Natural genetic variation in a multigenerational phenotype in C. elegans

Summary

While heredity mostly relies on the transmission of DNA sequence, additional molecular and cellular features are heritable across several generations. In the nematode Caenorhabditis elegans, insights into such unconventional inheritance result from two lines of work. First, the mortal germ line (Mrt) phenotype was defined as a multigenerational phenotype whereby a selfing lineage becomes sterile after several generations, implying multigenerational memory [1, 2]. Second, certain RNA interference (RNAi) effects are heritable over several generations in the absence of the initial trigger [3–5]. Both lines of work converged when the subset of Mrt mutants that are heat-sensitive were found to closely correspond to mutants defective in the RNAi-inheritance machinery, including histone modifiers [6–9]. Here we report the surprising finding that several C. elegans wild isolates display a heat-sensitive mortal germline phenotype in laboratory conditions: upon chronic exposure to higher temperatures such as 25°C, lines reproducibly become sterile after several generations. This phenomenon is reversible as it can be suppressed by temperature alternations at each generation, suggesting a non-genetic basis for the sterility. We tested whether natural variation in the temperature-induced Mrt phenotype was of genetic nature by building recombinant inbred lines between the isolates MY10 (Mrt) and JU1395 (non-Mrt). Using bulk segregant analysis, we detected two quantitative trait loci. After further recombinant mapping and genome editing, we identified the major causal locus as a polymorphism in the set-24 gene, encoding a SET- and SPK-domain protein. We conclude that C. elegans natural populations may harbor natural genetic variation in epigenetic inheritance phenomena.

Results

Many C. elegans wild isolates show a temperature-sensitive mortal germline phenotype

C. elegans is routinely cultured between 15 and 25°C [10]. At 25°C, the reference strain of C. elegans, N2, can be maintained for many generations (e.g. Figure 1B in [11]). While collecting new C. elegans wild isolates, we discovered that many display a multigenerational sterility phenotype (or mortal germ line, Mrt phenotype) when chronically exposed to temperatures such as 25°C. Figure S1 shows a multigenerational sterility assay at 25°C on 14 such wild isolates.

Figure 1. A subset of C. elegans wild isolates display a strong temperature-sensitive Mortal germ line (Mrt) phenotype.

(A) Experimental design for Mortal germline (Mrt) phenotyping. The phenotypic value is the number of generations (n) until sterility. (B) The Mrt phenotype of five wild C. elegans isolates was followed over 25 generations at different temperatures. The proportion of sterile replicates accumulating along generations is plotted. For QX1211, MY10 and JU775, n=20 lineages were scored at each temperature; for JU1395 and JU1171, n=20 lineages were scored at 25°C and 23°C and 10 at 18°C and 15°C. (C) The brood size along 30 single-individual lineages of MY10 and JU775 was manually counted from generation 1 to the generation when they became sterile at 23°C. The brood size decreases progressively in individual lineages in the generations preceding full sterility. (D) At generation 3, all MY10 animals raised at 15°C had a normal anatomy whereas those grown at 25°C were all sterile, with a reduced gonad and a characteristic dark intestinal color. Same scale for both panels. Bar: 1 mm. Eb, embryo. La, larvae. (E) Test of the reversibility of the MY10 Mrt phenotype with alternations of temperature at each generation (right panel). All ten replicates survived over 25 generations of such treatments, thus 13 non-consecutive generations at 25°C. With constant exposure to 25°C, all 50 lineages had a Mrt-value between 2 and 4 (left panel). See also Figure S1.

Focusing on three isolates with a strong Mrt phenotype and two non-Mrt isolates, we performed a multigenerational sterility assay at several temperatures ranging from 15°C to 25°C (Figure 1A-B). At 25°C, we observed a highly reproducible sterility phenotype for the three Mrt isolates after a few generations, with QX1211 (from California) showing the strongest phenotype, followed by MY10 (Germany), then by JU775 (Portugal). In contrast, the isolates JU1395 (France) and JU1171 (Chile) remained fertile throughout the 20 generations of the assay (Figure 1B). The Mrt phenotype of QX1211, MY10 and JU775 was quantitatively affected by temperature (Figure 1B). For example, at 25°C, MY10 lines were fully sterile after three or four generations; at 23°C their half-life was of four generations (range 3-6 generations), while at 21.5°C their half-life was of six generations with a wider range between four and eleven generations. Even at 20°C, the conventional culture temperature of C. elegans, up to 95% of the QX1211 lines, 75% of the MY10 lines and 40% of the JU775 lines were extinct after 20 generations (n=20 lines for each). At 15°C, all remained fertile over 20 generations. We conclude that C. elegans wild isolates differ quantitatively in the temperature-sensitive mortal germline phenotype and that the number of generations to sterility is highly sensitive to temperature.

We observed that the decline in fertility in the Mrt isolates was progressive, with a low brood size and a high male incidence in one or two generations prior to full sterility. In C. elegans where males are X0, high male incidence is the result of meiotic errors due to non-disjunction of the X chromosome, thus of meiotic "defects". We followed the lineage of single individuals (rather than averaging over three) of the MY10 and JU775 isolates and quantified their brood size over several generations at 23°C. In this experiment, we also observed a progressive decrease in brood size and a high male incidence in the generations before full sterility (Figure 1C).

In the final generations leading to sterility in MY10 at 25°C, we observed further germline abnormalities, such as sperm and/or oocyte differentiation defects, germline atrophy and meiotic chromosome pairing defects in oocytes (Figure 1D, Figure 2A-C, Figure S2A-C). The double-stranded breaks that normally occur transiently during prophase of meiosis I in the transition zone [12] were not resorbed and appeared more numerous in late pachytene of MY10 animals after three generations at 25°C (Figure 2D). These germline defects resemble previously reported temperature-sensitive mortal germline phenotypes in various mutants [7, 8, 13–15].

Figure 2. Germline integrity defects, chromosomal aberrations and 22G small RNA depletion in the final generations before or at full sterility.

(A) DAPI staining of JU1395 and MY10 animals at generation 2 (G2) of culture at 20°C or 25°C. Dashed lines indicate the distal germline arms. Wild type germlines display two symmetric arms (e.g. JU1395 25°C; MY10 20°C). In contrast, at 25°C, MY10 G2 displayed a disrupted gonad with medium to small/empty arms (middle right) or displayed germlines with endomitotic oocytes that endoreplicate in the absence of sperm (left). Emo, endomitotic oocyte. (B) At 25°C, MY10 G2 animals display various defects. Some MY10 G2 animals possess only oocytes, some only sperm and the majority of them have neither sperm nor oocytes (rightmost panel). Dashed lines indicate the germline arms, dotted circles surround single oocytes at diakinesis or the pool of sperm in spermathecae. (C) DAPI-stained late-G2-generation MY10 oocytes arrested in diakinesis grown at 20°C or 25°C. The dotted circle surrounds a single oocyte nucleus. The MY10 20°C oocyte displays the normal set of six paired chromosomes in diakinesis, whereas the MY10 25°C oocytes have more than six chromatin-DAPI stained bodies, indicative of meiotic pairing defects. (D) RAD-51 immunostaining in oocyte nuclei for MY10 and JU1395 G2 at 25°C. In JU1395, a few RAD-51 foci (red) are apparent in the transition zone (TZ) and the late zygotene/early pachytene (LZ/EP) zone and disappear in late pachytene nuclei (LP). RAD-51 foci in MY10 G2 animals persist until late pachytene and are more numerous. Scale bars: 2 μm. (E) Percentage of 22Gs (small RNAs of length 22 nucleotides, starting with a guanine) in the small RNA populations in the MY10 and JU1395 strains, across generations. The dots represent the mean over three replicates per strain and per generation, for generations 0, 1 and 2. The bars represent the standard deviations. The effect of the time (generation) on percentages of 22G was tested using a general linear model with a Poisson regression. ns stands for the non-significant. ** stands for a p-value < 0.005. See also Figures S2 and S3.

We monitored the secondary small siRNA populations, the major effectors of RNA interference pathways in C. elegans [16], using 5'-independent libraries, enriching for 5' triphosphate sRNAs [17, 18], in MY10 and JU1395: generation 0 was grown at 15°C and further generations were grown at 23°C until MY10 reached full sterility (Figure 2E and Figure S3). At 15°C, the compositions of small RNA populations, and specifically the proportions of 22G small RNAs (22-nucleotide RNAs with the first nucleotide a G, the major class of secondary siRNAs in C. elegans [16]), were similar in MY10 and in JU1395. In JU1395, the 22G proportion was stable over time (from the glm model, the effect of the number of generations was not significant, p = 0.967). In contrast, in MY10, the 22G small RNA pool was progressively depleted (p = 0.0034; Figure 2E, Figure S3). This progressive change of small RNA populations in MY10 likely corresponded to the germ line atrophy described above. However, it remained unclear whether the 22G depletion was a consequence or a cause of germ line atrophy and this would be interesting to dissect further in the future.

To test whether the sterility phenotype was caused by the accumulation of irreversible damage - for example in DNA - we submitted MY10 animals to a regime alternating between the temperatures of 15°C and 25°C at each generation, with a control at a constant temperature of 25°C (Figure 1E). Under this regime, all 10 temperature-alternation lines remained fertile after a total of 13 generations at 25°C, while all controls were sterile after four generations. This experiment rules out irreversible damage occurring at 25°C as the source of multigenerational phenotype formation.

These results overall demonstrate natural variation in the temperature-sensitive mortal germline phenotype among C. elegans wild isolates, thus a variation in multigenerational memory of the heat environment.

Crosses between MY10 and JU1395 yield two quantitative trait loci

To determine whether the natural variation in the Mrt phenotype could be mapped genetically, we crossed isolates MY10 (Mrt) and JU1395 (non-Mrt) in both directions to build Recombinant Inbred Lines (RILs), at 15°C (Figure 3A). We then phenotyped the 120 RILs in triplicates for 25 generations at 25°C (Data S1A). Figure 3B shows the distribution of phenotypic values in the RILs: some lines exhibited a phenotypic value that was intermediate between the parental values; in addition, the distribution was biased, with more lines exhibiting a strong MY10-like Mrt phenotype than the JU1395 non-Mrt phenotype. This distribution was not consistent with segregation of a single genetic locus, indicating the involvement of at least two loci – provided the difference in the parental Mrt phenotype was indeed of genetic nature.

Figure 3. Bulk segregant analysis of Recombinant Inbred Lines indicates two major loci.

(A) Crossing scheme for the production of the RILs. Only two chromosomes are represented. (B) Mrt-value distribution for the 120 RILs (mean of three replicates). Full data are provided in Data S1A. The two extremes of the distribution were incorporated in the bulk segregant analysis as shown. Because few lines had a fully non-Mrt phenotype, we included in the high-trait value pool weak mortal germline lines which resisted on average more than 22 generations. (C) MY10 allele proportions along the genome (six chromosomes and mitochondrial DNA) are in blue for the Mrt Pool and in red for the non-Mrt pool, respectively. Full data can be found in Data S1B. Log-odd values were calculated for 1.2 Mb as explained in Methods. The green lines indicate the thresholds for a significant difference between log-odd ratios at p=0.01. (D) Phenotyping of Near Isogenic Lines over 20 generations confirms the QTLs detected in C. The Mrt phenotype (5 replicates) is represented with a violin plot. The NIL genotype is represented on the right with a color code along the six chromosomes. Red: JU1395 background; blue: MY10; grey: unknown. Strain genotypes are provided in Data S2B. Wilcoxon rank sum test p-values are specified as ** for p < 0.005. See also Tables S1 and S2.

Using whole-genome sequencing, we next genotyped two pools of RILs corresponding to the extremes of the distribution in the Mrt phenotype (Figure 3B, Data S1), as well as the parental genome. We then assessed in each RIL pool the relative parental allele frequency across the genome. We thereby identified two chromosomal regions with a significant excess of MY10 allele in the Mrt pool compared to the non-Mrt pool, thus regions representing quantitative trait loci (QTL) for the Mrt trait. The strongest QTL on chromosome II was fully fixed for the MY10 allele in the Mrt pool and at 80% for the JU1395 allele in the non-Mrt pool. A second significant QTL was detected on chromosome V, where in both directions about 75% of the lines had the same parental allele. For both QTLs, the MY10 variant acted expectedly in the direction of decreasing the number of generations to sterility (Figure 3C). The mitochondrial DNA showed an equal proportion of both alleles in both pools, excluding an important role of the mitochondrial genotype or of other maternal effects.

We directly tested the effect of the two QTL regions by introgressing them in both directions (i.e. into both parents) to produce Near Isogenic Lines (NILs) (Figure 3D, Data S2). As expected, chromosome II replacements had a strong effect on the Mrt phenotype. The JU3178 line carrying the MY10 chromosome II in the non-Mrt JU1395 background became sterile after 5-6 generations at 25°C (Wilcoxon rank sum test with continuity correction between JU3178 vs. JU1395, p = 0.0034 and JU3180 vs. MY10, p = 0.0036). Conversely, the JU1395 chromosome II replacement in the MY10 background (line JU3180) partially suppressed the Mrt phenotype of MY10. Chromosome V replacement had a mild effect on the Mrt phenotype into the JU1395 background (JU3177 vs. JU1395, p = 0.0038) and no effect on the Mrt phenotype into the MY10 background (MY10 and JU3184, p = 0.11). The combination of chromosomes II and V had strong effects: the introduction of MY10 chromosomes II and V in the JU1395 background aggravated the Mrt phenotype compared to chromosome II alone (JU3212 vs. JU3178, p = 0.017); conversely, the introduction of chromosomes II and V of JU1395 in the MY10 background fully suppressed the Mrt phenotype (JU3181/JU3183 vs. MY10, p = 0.0003; Figure 3D). Altogether, the chromosome replacements validated the QTLs identified in the bulk segregant analysis. They also matched well the asymmetric distribution of RIL phenotypes (Figure 3B), which suggested that the combination of two JU1395 alleles was required for an immortal germline phenotype.

These results importantly further indicate that the natural variation in Mrt phenotype can be mapped on the genome.

A deletion in the gene set-24 underlies the major QTL on chromosome II

In order to identify the molecular nature of the QTL on chromosome II, we screened for recombinants on chromosome II after crossing MY10 and the NIL JU3180 (introgression mfIIR48, i.e. the JU1395 chromosome V in the MY10 background; Figure 3D). This further step of recombinant mapping resulted in a 1,574,560 pb interval (Figure 4A, Data S2B).

Figure 4. The major QTL is explained by a deletion in the set-24 gene in MY10.

(A) Recombinant mapping of the chromosome II QTL after a cross between JU3180 (see Figure 3D) and MY10 to obtain further recombinants on chromosome II. Red areas refer to JU1395 alleles/regions and blue areas refer to MY10 alleles/regions. Grey areas have an unknown genotype. Phenotyping of the recombinants over 20 generations is shown in the left panel. Dark and light grey represent two independent assays. Strain genotypes are provided in Data S2B. Wilcoxon rank sum test p-values between MY10 and respectively JU3220, JU3187, JU3219, JU3218 are specified as ** for p < 0.005. (B) Predicted set-24 gene structure in the N2 reference, with the protein domains indicated by different colors. The 5’UTR and two first exons of SET-24 are deleted in MY10 (mfP23). Position of the set-24 deletion alleles obtained by genome editing in JU3219 and JU1395 are indicated by the green triangle. (C) Phenotyping of the set-24 deletion alleles in JU3219. In the chromosome drawings on the right, red areas refer to JU1395 regions and blue areas refer to MY10 regions. The green stars correspond to the location of set-24 alleles mf123 (JU3253) and mf124 (JU3254). Wilcoxon rank sum test p-values are specified as * for p < 0.05. See also Data S2 and Figure S4.

Analysis of the molecular variants in this interval (Data S2C) suggested as a possible candidate a deletion in the set-24 gene in MY10, which we called mfP23. mfP23 encompasses a 5' upstream region and the two first exons of set-24, which deletes its entire predicted SET-domain (Figure 4B and Figure S4A-B). SET stands for the Drosophila melanogaster genes Su(var)3-9, Enhancer-of-zeste and Trithorax, and the SET domains are known to interact with histone tails [19, 20]. According to Interpro [21] predictions, the C. elegans genome codes for 34 SET-domain containing proteins. The set-2, set-25, set-32 and met-2 genes have been shown to encode methyltransferases that act on the balance between the modifications of different histone residues and thereby affect transgenerational RNAi inheritance [7–9, 22–28]. Moreover, loss of function in set-2 and set-32 also leads to a temperature-sensitive Mrt phenotype [7, 9, 22, 23, 26, 29]. Two SET-domain proteins, SET-5 and SET-24, in addition contain two SPK ('SET and PHD domain-containing proteins and protein kinases') domains of unknown activity. Little is known about SET-24, which is associated with no obvious phenotype in the C. elegans N2 reference strain [30] (but see [31] for a possible enhancement of the fertility defects of a spr-5 mutant). However, its putative histone binding ability and its expression in the germline (Wormbase.org) made the set-24 indel polymorphism a good candidate. To test whether the set-24(mfP23) deletion caused the Mrt phenotype in MY10, we introduced the full-length N2 version of the set-24 gene (YAC Y43F11A.5) into MY10 using extrachromosomal array transgenesis. The transgene partially but significantly rescued the strong Mrt phenotype of MY10 (Figure S4C, JU3310, JU3301 compared with JU3250, JU3249; Wilcoxon rank sum test with continuity correction, p = 0.0025). A weak rescue was expected as extrachromosomal arrays tend to be silenced in the C. elegans germline [32]. To confirm the effect of the set-24 indel, we conversely induced CRISPR/Cas9 deletions in the intact JU1395 allele of the set-24 gene in the NIL JU3219. These set-24(mf123) and set-24(mf124) alleles strongly enhanced the Mrt phenotype (Figure 4C). We thus conclude that the set-24 indel underlies the major QTL on chromosome II.

To understand the evolutionary history of the set-24(mfP23) allele, we investigated its presence in a set of 249 wild C. elegans isotypes whose genomes are available on the Caenorhabditis elegans Natural Diversity Resource (CeNDR) website [33]. The mfP23 allele was found to be specific to the group of isolates with a very similar genome sequence as the MY10 isotype, all sampled from the same compost heap in Roxel (Germany) in 2002 (Data S2D). Although the Roxel compost heap where MY10 was found was frequently sampled by the Schulenburg laboratory, isolates close to this isotype were not found again in Roxel, nor anywhere else [34, 35] (Data S2D). We thus confirmed the natural origin of the set-24(mfP23) allele, and showed that it is a rare allele.

Many more genetic polymorphisms may modulate the mortal germline phenotype of C. elegans wild isolates. Wild isolates such as QX1211 and JU775 do not carry the set-24 deletion (Figures 1 and S1), thus their strong Mrt phenotype is likely be explained by other genetic polymorphisms. Furthermore, our QTL mapping based on MY10 x JU1395 RILs uncovered a second QTL on chromosome V. In addition, a third QTL may be present on the left of chromosome II. Indeed, CRISPR/Cas9 deletions in the set-24 gene did not result in a Mrt phenotype when introduced into the background of JU1395 (Figure S4E). Comparing these knockouts with the phenotype of the JU3178 chromosome II replacement line (Figure 3D), it is possible that another locus on the left of the chromosome II QTL region (Figure 3C) interacts synthetically with the set-24 deletion to produce an intermediate value of Mrt phenotype in JU3178. That the chromosome II QTL may correspond to two molecular loci is consistent with the wide peak observed in the bulk segregant analysis. Although more replicates would be needed, Figure 4A (and Data S2) mapping results are also not inconsistent with an additional locus on the left of the original QTL peak that would have a weaker effect than set-24 in the MY10 background. Overall, these data suggest that the temperature-sensitive Mrt phenotype in C. elegans is likely influenced by many genetic polymorphisms.

Discussion

In the 20^th century framework of Mendelian genetics and the Modern Evolutionary Synthesis, genetic inheritance through DNA sequence variation was accepted as the sole mode of inheritance [36]. During the past few years, however, additional modes of inheritance have been uncovered in different organisms [27], including in C. elegans [3–5, 8, 9, 22, 24, 25, 27, 28, 31, 37–46]. It remains unclear whether these modes play an important role in nature and how they impact the course of evolution. In C. elegans, an alternative inheritance mode operates through 2° small interfering RNAs (siRNAs) obtained by amplification of 1° siRNAs at each generation. The inheritance of silencing through successive generations requires that 2° siRNAs are transferred into germline nuclei and impinge on histone modifications, thereby regulating gene expression across multiple generations [6, 8, 9, 22, 25, 27, 31, 39, 40]. This RNAi inheritance pathway relies on the Argonaute HRDE-1/WAGO-9 binding 2° siRNAs, MORC-1, histone-binding proteins such as HPL-2/HP1 and histone methyltransferases such as SET-2, SET-25 or SET-32 [7–9, 22–27]. Mutations in this inheritance system are specifically those resulting in a temperature-sensitive mortal germline phenotype (including for null alleles). In contrast, other mortal germline mutations in the reference N2 background affect other processes, for example DNA repair or chromosome pairing mechanisms [1, 47–49] and are generally not temperature-sensitive.

Here we found natural genetic variation in the temperature-sensitive mortal germline phenotype. The occurrence of this natural sterility phenotype at such relatively mild temperatures seems surprising. For example, the JU775 isolate and others with a Mrt phenotype (Figure S1) were found in or near Lisbon, Portugal, in the month of July, when temperatures rise well above 20°C. We hypothesize that the multigenerational sterility is likely suppressed by some unknown environmental factors in the wild (fluctuations in temperature may be such an environmental factor), so that the here reported Mrt phenotype only appears under particular laboratory conditions. A superficially similar phenotype was previously described at 23-24°C for a Bergerac strain studied by Brun [50, 51]. However, it turned out that this strain carried a temperature-sensitive allele in the zyg-12 gene that renders the strain immediately sterile at 25°C [52], independently of the high transposition activity that arose during culture of this strain [51, 53]. When we reassayed different derivatives of the Bergerac strain [51], we did not observe the same progressive phenotype as in our newly tested C. elegans isolates, for which fertility at high temperature is high at the beginning and progressively declines with a visible reduction in gonad size in the last generation (Figure 1C). We instead observed a very low brood size from the first generation at 23°C.

The deletion allele in the set-24 gene that we found as the main locus explaining the mortal germline phenotype in MY10 is a rare allele, which is not surprising as we had chosen MY10 as an isolate with an extreme Mrt phenotype, a choice that is likely to yield rare alleles. This derived allele may not have been maintained for a long time maybe due to counterselection at this locus. Because of the low outcrossing rate and high linkage disequilibrium in C. elegans [54–57], it may also have disappeared under counterselection for a linked locus, or under drift.

We distinguished here four levels of variation contributing to natural variation in the Mrt phenotype. First, the Mrt phenotype is sensitive to temperature and potentially to other environmental factors through chronic exposure. Second, the final sterility is caused by gross germ line abnormalities and germ line reduction in the last two generations. Whether this final sterility is caused by DNA damage, massive transposition or non-DNA sequence based phenomena such as deregulated gene expression remains unclear. Third, we show that the multigenerational nature of the phenomenon is caused by a reversible process, thus not by DNA damage accumulation or transposition that would occur within a generation at 25°C. Examples of possible processes that may explain the multigenerational nature of the phenotype are the accumulation of histone and siRNA composition modifications. Fourth, we demonstrate that variation in the Mrt phenotype among wild isolates of C. elegans at a given temperature can be attributed to genetic differences, as confirmed by QTL mapping and identification of the set-24 deletion.

In the N2 background, the temperature-sensitive Mrt phenotype corresponds to mutations in the RNAi inheritance pathway. The observed temperature-sensitive Mrt phenotype of C. elegans wild isolates was progressive and reversible, also suggesting an important non-genetic component in its accumulation across generations. Therefore, the diversity of the Mrt phenotype of C. elegans wild isolates observed under laboratory culture conditions provides an exciting model to test whether and how inheritance systems are modulated by natural genetic variation. In a now extended view of heredity in evolution, some forms of non-genetic inheritance may work in parallel to Mendelian heredity. Given this plurality of inheritance systems, how these systems themselves may vary in natural populations becomes a central question in biology. From the generally lesser stability of the variants, non-genetic inheritance may be particularly important on shorter timescales, depending on the frequency of environmental fluctuations, and may shape phenotypic evolution in this manner [58, 59].

Although more work is required to define the nature of the multigenerational memory and how it is affected by natural variation, our findings provide the first piece of evidence on genetically encoded variation for a multigenerational phenotype. The way in which SET-24 may influence histone modifications is still to be determined but its molecular nature converges with our current understanding of multigenerational inheritance in C. elegans. We conclude that natural populations harbor genetic variation in phenomena of multigenerational inheritance.

STAR Methods

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Marie-Anne Félix (felix@biologie.ens.fr).

Experimental Model and Subject Details

Caenorhabditis elegans animals were grown on 55 mm diameter NGM plates and fed with Escherichia coli OP50 bacteria [10], at indicated temperatures. Hermaphrodites were used, except for crosses where hermaphrodites were mated to males. We specificy the developmental stage of the animals when relevant.

The complete list of strains used in this study and their laboratory origin can be found in the Key Resources Table. The wild origin of C. elegans isolates can be found in [56]. JU isolates were isolated by our laboratory, while the others were obtained through the Caenorhabditis Genetics Center.

Method Details

Mortal germ line (Mrt) phenotyping

Prior to starting the phenotyping assays, adult hermaphrodites were bleached and their progeny grown for at least three generations at 15°C to produce the initial populations. Phenotyping assays begun with the transfer of three L4-stage hermaphroditic larvae (generation G₁) onto new NGM plates that were then incubated at the chosen temperature for the assay (generally 25°C, Figure 1A). Each generation G_n was then started with three L4-stage larvae picked in the progeny of generation G_n-1. When all three individuals of generation G_n were sterile, we took n as the value for the Mrt phenotype. Phenotyping assays were stopped after 20-25 generations and lineages that remained fertile were considered non-Mrt. Assays were replicated as indicated for each case either in methods or figure legends.

As the standard assay, three larvae were transferred at each generation, except if otherwise indicated. The original Mrt screen [1] was performed with transfer of six L1 larvae. We preferred to transfer L4 larvae when it is possible to differentiate males and hermaphrodites. We considered three as sufficient to avoid extinction due to manipulation or other hazards of life on a Petri dish such as drying on the side. Some assays focusing on specific lineages were performed by transferring a single larva, while the assays with GFP transgenes in Figure S4C were performed with ten larvae so as not to lose the extrachromosomal transgene.

Production of Recombinant Inbred Lines (RIL)

Recombinant Inbred Lines (RILs) were obtained starting from crosses between MY10 and JU1395 in both directions of crossing: MY10 males with JU1395 hermaphrodites, and MY10 hermaphrodites with JU1395 males. We genotyped the F1 progeny using niDf250del-F/R primers to identify heterozygote individuals. For each cross direction, one heterozygote hermaphrodite F1 was randomly chosen. 120 recombinant inbred lineages (60 in each direction of the cross) were then built by transferring one hermaphrodite per generation for 11 generations (Figure 3A). To prevent allele segregation biases which could emerge from temperature associated traits, animals were grown at 15°C throughout RIL production.

RIL phenotyping and pooling strategy

We measured the Mrt phenotype of each RIL at 25°C in triplicates for 25 generations. We decided to use a pooling strategy for the genotyping, focusing on the two extremes of the phenotypic distribution: a pool of 21 RILs with a non-Mrt phenotype and another pool of 22 RILs with a MY10-like phenotype. The 120 RILs, their Mrt-values and the RILs included in the QTL mapping are listed in Data S1A.

5'-end independent small RNA sequencing

Our aim was to monitor the small RNA populations along the progressive onset of fertility defects. We characterized the small RNA composition in JU1395 and in MY10, across generations and for three independent replicates each. Following the Mrt phenotyping design explained above, each generation was started with three L4 larvae from the previous generation, generation 0 (G0) was grown at 15°C and subsequent generations were grown at 23°C. As MY10 was sterile at generation 3 for two replicates and at generation 4 for one replicate, the siRNA monitoring was performed until generation 2 in triplicates for MY10 and JU1395. We added the siRNA data for the remaining replicate of MY10 that was fertile at generation 3 and sterile at generation 4. At each generation, adults were harvested more than 32 hours after the L4-to-adult molt. Each sample was immediately homogenized in Trizol (Invitrogen) and stored at −80 °C. RNA from each sample was extracted with isopropanol and chloroform (2.5:1). In order to obtain all small RNAs in a 5 -end independent manner (with and without triphosphate ends), samples were treated with 5 polyphosphatase (Epicentre/Illumina) for 30 min. Small RNA libraries were generated from 1 μg polyphosphatase-treated total RNA using the TruSeq Small RNA kit (Illumina), following the manufacturer’s instructions. Libraries were sequenced using the Illumina HiSeq 1500 to generate 36-nucleotide single-end reads.

Analysis of the small RNAs

Adaptors were removed from the Fastq files using the program Cutadapt v1. Fastq sequences were then trimmed to leave only reads of length between 16 and 33 nucleotides and were aligned using STAR (version STAR_2.4.2a) to the N2 reference genome (WS235 genome version ftp://ftp.wormbase.org/pub/wormbase/releases/WS235/species/c_elegans/c_elegans.WS235.genomic.fa.gz), reporting in a bam file only the best single alignment with up to one mismatch allowed. For each generated bam files, we extracted the counts of reads grouped according to their length and the identity of their first nucleotide. The reads with secondary alignment were not taken into account in our analysis. Data were normalized to the total number of reads with a length between 16 to 33 nucleotides. Plots of the proportions of siRNA grouped according to their length and their first nucleotide were generated to illustrate the small RNA population differences between generations.

DNA preparation and whole genome sequencing

Genomic DNA was extracted from mixed-stage growing populations of each RIL as described in [60]. DNA concentrations were quantified using a Qubit fluorometer with the dsDNA BR Assay Kit (ThermoFisher) and adjusted to 1 μg/μL. For pool sequencing, DNAs of RILs of a given pool (as specified in Data S1A) were mixed in equimolar proportions. We also prepared DNA from the parents, MY10 and JU1395. Four paired-end libraries were built using the Nextera DNA Library preparation kit (Illumina), following manufacturer's instructions and sequenced on an Illumina HiSeq1500 (100 bp paired reads).

Analysis of the parental genomes

Sequence quality was checked using FastQc. Reads were filtered for quality, mapped to the genome of the C. elegans reference N2 strain (WS243 genome version ftp://ftp.wormbase.org/pub/wormbase/releases/WS243/species/c_elegans/PRJNA13758/c_elegans.PRJNA13758.WS243.genomic.fa.gz) and variants called, following the first 7 steps of the mapping-by-sequencing pipeline (andalousian-map_Portable.sh) described in [60].

The Pindel software [61] was used to detect in homozygous indels in MY10 and JU1395 genomes. We used the bam format outputs from the above mapping pipeline as input files, specifying 250 as the expected average insert size. Parent-specific deletions were identified and those in the mapping interval manually checked using Tablet [62].

The Ensembl Variant Effect Predictor [63] and the WBcel235 annotation of the genome of reference were used to annotate impacts of the parent-specific SNPs and deletions (Data S2C).

Quantitative trait locus (QTL) mapping

As first described in [64], the purpose of bulk segregant analysis is to detect genomic regions where parental allele proportions deviate between groups of contrasted phenotypes, here between the Mrt and non-Mrt pools.

To do so, we first defined a set of SNPs (markers) between the two parents, MY10 and JU1395, along the genome. Using the compare function of the vcftools software [65], we listed the positions where SNP were homozygous and different between the two parental genomes. As a second step, we selected these positions in the genomic sequences of the Mrt and non-Mrt pools. For each pool and each marker, we extracted the total number of reads and the number of reads corresponding to the MY10 allele. Except for the mitochondrial DNA, we excluded positions where the total number of reads was below 20, above 250 (putative repeats) or where quality was not maximal. Proportions of MY10 alleles were calculated for each marker as the ratio of read counts with the MY10 allele divided by the total number of reads at this SNP. Data are provided in Table S1B. To minimize the effect of heterogeneous distribution of SNP along the chromosomes, we first calculated the mean allele frequencies on a 300kb window size with no overlap. We then displayed the distribution of MY10 allele proportions along chromosomes in a sliding window manner with a window size of 1.2Mb and step of 300kb. See below in 'Statistical analysis' for the log-odds ratio calculation.

Construction of Near Isogenic Lines (NIL)

Near Isogenic Lines were produced in order to confirm the QTL identified by bulk segregant analysis. We crossed MY10 and JU1395 to introduce chromosome II, chromosome V or both from one parent into the genetic background of the other parent through repeated backcrosses. All crosses were made at 18°C. Chromosomes were followed by PCR-based genotyping using the primers listed in Data S2A. These primers were designed to amplify regions neighboring MY10- or JU1395- specific deletions (ranging from 100 to 2520 bp). PCR products were separated on a 1.5% agarose gel to discriminate between alleles. We cannot rule out that the tips of the chromosomes outside the genotyped markers were from the other parent. NILs were genotyped over 20 generations.

To fine map the QTL, we crossed JU3180 (Figure 3D) and MY10 to obtain further recombinants of chromosome II in the MY10 background. The resulting recombinant and homozygous F3 progeny were phenotyped over 20 generations using the primers in Data S2A.

The genotypes of NILs and chromosome II recombinants are provided in Data S2B.

Genome editing

We used the CRISPR/Cas9 target design and reagents described in [66]. We targeted the second exon of the set-24 gene, with the guide sequence 5’-gtaacgcggcaagaactctaCGG -3’ with the final 'CGG' representing the PAM motif for the Cas9. We replaced the dpy-10 target site with the set-24 target site in the pjA58 plasmid from [67] using the site-directed mutagenesis kit (New England Biolabs) and the primers 5’- aagaactctaGTTTTAGAGCTAGAAATAGC-3’ and 5’- gccgcgttacAAGACATCTCGCAATAGG-3’. DH5-alpha bacteria were transformed with the modified pJA58 plasmid, pU6-set24-sgRNA2. Bacteria were then grown overnight at 37°C in liquid LB medium and plasmids were extracted using a PureLink Midiprep kit (Invitrogen). The set-24 target site in pU6-set24-sgRNA2 was confirmed by Sanger sequencing.

We then injected the following mix into gonads of JU3219 and JU1395 hermaphrodites: 100 ng/μL of the pU6-set24-sgRNA2 plasmid, together with 50 ng/μL of the Cas9 coding plasmid (Peft-3::Cas9-SV40 NLS::tbb-2 3′UTR; [66]) and 40 ng/μL of the pjA58 plasmid as co-injection marker. We singled the F1 progeny from plates with the highest number of animals displaying the Dpy phenotype. F1 were screened for set-24 editing by screening for deletions in a 107 bp region around the PAM site. The primers 5’-GAAACTCCACTGCATTGT-3’ and 5’-TTTTCCTCGGCAATACG-3’ were used to generated PCR products which were loaded onto a 3%-agarose gel to identify samples with a smaller PCR product length. Broods from independent P0 animals were found positive and rendered homozygous. Frameshifting deletions introduced in the exon 2 of set-24 were confirmed by Sanger sequencing. The resulting lines, JU3253, JU3254, JU3292, JU3293, were given the allele names mf123, mf124, mf130 and mf131, respectively and were phenotyped over 25 generations. Alleles sequences are specified in Figure S4B.

Quantification and Statistical Analysis

Statistical analyses

Statistical analyses were performed using R v3.3.2 [68]. The details on the number of replicates (animals or lines) and the tests can be found in the Results section and the Figure legends.

Bulk segregant analysis

In our design, the limiting factor for QTL mapping was the low number of pooled RILs pooled and not the read coverage. Indeed, we recovered a minimum of 82 reads per 300kb window, thus had a large excess of reads compared to the number of RILs in the pool (21 and 22). To test whether differences in MY10 allele proportions between the two pools were significantly different from expectations under a random distribution, we first calculated for all previously defined windows the log-odds ratio as: log(m₁/(21-m₁))/(m₂/(22-m₂))), m₁ being the MY10 allele proportion multiplied by the number of RIls (22) in the Mrt pool and m₂ the MY10 allele proportion multiplied by the number of RILs (21) in the non-Mrt pool. We calculated the threshold of significance (p=0.01) for log-odds ratios in a 2-tailed manner. Using the binomial law, we simulated log-odds ratios for 1 million randomized draws of the two pools.

Small RNA analysis

In order to test, in each genotype, the effect of the time (i.e. generation) on the percentage of 22G in small RNA population, we used the following general linear model with a Poisson regression: %22G =replicates + #generations ∕ replicates + ε.

Supplementary Material

Data Supplementary 2

Detailed genotyping of Near Isogenic Lines, of other recombinants and of the molecular variants in the QTL region. Related to Figure 3 and Figure 4.

A: List of primers used for genotyping the NILs and further recombinants.

B: Genotypes of Near Isogenic Lines and recombinants. Pairs of primers listed in A were used to genotype each NIL and further recombinant lines. For each marker, 'M' stands for MY10 allele and 'J' stands for JU1395 allele.

C: Molecular variants in the QTL region [II: 11056128.. 12630688] and annotation of their impact. Only homozygous parent-specific SNPs and deletions were retained. The low impact modifications are not shown.

D: Wild isolates with a genome haplotype close to that of MY10, sampled in July 2002 by H. Schulenburg's Lab from compost in Roxel.

Data Supplementary 1

QTL mapping data. Related to Figure 3.

A: List of Recombinant Inbred Lines and their Mrt phenotype values.

B: Pool sequencing output – Number of reads in the Mrt and the non-Mrt pools for each parental allele at different polymorphic sites along the genome.