Genome Integrity Is Regulated by the Caenorhabditis elegans Rad51D Homolog rfs-1

Abstract

Multiple mechanisms ensure genome maintenance through DNA damage repair, suppression of transposition, and telomere length regulation. The mortal germline (Mrt) mutants in Caenorhabditis elegans are defective in maintaining genome integrity, resulting in a progressive loss of fertility over many generations. Here I show that the high incidence of males (him)-15 locus, defined by the deficiency eDf25, is allelic to rfs-1, the sole rad-51 paralog group member in C. elegans. The rfs-1/eDf25 mutant displays a Mrt phenotype and mutant animals manifest features of chromosome fusions prior to the onset of sterility. Unlike other Mrt genes, rfs-1 manifests fluctuations in telomere lengths and functions independently of telomerase. These data suggest that rfs-1 is a novel regulator of genome stability.

MAINTAINING genome integrity is essential for the viability of the germ lineage. Whereas somatic cells are discarded each generation, the germline is passed from one generation to the next. Thus, accumulated damage in the germline stem cells is passed on to progeny, lowering overall fitness and survival. A number of mechanisms ensure the integrity of the genome, including DNA damage-repair genes, DNA replication factors, checkpoint genes, and telomere regulators.

The Rad51 paralog group consists of five proteins, Rad51B, Rad51C, Rad51D, XRCC2, and XRCC3, all of which are required for homologous recombination (HR) repair. A Rad51B-Rad51C-Rad51D-XRCC2 complex binds Y-shaped HR intermediates (Yokoyama et al. 2004) and a subcomplex of Rad51B-Rad51C promotes strand exchange by counteracting the effects of replication protein A (RPA) (Sigurdsson et al. 2001; Lio et al. 2003). A XRCC3-Rad51C complex binds Holliday junctions and has resolvase activity (Liu et al. 2004). Rad51D has also been shown to have a role in telomere-length regulation (Tarsounas et al. 2004) where it was proposed to serve a capping function.

Telomeres are nucleoprotein complexes that protect chromosome ends from degradation and end-to-end chromosome fusion. Telomeric repeats present several challenges for maintaining genome integrity. First, loss of repeats occurs as a natural consequence of semiconservative replication. Second, free chromosome ends are a substrate for the DNA damage repair (DDR) machinery (Smogorzewska and De Lange 2004). Third, replication through repetitive elements frequently causes fork stalling. Each of these factors can lead to loss of protection of the DNA end, induction of breakage-fusion-bridge cycles, genomic rearrangements or aneuploidy (McClintock 1939; Hackett et al. 2001), senescence, and apoptosis. Thus, mechanisms exist to ensure that each of these problems is addressed to maintain genome integrity. The reverse transcriptase telomerase adds repeats onto the ends of chromosomes. Specific end-binding proteins form a “cap” structure that sequesters the free DNA end from the DDR machinery. Replication and recombination proteins are localized to the telomere to promote fork progression through repeats and promote HR at an offending lesion, respectively.

In Caenorhabditis elegans, loss of telomere homeostasis also leads to replicative senescence and the onset of sterility after serial cloning, a phenotype known as mortal germline (Mrt). In addition to the catalytic subunit of telomerase, trt-1 (Cheung et al. 2006; Meier et al. 2006), the DNA damage checkpoint genes, mrt-2 and hus-1 (Ahmed and Hodgkin 2000; Hofmann et al. 2002) and their associated clamp loader hpr-17 (Boerckel et al. 2007) also cause loss of telomeric sequences and a Mrt phenotype. In these mutants, the onset of sterility is gradual, but is manifest in the entire population.

Mutations that increase the frequency of DNA damage or decrease the fidelity of damage repair also give Mrt phenotypes. These include mutations in the mismatch repair genes msh-2 and msh-6 that destabilize microsatellite repeats and increase mutation rates (Degtyareva et al. 2002; Tijsterman et al. 2002). Mutation of dog-1, a gene required for replication of G/C tracts (Cheung et al. 2002) and of high incidence of males-6 (him-6), the BLM syndrome helicase, also accumulate deleterious mutations that result in sterility (Youds et al. 2006). Furthermore, a class of “mutator” genes induces transposon hopping when defective and accumulate damage that can lead to loss of fecundity (Collins et al. 1987; Ketting et al. 1999).

During normal development, replication-blocking lesions arise and must be repaired. The rfs-1 mutation provides a unique opportunity to investigate the effect of such lesions on the development of a multicellular organism. The rad-51 short protein (rfs-1) gene has recently been shown to enhance the dog-1 mutant phenotype (Ward et al. 2007) and to interact with the DNA damage-repair genes rad-51 and brc-1 (Boulton et al. 2002). rfs-1 also appears to act upstream of him-6 in the HR repair of stalled replication forks and displays acute sensitivity to nitrogen mustards and cisplatin (Ward et al. 2007). These agents cause massive DNA damage, leading to immediate sterility of the exposed animals, suggesting that rfs-1 is important for maintaining fertility.

In this study, I show that rfs-1 is a novel Mrt gene. rfs-1 mutant sterility results in part from unchecked chromosome fusions. During extended passaging of rfs-1 mutant lines, both elongation and shortening of telomeric repeats are observed, suggesting that telomeric repeats are sensitive to the loss of rfs-1. In addition, changes in karyotype can be seen in rfs-1 mutant populations, suggesting that this protein has a more fundamental role in maintaining genome integrity.

MATERIALS AND METHODS

Stains and culture conditions:

C. elegans strains were maintained at 20° as described (Brenner 1974). The rfs-1(ok1372) strain was generated and provided by the C. elegans Gene Knockout Project. rfs-1(ok1372) was outcrossed 15 times and balanced with hT2[qIs48]. Several independent lines were established and tested for both ionizing radiation (IR) sensitivity and Mrt phenotypes. No differences between the subclones were observed. Sibling F₁ homozygotes were used to initiate the analyses of the Mrt phenotypes.

The following strains were kindly provided by the Caenorhabditis Genetics Center (University of Minnesota, St. Paul): wild-type Bristol N2, unc-86/him-15(e1416) also known as eDf25, rfs-1(ok1372), and vab-10(ok817)/hT2[qIs48], which was used as the source of the hT2∷GFP balancer. An outcrossed strain of trt-1(ok410) was kindly provided by Shawn Ahmed. It was outcrossed an additional five times to my lab's N2 stock prior to balancing it with hT2[qIs48]. To make the trt-1(ok410) rfs-1(ok1372) double mutant, F₂ rfs-1 males were crossed to trt/hT2 hermaphrodites. The trt-1/+; rfs-1/+ hermaphrodites were then crossed to hT2/+ males. Individual GFP⁺ hermaphrodites were plated and genotyped for the presence of the trt-1 and rfs-1 mutant alleles. The rfs-1, trt-1, and trt-1; rfs-1 stocks were all maintained as balanced heterozygotes for a minimum of five generations prior to initiating the Mrt studies. All assays were initiated with sibling F₁ homozygotes from these balanced lines.

Hatching rates and male production were assessed by individually plating 10–15 L4 larvae of each genotype and transferring the animals every 12 hr until the onset of sterility. The number of eggs was counted immediately after transferring and the number of adult worms was counted 2.5–3 days later.

Fitness assays:

Animals of genotypes rfs-1(ok1372), trt-1(ok410) rfs-1(ok1372), trt-1(ok410), and wild type (N2) were grown for successive generations using different growth conditions. These were labeled with the genotype and assigned an alphabetic line number. Under condition 1, six L4 homozygous animals from a single balanced heterozygote were plated on fresh Escherichia coli and grown to starvation. Thereafter, six L1 larvae were transferred to a fresh plate immediately poststarvation (∼1 week). Sterility was scored as the inability to starve out a plate after 6 weeks. Under conditions 2 and 3, six worms or one worm, respectively, were transferred to fresh E. coli each generation, thus keeping the animals continually in the presence of excess food. In both cases, sterility was scored as the inability to produce any progeny after 1 week and confirmed by the absence of progeny after 3 weeks. Brood sizes were assessed 4–6 days after plating. Sterility of the population was confirmed by testing at least five individuals per line from the generations preceding the onset of sterility. Fecundity of individual lines was assessed by individually plating animals, waiting 4–7 days, and counting all progeny.

Sensitivity to ionizing radiation:

Late L4 worms of each genotype were collected on fresh NGM plates with E. coli, subjected to different doses of IR, allowed to recover for 24 hr, and individually plated. Adults were removed after 12 more hours and the number of eggs/plate was counted. Hatching was assessed by counting the number of adults 48–72 hr later.

Telomere length:

Isolation of genomic DNA and Southern blotting and probing for telomeric sequences was performed as described (Ahmed and Hodgkin 2000). To confirm that the bands of the Southerns resulted from fully digested DNA, each DNA preparation was phenol extracted, EtOH precipitated, suspended in TE, and treated with proteinase K for 30 min. Samples were then reextracted three times with phenol, phenol/chloroform, and chloroform prior to precipitation, suspension, and redigestion. No differences in band migration were observed after this treatment (not shown). In Figure 5, each of the lines for each genotype was started from sibling F₁ worms. These were grown to starvation, chunked to 10-cm plates, and grown for ∼2 days before harvesting. Thus, the DNA represents mixed-age, but mostly gravid worms from the F₃ generation. Telomere-loss rate was calculated by measuring telomere lengths from the middle of the fastest migrating telomeric bands (see Figure 6) and dividing by the number of generations.

Figure 5.—

Telomere length in starting populations. Southern blot analysis of telomeric DNA from (A) 9 lines of wild type (N2) and 10 lines if rfs-1 and (B) 9 lines of trt-1 and 8 lines of trt-1 rfs-1.

Figure 6.—

rfs-1 effects telomere homeostasis. Southern blot analysis of telomeric DNA from two independent lines of (A) N2, (B) rfs-1, (C) trt-1 rfs-1 and trt-1, and (D) eDf25 at successive generations. (A) Wild-type N2 telomeres are stable over time. (B) rfs-1 telomeres fluctuate over time, showing generations of loss (line A) or loss, stabilization, and lengthening (line B). (C) trt-1 rfs-1 and trt-1 appear to shorten telomeres with similar kinetics. The large bands (arrowheads) seen in trt-1; rfs-1 line B, generation 7 and trt-1 line A, generation 10, may be the result of undigested DNA.

DAPI staining:

Gravid hermaphrodites were picked in M9 media on glass slide, fixed with a drop of Carnoy's solution (Sulston and Hodgkin 1988), washed once with M9, stained for 10 min in 500 ng/ml DAPI, and immediately mounted in Prolong Gold (Invitrogen, Carlsbad, CA). To examine germline morphology, animals were dissected as described (Dernburg et al. 1998) and stained with DAPI as above.

Microsatellite repeat lengths:

From the growth plates of individually transferred worms, populations of worms were harvested, lysed, and microsatellite repeats were amplified by PCR and sequenced using published primers for GT16, GT26, and AAT28 (Degtyareva et al. 2002).

Mapping chromosome fusions:

Worms carrying putative chromosome fusions were identified by karyotype analysis (described above), outcrossed, and strains bearing fusion chromosomes were established. The rfs-1 mapped strains had segregated males when heterozygous, suggesting the X chromosome was one of the partners in the fusion event. The fusions were mapped using single nucleotide polymorphisms (SNPs) with primers corresponding to the following cosmids (physical position), described in Wicks et al. 2001: chromosome (chr) I, F26C11 (−27.1) and C37A5 (23.05); chr II, T01D1 (−16.23) and Y51H1A (17.98); chr III, K02F3 (−26.1) and F54F12 (20.14); chr IV, Y66H1A (−24.79) and Y51H4A (25.57); chr V, F36H9 (−17.4) and F38A6 (27.08); and chr X, F28C10 (−19.5) and F59D12 (22.15). In brief, the fusion lines were crossed to males of the polymorphic Hawaiian strain in single pair matings. Crosses were monitored by the presence of ∼50% male cross progeny. F₁ (Hawaiian/N2) larvae were individually plated, allowed to lay eggs for two days, and genotyped for the presence of the Hawaiian marker to confirm that the F₁ animal was cross progeny. Individual F₂ L4 larvae were plated, and DNA was isolated from starved plates. For trt-1; rfs-1, those that had stable karyotypes (6 normal bivalents or 3 normal plus 1 large bivalent or 2 large bivalents) would give the most robust broods, whereas those with more complex karyotypes would give poor progeny viability. These F₂ broods were selected for mapping after karyotyping. Chromosome fusions of two chromosomes (rfs-1) were assessed by the reciprocal pattern of crossing over of two chromosomes since a crossover on one chromosome will suppress crossovers on its fused partner, but not on an unattached chromosome. The chromosome fusions of trt-1; rfs-1 line D were seen by the failure to obtain recombinants between markers on the same chromosome and by pseudolinkage to two other chromosomes.

unc-58(e665) reversion assay:

rfs-1; unc-58(e665) and unc-58(e665) homozygotes were grown as described previously (Harris et al. 2006). Approximately 100 10-cm plates per genotype were scored by eye for the presence of non-Unc or weak Unc revertants. Calculations of haploid genomes tested and reversion frequency were performed as described (Harris et al. 2006).

RESULTS

Identification of him-15 as allelic to rfs-1:

him-15 was identified in a mutant strain that increased the rate of X chromosome nondisjunction leading to the production of XO males from XX hermaphrodites in C. elegans, a phenotype known as Him (Hodgkin et al. 1979). him-15 was originally associated with unc-86 mutant strains bearing large deletions (>18 kb) of the upstream region (Finney et al. 1988), but was distinguished by separable loss-of-function alleles in the unc-86 gene (Finney and Ruvkun 1990). One of these deletions, eDf25, originally known as unc-86(e1416), removes part of the unc-86 promoter and an upstream operon with three genes (Figure 1A). The first two genes in the operon are uncharacterized; the third gene, rfs-1, is the only member of the Rad-51 paralog group found in C. elegans (Ward et al. 2007). This group contains the human proteins rad51B, rad51C, rad51D, XRCC2, and XRCC3, which form two known complexes that are implicated in DNA damage repair. Rfs-1 appears most similar to Rad51D (Ward et al. 2007). Thus, rfs-1 appeared to be a likely candidate to be the gene responsible for the Him phenotype.

Figure 1.—

rfs-1 is the unc-86-associated him-15 gene and is sensitive to DNA damage. (A) Schematic of the unc-86 region showing the adjacent operon that contains three genes, C30A5.4, C30A5.3, and rfs-1, respectively. The unc-86/him-15(e1416) deletion removes part of the unc-86 promoter and almost all the operon. rfs-1(ok1372) removes part of exon 1 and the rest of the rfs-1 locus, including the Walker A and B motifs. (B) Table of hatching rates and male production. (C) Percentage of progeny survival of N2 (wild type), rfs-1(ok1372) after exposure to increasing doses of ionizing radiation.

A deletion allele, rfs-1(ok1372), was isolated by the C. elegans deletion consortium and was used in these studies to determine the allelic relationship between rfs-1 and eDf25/him-15. As shown in Figure 1B, both rfs-1 and eDf25 as well as rfs-1/eDf25 transheterozygotes increase the production of males compared to wild type (Figure 1B). Neither mutation alone displays a dominant phenotype with regard to male production (not shown), thereby suggesting that these two mutations are allelic. Furthermore, wild-type worms grown on rfs-1 dsRNA producing bacteria [rfs-1(RNAi)] gave a mild, but reproducible, Him phenotype (Figure 1B) supporting the identification of rfs-1 as the eDf25-associated Him mutation. This locus is referred to as rfs-1, which most adequately describes the gene function.

rfs-1 is sensitive to DNA damage:

The mild Him phenotype in the rfs-1 null mutant is reminiscent of genes that are required for DNA damage repair. rfs-1, eDf25, rfs/eDf25 transheterozygotes, and control N2 worms were exposed to different doses of ionizing radiation and tested for the production of viable progeny. As shown in Figure 1C, rfs-1, eDf25, and the rfs-1/eDf25 transheterozygotes were significantly more sensitive to ionizing radiation than wild-type animals. The level of sensitivity to ionizing radiation is somewhat greater than seen in Ward et al. (2007), presumably because of differences in the sensitivity of different developmental stages to gamma irradiation: I irradiated L4 animals to obviate the egg-laying defect that appears in the eDf25 due to loss of unc-86 function; Ward et al. irradiated young adults that are reported to be less sensitive to ionizing radiation (Hartman et al. 1996). These results suggest that rfs-1 is required for DNA damage repair and further substantiate that the eDf25-associated Him mutation is allelic.

rfs-1 gives a mortal germline (Mrt) phenotype:

In observing the rfs-1 mutant animals after serial passaging, I noted an increase in the Him phenotype and a loss of fecundity in the mutants. This phenotype is reminiscent of a class of mutations in Mrt genes, named for the onset of a mortal germline phenotype. To determine whether rfs-1 is indeed a Mrt gene, I outcrossed the rfs-1 stock 15 times to wild type and thereafter maintained the stock as a rfs-1 heterozygote over the hT2 balancer chromosome. Brood sizes of the first generation homozygotes were near normal, with a low frequency of unhatched embryos (Figure 1B). To confirm that all the animals in the population were fertile, 200 homozygous rfs-1 worms each from generations 1 and 2 were individually plated and tested for fertility. All were fertile and had normal brood sizes. Forty lines each of rfs-1, started from F₁ siblings of a single rfs-1 balanced heterozygote, and wild-type N2 worms were then passaged every week and assayed for fecundity at each generation (condition 1, see materials and methods). In all 40 rfs-1 lines, as the animals were serially passaged, fecundity began to drop: brood sizes decreased and sterile animals arose in the population. By passage 20, all 40 of the rfs-1 mutant lines were completely sterile, whereas all of the wild-type animals were fully fecund (Figure 2A). Wild-type growth plates after 1 week were covered with hundreds of worms of mixed ages and were void of remaining bacterial food (Figure 2B). In contrast, an rfs-1 growth plate had very few animals or eggs and was still replete with food as shown by the tracks left in the E. coli lawn (Figure 2C). These results indicate that rfs-1 is a Mrt gene.

Figure 2.—

The mortal germline phenotype of rfs-1. (A) Number of viable lines of N2 (wild type), rfs-1(ok1372), and trt-1(ok410) at each passage until sterility. (B and C) Images of 1-week-old growth plates for a wild type (B) and a near-sterile rfs-1(ok1372) mutant at passage 15 (C). A few eggs, two adult worms, and tracks from the rfs-1 worms are visible in the E. coli food that remains on the plate. (D) Number of viable N2 and rfs-1(ok1372) lines at each generation when six animals are transferred each generation and maintained in the presence of sufficient E. coli (n = 20). (E) Number of viable N2, rfs-1(ok1372), and eDf25 lines when a single animal is transferred each generation in the presence of food.

The rfs-1 Mrt phenotype is dependent on growth conditions and selection:

Protocols for defining a Mrt phenotype differ between labs (Ahmed and Hodgkin 2000; Degtyareva et al. 2002; Youds et al. 2006), with the number of animals transferred each generation, the length of time between transfers, and the definition of sterility varying. To reconcile these differences, I analyzed wild type, rfs-1, and the telomerase mutant, trt-1, under different growth paradigms. These experiments led to the discovery that different growth conditions dramatically affect the Mrt phenotype in rfs-1.

As described above, under condition 1, six animals were grown for 1 week on a small agar plate with a lawn of E. coli OP50 as food (Ahmed and Hodgkin 2000). In 1 week, the six animals had reproduced and their progeny had consumed all of the food on the plate, so the F₂ worms were in a state of starvation when transferred to new plates. As stated above, wild-type animals grown under these conditions showed no loss of fecundity whereas rfs-1 went sterile by passage 20. As previously reported, trt-1 animals also go sterile with serial passaging (Meier et al. 2006) and all 10 lines of trt-1 were sterile by passage 14 (Figure 2A).

Given that starvation induces a stress response that is known to alter DNA damage kinetics and cell-cycle progression (Barzilai and Yamamoto 2004), it was conceivable that the Mrt phenotype in these strains arose as a consequence of starvation, rather than as a direct consequence of passaging. To distinguish between these possibilities, I tested whether similar results would be obtained when animals were passaged for the same number of generations without starvation. After plating, animals were allowed to reproduce and L4 larvae where transferred to fresh plates 3 days later (condition 2). Since each passage in the original protocol corresponds to ∼2 generations, I expected to observe complete loss of fecundity by generation 28 in trt-1 and by generation 40 in rfs-1. In wild type, no loss of fecundity was seen over 60 generations (Figure 2D). In contrast, 10 independent lines of trt-1 went sterile by generation 24 (Figure 4B). Surprisingly, when six animals per plate of rfs-1 were transferred every generation in the presence of excess food, the stocks remained fully fertile (Figure 2D).

Figure 4.—

rfs-1 brood sizes vary dramatically. (A and B) Plots of brood sizes at each generation for different rfs-1 lines that have been maintained by transferring a single adult worm each generation. Values on the y-axis are approximate to represent different brood-size classifications: very large (>225), large (176–225), medium-large (126–175), medium (76–125), medium-small (26–75), small (11–25), very small (1–10), and sterile. (A) Line O remains highly fecund throughout the course of the experiment. The other three lines demonstrate different trajectories prior to the onset of sterility. (B) Three examples of lines with extreme variability over the last 20 generations of the experiment. Note the transition from very low fecundity to high in line J1.

One trivial explanation for the difference in population dynamics under starved and nonstarved conditions is that strong selection for the healthiest animals is imposed under nonstarved growth conditions since the first progeny that appear each generation are chosen for transfer. No such selection occurs upon starvation since the animals are picked randomly from among hundreds that are arrested as L1 larvae. Several experiments argue against this hypothesis. First, trt-1 animals treated in the same way behave consistently under both growth conditions. Second, loss of fecundity was validated by retesting the fecundity of worms from the growth plate of the generation prior to sterility (growth conditions 1 and 3). Third, testing fecundity of numerous late-generation animals showed that a large percentage of the rfs-1 worms had lost fecundity under growth condition 1. The more likely interpretation for the differences in fecundity with growth condition is that starvation either induces or accelerates the onset of the rfs-1 Mrt phenotype.

I hypothesized that if rfs-1 had a low penetrance Mrt phenotype under normal growth conditions, any decrease in fertility would confer a selective disadvantage when six animals were plated each generation, as described above. Thus, I reasoned that the Mrt phenotype might reveal itself if a genetic bottleneck were imposed at each generation. Rather than transferring six animals per plate each generation, I repeated the experiment by transferring a single worm to fresh food each generation (condition 3). Again, in wild type, this resulted in no loss in fecundity (Figure 2E), whereas in trt-1 lines full sterility was observed by generation 24 (Figure 7B). In rfs-1, sterility also arose in the population. For example, at generation 16, 48 individuals from one of the rfs-1 lines were tested for fecundity: 37 worms gave large broods, 6 gave ∼50 progeny, 3 gave only 10–15 progeny, and 2 were sterile. After 36 generations, almost half of the 16 lines were sterile (Figure 2E). Thus, by imposing a bottleneck at every generation, I can observe loss of fecundity in rfs-1. A similar loss in fecundity is seen with eDf25 in which 2/10 lines were sterile by generation 36, further confirming that these mutations are allelic. No such effects were observed when wild-type populations were subjected to repeated bottlenecks. Thus, the Mrt phenotype is due to loss of rfs-1 and does not require stress to occur, although it can be exacerbated by stress.

Figure 7.—

Population dynamics of rfs-1 trt-1 mutants. (A) Plot of fecundity over time for 10 rfs-1 trt-1 double-mutant lines and 10 trt-1 lines. At each generation, the number of lines with a given brood size is shown by the different colored bars. Very large (>200 animals/plate), large (125–200), medium (50–125), small (10–50), very small (<10), sterile (no progeny after 2 weeks). (B) Number of viable lines per generation for rfs-1 and trt-1 rfs-1 mutants when either a single worm or six worms are transferred each generation. The double mutant gives a broader curve under both experimental conditions. (C) Plot of brood sizes within trt-1 rfs-1 lines prior to and at the generation at which survivors appear. Color key is as in A. (D) DAPI-stained diakinetic nuclei from rfs-1 trt-1 lines prior to (line D, generation 20; line G, generation 26) and after (line D, generation 22; line G, generation 40; line C, generation 24) the formation of survivors.

rfs-1 has characteristics of telomere defective mutants:

At the molecular level, several different defects can cause the mortal germline phenotype: microsatellite instability, unrepaired DNA damage, mutagenic load (from replication errors, defective DNA damage repair, or transposition), epigenetic loss of germline identity, and loss of telomeres. I assessed each of these features in the rfs-1 mutant to determine the nature of the sterility.

To assay for microsatellite instability, I clonally grew 40 independent lines of wild type or rfs-1 for 25 generations. Genomic DNA from each clone was isolated every 5 generations and three microsatellite loci were examined. Unlike the mismatch repair genes msh-2 and msh-6, which have been shown to readily accumulate changes in the microsatellite repeat length (Degtyareva et al. 2002; Tijsterman et al. 2002), rfs-1 mutants and control wild-type animals showed no changes in AAT28, GT14, and GT26 repeat lengths when these lines were assayed for changes at generations 5, 10, 15, 20, and 25 (Table 1).

TABLE 1.

Microsatellite repeats are not affected by loss of rfs-1

	Generation 5		Generation 10		Generation 15		Generation 20		Generation 25
Microsatellite	N2	rfs-1	N2	rfs-1	N2	rfs-1	N2	rfs-1	N2	rfs-1
GT14	0/38	0/28	0/35	0/36	0/38	0/31	0/36	0/29	0/38	0/27
GT26	0/35	0/28	0/31	0/28	0/36	0/28	0/36	0/31	0/35	0/28
AAT28	0/30	0/32	0/32	0/30	0/37	0/28	0/31	0/30	0/36	0/24

Number of microsatellite alterations/number of samples tested.

To determine whether unrepaired DNA damage or mutagenic load leads to the onset of the Mrt phenotype, I examined each of the independent lines for the appearance of spontaneous mutations. During early generations, no mutations were observed. In later generations, as the lines began to show loss of fecundity, dumpy, uncoordinated, and roller animals became apparent in the population. However, progeny testing of these animals indicated that none of these mutations bred true (data not shown), suggesting that the phenotypes were likely due to somatic mutation or transient genome instability, rather than to germline changes that define the Mrt phenotype.

To further test whether rfs-1 has a mutator phenotype, I assayed whether the reversion rate of unc-58(e665) was increased in the rfs-1 mutant animals. This assay was developed by Harris et al. (2006) to quantify the Mutator phenotypes of mrt-2 and clk-2. Both extragenic and intragenic suppressors can be obtained, thus the spontaneous rate of reversion of this locus is higher than many others. Double mutants between rfs-1 and unc-58(e665) were made and tested for revertants. No difference between wild type and rfs-1 was observed (Table 2). These results strongly suggest that rfs-1 does not have a Mutator phenotype.

TABLE 2.

unc-58(e665) spontaneous revertant frequencies

unc-58(e665) background	Trial	Plates with revertants/total plates	Mutation frequency
Wild type	1	1/50
Wild type	2	0/50	1.1 × 10−6
rfs-1	1	0/52
rfs-1	2	1/56	1.0 × 10−6

To assess for telomere defects, I examined the germlines of rfs-1 animals from the passage prior to the onset of complete sterility. In C. elegans, the germline is a syncytium of nuclei, spatially organized with the stages of meiosis ordered distal to proximal (Figure 3A). Prior to fertilization, these nuclei cellularize and arrest in diakinesis of prophase I where the six condensed chromosomes appear as strong DAPI staining spots (Figure 3I). Oocytes are fertilized as they are pushed through the spermatheca and they accumulate transiently in the uterus, until deposition around the 8-cell stage (Figure 3, C and D). In rfs-1 animals, the germlines are usually full of eggs and sperm, indicating that germ cells are not lost due to changes in cellular identity or apoptosis. In ∼15% of rfs-1 animals just prior to the onset of sterility, aberrantly sized sperm, prematurely cellularized nuclei, a decrease in number of pachytene nuclei, and in rare animals (<5%), a complete loss of germline nuclei are observed (Figure 3B). In addition, accumulation of unlaid eggs can be seen in the uterus (Figure 3, E and F). All of these phenotypes are reminiscent of trt-1 and mrt-2 mutants (Ahmed and Hodgkin 2000; Meier et al. 2006) raising the possibility that loss of rfs-1 destabilizes telomeric repeats. Consistent with defects in telomere function, in the generation prior to sterility, greater than 95% of rfs-1 mutant worms manifest hallmarks of chromosome fusions. These include chromatin bridges, large agglomerates of genetic material in developing eggs (Figure 3, E–H), and aberrant numbers of DAPI staining bodies at diakinesis (Figure 3, J and K). These defects were also seen in eDf25 (Figure 3L and data not shown). Similar defects could also be seen in mutants that more generally cause genome instability by formation of translocations and large inversions (Sigurdson et al. 1984). The experiments below were aimed at distinguishing between a role in telomere maintenance or general genome stability.

Figure 3.—

rfs-1 shows hallmarks of telomere loss. (A–J) Images of wild-type worms and chromosomal abnormalities in late generation rfs-1(ok1372) worms as shown by DAPI staining and DIC imaging. (A) Wild type. (B) rfs-1 sterile worm with no obvious germ cells. (C) DIC and (D) DAPI images of wild-type spermatheca (lower) and just fertilized, one- and two-cell embryos in the uterus. (E and F) DIC and DAPI image of mitotic catastrophe in rfs-1 embryos. (G and H) DIC and DAPI image of ∼8- and 32-cell stage, rfs-1 embryos showing a chromatin bridge between two mitotic nuclei. (I and J) Germline nuclei at diakinesis of meiosis I. Six DAPI spots are seen in wild type (I). Five spots can be seen in rfs-1 nuclei (J). (K and L) Examples of changes in karyotype that increase the number of DNA spots at diakinesis in rfs-1 (K) and eDf25 (L). Note the smaller size of several of the spots, suggesting these are chromosome fragments.

Genomic instability is a feature of rfs-1:

Insight into the effects of rfs-1 mutation on genome stability comes from an examination of population dynamics in rfs-1. Fecundity over time in representative lines of individually plated rfs-1 stocks (condition 3) are shown in Figure 4, A and B. Half of these lines were still highly fecund after 35 generations (Figure 2E and Figure 4, rfs O). In the half that lost fecundity, the brood-size dynamics were extremely variable. Several lines rapidly transitioned from fully fertile to sterile (Figure 4A, rfs F and H) not unlike the trt-1 single mutant (Cheung et al. 2006; Meier et al. 2006). Other lines slowly declined, remaining subfertile for many generations prior to the onset of sterility (e.g., rfs E). Among the viable lines, one line showed a decline and apparent stabilization at low fertility for a number of generations, followed by rescue of fecundity to near normal levels (Figure 4B, rfs J1). Two other lines (rfs D and rfs G) displayed several incidences of transitions from low to high to low fecundity. Similar fluctuations were seen in the eDf25 strain (data not shown).

To further characterize these rfs-1 lines, karyotype analysis was performed. Unlike in the generation preceding sterility, in the stabilized rfs-1 lines, chromatin bridging was not observed. Rather, changes in the number of DAPI staining bodies at diakinesis were observed including increases, decreases, and no changes. When an increased number of DAPI staining bodies was observed, several were smaller and less intensely stained, suggesting they are chromosome fragments (see Figure 3, K and L). No change after a period of low fecundity may reflect the formation of balanced translocations, although this was not directly tested. A decrease in the number of spots can be indicative of full chromosome fusions. To determine if full chromosome fusions did occur, three of these rfs-1 lines were outcrossed to wild type, and putative fusion chromosomes were isolated on the basis of the increased segregation of males in the heterozygous population. A homozygous (non-Him) line with a stable altered karyotype of five chromosomes was established for one of these lines. Mapping of this line was performed by crossing to the polymorphic Hawaiian strain and screening single nucleotide polymorphisms at the ends of each chromosome for cosegregation. In 48 meioses assayed, no recombinants were detected between the XR and IVR markers, whereas all other pairs of markers assorted independently. These changes in chromosome number indicate that genomic instability results from the loss of rfs-1 function.

Telomere lengths are altered in rfs-1 mutants:

The karyotypic changes and the presence of chromatin bridges at the onset of sterility could be explained by a loss of telomeric DNA. Similarly, the stabilization of populations could be explained by healing of critically short telomeres. Shortened telomeres lose critical binding proteins that protect the chromosome end from fusion. As mentioned above, I was able to identify an end-to-end chromosome fusion on the basis of genetic linkage between terminal SNPs on chromosomes IV and X. This genetic data suggested that rfs-1 might be defective in telomere homeostasis. To evaluate whether telomeres were affected by loss of rfs-1 function, I assessed telomere lengths in the starting stains and over the course of serial passaging. Shown in Figure 5, A and B, are representative lines from wild type (N2), rfs-1, trt-1, and trt-1; rfs-1 (discussed below). In wild type (N2), telomeres run as differentiable smears between 3 and 8 kb (Figure 5A), with occasional larger telomeric bands of 10–12 kb (Figure 6A). Telomeres from the trt-1, rfs-1, and trt-1; rfs-1 lines all displayed a broader range of telomere lengths than wild type. In rfs-1, telomeres appear slightly longer than in wild type as seen by the lowest telomeric bands (Figure 5A). This result supports the conclusion that rfs-1 is fully recessive and is not predisposed to telomeric catastrophe due to abnormally short telomeres at the onset of these experiments. Nevertheless, I cannot rule out the possibility that a very small population of short telomeres exists in the rfs-1 strains, but are undetectable due to the limits of hybridization in Southern blotting. Telomeres in trt-1 are slightly different than rfs-1 as trt-1 has both longer and shorter telomeres and each individual telomeric band appears more discrete. The very short telomeres in trt-1 are thought to contribute to genome instability and sterility.

In wild type, telomere lengths are stable over many generations (Figure 6A). In trt-1, telomeres shorten in successive generations usually becoming discrete, lower molecular weight bands (Cheung et al. 2006; Meier et al. 2006; Figure 6C). In rfs-1 mutant lines, telomeres differ from both wild type and trt-1. Although successive generations of telomeric shortening can be seen in rfs-1 mutants (Figure 6B, rfs-1 line A), the loss of telomere sequences in rfs-1 line A corresponds to a net change of ∼700 bp over nine generations, or ∼75 bp/generation (Figure 7B). This rate of shortening is detected in the population of shorter telomeres since small changes in migration of the higher mobility telomeric fragments are difficult to measure. The rate of telomeric shortening in rfs-1 is slower than the 122 bp/generation in trt-1 (Cheung et al. 2006; Meier et al. 2006), suggesting that the loss of repeats in rfs-1 is unlikely to be due to (a complete) loss of telomerase function.

Another difference between rfs-1 and the other known regulators of telomere length became apparent from Southern analysis of another independent line, rfs-1 line B (Figure 6B, right). In this line, telomere lengths fluctuate from generation to generation with periods of loss (e.g., generations 13–20) and addition (e.g., generations 20–26). These results indicate that rfs-1 regulates telomere lengths in a manner distinct from telomerase. It should be noted that in both rfs-1 lines shown, the changes in telomere length are occurring well before any loss of fecundity is observed. These observations suggest that rfs-1 is acting in a fundamentally different manner at the chromosome ends than other telomere binding proteins.

rfs-1 affects the onset of sterility of the telomerase mutant:

The observations that rfs-1 mutants show both lengthening and shortening of telomeres and that the rfs-1 phenotype responds differentially to growth conditions suggested that rfs-1 acts independently of telomerase to affect telomere lengths. To genetically test this hypothesis, I created a trt-1; rfs-1 double-mutant strain and assayed for the onset of the Mrt phenotype under normal growth conditions (conditions 2 and 3). As described above, rfs-1 mutants show the Mrt phenotype only when subjected to a bottleneck. In contrast, the trt-1 lines went sterile under both conditions 2 and 3: fecundity declined rapidly over 6 generations, resulting in complete sterility by generation 24 (Figure 7A). The trt-1; rfs-1 lines behave unlike either parental strain. They are Mrt under all conditions tested, but display a statistically broader range of fecundity (P < 0.02) than trt-1. Several trt-1; rfs-1 double-mutant lines went sterile well before any loss of fecundity was seen in trt-1 (Figure 7A). However, the last of the viable trt-1; rfs-1 lines has persisted past generation 70, well beyond the extent of the trt single mutant. The broader curves for trt-1; rfs-1 fecundity were evident whether the lines were subjected to a bottleneck or not (Figure 7B). Thus, genetically, rfs-1 both enhances and suppresses the trt-1 mortal germline phenotype. Therefore, rfs-1 must be acting, at least partially, in a trt-1-independent fashion.

The ability of rfs-1 to partially suppress trt-1 could be explained if when making the double mutant, rfs-1 provided longer telomeres than were in the initial trt-1 strain. As discussed above, bulk telomere lengths in the starting strains of rfs-1 and trt-1 were nearly identical as assessed by Southern analysis (Figure 5B). Although I cannot rule out the possibility that a subpopulation of telomeres in the double mutant were very short or very long leading to an earlier or later demise, I think this interpretation is unlikely for two reasons. First, the dynamics of the suppressed trt-1; rfs-1 lines are qualitatively different than the trt-1 lines themselves. Suppression occurs in lines that reduced viability and leads to a substantial increase in population size. In trt-1 mutants, the decline in viability is gradual and continuous. Second, the persistance of the rfs-1 line D to generation 70 suggests that suppression occurs by an active mechanism. Generation 70 is more than double the generation span of the trt-1 single mutant. In addition, this is well beyond the point when the bulk telomeres should hit “crisis” on the basis of the rate of telomere attrition acting on the average band size from the Southern analyses. In fact, the bulk telomere lengths in the starting trt-1; rfs-1 strain were shorter than in trt-1 and included several populations of very short telomeres and, overall, many fewer discrete long bands (Figure 5B). Thus, the enhanced lethality of trt-1; rfs-1 is easily explained by telomere lengths in the starting population, whereas the suppression of trt-1 lethality is not.

Alternatively, the trt-1; rfs-1 phenotypes may indicate that the two genes act independently at telomeres. In this scenerio, rfs-1 would ameliorate the deleterious effects of telomeres critically shortened by absence of trt-1. rfs-1 could function stochastically on such telomeres to either lengthen them or induce chromosome fusions that “hide” them. This hypothesis predicts that the fluctuations in population dynamics seen in rfs-1 should also be present in trt-1; rfs-1. I monitored brood sizes of trt-1; rfs-1 worms grown under condition 2 (Figure 7C). Two of the 10 trt-1; rfs-1 lines, A2 and D, showed evidence of rescue in these experiments. Each of these lines was struggling along with very low brood sizes (red, orange, and yellow bars) in generations 19 and 21, respectively. In both cases, in the subsequent generation, several animals with significantly increased fertility appeared in the population (turquoise and blue bars). The fecundity of these was high in only a very small percentage of the animals and for only a generation or two. In fact, when I individually plated the F₁ progeny, they still exhibited a wide range of fertility (not shown). Line A2 survivors ultimately became sterile. In contrast, line D survived for many generations with low fertility and showed a second instance of increased fertility followed by low fertility. The line is still currently being maintained past generation 70. Thus, these increases in fecundity are transient and variable. In contrast, last-stage trt-1 mutants never exhibited any rescue events. Together with the observation that the rfs-1 single mutant displays fluctuations in fecundity, these data suggest that the rescue is a unique feature due to loss of rfs-1. rfs-1 may promote genome rearrangements that elicit the transient increases in fecundity that rescue the dysfunction caused by loss of trt-1, rfs-1, or both.

To further characterize the rescue events in the trt-1; rfs-1 double mutants, the germlines of the rescued animals were dissected, stained with DAPI, and examined for meiotic and mitotic defects. In these animals, the germlines were mostly wild type. The one obvious abnormality was the number of DAPI staining bodies at diakinesis (Figure 7D). In wild type, each nucleus has 6 DAPI spots, representing each of the chromosome pairs. Prior to rescue, in trt-1 rfs-1 line D, generation 20, each nucleus varied in the number of DAPI staining bodies. In the rescued animals, each nucleus had a uniform, although decreased, number of DAPI spots indicative of chromosome fusions. In the case of line D survivors, only 2 DAPI spots were seen, suggesting that the chromosome complement had fused into two large chromosomes. When outcrossed to wild type, these chromosomes were able to establish stable karyotypes with four chromosomes, suggesting they are near complete fusions of three chromosomes each. Mapping of line D (see materials and methods) revealed that no recombinants of chromosome IV or chromosome I with the Hawaiian strain chromosome were recovered (n = 64), suggesting that these chromosomes have fused at both ends to other chromosomes. Further, this analysis indicated that chromosome IV showed pseudolinkage to the IIL and VR markers (59/64 and 62/64 nonrecombinants, respectively) and that chromosome I showed pseudolinkage to XL and IIIR markers (58/64 nonrecombinants each). Examination of line D after the second rescue event showed yet another altered karyotype with two large DAPI staining bodies and one small DAPI spot in each nucleus (not shown). Another survivor displayed five DAPI spots in each nucleus (Figure 7D, line C, generation 24), indicating that a chromosome fusion between just two chromosomes occurred. This fusion was likely between the X and an autosome as the number of males increased significantly in the generation immediately after rescue. The same number of foci in every diakinetic nucleus in lines C and D survivors suggests that these chromosome complements became fixed in the population and further suggests that the event that resulted in this karyotype may have been the stabilizing factor that led to an increase in fertility.

A second class of long-lived lines was also seen, as shown by line G (Figure 7D). In contrast to the other lines, line G had four DAPI staining bodies both prior to and after the rescue event (at generation 28). This class of long-lived trt-1; rfs-1 mutant lines could arise from telomere-lengthening events such as occur by ALT pathways in other organisms. Unfortunately, due to the low fecundity of the line prior to rescue and the paucity of markers for each chromosome end, I was not able to examine the telomeric DNA directly to ascertain repeat lengths prior to and after rescue. In addition, since these rescue events are rare and mutation in the C. elegans genes required for recombination give sterile phenotypes, ascertaining whether these events are bona fide recombination-based phenotypes awaits further characterization.

When I compared telomere lengths from the trt-1; rfs-1 double mutant and trt-1 single mutants over successive generations by Southern blot analysis (Figure 6C), the rate of telomere shortening did not seem to be significantly different between the two strains with an average close to the published rate for trt-1, ∼122 bp/generation. Thus, it appears that trt-1 is epistatic to rfs-1 for establishing the rate of telomeric shortening. These results underscore the difference in behavior of the single and double mutants in regard to lethality and reinforce the conclusion that rfs-1 acts fundamentally differently from other Mrt mutants to regulate genome stability.

DISCUSSION

I have shown that the rfs-1 mutant displays a Mrt phenotype with progressive sterility, telomere attrition, chromatin bridges, and chromosome fusions. Like the Mrt genes, mrt-2, hus-1, and trt-1, mutants of rfs-1 show chromatin bridging in almost all animals in the generations immediately preceding sterility, suggesting that these are the cause of sterility in rfs-1 (Smelick and Ahmed 2005). Despite similarities with these telomeric genes, rfs-1 is distinct from this class of Mrt mutants in three ways. First, in addition to telomere erosion in the rfs-1 mutant, telomere lengthening can also be observed (Figure 6). Second, rather than showing progressive loss of fertility, the rfs-1 mutants display variability in overall fecundity from generation to generation (Figure 4). Third, rfs-1 can have stable karyotypic changes well before the onset of sterility. rfs-1 is also unlike the classes of Mrt genes that display microsatellite repeat instability and mutator phenotypes. These results suggest that rfs-1 is a novel Mrt gene that is required for telomere homeostasis and genome stability.

Telomeric repeats in rfs-1 show periods of shortening, lengthening, or relative stasis. Such changes in telomere length are consistent with the known role for rfs-1 in homologous recombination repair of stalled replication forks (Ward et al. 2007). Telomeric repeats have long been known to cause fork stalling due to their repetitive nature (reviewed in Mirkin 2006). The repetitive nature of the telomere can lead to sequence misalignment and a gain or loss of repeats. Repeat misalignment and pairing with a nonhomologous chromosome or with an internal telomeric sequence can result in the formation of telomere length changes, fused chromosomes, or inversions, respectively. Thus the apparent contradiction that rfs-1 can both enhance and suppress the trt-1 sterility phenotype can be explained by HR-dependent repair of telomeres. The early onset sterility in trt-1; rfs-1 could be due to the additive affects of telomere loss by end attrition and by misalignment and deletion during replication fork repair. The increased survival of a trt-1; rfs-1 line could occur if the repair of a stalled fork led to telomere repeat lengthening or to a stable chromosome fusion, thus eliminating a short and potentially fusagenic telomere. In fact, I observed telomere proximal fusions in both the rfs-1 background and in the trt-1; rfs-1 mutant background. Thus, rfs-1 may have a role in telomere length homeostasis through its role in promoting replication fork reinitiation.

rfs-1 is the first C. elegans replication/repair factor for which both lengthening and shortening of telomeres has been observed. Several other replication factors have been shown to have roles in telomere length regulation, including RPA and yeast ELG1 (Smith et al. 2000; Kanellis et al. 2003). RPA is intriguing because despite a hyperrecombination phenotype, only telomere attrition has been observed. RPA, like rfs-1, helps prevent replication fork collapse at telomeres. RPA also leads to premature senescence in the absence of telomerase (Smith et al. 2000), which may be analogous to the enhancement of trt-1 sterility seen in the trt-1; rfs-1 double mutant. Since recombination should lead to both increases and decreases in telomere length, it will be interesting to further investigate whether there is a fundamental difference between the activity of rfs-1 and these other genes in telomere repeat replication or whether the failure to observe telomere lengthening in other mutants was a consequence of experimental design.

Several lines of evidence suggest that rfs-1 has a general role in genome intregity. First, the strong sensitivity of rfs-1 to replication-inducing lesions (Ward et al. 2007) suggests that stalled replication forks anywhere in the genome can be a substrate for Rfs-1. However, rfs-1 does not confer a Mutator phenotype, suggesting either that the number of replication stalling lesions is very low in C. elegans or, more likely, that the majority of these repair events are repaired with high fidelity off the homologous chromosome. Second, rfs-1 increases the frequency of intrachromosomal deletions in poly(G) tracts in the dog-1 mutant (Ward et al. 2007). Nevertheless, rfs-1 did not by itself create deletions in poly(G) repeats, suggesting that these repeats are replicated efficiently. Third, fluctuations in rfs-1 fecundity and the appearance of stable karyotypic changes are reminiscent of the formation of balanced translocations (Rosenbluth et al. 1985). These changes could occur anywhere in the genome, but given the prevalence of telomere-like repeats, it is likely that at least some of these stable changes in karyotype directly involve telomeric DNA and internal telomeric repeats.

My data, together with a growing literature on Rad54 and Rad51D, suggest that both telomerase and recombination are essential for telomere homeostasis (Tarsounas et al. 2004). It is intriguing to consider that since rfs-1 is the sole C. elegans Rad51 paralog (Ward et al. 2007) that the conserved function of these family members is recombination repair and that rad51D become specialized to meet the challenges imposed by repetitive sequences in the genome. Differences in behaviors of the Rad51D homologs (with regard to cell viability) may reflect differences in the specialization of the gene function in the different organisms or a more general difference in telomere regulation. In addition to a role for Rad51D at the telomere, XRCC3 and Rad51C are in a complex that binds chicken-foot structures that are found at regressed replication fork intermediates (Griffith et al. 1999). They act to reinitiate replication without significant loss of telomeric sequence. It will be interesting to determine whether rfs-1 has retained this ability to resolve regressed forks. Alternatively, rfs-1 may be required to recruit such factors to the telomere. It is tempting to speculate that rfs-1/Rad51D might function with TRF2 (or its worm ortholog), which has been implicated in both replication reinitiation and telomere end-structure formation (Fouche et al. 2006).

Mrt phenotype of rfs-1 is dependent on growth conditions. When allowed to live without food for several days each generation, rfs-1 populations displayed sterility within 20 passages (the equivalent of 40 generations). When maintained continually in the presence of food, the Mrt phenotype arose only when a significant bottleneck was imposed on the population (Figure 2). The simplest model to explain these observations is that stress, at least in the form of starvation, exacerbates the rfs-1 phenotype. Whether starvation increases the incidence of replication fork stalling or affects the processing of stalled intermediates will be of interest to resolve. Furthermore, it will be interesting to determine whether starvation acts directly on the presumptive germ cells (in the starved L1 larvae) or whether it induces a systemic response that leads to later changes in the adult germline much in the same way that life span is affected by stresses during larval development (Dillin et al. 2002).

Alternatively, rfs-1 may accentuate the effect of stress on telomere length. Over the past 10 years, it has become increasing clear that oxidative stress can induce telomeric shortening (Liu et al. 2002, 2003; Von Zglinicki 2002; Tchirkov and Lansdorp 2003) although the mechanism by which this occurs is not fully understood. Oxidative stress can both increase DNA lesions in telomeric DNA (Petersen et al. 1998; Von Zglinicki et al. 2000) and also disrupt the association of the essential telomere binding proteins, TRF1 and TRF2, which protect the ends from fusion (Opresko et al. 2005). Other stresses including hyperoxia and heavy metal exposure also induce telomeric shortening (Von Zglinicki et al. 1995; Liu et al. 2003). In vivo studies have also suggested that stress can accelerate telomere shortening, although the heterogeneity of telomere lengths in normal tissues has precluded definitive proof of these events (reviewed in Von Zglinicki and Martin-Ruiz 2005). My observation that the onset of the rfs-1 Mrt phenotype is accelerated under conditions of starvation and that it may be related to telomeric dysfunction or to a more general effect on genome stability suggests that C. elegans might be a good model system to address the functional link between stress and telomere length regulation or genome stability in the context of the whole organism.