same size order, irrespective of whether chromosome II or chromosome IV was labelled. Therefore, chromosome loss from old mother cells seems fully explained by a rise in sister-chromatid non-disjunction with age and the resulting segregation of chromosomes asymmetrically to the bud. Thus, we examined the mechanism behind this peculiar mode of directed chromosome mis-segregation.

Figure 2

In budding yeast, the pre-existing (old) SPB, i.e., the SPB that the cell inherited from the previous mitosis, segregates to the bud in ∼95% of mitoses, i.e., in a highly biased manner ^33–35. This can be easily monitored when stable SPB components, such as the core SPB protein Spc42, are labelled with mCherry, the maturation of which is relatively slow (∼45 minutes half-time) relative to the duration of yeast mitosis (∼50 minutes from spindle assembly to completion of cytokinesis). Therefore, in cells expressing Spc42-mCherry as a reporter, the old SPB (Figure 2A, red arrowhead) shines brighter than the newly synthesized one. Given the convergent bias of SPB and sister-chromatid inheritance by the bud of old cells, we asked whether in old mother cells it is the old SPB that drives the non-disjoining sister-chromatids to the bud. In other words, we wondered whether the co-segregating sister-chromatids attach to and co-segregate with the old SPB. To test this possibility, we examined the partition of the labelled chromosomes relative to the old and new SPBs in old mother cells (t=24h, Figures 2D, E). In about 10% of old cells, SPB inheritance is inverted, i.e., the old SPB mis-segregates to the mother cell ^33,34. Analysis of these events indicated that the non-disjoining sister-chromatids followed the old SPB to the mother cell in 80% of the cases instead of going to the bud (Figures 2D, E). We conclude that chromosome loss in aging mother cells is due to preferential attachment of sister-chromatids to the old SPB, which then partitions them asymmetrically to the bud.

In budding yeast, both sister chromatids initially attach in a syntelic manner to the old SPB upon replication, as this SPB is the first to be available. These initial syntelic attachments are then dissolved by the aurora B kinase to allow sister-chromatids to reorient and ultimately reach bi-orientation and partition symmetrically in mitosis ^36,37. As a consequence, failure to correct initial attachments errors causes sister-chromatids to co-segregate with the old SPB to the bud ³⁸. Therefore, we hypothesized that the loss of chromosomes by mother cells to their daughters might be due to the error correction pathway being impaired in old yeast cells. To test this idea, we analysed the effects of reducing Ipl1/aurora B function during aging. Cells carrying the temperature sensitive ipl1-321 mutant allele and grown at a permissive temperature (27°C) showed no detectable chromosome loss at a young age (Figure 2F). However, they started losing chromosomes prematurely during replicative aging (Figure 2F), resulting in a shortened lifespan (Sup Fig 1A). Strikingly, the chromosome loss rate of the old ipl1-321 mutant cells (t=24h) was undistinguishable from that of the wild type cells, indicating that the effects of age and reducing Ipl1 activity were not additive (Figure 2F). This suggests that age and the ipl1-321 mutation affect the same pathway. We concluded that old cells lose chromosomes due to defects in the pathway dedicated to correcting chromosome attachment errors.

ERC accumulation drives chromosome loss in aging

To identify mechanisms behind the deterioration of mitotic error correction with age, we investigated whether chromosome loss is causally linked to other aging events, particularly to the displacement of the basket from NPCs⁴. Given that basket displacement is triggered by accumulating ERCs and their SAGA-dependent anchorage to NPCs^4,24,25, we first investigated whether ERC accumulation promotes chromosome loss. Recruitment of the deacetylase Sir2 to rDNA repeats is one of the mechanisms by which cells keep the rate of ERC formation low. Therefore, the sir2Δ mutant cells show an increased rate of ERC formation, accumulate these circles quicker and prematurely compared to wild type cells, and show a reduced longevity^19,39. Supporting the notion that ERC accumulation might promote chromosome loss, these mutant cells also showed a strong increase in chromosome loss, prematurely in life (Figure 3A and Sup 2A). To test further whether ERC accumulation could promote chromosome loss, we asked whether reducing the rate of ERC formation delays it. Recombination in the rDNA locus and the consequent excision of ERCs is stimulated by the presence of a replication fork barrier between rDNA units. While these barriers prevent the transcription machinery to collide with replication forks coming from neighbouring rDNA units, the stalling forks are fragile, causing high rates of double strand breaks. Therefore, removing these fork barriers, for example by inactivating the fork barrier protein Fob1, drastically decreases the rate at which ERCs form and accumulate, stabilizes the nuclear baskets of NPCs and prolongs the life span of the cells^4,40. Thus, we asked whether inactivating Fob1 had any effect on chromosome loss in old cells. Analysis of chromosome II presence in fob1Δ mutant cells trapped in our microfluidics platform demonstrated that only very few of them had lost the TetO signal even after dividing 18 times or more (t=24h; Figure 3A; Sup Fig 2A). We concluded from these data that chromosome loss correlated remarkably well with ERC levels, suggesting that ERC accumulation and hence, possibly the displacement of the basket from NPCs might promote chromosome loss. Further supporting this notion, removal of the SAGA subunit Sgf73, which mediates ERC attachment to NPCs and the subsequent displacement of their nuclear basket, phenocopied the effect of the fob1Δ mutation²⁴. Unlike the wild type cells, the sgf73Δ mutant cells failed to show any significant increase in chromosome loss at any age (Figures 3A Sup Fig 2A). Thus, our data suggest that ERC formation and anchorage to NPCs triggers chromosome loss in old yeast cells.

Figure 3

Nuclear basket remodelling drives chromosome loss

We reported that SAGA-dependent anchorage of ERCs to NPCs leads to the displacement of the NPCs’ nuclear basket²⁴. To investigate whether basket displacement promotes chromosome loss in some manner, we sought next to genetically destabilize the basket and determined whether this accelerated aging and the concomitant rise of chromosome instability. To do this, we disrupted the MLP1 gene, encoding the major isoform of the core basket protein TPR in yeast, and monitored the effect of that mutation on chromosome II stability in aging cells. In stark contrast to the sgf73Δ and fob1Δ single mutant and the wild type cells, the mlp1Δ single mutant cells exhibited a rapid increase in their chromosome loss as they aged, along with a shortened lifespan (Figures 3A-B Sup Fig 2A). Thus, our data suggest that displacement of the basket from NPCs as the cells age promotes chromosome loss through some yet unknown mechanism.

Introns in genes of the aurora B pathway drive asymmetric chromosome partition in aging

The displacement of the nuclear basket from NPCs upon ERC anchorage weakened the molecular pathway that corrects chromosome attachment errors (Figure 3A). This in turn caused old mother cells to lose chromosomes to their buds. We wondered about the molecular mechanism linking asymmetric chromosome partition to basket displacement. The basket of the nuclear pore has been involved in various processes, such as mRNA export ⁴¹, protein quality control ¹¹, docking of environmentally regulated genes to the nuclear periphery upon their induction ¹⁵, as well as the retention in the nucleus of faulty mRNAs, such as unspliced pre-mRNAs ^10,42, as well as docking Mad1 and Mad2, two key components of the mitotic Spindle Assembly Checkpoint(SAC), to the NPC^43,44.

The SAC components Mad1 and Mad2 bind Mlp1 and Mlp2 as well as Nup60 at the basket of the NPC⁴⁴. However, basket dissociation, which delocalizes Mad1/2 from the nuclear periphery, does not appear to interfere with their function ⁴⁴. Nevertheless, we investigated whether SAC dysfunction could account for aging phenotypes caused by basket dissociation and for the premature aging phenotype of the mlp1Δ mutant cells. Arguing against this hypothesis, the mad1Δ and mad2Δ single mutant cells exhibited no change in replicative longevity compared to the wild type cells and did not phenocopy mlp1Δ mutant cells (Sup Fig 3A). Therefore, we searched for other mechanisms through which basket displacement could trigger chromosome loss.

While perusing through the list of genes involved in chromosome segregation (Figure 4A) we noticed that, surprisingly, only three of them contain an intron, namely NBL1, MCM21 and GLC7. Strikingly, these three genes function with Ipl1/aurora B in preventing syntelic chromosome attachment. Nbl1/borealin is a component of the chromosome passenger complex of which aurora B/Ipl1 is the catalytic subunit. Mcm21/CENP-O is a docking receptor for Ipl1/aurora B at kinetochores. Glc7/PP1A is the protein phosphatase 1, which counteracts aurora B by dephosphorylating its targets at the outer-kinetochore ^45–47. Intrigued by this remarkable convergence, we wondered whether asymmetric chromosome partition in aging could stem from the presence of these three introns in error correction genes. Indeed, we rationalized that displacement of the basket from NPCs could potentially affect the regulation, export and hence expression of intron-containing transcripts, specifically, thereby affecting the function of the NBL1, MCM21 and GLC7 genes.

Figure 4

To investigate the possibility that the introns of GLC7, MCM21, NBL1 contribute to the rise of chromosome loss in old cells, we removed them from each of the three genes (-Δi alleles) and imaged the mutant cells throughout their replicative lifespan. Removing their introns individually or all three simultaneously (3xΔi strain) had no detectable effects on chromosome segregation in young cells (Figure 4B). In sharp contrast, simultaneously removing all three introns nearly completely suppressed chromosome loss in old cells (Figure 4B-C). Removing them individually or in pair did so as well, albeit the effects were milder. Removing the introns of the two α-tubulin genes TUB1 and TUB3 simultaneously had no effect, indicating that the effects observed with the introns of NBL1, MCM21 and GLC7 were specific (Figure 4B). The effects observed were similar for chromosome II and chromosome IV, indicating that the role of these introns was not chromosome specific (Sup Figure 3B). Furthermore, removing these three introns also suppressed the occurrence of asymmetric anaphases in old cells (Figure 4C), supporting the view that asymmetric chromosome segregation and chromosome loss go hand-in-hand and are under the control of these three introns. Strengthening this conclusion, removal of these introns also suppressed chromosome loss in the ipl1-321 mutant cells (Sup Figure 1B).

Remarkably, simultaneous removal of MCM21, NBL1 and GLC7 introns had also a substantial effect on the longevity of the cells, extending their replicative lifespan up to ∼20% (Figure 4D). A similar effect was observed in the ipl1-321 mutant cells (Sup Fig 1A). This suggests that chromosome loss is the first direct cause of death for many yeast cells upon aging, although other mechanisms eventually take over when chromosome loss is abrogated.

Basket destabilization causes intron-dependent chromosome loss in young cells

Thus, our data indicate that basket dissociation from NPCs promotes chromosome loss through an intron-dependent dampening of the error correction pathway. To test this idea more in depth, we asked whether mutations affecting the nuclear basket also promote chromosome loss outside of the aging context. Thus, we characterized chromosome segregation in young mlp1Δ cells (Figure 5A-C). Analysis of chromosome II segregation indicated that the mlp1Δ mutant cells not only exhibited a four-fold increase in their chromosome mis-segregation rate compared to the young wild type cells, but ∼80% of these mis-segregation events followed the old SPB, as in old cells (Figure 5C). Furthermore, this effect was suppressed upon removal of the introns of MCM21, NBL1 and GLC7(Figure 5B). Therefore, although to a lower extent than in old cells, defects of the nuclear basket result in asymmetric chromosome partition also in young cells. Like in aging, these effects are mediated through the introns of MCM21, NBL1 and GLC7. Interestingly, while the lifespan of the mlp1Δ mutant cells was shortened compared to wild type, removal of these three introns rescued chromosome loss and partially restored the longevity of the cells (Figure 5D-E). We concluded that intron-dependent dampening of the error correction pathway is one, even if not the only one, progeriatric effects of the mlp1Δ mutation. Possibly, other introns contribute to other aging phenotypes. Thus, our data indicate that basket displacement promotes chromosome loss in young mutant cells and old wild type mother cells.

Figure 5

Old cells leak pre-mRNA to the cytoplasm

Furthermore, our data suggested that the role of the nuclear basket in enforcing proper chromosome segregation involved its function in mRNA processing and quality control, and particularly in some process related to splicing or to intron function. Thus, to explore this possibility, we next investigated whether old mother cells show defects in pre-mRNA quality control. Indeed, basket mutant cells fail to retain unspliced pre-mRNAs in the nucleus ^10,42. We therefore reasoned that old yeast mother cells might show similar defects.

As a first proof of principle, we went on to characterize the localization of intronic RNA sequences in young and old cells, using single molecule RNA fluorescence in situ hybridization (smRNA FISH). For these studies, we focused on three introns, namely those of the DBP2, YRA1 and GLC7 genes. Indeed, most yeast introns are short (50-150 nucleotides) and well below the size detectable by smRNA FISH (>500 nts, ideally >1 kb). In the yeast genome, only 20 introns are longer than the required minimal size. Thus, we chose to develop probes for the long introns, namely those of DBP2 (766 nts), and YRA1 (1002 nts), and for obvious reasons for GLC7 (525 nts). The introns of NBL1 (66 nts) and MCM21 (88 nts) are too short for this method. Cohorts of old cells were collected using the Mother Enrichment Program (MEP)⁴⁸ followed by FACS sorting to reach high homogeneity. By inducing cell death in the daughter cells specifically, the MEP ensures that aging mother cells are diluted linearly instead of exponentially in the population, greatly simplifying the isolation of very old cells. FACS sorting ensured that the cohorts were free of young cells and young cell debris. With each of these three series of smRNA FISH probes, we detected one or two nuclear foci in young cells. Only very rarely did we observe a focus in the cytoplasm (<3% of the cells). No foci were detected in the corresponding Δi strains, demonstrating probe specificity (Sup Fig 4). In contrast, for each of the three introns tested, smRNA FISH detected intronic RNA sequences in the cytoplasm of the majority of old cells. For example, in the case of GLC7, ∼70% of smRNA FISH foci were found in the cytoplasm of old mother cells, compared to ∼3% in young cells (Figures 6A-B Sup Fig 4). Together, these data indicated that 3 of 3 tested introns showed a strong tendency to leak to the cytoplasm in old mother cells. However, this first test did not clarify whether the full pre-mRNA or only the intron escaped from the nucleus.

Figure 6

To address this question very directly, we turned to a functional assay and monitored the translation of an unspliced pre-mRNA reporter in the cytoplasm. This fluorescent reporter construct expresses GFP upon proper splicing of the nascent transcript, or mCherry, coded in the intron, when the unspliced pre-mRNA leaks out of the nucleus and hence, becomes translated (Figure 6C) ⁴⁹. Cells carrying this construct integrated at the TRP1 locus were trapped and imaged in our microfluidics platform. The fluorescence levels of the two reporter fluorophores as well as the number of completed budding events of the recorded cells were measured as they aged. Analysis of the fluorescence signal established that the mCherry/GFP ratio increased in old compared to young cells, demonstrating elevated translation of the pre-mRNA in the cytoplasm of aged cells compared to young ones (Figures 6D-E).

Consistent with leakage and translation of this reporter depending on ERC accumulation, the mCherry/GFP ratio increased prematurely and to a higher level in the sir2Δ mutant, compared to the wild type cells. In reverse, the fob1Δ mutant cells showed a lowered mCherry/GFP ratio, which did not increase much during their lifespan (Figures 6D-E). Thus, we conclude that ERC formation and accumulation promotes pre-mRNA leakage during yeast aging, as expected from their effect on the NPCs’ nuclear basket.

pre-mRNA leakage induces asymmetric chromosome partition

Given the results above, we wondered whether chromosome loss in old cells is linked to their failure to retain pre-mRNAs in the nucleus. Thus, we asked whether inducing pre-mRNA leakage would alone be sufficient for triggering chromosome loss. The two yeast SR-protein Gbp2 and Hrb1 have been implicated in the retention of pre-mRNAs in the nucleus by binding intron-containing transcripts and delaying their export until they are properly spliced or degraded^50,51.Therefore, hrb1Δ gbp2Δ double mutant cells leak pre-mRNA to the cytoplasm at a higher rate than wild type cells ^10,50–52. Thus, we asked whether these double mutant cells showed any tendency to lose the labelled chromosome II. As a control, we asked whether the snu66Δ mutation, which affects the spliceosome itself directly ^53,54, lead to similar phenotypes.

The young spliceosome-defective, snu66Δ mutant cells did exhibit a five-fold elevated level of chromosome mis-segregation compared to young wild type cells. However, this mis-segregation was not biased towards any specific SPB (Figures 7A-B), suggesting that it involved a distinct mechanism than that observed in old cells and in young cells lacking the basket protein Mlp1. In contrast to the snu66Δ mutant, the hrb1Δ gbp2Δ double mutant cells, showed an even stronger increase in their rate of chromosome mis-segregation, which was more than ten-fold above that of young wild type cells. Furthermore, sister-chromatid mis-segregation was strongly biased towards the old SPB. Removing the introns of GLC7, MCM21 and NBL1 suppressed this chromosome segregation defect (Figure 7A). Therefore, pre-mRNA leakage is sufficient to induce asymmetric chromosome partition and is contingent on the presence of introns in the error correction genes NBL1, MCM21 and GLC7. The fact that mutations affecting the spliceosome did not cause such an asymmetric segregation of chromosome to the bud indicates that the effects observed in old cells cannot be explained simply by splicing defects.

Figure 7

We conclude that the loss of the NPCs’ basket and subsequent leakage of pre-mRNAs impairs the correction of syntelic attachments by the Ipl1/aurora B pathway, resulting in asymmetric partition of sister-chromatids and chromosome loss with age. Basket displacement and pre-mRNA leakage inhibited error correction by some yet unknown mechanism involving the introns of the NBL1, MCM21 and GLC7 genes, specifically.

Discussion

Aging manifests itself through a broad diversity of conserved cellular phenotypes but except in few cases we know little about how these relate or are causally linked to each other^1–3. In this study we investigated whether NPC remodelling in old mother cells contributes to the emergence of other aging phenotypes, taking chromosome loss as a study case⁴. We show that NPC remodelling is both necessary and sufficient for the raise of chromosome loss with age and provide insights into the mechanisms involved. Indeed, interventions that promote displacement of the NPCs’ basket and basket defects in general promoted non-disjunction and mis-segregation of sister-chromatids to the bud, and hence chromosome loss in the old mother cell. These interventions include mutations that stimulate ERC formation and removal of the main yeast TPR isoform, Mlp1. In reverse, interventions that prevent NPC remodelling, such as delaying ERC accumulation, their anchorage to NPCs and the recruitment of SAGA to NPCs, restored proper chromosome segregation in old cells and prevented chromosome loss. Thus, NPC remodelling appeared as the pivotal point for triggering chromosome loss in old cells.

Two potential caveats should be considered. First, the sir2Δ and fob1Δ mutations have each more effects than only influencing ERC formation. The sir2Δ mutant cells lose silencing of the hidden mating type loci, genetically behaving as diploids, fail to silence subtelomeric regions and act through some yet unknown mechanism on the maintenance of proper proteostasis^55–58. Likewise, the fob1Δ mutation renders the rDNA locus unstable, which might interfere with rRNA transcription and ribosome biosynthesis⁵⁹. However, their only clear common denominator is that they both affect ERC formation, and hence NPC remodelling even if in opposite manners^39,40. Moreover, the sgf73Δ mutation phenocopies the effects of the fob1Δ mutation. Here again the only known functional overlap between Fob1 and Sgf73 resides in their requirements for the accumulation of ERCs in the mother cell, and indirectly (Fob1) or directly (Sgf73), for their anchorage to NPCs and the consequent displacement of NPCs’ basket^4,24. In fitting, the fact that destabilizing the basket also promotes the premature onset of chromosome loss reinforces the notion that it is indeed the displacement of the basket from NPCs that triggers chromosome loss upon ERC accumulation.

A second possible caveat is that chromosome loss could be linked to aging in some other manner. The effects of the diverse perturbations listed above would promote chromosome loss indirectly, via the effects that they have on cellular aging. To address this possibility, we investigated the molecular mechanisms that link basket displacement and chromosome loss. Our data reveal that chromosome loss is due to the non-correction of syntelic chromosome attachments in old cells. In young cells, such corrections are enforced by the kinase aurora B and its accessory factors³⁶. Our data also show that the inhibition of correction in old cells requires the presence of three introns, those standing in the genes NBL1, MCM21 and GLC7. These three genes are the only genes of the chromosome segregation machinery to have kept an intron. Strikingly, the proteins Nbl1, Mcm21 and Glc7 are all involved in the error correction pathway, together with aurora B/Ipl1^45–47. Removing their introns sufficed for restoring proper attachment correction and chromosome segregation in old cells and for suppressing the premature chromosome loss phenotype of the ipl1-321 mutant cells grown at semi-permissive temperature. Intron removal delays cell death but does not abrogate aging. Furthermore, the suppression of chromosome mis-segregation upon removal of these introns is not only observed in old wild type cells but also in young mlp1Δ mutant cells. Thus, intron removal separates aging and chromosome loss, indicating that the action of these introns is part of the proximal cause of chromosome loss and is independent of age. Therefore, our data indicate that there is a direct, intron-dependent link between basket displacement from NPCs and chromosome loss. We suggest that this link relates to the role of the nuclear basket of the NPC in inhibiting the export of unspliced pre-mRNAs out of the nucleus, and hence, in mRNA quality control. The fact that another, independent set of mutations that also interfere with pre-mRNA retention in the nucleus, such as the gbp2Δ hrb1Δ double mutant cells, also increase the rate of chromosome loss through the same three introns strengthens this conclusion. Thus, together our data establish that displacement of the nuclear basket from NPCs in old mother cell triggers chromosome non-disjunction and loss.

This conclusion has several consequences and opens new questions. The key question stems from our observation that releasing pre-mRNA from the nucleus relaxes the control that cells impose on chromosome attachment. The question is, how does this work? Does it involve the translation of the pre-mRNAs in the cytoplasm? Does it come about through the correlated loss of properly mature mRNAs and hence, the reduced expression of their products? Two arguments speak against these interpretations. First, if any of these two options were to account for the effect of pre-mRNA leakage, mutations affecting splicing efficiency would lead to the same effects. However, tempering with the spliceosome does not affect the correction of chromosome attachment errors to any extent and specificity close to the effects of pre-mRNA leakage. Second, since Mcm21 and Glc7 have opposite effects on chromosome correction it is unclear how impairing the expression of both would lead to the synergistic effects that we observe between them. Finally, analysis of RNA-seq data from replicatively aged cells indicate that if aging indeed impairs the splicing of some transcripts, such as those encoding ribosomal proteins, it stimulates that of others⁶⁰. This suggests that aging has a non-univocal effect on intron-containing transcripts. It will be important to determine whether basket displacement contributes to these effects, how it does so and whether this could explain the effect it has on the error correction pathway.

Our observations that basket displacement has such profound and highly specific impacts on processes functionally very distant from that of the NPC, such as chromosome segregation, has several consequences. First, it means that NPC remodelling could affect many more cellular functions than only chromosome segregation. Interestingly, beyond mRNA quality control the basket of the NPC is also involved in protein quality control and the activation of environmentally regulated genes, two processes that are indeed impaired in old cells^1,2. Thus, basket displacement might contribute to these phenotypes in aging cells. It is also very tempting to speculate that the role of the basket in mRNA quality control could affect more than only chromosome segregation. Indeed, while budding yeasts have kept only a subset of the introns of their ancestors, these introns are not distributed randomly and seem to be enriched in very suggestive genes, such as genes coding for ubiquitin-conjugating enzymes and proteasome subunits, components of the mitochondrial respiratory chain and ribosomal proteins⁶¹. Since decay of the ubiquitin-mediated protein degradation, mitochondrial membrane potential and function and of ribosome assembly are classical and ubiquitous hallmarks of aging^1–3, it is possible that basket displacement in old cells contributes to the emergence of these aging phenotypes as well.

A second type of consequences emerging from our observations stems from the fact that displacement of the NPCs’ basket is also observed in stressed cells and not solely in the context of aging. For example, heat shock also causes basket displacement in yeast⁶². Therefore, it is possible that the downstream effects we observe in old cells also take place during stress response. In this respect, it is interesting to note that heat shock has been associated with a strong raise in aneuploidy in yeast^63–65. Furthermore, aneuploidy is a frequent response to environmental stresses and a primary mechanism for the emergence of stress resistant clones in fungi⁶⁶. Therefore, it is tempting to speculate that the chromosome mis-segregation resulting from basket displacement might be a selected response to stress. It might indeed be selectively advantageous for cells to have means to promote karyotype variations when their survival is in question. It is therefore worth considering that a similar process, perhaps using similar mechanisms, underlies the aneuploidy of cancer cells.

Third, the role of basket displacement in the emergence of aging phenotypes is likely not a peculiarity of yeast. Using proteomics, basket displacement has been inferred to take place in aging cells such as the hepatocytes of old mice and humans^6,67. Little is known about whether the nuclear basket plays any similar role in pre-mRNA retention in the nucleus of metazoans as of yeast cells, but it is still striking that intron-retention is one of the hallmarks of aging on the metazoan transcriptome, and that pre-mRNA processing genes are important in mitotic stability^68–71. Such intron-retention would be expected if some transcripts tend to escape the nucleus prematurely. More striking maybe, the perhaps most ubiquitous effect of the expression of the progeriatric variant of lamin A, progerin, across cell types is the displacement of the nuclear basket protein TPR from NPCs⁸. We know still very little about the impact of basket displacement in the etiology of the Gilford-Hutchinson Progeria Syndrome, but our yeast studies suggest that it might be relevant. If so, investigating the potential effects of progerin on pre-mRNA leakage might become an essential step towards understanding and perhaps mitigating the effects of progerin expression.

In a broader sense, our findings might bring fundamental insights towards our understanding of aging. For decades, the role of ERCs in aging has been perceived as a yeast peculiarity. However, the effect it has on NPCs may close the gap. On one hand, it seems to illuminate what might be the commonality between ERC-driven aging in yeast and other aging processes in other eukaryotes. Indeed, the basket-centric view of aging that we propose here would explain how ERCs actually cause aging in yeast and indicate that while ERC accumulation might be an idiosyncrasy of yeast, the downstream events are not. On the other hand, this idea open two corollary questions. 1) What is the biological significance of eccDNAs promoting basket displacement in yeast? 2) What causes basket displacement in other aging systems. We are intrigued by the idea that ERCs and any replicating eccDNA for that matter are the closest mimics that laboratory yeast strain may encounter of pathogenic DNA of exogenous origin. In that perspective, basket displacement and the ensuing effects might have first appeared as a response to infections. In other words, yeast aging might reflect the activity of a death pathway that normally functions in pathogen control. It might represent some fungal version of what an auto-inflammatory or auto-immune response in metazoans would be. Therefore, it will be interesting to investigate whether exogenous or non-chromosomal DNA, such as possibly that of viruses, also triggers the displacement of the nuclear basket from NPCs in yeast and other systems, and whether this contributes to either or both of anti-viral defence and aging. If so, aging might have emerged as an anti-viral defence.

Author Contributions

Conceptualization: MM, ACM, YB Investigation: MM, JM, ACM Formal Analysis: MM,JM, ACM,CD Writing-original draft: YB,MM Writing – review & editing: MM,JM,JP,SAE,YB Resources: JP,SSL Supervision: YB Funding acquisition: YB

Acknowledgements

We acknowledge Dr.Dan Jarosz, Dr. Gabriel Neurohr, Dr. Madhav Jagannathan, Dr. Patrick Meraldi, Dr Anna Marzelliusardottir for critical reading and suggestions on the Manuscript. We acknowledge the ScopeM facility at the Institute of Biochemistry, ETH for their technical assistance. We acknowledge Elisa Dultz for providing the Chromosome II strain with the POA1::TetO array.

Funding

MM: grant 31003A-105904 from the Swiss National Science Foundation (SNSF) and the ETH postdoctoral fellowship 19-1 FEL-10

JM: grant 31003A-105904 from the Swiss National Science Foundation (SNSF)

JP: CIHR (Canadian institute of health research) 201809PJT-407549-BMB-CFDA-53473.

SSL: Scope M facility of ETH Zurich, Institute of Biochemistry

SAE: CIHR (Canadian institute of health research) 201809PJT-407549-BMB-CFDA-53473.

YB: grant 31003A-105904 from the Swiss National Science Foundation (SNSF)

Declaration of interests

Authors declare that they have no competing interests.

Data and materials availability

All the raw data generated in the manuscript will be deposited in the Mendelay online repository.

Materials and methods

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Yves Barral (yves.barral@bc.biol.ethz.ch).

Materials availability

This study generated ∼50 novel yeast strains, derived either by crossing or transformation. All strains are available to the public upon request

Experimental Model Details Strains and plasmids

All yeast strains and plasmids used in this study are listed in the Supplementary file1. Strains are derived from S288C background. For every strain derived from a cross, an appropriate control strain was always selected from the same tetrad. Protein C-terminal yeGFP-tag and knock-out strains were generated using methods described in Janke et al 2004 ⁷². All cultures were grown using standard conditions at 160 rpm, in YPD or synthetic drop-out medium supplemented with 0.1% BSA for aging chips (SD-medium; ForMedium, Norfolk, UK) at 25°C or 30°C.

Intron-less strains

Intron deletions were conducted in two independent colonies of the strain JPY10I, as outlined previously in Parenteau et al 2008 ⁷³. Briefly, each intron was individually deleted using a modified two-step process (pop-in / pop-out), ensuring the absence of any additional sequence or markers post-deletion. The deletions were done in diploid strains, confirmed via PCR, and three distinct haploid strains bearing the confirmed deletions were chosen for subsequent analyses. Combinations of different Δ-intron strains have been derived by crossing and validated by PCR for the absence of the intron.

Chromosome and Plasmid reporter strains

Multiple chromosomes and minichromosome visualization strains have been used in this study: 1) Chromosome II reporter strain, with the centromere proximal TetO array (POA1 locus) strain was kindly provided by Elisa Dultz from the Weis Lab, ETH, institute of Biochemistry (internal reference KWY3759). 2) Chromosome IV TetO (Centromere proximal) and LacO (centromere distal) were acquired from strains used in Neurohr et al 2011 ²⁹, originally derived from Vas AC et al 2006 ⁷⁴ 3) Ubi-GFP minichromosome (PYB 2665) was derived by cloning UBIYdkGFP* from Houser et al 2012 (PNC 1136)⁷⁵ into a PYB 2640 backbone by homologous recombination in yeast. 4) A minichromosome reporter containing the TetO array was previously described in Denoth-Lippuner et al 2014 ²⁴.

Materials and methods

Microscopy

For confocal fluorescent microscopy, yeast cells were precultured in YPD, and then washed in synthetic drop-out medium. One ml of cells from exponential growing cultures with OD <1 was concentrated by centrifugation at 1500 rcf, washed in SD complete, and resuspended in ∼10 µl of low fluorescent SD-medium, spotted on a round coverslip and immobilized with a SD/agar patch. The cells were imaged in >15 z-stacks slices with 0.3 μm spacing, with a 100×/1.4 NA objective on a DeltaVision microscope (Applied Precision) equipped with a CCD HQ2 camera (Roper), 250 W Xenon lamps, Softworx software (Applied Precision) and a temperature chamber set to 30°C.

Aging microfluidic platform

Chromosome segregation and protein intensity during aging were assessed using the high-throughput yeast aging analysis (HYAA) microfluidics dissection platform ³¹. The PDMS (polydimethylsiloxane) microchannel is made by soft-lithography and bonded on the 30 mm micro-well cover glass in the 55 mm glass bottom dish (Cellvis, CA, USA). For the lifespan analyses, a chip with a new cell trapping design was used ⁴, to ensure excellent retention of old cells.

To start the aging experiment, yeast cells were pre-cultured for 24 hr in SD-full media supplemented with 0.1% Albumin Bovine Serum (protease free BSA; Acros Organics, Geel, Belgium). Young cells from an exponentially growing culture were captured in the traps of the microfluidic chip; the chip was continuously flushed with fresh medium at a constant flow of 10 μl/min, using a Harvard PHD Ultra syringe pump (Harvard Apparatus, Holiston, MA, USA) with 60 ml syringe per lane, with inner diameter 26.7 mm (Becton Dickinson, Franklin Lakes, NJ, USA). Bright field images in a single z focal plane were recorded every 15 min throughout the duration of the entire experiment to measure replicative age of cells. To record fluorescent signals, images with a florescent lamp and at least 13x z-stacks (0.5 μm step) were acquired in 12h intervals. For imaging we used an epi-fluorescent microscope (TiE, Nikon Instruments, Tokyo, Japan) controlled by Micro-Manager 1.4.23 software ⁷⁶, with a Plan Apo 60 × 1.4 NA objective. For fluorescence illumination of the GFP and mCherry labelled proteins, a Lumencor Spectra-X LED Light Engine was used. Z stacks of >13 slices with 0.5 μm spacing were recorded for fluorescent imaging every 12 hours in order to cover the entire volume of the nucleus during the aging process. Transmitted light imaging was done in a single Z plane to minimize the light exposure of the cells. The age of the cell was defined by the number of daughter cells that emerged during the budding cycles (CBE).

Chromosome loss assessment

For the assessment of chromosome and mini-chromosome loss and mis-segregation in aging, trapped cells in the chip that had the labeled chromosomes at the initial time point of 0h (∼virtually all cells) were followed in transmitted light channel and fluorescent images were acquired every 12 hours. Only cells where the entire volume of the nucleus was in focus and imaged, and where SPB was clearly visible were considered in the analysis. The background fluorescence was subtracted in FIJI using a subtract background function. The aged nuclei were examined by scrolling through the z stack (>13 stacks, 0,5um step) in both channels, to validate the presence of the SPBs and TetR foci in the mother or the daughter cell. “Chromosome loss” refers to two quantified events: 1) The absence of the chromosome from a mother or a daughter cell compartment during anaphase 2) A G1 mother cell that just underwent its final division and is clearly viable, with a present SPB dot, but no chromosome dot. When “Chromosome loss” is listed, it pertains to the sum of these two events. When “chromosome loss in anaphase” is listed, it only pertains to the loss frequency observed during anaphase. Cells which lost chromosomes due to abnormal mitotic events in the last replicative division, such as SPB overamplification or full nuclei migration into the bud (Combined ∼10% of all cells in the final replicative division) were discarded from the quantifications. All cells were examined in the DIC channel for signs of disrupted morphology indicating cell death and followed up in DIC to confirm cell death after chromosome loss. For the assessment of minichromosome loss in aging (TetO array or ubi-GFP minichromosome) the yeast cultures were grown in SD-LEU or SD-URA selective media enabling minichromosome maintenance in all starting young cells. When loaded onto an aging chip cell were transferred to SD complete media, allowing them to lose the minichromosome during aging without inducing cell death. As the cells were grown in selective media, the minichromosome loss rate at 0h is N/A. In general, aging chips are not suitable for measurement of chromosome loss rate in young cells, as cells within the young population that are missegregating chromosomes are usually stressed and enlarged, and thus cannot be loaded into the microfluidic traps intended for the normal young cells. For chromosome missegregation rates in young cells, we have used classical confocal imaging on a patch of SD agar, as described in the Microscopy section.

Measurements of chromosome loss per CBE

To measure chromosome loss probability per Completed Budding Event (CBE), we utilized at least 150 cells per genotype and quantified the number of cells that lost the reporter chromosome during a specific replicative division. The number of losses at a specific CBE was then divided by total number of cells which passed though mitosis at that CBE number to give the probability of chromosome loss per CBE. Categories of 5x CBE and its mean value of chromosome loss were plotted along with its SD.

Mother Enrichment Program (MEP) with FACS and RNA FISH

Mother Enrichment Program was performed according to the standard procedure, using the strains acquired from the Gotchling Lab ⁴⁸. Cells were labeled with Biotin and aged for 28h, after which they were fixed in 4% PFA for 30 minutes. The cells were then stained with streptavidin conjugated with a FACS compatible fluorophore (647nm) and sorted using FACSAria III at the Flow Cytometry Core Facility at ETH Zurich to enrich for old mother cells. After FACS sorting, single molecule RNA FISH with Stellaris® probes designed for introns of GLC7, YRA1, DBP2 was performed as described in Rahman et al 2013 ⁷⁷.

Quantification and Statistical analysis

Statistical analyses were performed using GraphPad Prism 10 software

Resources table