A fln-2 mutation affects lethal pathology and lifespan in C. elegans

Differences in genetic background in model organisms can have complex effects on phenotypes of interest. We previously reported a difference in hermaphrodite lifespan between two wild-type lines widely used by C. elegans researchers (N2 hermaphrodite and male stocks). Here, using pathology-based approaches and genome sequencing, we identify the cause of this difference as a nonsense mutation in the filamin gene fln-2 in the male stock, which reduces early mortality caused by pharyngeal infection. We show how fln-2 variation explains previous discrepancies involving effects of sir-2.1 (sirtuin deacetylase) on ageing, and show that in a fln-2(+) background, sir-2.1 over-expression causes an FUDR (DNA synthesis inhibitor)-dependent reduction in pharyngeal infection and increase in lifespan. In addition we show how fln-2 variation confounds effects on lifespan of daf-2 (insulin/IGF-1 signalling), daf-12 (steroid hormone signalling), and eat-2 (putative dietary restriction). These findings underscore the importance of identifying and controlling genetic background variation.

R esearch using invertebrate model organisms has revealed much about how genotype specifies many aspects of phenotype, including ageing. Studies on the short-lived nematode Caenorhabditis elegans, in particular, have led to the discovery of signalling pathways that affect lifespan, not only in nematodes but also in higher organisms, including humans 1,2 . These include the insulin/IGF-like signalling (IIS) and the target of rapamycin (TOR) pathways 3,4 . In recent years, the importance of genetic variation in C. elegans strain backgrounds has become increasingly clear, partly arising from inadvertent selection acting on the wild type under laboratory conditions and/or random mutation accumulation [5][6][7] . Such variation may confound the effects of genes of interest.
First isolated more than 60 years ago, the laboratory strain N2 (Bristol) is the standard reference strain for C. elegans research 6 . N2 is distributed by the Caenorhabditis Genetics Center (CGC) either as a hermaphrodite only stock (which we will refer to here as N2H) or as a male stock (referred to at the CGC as N2 male, which we will refer to as N2M). We previously noted that hermaphrodite lifespan is slightly but consistently longer in N2M than N2H (+11% median lifespan) 8 . As both stocks have been used as wild type, it is possible that the stock difference may have confounded prior studies of gene effects on lifespan.
Ageing limits animal lifespan by promoting multiple pathologies that increase mortality rate. Using necropsy analysis, we recently identified death with a swollen, infected pharynx (P [big P] death) as one form of mortality in C. elegans 9 . Under standard laboratory culture conditions 10 , initial intra-pharyngeal infection occurs in early adulthood during a period of fast pumping and mastication of the bacterial food source, and later accounts for 40% of N2H deaths, largely those occurring prior to mean lifespan. The remainder of the population die with an uninfected but often atrophied pharynx (p [small p] death). Changes in the shape of survival curves of different C. elegans strains can be understood in terms of altered frequency and/or timing of P and p death (mortality deconvolution) 9 .
In this study, we use mortality deconvolution to investigate the cause of the difference in lifespan between N2H and N2M, and the question of which of these N2 lines is wild type. We report that the longer lifespan of N2M is attributable to reduced P frequency that results from a single nonsense mutation in the X-linked gene fln-2, indicating that N2H is wild type and N2M mutant. We demonstrate the importance of controlling for fln-2 genotype by illustrating how it can confound investigations, using several examples of genes that have been the focus of many studies of ageing in C. elegans: sir-2.1, daf-2, daf-12 and eat-2.

Results
N2M longevity is due to reduced P death frequency. To investigate the causes of the lifespan difference between N2H and N2M, we performed mortality deconvolution on the two lines, comparing timing and frequency of P and p death, using combined mortality and necropsy analysis 9 . Overall mean lifespan of N2M hermaphrodites was 16.6% longer than N2H (22.1 days vs. 18.7 days, p < 0.0001, log rank test)( Fig. 1a; Supplementary  Table 1), consistent with previous findings 8 . Necropsy analysis revealed that 37% of N2H worms died with P, consistent with previous findings 9 , but this was the case in only 13% of N2M worms (p = 0.0031, Student's t test) (Fig. 1b). By contrast, the lifespans of both the P and p subpopulations were similar in the two lines ( Fig. 1a; Supplementary Table 1). Thus, the lifespan difference between N2H and N2M is attributable to reduced P death frequency in the latter. This predicts that blocking E. coli proliferation with antibiotics or UV radiation, which completely prevents P death 9 , should abrogate the difference in lifespan between N2H and N2M, and this proved to be the case ( Fig. 1a; Supplementary Fig. 1a; Supplementary Table 1).
To further investigate the difference between N2H and N2M we compared other characteristics of the two lines. No difference in self-fertilised brood size was detected, consistent with previous findings 8 ( Supplementary Fig. 1b). Mutants with reduced pumping rate often show reduced P frequency 9 , raising the possibility that pumping rate might be reduced in N2M, but it was not ( Supplementary Fig. 1c). Moreover, the difference between N2H and N2M is not attributable to effects of parental mating in the latter: hermaphrodite progeny of mated N2H still show higher frequency of P death than N2M ( Supplementary  Fig. 1d). Thus, N2M's intrinsic resistance to pharyngeal E. coli infection accounts for its longer lifespan.
We previously reported evidence that pharyngeal infection occurs in two stages: initial bacterial invasion occurs during early adulthood, apparently due to activity-dependent damage to the pharyngeal cuticle, while major infection occurs later in some animals when bacteria escape cyst-like intracellular structures and proliferate 9  To test whether each stage can be selectively suppressed, worms were fed with E. coli OP50 expressing a red fluorescent protein (OP50-RFP), which renders initial bacterial invasion visible as red fluorescent puncta within pharyngeal tissue 9 (Fig. 1c). Culture on antibiotic-treated E. coli did not suppress initial bacterial invasion of the pharynx (presumably because antibiotics do not suppress grindergenerated mechanical damage and perforation of the pharyngeal cuticle 9 ) but did prevent later bacterial proliferation, while in pumping-defective phm-2 mutants initial invasion was fully suppressed (Fig. 1c), perhaps due to a combination of reduced mechanical damage and bacterial lawn avoidance 11 . Thus, it is possible to selectively suppress the invasion and proliferation stages.
We then asked: which stage is suppressed in N2M? Initial bacterial invasion was observed in 40% of the population in N2M, compared to 75% in N2H as early as day 4 of adulthood; yet in both strains about half of those showing early signs of invasion die with a swollen, infected pharynx (Fig. 1d). These results suggest that N2M is resistant to pharyngeal cuticle perforation and/or initial bacterial invasion rather than to later bacterial proliferation.
N2M longevity is caused by fln-2(ot611). Reduced P frequency in N2M likely reflects a different genotype. Analysis of F1 hybrids from a N2H/N2M cross showed that the N2M infrequent P trait is recessive (Fig. 1e). We also noted, during backcrossing with N2M of a strain with high frequency of P, that P frequency dropped in a single round of outcrossing from high to low P frequency (Fig. 1f), suggesting effects of a single gene.
To identify the causative variant, whole genome sequencing was used. An initial comparison between genomic sequences of the two N2 lines revealed 4259 single nucleotide polymorphisms (SNPs) and small insertions and deletions (INDELs). Most of these mutations likely accumulated during separate propagation of N2H and N2M before they were both frozen in 1975 8 . In order to map the mutation affecting P frequency, we performed Variant Discovery Mapping 12 . For this, N2H and N2M were crossed and the progeny of individual F2 recombinant lines with the N2M phenotype pooled and subjected to genome sequencing (Fig. 2a). Parental allele enrichment peaked near the 10 Mb position on LGX implying the presence of the causal variant in this vicinity (Fig. 2b, c; Supplementary Figs. 3 and 4a). Variants within this region co-segregated with low P frequency in the aforementioned  backcrossed strains (Fig. 1f), as established by sequencing (16 strains sequenced). The 3 Mb region near the parental allele enrichment peak contained four genetic differences predicted to alter protein sequence, including three missense and one nonsense mutation ( Supplementary Fig. 4b). The nonsense mutation was in fln-2, which encodes a filamin that is localised, among other places, to the pharyngeal cuticle 13 . We therefore focused on fln-2. FLN-2 is a distant ortholog of filamin-A (FLNA) in humans, which links actin filaments to membranes and is involved in cytoskeletal remodelling and mechanical force sensing. The variant in fln-2 results in a stop codon at Y800 in N2M, downstream of the N-terminal actin-binding domain in the second immunoglobulin(Ig)-like domain 13 (Fig. 2d). Of the 16 isoforms that fln-2 encodes, the premature stop codon affects six, all of which contain the actin-binding domain ( Supplementary  Fig. 4c). We checked the fln-2 allele in other C. elegans wild isolates available at the Caenorhabditis elegans Natural Diversity Resource (CeNDR) 14 , and found that none of the wild strains contain the Y800* allele. fln-2 orthologs in a selection of other nematode species also encode the full-length fln-2(+) allele and not the Y800* equivalent ( Supplementary Fig. 4d). We therefore conclude that the fln-2 variant in N2M is a nonsense mutation. This mutation was previously observed 6 out of 17 sequenced EMS-mutagenized genomes 15 and designated fln-2(ot611) (formerly flnb-1); on the assumption that strains containing the mutant allele were derived from N2M, we will refer to the variant identified here as fln-2(ot611).
To test whether fln-2(ot611) causes the N2M infrequent P trait, we first tested effects of the mutation fln-2(ok1305), which contains a deletion in the N-terminal region. Although it showed infrequent P death ( Supplementary Fig. 5a), the strain RB1240 also contained ot611, making this result difficult to interpret. Another fln-2 mutant allele, fln-2(tm4687), which does not contain ot611, has a deletion in the C-terminal region affecting all fln-2 isoforms. This also caused low P frequency ( Supplementary  Fig. 5b), demonstrating that variation at fln-2 can affect P incidence. Additionally, we introduced the Y800* mutation into   Table 2). Taken together, these results indicate that fln-2(ot611) is the cause of the reduced P frequency of N2M.
Tests on four C. elegans wild isolates showed a P death incidence similar to that of N2H ( Supplementary Fig. 5c), suggesting that this relatively high P frequency is a species-typical (i.e. wild-type) characteristic of C. elegans.
Discrepant effect of sir-2.1 attributable to fln-2 variation. Given the effect of the fln-2(ot611) allele on pharyngeal pathology and lifespan, and the fact that the Gems lab routinely used N2M as the wild-type for backcrossing 8 , we set out to revisit some earlier, puzzling results. In budding yeast (Saccharomyces cerevisiae) over-expression of the NAD-dependent histone deacetylase SIR2 (sirtuin) increases replicative lifespan 18 . There have been conflicting reports about whether or not increasing sirtuin levels in C. elegans increases lifespan. In C. elegans, independent transgenic lines with sir-2.1 over-expression (oe) generated in two studies were long lived 19,20 . In the case of one high copy number sir-2.1 over-expresser (transgene array geIs3, hereafter referred to as sir-2.1(oe)) 19 it was subsequently reported that longevity of the original strain (LG100) was attributable to a second site dyf gene mutation, and that sir-2.1(oe) no longer increased lifespan after backcrossing with N2M 21 . However, in another study that confirmed the presence and action of the second site mutation, a residual, smaller lifespan-increasing effect of sir-2.1(oe) was seen after outcrossing with N2 (line not specified) 22 .
We asked: could this discrepancy be attributable to effects of N2H vs. N2M background differences? Genotyping of the strains used by Viswanathan and Guarente (2011) 22 and Burnett et al. 21 confirmed that the former were fln-2(+), and the latter fln-2(ot611), presumably reflecting the different N2 lines used in backcrossing of the original sir-2.1(oe) line. Consistent with both previous reports, sir-2.1(oe) resulted in increased lifespan in a fln-2(+) background (+27% mean lifespan, p < 0.0001, log rank test), but not a fln-2 (ot611) background (p = 0.96, log rank test) (  Table 4) under culture conditions as previously described, in which the thymidylate synthase inhibitor 5-fluorodeoxyuridine (FUDR) is added to block progeny development (see Methods and below).
In the previous studies, FUDR was used to block progeny development: 1.6 mM in the 2001 study 19 , and 10 to 50 μM 21 or an unspecified concentration 22 in the 2011 studies. Therefore in the trails described above FUDR was included (10 μM in Fig. 3a-d, and 50 μM in Supplementary Fig. 6a, b). Since impacts of many interventions on lifespan are affected by the presence of FUDR [23][24][25][26] , and since fluoropyrimidines have complex effects both on E. coli and C. elegans 27 , we also tested effects of sir-2.1(oe) on P frequency and lifespan (overall and P or p alone) in the absence of FUDR. Here no effect was detected in either fln-2 background, in either the original strains or the outcrossed strains (2 trials each) (Fig. 3e, f; Supplementary Fig. 6d, e; Supplementary  Tables 6 and 7).
We further tested for interaction between FUDR and sir-2.1 over-expression using a previously described low copy number transgene array pkIs1642 20 . It was previously reported that longevity resulting from pkIs1642 sir-2.1(oe) was abrogated by outcrossing 21 . Here both original and outcrossed strains were fln-2(+). Nonetheless, we observed that the longevity of the nonoutcrossed pkIs1642 strain was, again, FUDR dependent (Supplementary Fig. 7; Supplementary Table 8). However, here longevity resulted from increased p lifespan rather than reduced P frequency. For more details, see Supplementary Fig. 6.
The FUDR dependence of the reduced P death caused by sir-2.1 (oe) suggests that SIR-2.1 and FUDR act synergistically to protect against bacterial infection of the pharynx. To explore the mechanism of this protection we compared the progression of pharyngeal infection in OP50-RFP-fed worms with either sir-2.1 (oe), FUDR, or both. Without FUDR, as expected, sir-2.1(oe) had little effect on either the initiation or subsequent spread of E. coli infection; with FUDR, the appearance of intracellular RFP puncta was unaffected by sir-2.1(oe), but the progression of infections in older animals was suppressed, with fewer animals showing widespread RFP (Fig. 3g). Notably, FUDR alone delayed the appearance of RFP puncta, though it had little effect on the frequency of P (Fig. 3g). Thus, FUDR and sir-2.1(oe) affect pharyngeal infection by different mechanisms: FUDR delays initial invasion, while SIR-2.1 suppresses later spread of infection. This suggests that FUDR-induced resistence to bacterial pathogenicity enables SIR-2.1 to reduce later E. coli proliferation, leading to synergy between the two interventions. Consistent with this, pretreatment of C. elegans with FUDR increases resistance to the Gram-negative bacterial pathogen Pseudomonas aeruginosa 30 .
fln-2 variation confounds effects of daf-12 on lifespan. The daf-12 gene encodes a steroid hormone receptor, and has been shown to have complex effects on different daf-2 insulin/IGF-1 receptor mutants [31][32][33] . For example, daf-2(m41) longevity (the Age phenotype) is partially suppressed by daf-12(m20) under certain conditions 31 . We noted that daf-12(m20) shortens daf-2(m41) lifespan by increasing early mortality 31 , and that the daf-12 and fln-2 loci are closely linked on LGX (at X:2.36 ± 0.007 cM and X:1.18 ± 0.015 cM, respectively). This raised the possibility that some observed effects of daf-12 on daf-2 Age might be attributable to the differences in P frequency due to variation at the fln-2 locus.
fln-2(ot611) masks eat-2 Age effect. Dietary restriction (DR) increases healthspan and lifespan in many model organisms 34 . In C. elegans, some Eat mutants with reduced feeding rate (e.g. eat-2) show increased lifespan, which has been interpreted as resulting from DR 35  More recently we showed that many Eat mutants show reduced P frequency, which may reflect reduced mechanical damage to the pharynx otherwise caused by high pumping rate in early life 9 . Mortality deconvolution of eat-2 populations showed that the lifespan increase is largely attributable to reduced P death frequency. This predicts that eat-2 should increase lifespan more in a N2H background than a N2M background, due to the higher P frequency in the former. To test this, we performed mortality deconvolution on eat-2(ad1116) mutants in fln-2(+) and fln-2(ot611) backgrounds. As expected, P death frequency was reduced by either fln-2 or eat-2 (Fig. 5a).

Discussion
The findings reported here reveal a major confounding variable in studies of ageing in C. elegans: variation at the fln-2 locus between two widely used N2 lines. We demonstrate how this has actually confounded research on C. elegans ageing, distorting experimental results, and causing confusion and controversy, issues that can now be readily resolved. These results also provide further evidence of the power of mortality deconvolution and, more broadly, pathology-centred approaches to solve complex problems in biogerontology.
The advance of science often requires a process of over-coming limitations within a field, both conceptual and technical. The latter includes identification and control of confounding variables, which, over time, can lead to accumulation of unsolved questions and conflicting results. This is a particular problem with model organism biogerontology, since lifespan as a trait is particularly sensitive to uncontrolled variables 36,37 . For C. elegans, these include variables affecting genetic background, such as that between N2H and N2M, and effects of variation at fln-2 as described here. It also includes a range of environmental variables; since the introduction of a standardised C. elegans plate culture system 10 , several modifications have been introduced, including the use of FUDR to block progeny production, and antibiotics or UV irradiation to block E. coli proliferation. It is becoming increasingly clear that the effects of many interventions that alter C. elegans lifespan reflect, at least partially, changed interactions between C. elegans and E. coli 9,38,39 . Thus, variables affecting C. elegans-E. coli interactions, including fln-2 allele variation and use of different culture conditions, either introduce confounding variables, or make it difficult to relate results from different research groups to one another.
Variation at fln-2 can confound studies of the genetics of lifespan, as illustrated here using three cases, involving sir-2.1, daf-2 and daf-12, and eat-2. It is therefore possible that variation at fln-2 has also confounded other studies of C. elegans ageing.
Our results resolve a controversy surrounding effects of sir-2.1 over-expression in C. elegans lifespan 21,22,40 . This provides a useful lesson in how science can be impeded by confounding variables, in this case three of them. The first was the presence of a dyf mutation which increased lifespan in a sir-2.1(oe) strain 21,22 . The second was the presence of fln-2(ot611) in strains backcrossed with N2M, which suppressed the effect of sir-2.1(oe) on lifespan 21 . The third was the use of FUDR, which as we show here is required for sir-2.1(oe) to increase lifespan; it is notable that FUDR-dependent induction of increased lifespan by hypertonic stress is partially sirtuin dependent 24 . Does this FUDRdependence invalidate the conclusion that sir-2.1(oe) protects against ageing? This question raises important issues about the use of C. elegans as a model for understanding human ageing. We have argued elsewhere 41,42 for a reconceptualization of the relationship between ageing and lifespan, as follows. Lifespan is not a read out of one process, i.e. ageing, but rather a function of some (but not all) senescent pathologies that arise as animals age. Importantly, the identity of the pathologies that limit lifespan will vary not only between species, but within species depending upon genotype and environment 41,42 . According to this view, the key objective for understanding ageing is to understand the etiologies of any pathology, whether they affect lifespan or not. In terms of sir-2.1(oe), FUDR and lifespan, if the use of FUDR enables studies that yield useful insights about senescent pathology, then this is a good model to study.
We showed here that under standard culture conditions (proliferating E. coli, no FUDR), previously reported suppression of daf-2(m41) Age (25˚C) by daf-12(m20) is likely to have been partially attributable to variation at fln-2 (Fig. 4a-c). There is other evidence that signalling via the DAF-9/DAF-12 pathway can promote daf-2 Age. For example, at 22.5˚C daf-9(rh50) shortens daf-2(e1368) lifespan 43 , and daf-2(e1368) Age can be enhanced by treatment with exogenous DAF-12 ligand (Δ 4 -dafachronic acid) 44 . However, our results imply that clearer  Table 11 for full statistics, and Supplementary data 1 for raw mortality data. Source data are provided as a Source Data file. c eat-2(ad1116) causes a small but significant increase in p sub-population lifespan in both fln-2(+) and fln-2(ot611) backgrounds. See Supplementary data 1 for raw mortality data. d Effects of fln-2(ot611) on lifespan in eat-2(ad1116) mutants disappears after treatment with antibiotics (200 μg/ml carbenicillin). See Supplementary data 1 for raw mortality data understanding of how the IIS and DAF-12 steroid signalling pathways interact to affect ageing could be achieved by enhanced control of confounding variables, including fln-2 and culture condition variables (as listed above). These results reveal that the greater lifespan of N2M is attributable to fln-2(ot611). The presence of the fln-2(+) allele in N2H and all C. elegans wild isolates examined demonstrates that N2H is the wild type and N2M a long-lived mutant, in contrast to a previous suggestion 8 . This is an argument for discontinuing the use of N2M as wild type for most purposes. The mechanisms by which fln-2(ot611) protects against pharyngeal infection remain to be explored. Comparison between the two N2 strains raised on fluorescent E. coli suggests that fln-2(ot611) is resistant to cuticle perforation and/or bacteria invasion. The presence of FLN-2 protein in the pharyngeal cuticle 13 is consistent with a role as a direct, cell autonomous determinant of susceptibility to mechanical damage and/or bacterial invasion.
Lifespan assays and necropsy analysis. Unless otherwise specified, lifespan assays were conducted at 20˚C and without FUDR, and animals were transferred daily during the egg laying period, and every 3-7 days thereafter. Animals lost due to causes other than senescence (e.g. desiccation on the side of the plate, internal hatching, extensive rupture through the vulva) were censored. For lifespans performed at 25°C (daf-2 daf-12 experiments), worms were raised at 20°C, and transferred to 25°C at L4 stage. For trials performed in the presence of FUDR, L4 stage animals were transferred to seeded NGM plates containing the specified concentration of FUDR, and animals were transferred every 3-7 days to avoid hypertonic stress 24 . For antibiotic treatment, 200 μg/ml carbenicillin was added to two-day old lawns of OP50, and nematodes were transferred to antibiotic plates as L4 larvae. Mortality and pharyngeal status of corpses were scored every 2-3 days. For lifespan analysis on UV-irradiated E. coli OP50, bacteria were inoculated onto NGM plates and allowed to grow overnight at room temperature before exposure to UV light (Stratagene UV Stratalinker 2400) for 5 min, 2 days prior to use (to detect and exclude plates with bacterial growth). Bacteria-free (alkaline hypochlorite-treated) C. elegans eggs were then plated on irradiated E. coli lawns. Lifespan assays were repeated at least twice, and raw mortality data are provided in the Ziehm tables (Supplementary Data 1). Survival plots were generated in GraphPad Prism using combined lifespan data from multiple trials, with L4 as day 0 on the time scale.
OP50-RFP infection analysis. An overnight culture of OP50-RFP in LB supplemented with tetracycline was washed three times in M9 to remove tetracycline, and seeded onto NGM plates two days before use. L4 worms cultured on OP50 bacteria were transferred to OP50-RFP plates, and transferred daily during the reproductive period. Before imaging, live worms were mounted on 2% agar pads and anesthetised using 0.2% levamisole. Fluorescence within pharyngeal tissue were scored at 630x magnification using a Zeiss AxioImager Z2 microscope with a rhodamine filter. A different cohort of worms was imaged for each time point so as to avoid recovering worms from levamisole.
Test cross to confirm single-locus association. Strain GR1307 daf-16(mgDf50) was backcrossed with either N2H or N2M males, and F1s allowed to self. Single F2 animals were picked at random and the daf-16 allele genotyped, and F3 populations of single daf-16 homozygous animals were scored in necropsy analysis and used for further backcrossing. The same analyses were performed for a single F4, F6, and F8 animal of each F2 line. All progeny of the backcrossed F2, F4, F6, and F8 animals used in lifespan were frozen and subsequently genotyped with respect to the fln-2 allele.
Whole genome sequencing. Genomic DNA was extracted using Gentra Puregene Tissue Kit (Qiagen), and sequenced with an average coverage of 100× (Center for Applied Genomics, The Hospital for Sick Children, Toronto, Canada). N2M genome sequence is available on ENA with the accession number PRJEB34655 [https://www.ebi.ac.uk/ena/browser/view/PRJEB34655], and will be incorporated in the N2 genome sequence on CeNDR 14 . For variant discovery mapping, N2H hermaphrodites and N2M males were allowed to mate overnight, after which males were removed. F1 hermaphrodites were allowed to self-fertilise, and a total of 250 F2 animals were singled and allowed to self-fertilise. From each plate, 50-60 F3 progeny were transferred to plates containing 10 μM FUDR, allowed to age, and then scored for pharyngeal phenotype. 50 of the populations displaying lowest frequency of P death were re-analysed and 25 populations that showed less than 20% P death in both tests were selected. F3 and F4 genomic DNA from selected populations were then extracted and pooled to be sequenced. The sequencing data were processed in CloudMap using the variant discover mapping method, which mapped the causal variant to the vicinity of the 10 Mb position on chromosome X.
Statistics. Survival analyses of whole populations (i.e. P and p deaths) were conducted using GraphPad Prism, including right censoring of animals lost due to causes other than senescence. To test for statistically significant differences between survival curves, the non-parametric log rank test was used. In the case of P or p subpopulations, due to the absence of information about corpse type, deaths due to causes other than senescence were not censored but rather excluded altogether from statistical analysis. Sample sizes and statistics for each experiment are provided in the lifespan tables in the Supplementary file. An unpaired Student's t test was used to compare frequencies of P death between different strains and/or conditions.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All statistics relating to survival analysis are provided in the Supplementary tables. Raw mortality data are provided in Ziehm tables 9 in Supplementary Data 1. Source data relating to mortality analysis are provided as a Source Data file. N2M genome sequence is available on ENA with the accession number PRJEB34655.