Substantial rDNA copy number reductions alter timing of development and produce variable tissue-specific phenotypes in C. elegans

Abstract

The genes that encode ribosomal RNAs are present in several hundred copies in most eukaryotes. These vast arrays of repetitive ribosomal DNA (rDNA) have been implicated not just in ribosome biogenesis, but also aging, cancer, genome stability, and global gene expression. rDNA copy number is highly variable among and within species; this variability is thought to associate with traits relevant to human health and disease. Here we investigate the phenotypic consequences of multicellular life at the lower bounds of rDNA copy number. We use the model Caenorhabditis elegans, which has previously been found to complete embryogenesis using only maternally provided ribosomes. We find that individuals with rDNA copy number reduced to ∼5% of wild type are capable of further development with variable penetrance. Such individuals are sterile and exhibit severe morphological defects, particularly in post-embryonically dividing tissues such as germline and vulva. Developmental completion and fertility are supported by an rDNA copy number ∼10% of wild type, with substantially delayed development. Worms with rDNA copy number reduced to ∼33% of wild type display a subtle developmental timing defect that was absent in worms with higher copy numbers. Our results support the hypothesis that rDNA requirements vary across tissues and indicate that the minimum rDNA copy number for fertile adulthood is substantially less than the lowest naturally observed total copy number. The phenotype of individuals with severely reduced rDNA copy number is highly variable in penetrance and presentation, highlighting the need for continued investigation into the biological consequences of rDNA copy number variation.

Introduction

One of the fundamental constraints on cell growth and proliferation is ribosome biogenesis (Warner 1999; Rudra and Warner 2004; Bremer and Dennis 2008; Chaillou et al. 2014). Resources for ribosome production make up not only the majority of a cell's transcriptional output (Warner 1999) and a large fraction of its protein content (Liebermeister et al. 2014; Reuveni et al. 2017), but also an appreciable amount of genomic real estate (Henderson et al. 1972; Gonzalez and Sylvester 1995; Warner 1999; Long et al. 2013; Morton et al. 2020; Nurk et al. 2021). The genes that encode the 4 eukaryotic ribosomal RNAs (rRNAs) are present in hundreds or even thousands of copies. In humans, the 45S transcript (which will be processed into the 18S, 5.8S, and 28S rRNAs) is present per haploid genome in a reported 114–598 copies, depending on the individual (Henderson et al. 1972; Worton et al. 1988; Stults et al. 2008; Hall et al. 2021). These arrays (rDNA) amount to ∼5–25 Mb of DNA, representing the majority of the small arms of the 5 acrocentric chromosomes, and remain one of the last regions of the human genome to be fully resolved (Miga et al. 2020; Nurk et al. 2021). Recent evidence implicates rDNA copy number loss in certain cancers (Salim et al. 2017; Wang and Lemos 2017; Xu et al. 2017; Hosgood et al. 2019; Feng et al. 2020; Valori et al. 2020) making directed study of the effects of copy number change increasingly relevant to human health.

In model organism eukaryotic genomes, the rRNA genes are similarly organized into large repetitive arrays of variable copy number. Saccharomyces cerevisiae has an rDNA array consisting of anywhere from 91 copies (or even lower, by some estimations) to 304 copies; Caenorhabditis elegans isolates have between 68 and 418 copies of the 45S rRNA genes (per haploid genome) (Files and Hirsh 1981; Albertson 1984; Ellis et al. 1986; Thompson et al. 2013; Morton et al. 2020; Sultanov and Hochwagen 2022). In humans, plants, flies, and yeast, 50–90% of these copies are heterochromatically silenced (Conconi et al. 1989; Haaf et al. 1991; Dammann et al. 1993; Jackson et al. 1993; Ye and Eickbush 2006; McStay and Grummt 2008; Gagnon-Kugler et al. 2009; Chandrasekhara et al. 2016; Rabanal et al. 2017). Despite the pervasiveness of large rDNA arrays among eukaryotes and the observed extreme copy number variation within species, it remains largely unresolved what benefit such expansive arrays might confer and what consequences might result when organisms deviate from their established species-specific copy number range.

In the unicellular model yeast, substantially reducing the rDNA copy number (from ∼150 to below 30 copies) reduces growth rate (French et al. 2003; Sanchez et al. 2019), presumably through rRNA transcriptional needs having exceeded the packing limit for RNA polymerases per gene (French et al. 2003). However, reductions in rDNA copy number that are modest enough to maintain a wild-type level of rRNA production nevertheless associate with other defects: issues with genome replication and increased mutagen sensitivity (Kobayashi 2008; Ide et al. 2010; Kwan et al. 2021).

The minimum rDNA requirements of multicellular organisms are far less well-characterized. In Drosophila melanogaster, nonlethal rDNA copy number reductions result in shortened scutellar bristles and delayed development (Ritossa et al. 1966); the reported copy number threshold for this bobbed phenotype varies anywhere from 90% (Paredes and Maggert 2009a) to under 50% of wild-type copy number (Ritossa et al. 1966; Tartof 1973; Long and Dawid 1980; Ye and Eickbush 2006). Drosophila studies are, however, confounded by the fact that 32–77% of fly rDNA repeat units contain a retrotransposon interruption (Jakubczak et al. 1992), the consequences and regulation of which are poorly understood. Drosophila rDNA has the additional complications of being located on the sex chromosomes, and subject to nucleolar dominance (Greil and Ahmad 2012; Warsinger-Pepe et al. 2020). C. elegans as a model system offers the advantages of a single 45S rDNA array on an autosome without retrotransposon interruption, making it ideal for the investigation of how different rDNA copy numbers interact with different tissue types and developmental stages.

Recently, researchers deleted the entire 45S rDNA array in C. elegans (Cenik et al. 2019) and discovered that worms homozygous for this deletion are capable of completing embryogenesis, sustained solely by maternally supplied ribosomes. These worms developmentally arrest in the first larval stage post-hatching (L1). Mosaic animals with both wild-type and ribosome-deficient cells also arrest at L1, suggesting that impairment of ribosome biogenesis in only a subset of cells is enough to arrest whole-organism development (Cenik et al. 2019). Beyond this complete elimination of rDNA, the minimal requirements of rDNA copy number on phenotype have not been further explored.

We used this rDNA deletion and other low-copy number alleles to explore the consequences of rDNA copy number reduction on development and fitness of C. elegans. We find that ∼11 rDNA copies (per diploid genome) can be sufficient for development to adulthood, with variable penetrance. Adults with only ∼11 rDNA copies were sterile and displayed abnormal phenotypes in post-embryonic dividing tissues such as the vulva and germline. We genetically engineered another low-copy rDNA allele of ∼22 copies and find that worms with a total of ∼22 rDNA copies achieve some measure of fertility after a substantial delay in development. A total of ∼70 copies of rDNA, half the lowest total amount observed in any wild isolate, was also sufficient for fertile adulthood but still subtly disrupted developmental timing. For higher copy number alleles, worms haploid for rDNA appeared indistinguishable from worms with rDNA on both chromosomes, suggesting that absolute copy number rather than number on each chromosome affects developmental timing. Our results demonstrate that the lower boundary of rDNA copy number sufficient for organismal development is not a hard threshold but rather a continuum. These results also support tissue-specific needs of rDNA abundance and point to specific tissue types and biological processes that are most vulnerable to reduced rDNA copy number.

Methods

Strains

All N2 worms used were strain VC2010. PD2621 and PD2620 were propagated by picking individual hermaphrodites to 4 single plates; one of these plates was selected for the next round of propagation based on the proportion of Unc worms (one quarter of the progeny are expected Unc (unc-54/unc-54)). All strains were maintained and all assays performed at 20 °C. Near isogenic lines: line SEA51 contains the transgene mIs13[myo-2p::GFP + pes-10p::GFP + F22B7.9p::GFP] on chromosome I, in proximity to the rDNA, backcrossed into N2 strain VC2010 6 times; this strain was observed to have ∼133 rDNA haploid copy number (Fig. 1b). We used this marker in a cross strategy to generate strain SEA305 by crossing the rDNA region from wild isolate MY16 (∼73 haploid rDNA copies) into SEA51 8 times, selecting non-GFP worms, followed by 6 generations of single-worm selfing. The final backcross was with a SEA51 hermaphrodite to restore wild-type mitochondrial DNA. Strain SEA302 was generated in the same way, with rDNA introgressed from wild isolate JU775 (∼81 haploid rDNA copies). Both strain constructions involved rounds of selfing between each round of backcrossing. Strain SEA300 was generated by introgressing the ∼417-copy rDNA array from wild isolate MY1 into the N2 background (SEA51) with creation first of a homozygous recombinant inbred line, then an additional 6 rounds of crossing to SEA51, a final 6 rounds of selfing, and a final cross from male into SEA51 hermaphrodite to restore N2 mitochondrial DNA. Outside of minor fluctuation that could also be attributed to CHEF measurement error (Morton et al. 2020), rDNA copy numbers remained stable through this described line construction, which in the case of SEA300 amounted to over 40 worm generations [MY1 rDNA copy number = 418 (Thompson et al. 2013), SEA300 rDNA copy number = 417 (Fig. 1d)]. Strain SEA296 was generated by introgressing N2 rDNA (from strain SEA51) into the wild isolate MY1 background, first with creation of a recombinant inbred line, then with 6 additional backcrosses to MY1 followed by 6 selfing generations, 6 more backcrosses to MY1 and 6 more selfing generations. During this propagation, a spontaneous reduction in rDNA copy number occurred, reducing the 45S rDNA from ∼130 copies to ∼64 copies, observed by pulsed-field gel electrophoresis (Fig. 1b). It should be noted that this reduction occurred in only 1 of 3 sibling strains generated in parallel using this cross strategy. unc-54: The allele of unc-54 used (strain PD2621) was documented as e1152 but Sanger sequencing for genotyping during the course of the experiments showed a stop codon at Q816 and a G704E mutation, rather than the expected e1152 mutations (G852R, K853M). As the stop codon results in a null allele [homozygous Unc phenotype and in accordance with other documented nonsense mutations in this region (Dibb et al. 1985; Bejsovec and Anderson 1990)], this allele served as well as e1152 for the purposes of our experiments. A summary of strain genotypes and estimated rDNA copy numbers is available in Supplementary Table 6.

Fig. 1.

Supplying ∼11 copies of rDNA rescues ΔrDNA L1 arrest. a) The C. elegans rDNA array (black) lies at the end of the right arm of chromosome I (not to scale). In N2, the 7.2 kb rDNA repeat unit is present in ∼100 copies (array size 0.72 Mb). Strain PD2621 is heterozygous for a complete deletion of the rDNA array (Cenik et al. 2019). The eDf24 deletion allele deletes all but ∼11 copies of the rDNA as well as ∼70 kb of adjacent sequence (Supplementary Fig. 1) and is embryonic lethal when homozygous (Ahnn and Fire 1994). b–d) Strain rDNA copy numbers were determined by embedding worm genomic DNA in agar plugs and digesting to separate the 45S rDNA array from the rest of chromosome I; bands were resolved using pulsed-field gel electrophoresis and detected with an rDNA-specific labeled probe (see Methods). Ladders were visualized with ethidium bromide staining and band sizes are indicated to the left. The calculated rDNA copy number of a band is indicated in above the band, determined by band size divided by rDNA unit size (7.2 kb). b) rDNA bands were separated with run conditions such that rDNA arrays between ∼50 and 200 copies should be resolved. The band sizes indicated for CB2769 and CB2778 are close to or beyond the range of confident calls for this gel and should be taken as approximate sizes only. CB2769 is run with more appropriate separation conditions in panel D. c) rDNA bands were separated with run conditions such that rDNA arrays between ∼5 and 50 copies should be resolved. Carrot indicates large bands running at the limit of separation (see panel b for size separation of these bands). d) rDNA bands were separated with run conditions such that rDNA arrays between ∼150 and 430 copies should be resolved (see Methods). e) Self-fertilized PD2621 worms (unc-54/ΔrDNA) produce wild-type, Unc (“Uncoordinated,” nonmoving, unc-54/unc-54), and larval stage 1 (L1) arrest phenotypes. Results concur with those reported in Cenik et al. (2019). Diploid rDNA copy numbers (sum of both alleles) are provided based on estimates from b, c and d. f) A cross strategy was used to generate ΔrDNA/eDf24 worms (diploid rDNA number = 11). eDf3 is a deletion allele that encompasses unc-54 but does not overlap eDf24 (Ahnn and Fire 1994). Diploid rDNA copy numbers are provided below genotypes. g, h) Phase contrast images under a 10× objective are presented of both PD2621 adults (diploid rDNA number = 129) and ΔrDNA homozygous or heterozygous larvae, all at D1 (3 days after egg laid). Scale bar = 100 µm. g) Arrow indicates arrested L1 of genotype ΔrDNA/ΔrDNA (diploid rDNA number = 0). h) Arrow indicates worm of genotype ΔrDNA/eDf24 (diploid rDNA number = 11), estimated to be at larval stage 2 (L2). Oblong objects are embryos from the PD2621 adult. i, j) 40X DIC images of D0 ΔrDNA/ΔrDNA (i) or ΔrDNA/eDf24 (j) worms. Scale bar = 20 µm. D0 Assignment of “L2” was on the basis of size as assessed under a stereoscope, and, for those imaged under high magnification DIC, expanded gonad. Assignment of genotype ΔrDNA/eDf24 was based on progeny proportions (Supplementary Table 1) and PCR (Supplementary Table 2).

CRISPR generation of 22-copy rDNA allele

The rDNA array was targeted for CRISPR cutting in order to generate alleles with reduced rDNA copy number. A co-CRISPR strategy was used to select for successfully CRISPR edited worms (Ward 2015). This strategy uses CRISPR targeting of 2 loci in the same animal to enable screening. One locus is a temperature-sensitive lethal mutation (pha-1(e2123)); successful editing at this site will restore wild type and allow growth at the restrictive temperature (25 °C). The other locus is the target of interest (rDNA). Worms with successful restoration of the wild-type allele of pha-1 will be enriched for editing of the co-targeted rDNA locus.

Strain GE24 (pha-1(e2123)) was outcrossed to laboratory strain VC2021 4 times (in addition to the 3 reported outcrosses) to generate strain SEA348. These worms were injected with an injection mixture of Cas9 nuclease, tracrRNA, crRNAs, and single-stranded DNA repair template. To prepare this injection mixture, the following were combined: Cas9 Nuclease V3 (IDT #1081058) at 250 ng/μL, tracrRNA (IDT #1072532) at 100 ng/μL, crRNA targeting pha-1 (Ce.Cas9.PHA-1.1AE) (Ward 2015) at 2.5 pmol/μL, and crRNA targeting rDNA (rEM4) (Supplementary Table 6) at 5 pmol/μL. This was brought up to a volume of 10 μL with TE pH 7.5 (IDT #11-01-02-02). The rDNA crRNA rEM4 was designed using the tool http://genome.sfu.ca/crispr/ and targets a PAM in the internal transcribed spacer of the rDNA between the 5.8S and 26S genes. The above components were combined, pipetted up and down 20 times, and incubated 37 °C for 15 min. Repair templates and an injection marker were then added to the mixture: pha-1 repair template (EM524) at 110 ng/μL, rDNA repair template (EM655) at 110 ng/μL, and pRF4 (rol-6) injection marker plasmid at 40 ng/μL. The mixture was brought up to 20 μL total volume with nuclease-free water, spun at 14,000 rpm for 5 min, and the top 12 μL were moved to a new tube for use in injection. Repair templates were ordered as PAGE-purified single stranded DNA (IDT) and introduced mutations in the PAM sites (Supplementary Table 6).

Forty SEA348 pha-1(e2123) worms were injected with the above mixture, allowed to recover at 16 °C for 2 h and then put at 25 °C, the restrictive temperature for pha-1(e2123) growth. Viable F2 worms (nonroller) were picked to individual plates and put at 20 °C to grow up into populations that could be prepared for pulsed-field gel electrophoresis and Southern blotting (section below) in order to quantify copy number of rDNA array alleles present. Individual worms were picked to new plates from lines with indicated rDNA copy number changes. These worms were allowed to lay eggs and then were picked into lysis buffer for single worm PCR and Sanger sequencing of the pha-1 and rDNA loci, to confirm homozygosity of the pha-1 repair and confirm presence of the mutated PAM in the rDNA (Supplementary Table 6 for primers). The line SEA354 was determined to have a single band at 22 copies of the rDNA (Supplementary Fig. 4). By Sanger sequencing chromatogram, all of the rDNA sequence in this line has the mutated PAM site (Supplementary Fig. 4). This 22-copy rDNA allele was determined not to be an extrachromosomal array of rDNA: in a pulsed-field gel run without sample SwaI digestion, the 22-copy rDNA array runs at the limit of detection rather than at 22-copy size (data not shown). This is consistent with the array being chromosomal and requiring SwaI digestion to separate the rDNA from the rest of the chromosome; an extrachromosomal array of rDNA would run at the same size regardless of SwaI digestion.

SEA354 was then outcrossed twice to line SEA51 (N2-background strain with the mIs13 GFP transgene integrated near the rDNA locus) and single-worm propagated 4 times to generate strains SEA379, SEA380, SEA381, SEA382, and SEA383. These strains were characterized again for rDNA copy number via CHEF gel and Southern blot (described below). Four of the 5 outcrossed strains retained the preoutcross rDNA copy number (within the measurement error of pulsed-field gel analysis, accounting for the 22 vs 21 copy measurement discrepancy between Supplementary Fig. 4, b and c), while one (SEA379) exhibited an increase in copy number to ∼50 copies. Strains SEA380, SEA382, and SEA383 were used in subsequent experiments (Supplementary Table 3).

Crosses

All crosses were performed on 3 cm NGM plates seeded with OP50, maintained at 20 °C. Crosses were set up with proportions of 3 L4 hermaphrodites to 10 males. Hermaphrodites and males were allowed 2 days for mating to occur, after which hermaphrodites were transferred to new individual 3 cm plates and allowed to pulse lay embryos for 6 h (unless otherwise stated). Approximately 24 h after the end of the pulse lay, plates were scored for the presence of unhatched embryos in indicated crosses. Approximately 48 h after the end of the pulse lay, plates were scored for progeny proportions of L1, L2, Unc, male, and moving L4/adult hermaphrodite phenotypes (Supplementary Tables 1 and 3). Absence of males or infrequency of males was taken as a sign of failed mating and progeny were considered the product of self-fertilization. Each cross has an assigned experiment ID, provided in Supplementary Tables 1 and 3 and used for reference in subsequent tables.

The parental source of the ΔrDNA allele (PD2621 in all experiments except experiment ID 020320, which used PD2620) is maintained as a heterozygote over a null allele of unc-54, which has a recessive Unc (“Uncoordinated,” nonmoving) phenotype (Cenik et al. 2019). To generate worms with a diploid rDNA number of ∼11 (genotype ΔrDNA/eDf24) we crossed strain PD2621 (unc-54/ΔrDNA) to strain CB2769 (eDf3/eDf24). eDf24 does not overlap unc-54, but is maintained over a deficiency allele that does: eDf3 (Ahnn and Fire 1994). We mated unc-54/ΔrDNA hermaphrodites to eDf3/eDf24 males. Mated hermaphrodites were allowed to pulse lay embryos for 6 h and progeny were scored for stage 2 days later. We observed the predicted proportions of moving worms (L4/adult hermaphrodites and males) expected from genotypes unc-54/eDf24 and ΔrDNA/eDf3, as well as expected proportions of nonmoving Unc worms (unc-54/eDf3) (Table 1; Supplementary Table 1; Fig. 1f). These worms were genotyped by PCR to confirm presence of both rDNA and ΔrDNA (Supplementary Table 2), supporting the progeny proportion-based determination that their genotype is ΔrDNA/eDf24 (diploid rDNA number = 11). Worms with ∼11 rDNA copies have about 5% of the total genomic rDNA of a wild-type worm (Fig. 1b).

Table 1.

A cross between PD2621 (unc-54/ΔrDNA) hermaphrodites and eDf3/eDf24 males reveals a category of offspring that appear stage L2 at 48 h.

	N	Progeny proportions at 48 h					Unhatched embryos (%)
		L4/Adult hermaphrodite (%)	Male (%)	Unc (%)	L1 (%)	L2 (%)
PD2621×eDf3/eDf24 cross	41	36.6	29.3	17.1	7.3	9.8	0
	45	24.4	26.7	22.2	6.7	17.8	2.2
	20	20.0	25.0	40.0	5.0	10.0	0
	50	34.0	8.0	30.0	12.0	16.0	0
	39	28.2	51.3	12.8	0.0	5.1	2.6
Average		28.6	28.0	24.4	6.2	11.7	1.0
PD2621 unmated	50	44.0	0	38.0	16.0	0	2.0
	46	50.0	0	17.4	32.6	0	0
	43	41.9	2.3	32.6	23.3	0	0
	41	56.1	0	17.1	26.8	0	0
Average		48.0	0.6	26.3	24.7	0	0.5

Raw numbers are provided in Supplementary Table 1, experiment ID 111819. Unhatched embryos were scored at 24 h-post-end of pulse lay, other categories were scored at 48-h-post-end of pulse lay. Each row represents the progeny of a single hermaphrodite. N = total number of progeny. Averages presented are the average of the above percentages.

To generate worms with 22 total rDNA copies per diploid cell, we mated males of one of 3 outcrossed strains of 22-rDNA homozygotes (SEA380, SEA382, or SEA383) to PD2621 (unc-54/ΔrDNA) hermaphrodites and scored progeny proportions 2 days after eggs were laid (Table 2, Supplementary Table 3; Fig. 3). Worms with 44 total rDNA copies used in Fig. 3 were the progeny of mating a homozygous 22-rDNA male to a homozygous 22-rDNA hermaphrodite, and were stage-scored and imaged blind (blinded collectively with matings between 22-rDNA homozygous males and PD2621 hermaphrodites) (see Blinding in Supplementary Methods). Mated worms were allowed to pulse lay embryos for 4 h. F1 worms were scored for stage at a timepoint between 49.5 and 55 h after the end of the pulse lay (Supplementary Table 3) (experiment IDs 052322, 053022, and 060222 were all scored blind to parental genotype). To determine early life fertility (Supplementary Fig. 5, a–c), F1 worms from 3 of these cross trials were allowed to lay eggs for a 6 h period on D1. Following this period, the ΔrDNA/22-rDNA worms were picked to new individual plates and allowed to lay eggs for another 2 days. All progeny (hatched larvae and unhatched embryos) were counted from each pulse lay. Ten worms were used per genotype per trial.

Table 2.

A cross between PD2621 (unc-54/ΔrDNA) hermaphrodites and SEA305 (73-rDNA/73-rDNA) males reveals rescue of larval arrest.

	N	Progeny proportions at 48 h					Unhatched embryos (%)
		L4/Adult hermaphrodite (%)	Male (%)	Unc (%)	L1 (%)	L2 (%)
PD2621×SEA305	33	45.5	54.5	0	0	0	0
	44	38.6	59.1	0	0	0	2.3
	36	44.4	50.0	0	0	0	5.6
	37	43.2	54.1	0	0	0	2.7
	45	55.6	44.4	0	0	0	0
	39	56.4	41.0	0	0	0	2.6
	43	58.1	39.5	0	0	0	2.3
	35	45.7	51.4	2.9	0	0	0
	34	44.1	38.2	5.9	11.8	0	0
Average		48.0	48.0	1.0	1.3	0	1.7
PD2621 unmated	38	28.9	0	34.2	36.8	0	0
	48	50.0	0	27.1	22.9	0	0
	36	47.2	0	25.0	27.8	0	0
Average		42.1	0	28.8	29.2	0	0

Raw numbers are provided in Supplementary Table 3, experiment ID 061220. Unhatched embryos were scored at 24 h-post-end of pulse lay, other categories were scored at 48 h-post-end of pulse lay. Each row represents the progeny of a single hermaphrodite. N = total number of progeny. Averages presented are the average of the above percentages.

Fig. 3.

Worms with ∼22 rDNA copies exhibit delayed development. a) The available alleles are diagrammed (not to scale), including the new ∼22-copy rDNA allele generated through CRISPR-Cas9 (Supplementary Fig. 4). b) A cross strategy between PD2621 worms (unc-54/ΔrDNA) and homozygous 22-rDNA worms was used to generate worms with the genotype ΔrDNA/22-rDNA. c) Control crosses were performed between 22-rDNA homozygous males and hermaphrodites (diploid rDNA number = 44). d, e) Phase contrast images under a 10× objective are presented for genotypes unc-54/22-rDNA (diploid rDNA number = 151) and ΔrDNA/22-rDNA (diploid rDNA number = 22) (d) as well as homozygous 22-rDNA/22-rDNA (diploid rDNA number = 44) worms (e), all at D0 (2 days after eggs laid). Scale bar = 100 µm. f–h) 40X DIC images of unc-54/22-rDNA (f), ΔrDNA/22-rDNA (h), or 22-rDNA/22-rDNA (g) worms imaged on D1. Scale bar = 20 µm. i) Male worms homozygous for the 22-rDNA array were mated to hermaphrodite unc-54/ΔrDNA worms or to 22-rDNA homozygous worms to generate worms with diploid rDNA numbers of 22, 44, or 151. These F1 worms were scored for developmental stage (L3 being the least developed and adult being the most developed) 2 days after eggs were laid. Across all trials, n = 131, n = 247, and n = 155 for diploid rDNA numbers of 22, 44, and 151, respectively. Developmental stage was scored blind to genotype of the worms (Supplementary Table 3).

To generate worms with 73 total rDNA copies per diploid cell, we mated male SEA305 (73-rDNA homozygotes) to PD2621 (unc-54/ΔrDNA) hermaphrodites and scored progeny proportions 2 days after eggs were laid (Table 2; Supplementary Table 3; Fig. 4). Stage scoring was performed as described below.

Fig. 4.

A low-copy rDNA array (73-rDNA) rescues ΔrDNA L1 arrest. a) The available alleles are diagramed (not to scale). b) Self-fertilized PD2621 worms (unc-54/ΔrDNA) produce wild-type, Unc, and larval stage 1 (L1) arrest phenotypes. Diploid rDNA copy numbers (sum of the 2 alleles) are provided based on Fig. 1, b–d. c) A cross strategy between PD2621 and SEA305 (73-rDNA/73-rDNA) was used to generate worms with the genotype ΔrDNA/73-rDNA. d–g) Representative DIC images of D1 adult hermaphrodites for the following genotypes: d) wild-type N2 [diploid rDNA number = 219 (Fig. 1b)]. e) SEA305 (73-rDNA/73-rDNA) (diploid rDNA number = 146). f) unc-54/73-rDNA (diploid rDNA number = 202) resulting from a cross of PD2621×SEA305 (see Supplementary Table 2). g) ΔrDNA/73-rDNA (diploid rDNA number = 73) resulting from a cross of PD2621×SEA305 (see Supplementary Table 2 for genotyping). Scale bar = 20µm.

Stage scoring

For assays addressing developmental time of natural rDNA allele heterozygous worms (ΔrDNA/73-rDNA, ΔrDNA/81-rDNA, ΔrDNA/64-rDNA, and ΔrDNA/417-rDNA) worms, crosses were performed as above. Mating success was determined by a high occurrence of F1 males (Supplementary Table 3); absence of males or presence of more than 2 Unc worms was taken as a sign of failed mating. On the second day following the pulse lay, moving F1 hermaphrodites were picked to individual wells of 24-well NGM plates seeded with OP50 and scored for developmental stage at various time points, ranging from 43 to 56 h after the end of the pulse lay (see Supplementary Table 5). Developmental stage was assigned primarily based on vulva morphology and gravidity. Staging data was collected across several time points, only a subset of which were used in the determination of which worms were developmentally the oldest and youngest in the cohort (Supplementary Table 5); the time points and criteria to use were decided before unblinding.

Three days after stage scoring, the F2 progeny of these individuals were scored for presence or absence of nonmoving (Unc) animals, from which the genotype of the F1 (unc-54/73-rDNA vs. ΔrDNA/73-rDNA) was inferred. Before unblinding, some wells were eliminated due to an insufficient number of progeny to confidently call “absence of Unc progeny.” [In some cases of insufficient progeny (4–18 F2 worms) the F2 worms were allowed to reproduce and the F3 generation was screened for Unc.] Presence of 2 or more males in the F2 population was also grounds to eliminate that well (implying that mating happened between F1 siblings before the F1 was picked to the individual well). Because the F1 stage scoring was done 3 days before F2-based genotype scoring, stage scoring was by default blind to genotype information (additional blinding measures were taken for comparisons between genotypes. See Blinding in Supplementary File 1).

Imaging and PCR genotyping

Worms were immobilized with 1 mM levamisole and placed on 2% agarose pads. Worms were imaged on a Zeiss Axioplan 2 (software AxioVision V 4.8.2.0) under 40X DIC unless otherwise stated. After imaging, ΔrDNA/eDf24 worms were recovered from imaging pads into lysis buffer [5% Proteinase K (Sigma P4850) in 1X TaqPCR Rxn Buffer (Invitrogen Y02028)], lysed (freeze-thawed at −80 °C 3 times, followed by incubation at 60 °C for 1 h and 95 °C for 15 min), and used as template in PCR (4 µL in 25 µL reactions in most cases). PCR primers are provided in Supplementary Table 6, and PCR results are presented in Supplementary Table 2. All worms recovered from imaging pads were used for lysis and not recovered for further growth; time course images are separate individuals. Genotype confirmation of imaged ΔrDNA/73-rDNA hermaphrodites was done both by progeny assessment (lack of unc-54 homozygous progeny) and with PCR detection of ΔrDNA and restriction fragment length polymorphism assessment of 73-rDNA (Supplementary Table 2). The latter was made possible by the presence of nucleotide variation in wild isolate MY16 DNA linked to the 73-rDNA allele. RFLP genotyping was performed on PCR products generated using 1.5 µL worm lysate in 20 µL reactions using Quickload Taq Master Mix (NEB M0271S) (primers in Supplementary Table 6); after amplification, 0.5 µL of MnlI was added to the reaction and incubated 1 h at 37°C. This enzyme cuts the PCR product from the N2 genotype into 2 bands (∼380 and 500 bp) but does not cut a MY16-genotype product.

The 22-copy allele of rDNA has a mutation in the PAM site as a result of the CRISPR-based allele generation approach (Supplementary Table 6; Supplementary Fig. 4). Heterozygous ΔrDNA/22-rDNA worms genotyped by PCR were lysed as above and the region of the mutation in the rDNA was amplified and Sanger sequenced (Supplementary Table 2). Worms of genotype ΔrDNA/22-rDNA show a chromatogram indicating that all copies of rDNA present have the mutated PAM site (Supplementary Fig. 4e) (Davis and Jorgensen 2022 for visualization). Worms of genotype unc-54/22-rDNA show heterozygosity at the PAM site, with the majority of rDNA copies wild type. These worms were also PCR genotyped for the presence or absence of a ΔrDNA-specific band.

Pulsed-field gel electrophoresis

Worms were grown to starvation on NGM + OP50 plates, washed into 15 mL conical tubes in 1X M9 [42 mM Na₂HPO₄, 22 mM KH₂PO₄, 22 mM NaCl, 1 mM MgSO₄]. Worms were washed once in 10 mL 1X M9 and once in 10 mL autoclaved, double glass distilled water. Worm pellet and residual water volume were estimated, with worms making up ∼half of the total volume. An equal volume of melted 42 °C 1% SeaPlaque GTG agarose was added to the worm pellet, mixed by pipetting up and down, and immediately transferred to agarose plug molds (∼80 µL per mold) kept on ice. Agarose plugs were solidified at 4 °C for at least 10 min, then extracted into 2 mL tubes and incubated in 300 µL TEL [9 mM Tris, 90 mM EDTA pH 8, 10 mM levamisole] on ice for 30 min. Supernatant was removed, then 300 µL lysis buffer [1% SDS, 1 mg/mL Proteinase K (Sigma-Aldrich P4850), 8 mM Tris, 80 mM EDTA pH 8, 1 mM levamisole] was added to the plugs. Plugs were incubated at 50 °C for ∼24 h to lyse the worms. After digestion, the plugs were transferred into 24-well plates, supernatant removed, and plugs washed in 300 µL TE [10 mM Tris, 1 mM EDTA]. Plugs were washed at least 8 times for at least 15 min each in 300 µL TE, with at least one wash taking place at 4 °C overnight. Plugs were then stored in TE at 4 °C until used.

Plugs were prepared for CHEF gel electrophoresis as follows: Plugs were equilibrated in 1X NEB 3.1 buffer either overnight at 4 °C, or on the day of digestion by soaking plugs for 1 h in 1X NEB 3.1 buffer then replacing with fresh 1X NEB 3.1 buffer in a 24-well plate on ice. Approximately one-fourth of the plug was cut with a razor blade and transferred to a parafilm-wrapped slide. The rDNA was cleaved from the rest of chromosome I using ∼3 µL SwaI per plug, which cuts 3,927 bp upstream of rrn-3.56 and not at all within the rDNA. Plugs were placed in a small humid chamber in a 25 °C incubator and incubated for 4 h before loading into the CHEF gel. CHEF gels were prepared by placing digested plugs and ladders onto the teeth of gel combs. 0.8% agarose prepared in 0.5X TBE at 55 °C was poured around the plugs and allowed to solidify. Gel was transferred to a CHEF gel box (Bio-Rad CHEF DRII). Run conditions were as follows: For long rDNA (∼200 copies or above) (Fig. 1d): 100 V for 68 h, 14 °C, switch times = 300–900 s. For medium-length rDNA (∼70–200 copies) (Fig. 1b): 165 V for 66 h, 14 °C, switch times = 47–170 s. Short rDNA (<70 copies) (Fig. 1c; Supplementary Figs. 4 and 7): 200 V for 24 h, 14 °C, switch times = 15–25 s. After run completion, gel was stained by soaking in ethidium bromide (0.3 µg/mL in 0.5X TBE) to visualize the ladders and imaged on a Bio-Rad GelDoc XR+.

The following ladders were used in the pulsed-field gel electrophoresis: Fig. 1b ladder: Yeast Chromosome PFG Marker (S. cerevisiae) (NEB #N0345; discontinued); Fig. 1d ladder: Hansenula wingei chromosomes (maximum size 3.13 Mb, Bio-Rad 170-3667); Fig. 1c ladders: 48.5–1,000 kb lambda ladder (Bio-Rad 1703635) (Fig. 1c, right) and 5 kb ladder (Bio-Rad 170-3624) (Fig. 1c, left), which was mixed with 6X DNA loading buffer and loaded as a liquid into well of the short rDNA CHEF gel (all other ladder were loaded as agarose-embedded plugs). Supplementary Fig. 4b and c used 48.5–1,000 kb lambda ladder. Supplementary Fig. 7 used both lambda and 5 kb ladders.

Southern blotting was performed as outlined (Tsuchiyama et al. 2013) and previously described (Morton et al. 2020). Each CHEF gel was washed twice for 10 min each in 0.25N HCl, followed by 2 washes for 15 min each in 0.5N NaOH, 1 M NaCl, then 2 washes for 15 min each in 0.5 M Tris, 3 M NaCl. DNA was then transferred from the gel to a nylon membrane (Perkin Elmer GeneScreen Hybridization Transfer Membrane) and crosslinked using a Stratagene Stratalinker UV Crosslinker in preparation for radioactive probe hybridization. Membranes were Southern blotted for C. elegans rDNA using a probe created from an 850 bp PCR product that overlaps the first 300 bp of rrn-1 (18S) (Supplementary Table 6). The PCR product was purified with the Zymo Clean & Concentrator kit (D4013) before radioactive labeling.

Band measurement: Ladder bands were measured in cm distance from the bottom of the loaded agarose plug (or from the bottom of the well in the case of the liquid loaded 5 kb ladder). Southern blot bands were measured in centimeter distance from the bottom of the plug; rDNA copy number was estimated by graphing Southern blot rDNA band distance against measured ladder band distances to get band Mb size and dividing by 7.2 kb rDNA unit size. In Fig. 1b gel, the uppermost band represents chromosome sizes 1,900 + 1,640 and was not used; only ladder bands 1,100 kb and below were used in the determination of sample band sizes.

Whole genome sequencing

Nextera libraries were produced as previously described with the Illumina Nextera DNA Sample Preparation kit (FC-121-1030) (Morton et al. 2020). Briefly, 10 ng of genomic DNA in 9 µL total volume was tagmented with 1 µL Tn5 for 8 min at 55 °C. The reaction was stopped by the addition of 10 μL 5 M guanidine thiocyanate and incubated at room temperature for 3 min. Tagmented DNA was purified with AMPure XP beads (Beckman Coulter A63881): 15 μL room temperature AMPure beads and 25 μL binding buffer [20% PEG8000, 2.5 M NaCl] were added to the reaction and mixed by pipetting. Bead–DNA mixture was incubated at room temperature for 10 min then placed on a magnet for 5 min. Supernatant was removed, and beads were washed twice with 150 µL 70% ethanol. After the second wash, beads were allowed to dry slightly, then DNA was eluted in 20 µL Qiagen EB elution buffer. Libraries were amplified with Kapa HiFi 2x master mix as follows: 10 µL tagmented DNA was mixed with 7.5 µL Kapa master mix and 2.5 µL each of Illumina barcode index primers. Amplification conditions were: 72 °C 3 min, 98 °C 30 s, 98 °C 10 s, 63°C 30 s, 72°C 40 s, with the latter 3 steps repeated 6 times. Amplified libraries were purified with Ampure XP beads: 30 µL beads were added to the PCR reaction, incubated at room temperature for 5 min, and incubated on the magnet for 5 min. Supernatant was removed and beads were washed twice in 200 µL of 80% ethanol. After the second wash, beads were dried slightly and resuspended in 32.5 µL resuspension buffer from the Illumina kit. Libraries were sequenced on a Nextseq 550.

Sequencing data analysis

Genotyping NILs

Reads from the Nextseq 550 were demultiplexed with bcl2fastq and aligned to the WS230 genome with bowtie2. The wild isolate parental strains for the NILs were previously sequenced by our lab and are uploaded under PRJNA565452 at SRA (Morton et al. 2020). For variant calling, FASTQ files were aligned to the WS230 genome with bowtie2 (version 2.2.3) and converted to bam files with samtools (version 1.4) (Li et al. 2009; Langmead and Salzberg 2012; Langmead et al. 2019). Bam files were sorted with samtools and read group information added with Picard AddOrReplaceReadGroups function (Picard Tools—By Broad Institute). All bams for the same strain were then merged with the samtools merge function. Variants were called with GATK (version 3.7) Haplotype Caller with the following specifications: emitRefConfidence GVCF –variant_index_type LINEAR –variant_index_parameter 128000 –forceActive –dontTrimActiveRegions (Poplin et al. 2018). Output GVCF files were genotyped with GATK GenotypeGVCFs function. The output VCF files were filtered to omit poorly mapping positions (Thompson et al. 2013). All variants for each parental strain were compiled into a list file in the format chromosome:position.

NILs were genotyped as the parental strains were through variant calling with GATK. NIL variants were called with GATK Haplotype Caller in the same manner as the parents, except positions were restricted to those present in the list file for the wild isolate rDNA donor using the -L function. Breakpoints were determined by manual inspection of VCF files.

Analysis of CB3740 [eDf24 I; eDp20 (I,II); mnT12 (IV,X)]

Reads from the Nextseq 550 were demultiplexed with bcl2fastq version 2.20 and aligned to the WS235 genome with bowtie2 version 2.2.3. Reads were sorted with samtools version 1.9, and read depth was calculated with samtools depth function, using the flag –d 0 to remove a maximum depth threshold. Select regions of chromosome I were analyzed further in R, normalizing the read depth of CB3740 to N2 and plotting per-base read coverage (Supplementary Fig. 1; Supplementary File 2).

Statistics

Statistical tests were performed in RStudio (version 1.2.5001): Pearson's Chi-square test with Yates continuity correction (chisq.test, for data with one degree of freedom) and Fisher's exact tests (fisher.test).

Results

Supplying ∼11 copies of rDNA rescues the L1 arrest of worms lacking rDNA

The laboratory strain of C. elegans, N2, has approximately 200 copies of the 45S rDNA, 100 per haploid genome, tandemly repeated on the far-right arm of chromosome I (Files and Hirsh 1981; Albertson 1984) (Fig. 1, a and b). The minimal number of rDNA copies required to complete development is unknown, but is likely well below the wild-type diploid ∼200 copy amount. Wild C. elegans isolates have been described with ∼140 diploid copies (70/70) (Thompson et al. 2013) and at least one deficiency allele with severe rDNA reduction has been generated (Albertson 1984; Ahnn and Fire 1994). This deficiency, eDf24, deletes most of the rDNA array (Albertson 1984) along with ∼70 kb of adjacent DNA (Supplementary Fig. 1), and is embryonic lethal when homozygous (Ahnn and Fire 1994). We used pulsed-field gel electrophoresis to determine the copy number of the remaining rDNA in eDf24 and find it to be ∼11 copies (Fig. 1c). eDf24 heterozygotes [≥140 total 45S rDNA copies (Fig. 1, b and c)] are viable and fertile (Fig. 2a), indicating hemizygosity of the ∼70 kb deleted region does not prohibit growth. C. elegans individuals heterozygous for a complete deletion of the 45S rDNA (≥129 total 45S rDNA copies) are also viable and fertile, but worms homozygous for a complete deletion of the 45S rDNA (ΔrDNA) developmentally arrest in the first larval stage (Cenik et al. 2019) (Fig. 1, a, e, g, and i).

Fig. 2.

Worms with ∼11 rDNA copies exhibit slow development into sterile adults. a) DIC control images of D1 adult hermaphrodites of N2, PD2621 (unc-54/ΔrDNA), or eDf3/eDf24. D1 = first day of adulthood for a wild-type worm (i.e. 3 days after egg laid). Carrot indicates vulva. Schematic depictions of worm images are shown to the right for clarification of worm anatomy. b) DIC images of arrested L1s (ΔrDNA/ΔrDNA) resulting from PD2621 self-fertilization. L1s picked on D0 (2 days after egg laid) were imaged on D1, D3, and D6. c) DIC images of ΔrDNA/eDf24 (diploid rDNA number = 11). Approximately half of worms picked as L2 on D0 developed further [Supplementary Table 1 (experiment ID 092820, 101220, 112721, 120421, 120821, 032422)]. DIC images of worms that were picked as L2 on D0 are shown for D1, D2, D3, D4, D5, and D6. Carrot indicates a possible prevulva structure and arrow indicates putative oocyte. After imaging each day, worms were recovered from the slide into PCR buffer for genotyping PCRs (Supplementary Table 2). Scale bar = 20 µm.

To determine the effect of severe rDNA copy number reduction on organismal health and development, we generated animals with a total of ∼11 rDNA copies (∼5% of the wild-type copy number) by mating a balanced strain with a complete rDNA deletion (PD2621) to a balanced strain with eDf24 (Fig. 1, see Methods). Mated worms were given 6 h to lay eggs; the eggs were then given 2 d to hatch and develop, at which time they were assessed for how far they had progressed through the 4 larval stages of worm development. In the case of worms with ∼11 rDNA copies, larvae at approximately the second larval stage of development (L2) were observed, instead of the L1 larval arrest observed in animals lacking the 45S rDNA (ΔrDNA/ΔrDNA) or the embryonic arrest observed with eDf24/eDf24 (Table 1; Supplementary Table 1; Fig. 1, f, h, and j). L2 worms were genotyped by PCR (Supplementary Table 2) to confirm that this phenotype results from a ΔrDNA/eDf24 genotype (diploid rDNA number = 11). Twelve experiments, with 3–9 mated individuals each, were performed with similar results (Supplementary Table 1). This result was further validated by a cross between an independently generated ΔrDNA allele (strain PD2620) and eDf24 balanced over a different deletion, eDf12 (Supplementary Table 1). The presence of an allele containing ∼11 copies of 45S rDNA was thus sufficient to overcome the L1 arrest phenotype seen in the complete absence of rDNA, and the embryonic lethality observed in eDf24 homozygotes is not a consequence of reduced rDNA copy number.

Reducing rDNA copy number to ∼5% of wild type permits variable development into sterile adults with abnormal morphology

Approximately 48–54 h post-egg lay, wild-type C. elegans are in the last larval stage (L4) of development or in early adulthood. Although ΔrDNA/eDf24 worms were still in larval stage L2 at this time point (termed D0), it was unclear if this represented an arrest or simply a developmental delay. In the case of ΔrDNA homozygotes, the L1 arrest is a terminal phenotype; maternally provided ribosomes support development through embryogenesis, but worms unable to produce their own ribosomal RNA are unable to advance beyond L1 (Cenik et al. 2019).

The ΔrDNA/eDf24 genotype allowed us to ask if ∼11 copies of endogenous 45S rDNA can support development into adulthood. We picked putative ΔrDNA/eDf24 worms (L2 on D0) and imaged worms 1–6 days later (D1–D6). After imaging each day, worms were lysed for PCR genotype confirmation (Supplementary Table 2). Three days post-egg lay, wild-type C. elegans are in their first full day of gravid adulthood (D1) (Fig. 2a). ΔrDNA homozygotes (diploid rDNA number = 0) arrest as L1 and remain so through D6 (Fig. 2b). Of 225 ΔrDNA/eDf24 worms (diploid rDNA number = 11) followed to D6, 17.3% died and 31.1% remained arrested around the L2 or L3 stages. The remaining 51.6% of F1s identified as L2 on D0 developed further over the course of 6 subsequent days (Fig. 2c; Supplementary Fig. 2; Supplementary Table 1). The hermaphrodites among these worms showed germline development and possible oocytes (Fig. 2c; Supplementary Fig. 2, a and b) by D5 and D6, consistent with benchmarks of adulthood, albeit on a substantially delayed timeline compared to wild type. Male ΔrDNA/eDf24 worms were also identified beginning on D3, with visible germline development and tail structure defects ranging widely from grossly to mildly deformed (Supplementary Fig. 3; Supplementary Table 2). Males with similarly described defects have impaired mating ability (Hodgkin 1983), thus gamete viability in these males was not determined.

There is some evidence that the requirements for rDNA content can differ among tissues (Ritossa et al. 1966; DeSalle et al. 1986; Spring et al. 1996). We asked if the effects of severe rDNA copy number reduction show tissue-specificity, facilitated by the thoroughly characterized development and tissue types of C. elegans. All observed ΔrDNA/eDf24 hermaphrodites were sterile, establishing the sensitivity of germline development to rDNA supply.

Some additional tissues exhibited striking phenotypes. The majority of ΔrDNA/eDf24 individuals that developed beyond L2 had no identifiable vulva (Fig. 2c; Supplementary Fig. 2, a, b, and k). Those ΔrDNA/eDf24 worms with apparent vulva exhibited severe vulval malformation (18 of 57 imaged D5 or D6 hermaphrodites Supplementary Fig. 2, g–k). Some evidence was also observed for defects in cuticle development: improper cuticle shedding was observed in one imaged ΔrDNA/eDf24 D6 male (Supplementary Fig. 2f) and 2 underdeveloped ΔrDNA/eDf24 D5 hermaphrodites (Supplementary Fig. 2e). Out of 57 imaged developed ΔrDNA/eDf24 hermaphrodites, 2 exhibited mid-body constrictions, which may be an indication of cuticle molting defects, though not conclusively (Supplementary Fig. 2, c and d). The vulva, the adult cuticle, and the male tail represent late-developing structures that require post-embryonic cell division. The observed defects in these tissues indicate the insufficiency of 11 rDNA copies for full post-embryonic development. Overall, we observe that severe reduction in rDNA copies exhibits highly variable penetrance of morphological phenotypes and tissue-specific abnormalities. An rDNA supply of ∼11 copies supports some degree of post-embryonic development, but is insufficient for fertility.

A new rDNA copy number allele was engineered with CRISPR-Cas9

In order to probe the minimum rDNA requirements for generational viability in C. elegans, we used the CRISPR-Cas9 system to reduce rDNA copy number (Cenik et al. 2019; Ghanta and Mello 2020). We targeted a site in the 45S rDNA array with the intention that cutting would happen in many, if not all, of the copies in the array (Supplementary Fig. 4a). A repair template with the PAM mutated was provided and worms with successful editing (Ward 2015) were screened for altered rDNA copy number using pulsed-field gel electrophoresis followed by Southern blotting for rDNA sequence (Supplementary Fig. 4, b and c). An allele estimated to contain ∼22 copies of the 45S rDNA was obtained (22-rDNA) (see Methods).

Reducing rDNA copy number to ∼10% of wild type allows development to fertile adulthood, on a delayed timescale

By examining the 22-rDNA allele in both homozygosity and hemizygosity, we assessed the effects of reducing rDNA copy number to approximately 10% or 20% that found in wild type. Animals were again generated with a cross strategy utilizing the complete rDNA deletion allele to produce worms with a genomic total of ∼22 rDNA copies (genotype ΔrDNA/22-rDNA) (Fig. 3, a–c). These worms exhibited a substantial developmental delay less severe than that of ∼11-copy rDNA worms. Two days following egg lay, most ΔrDNA/22-rDNA worms were at approximately the third larval stage (L3), while their siblings with a fully intact rDNA array (unc-54/22-rDNA) were either fourth larval stage (L4) or adult (Fig. 3, d and i; Supplementary Table 3). Worms homozygous for the 22-rDNA allele (diploid rDNA number = 44) grown in concurrent crosses were 90% L4 stage at this timepoint (Fig. 3, e and i). At the following timepoint, D1, when wild-type worms are fully gravid, ΔrDNA/22-rDNA worms showed fewer internal embryos and reduced early-life fertility (Fig. 3, f and h; Supplementary Fig. 5, a–c). Out of 30 ΔrDNA/22-rDNA worms assessed, 29 eventually produced progeny (Supplementary Fig. 5, a–c), indicating that ∼22 rDNA copies are sufficient to support reproduction.

Worms with ∼22 total rDNA copies also exhibited egg laying defects later in life. At D6, presence of internal late-stage embryos or hatched worms was enriched in ΔrDNA/22-rDNA worms compared to unc-54/22-rDNA worms (Fisher's exact P = 0.002 and 0.0002) (Supplementary Fig. 5, d–g).

Male anatomy of ∼22-copy rDNA worms did not show obvious defects (Supplementary Fig. 6) and fertility was assessed by combining 10 males of genotype ΔrDNA/22-rDNA with 3 homozygous unc-54 hermaphrodites. After 2 days, the males were recovered for genotype confirmation. The progeny of the hermaphrodites were screened on subsequent days for both ability to move and presence of F1 males, indicating products of a successful mating. Three trials were performed for which all 10 males were successfully recovered and genotyped post-mating; all 3 trials produced both moving and male worms, indicating that 22-copies of rDNA can support fertility in males (Supplementary Table 2; Supplementary Methods).

Low copy number rDNA alleles exhibit mutability in copy number

The developmental timing of worms with ∼22 or ∼44 total rDNA copies spanned a phenotypic distribution. We asked if 22-copy array instability might be observed when comparing worms at the extremes of this distribution. We examined variation within broods of worms either homozygous or hemizygous for the 22-rDNA allele. Single F1 worms were identified as being developmentally advanced or developmentally delayed, either through stage at D0 or through degree of early life fertility. Populations were grown from these single worms in order to allow pulsed-field gel-based rDNA copy number estimation. We observed copy number variation among some of the descendants of the 22-copy rDNA line (Supplementary Fig. 7). Five sets of paired worms were compared; in all 5 sets the developmentally delayed sample retained a band at ∼22 rDNA copies. In 4 of the 5 sets, the developmentally advanced worm displayed an additional band at a higher copy number than that of the paired developmentally delayed worm (Supplementary Fig. 7), suggesting that the original developmentally advanced F1 worm may have possessed an expanded rDNA allele. Although on their own these results do not demonstrate causality between rDNA copy number and developmental phenotype, collectively our data are consistent with a model where increasing the copy number of short rDNA alleles permits progressively faster developmental timing.

Reducing rDNA copy number to ∼33% of wild type allows development to fertile adulthood

The shortest verified naturally occurring C. elegans 45S rDNA array is approximately 70 copies (Thompson et al. 2013; Morton et al. 2020), present, of course, in diploid in the native strain. Given the phenotypes of artificially generated rDNA deletion alleles, we wanted to investigate the phenotype of worms hemizygous for a low-copy natural allele. We introgressed this array from wild isolate MY16 into the N2 background to create a near isogenic line (SEA305) with ∼70 copies of the 45S rDNA. We used pulsed-field gel electrophoresis to confirm this strain's haploid copy number at 73 (73-rDNA, Fig. 1b). We used this strain to generate worms with 73 total rDNA copies per diploid cell (ΔrDNA/73-rDNA) and scored worms 2 days after eggs were laid (Fig. 4). The ΔrDNA/73-rDNA animals (diploid rDNA number = 73) display neither the larval arrest observed in ΔrDNA homozygotes nor the severe developmental delay observed in ΔrDNA/eDf24 worms (Fig. 4; Table 2; Supplementary Table 3). Although ΔrDNA/73-rDNA worms have only ∼33% of the total rDNA copies available to wild-type N2 worms, these worms develop into morphologically wild-type, fertile adult hermaphrodites (Fig. 4g; Supplementary Table 2). Both hermaphrodites (Fig. 4g) and males (Supplementary Fig. 6) show visible germlines.

rDNA copy number reduced to ∼33% of wild type confers a developmental timing phenotype

Although 73-rDNA heterozygous worms showed no apparent morphological defects, we investigated if development was delayed in these worms. We examined the progeny of a cross of unc-54/ΔrDNA and 73-rDNA/73-rDNA animals laid over a 6-h time window (Fig. 5a). At 2 time points between 46 and 53 h after the end of the pulse lay, all hermaphrodite progeny (F1) from successfully mated (P0) worms were scored for developmental stage, based on developing vulval morphology and presence or absence of F2 embryos. Mating success was determined by the presence of male progeny (Supplementary Table 3). Based on stage scores, worms in the F1 cohort were classified as “young”, “old”, or developmentally in the middle of the cohort (Supplementary Tables 4 and 5). All of these F1 worms were either near wild-type in their rDNA copy number (unc-54/73-rDNA) or moderately reduced in their rDNA copy number (ΔrDNA/73-rDNA). Genotype was determined after stage scoring was completed. We hypothesized that if moderate reduction in rDNA copy number affects developmental timing, these 2 genotypes would not be equally represented in the old and young categorized groups. We found this to indeed be the case. The developmentally youngest subset of worms was enriched for having a genotype with reduced rDNA copy number (ΔrDNA/73-rDNA, diploid rDNA number = 73) while the developmentally more advanced subset was enriched for a genotype with wild-type rDNA copy number (unc-54/73-rDNA, diploid rDNA number = 202), in 2 independent replicates (Fisher's exact P < 0.0001 and = 0.0007) (Fig. 5, b and c; Supplementary Table 4).

Fig. 5.

Worms with rDNA copy number reduced to ∼33% of wild type show a developmental timing phenotype. a) Hermaphrodite PD2621 (unc-54/ΔrDNA) worms were mated to SEA305 (73-rDNA/73-rDNA) males. After 2 days, mated hermaphrodites were allowed to pulse lay embryos for 6 h. At 2 time points between 46 and 53 h after the end of the pulse lay, all hermaphrodite F1 progeny were scored for developmental stage, and worms at the developmental extremes of youngest and oldest in the cohort were identified (Supplementary Tables 4 and 5). Three days after the stage scoring of the F1 worms, the F2 progeny were used to retroactively assign the genotype of each F1. Outcross was confirmed by the presence of males in the F1 population (see Supplementary Tables 2 and 3). b) Trial one (experiment ID 061220, see Table 2, Supplementary Table 4). n = 154 F1s, of which 35 were categorized as “young” and 28 as “old.” Genotype was determined 3 days after categorization. Young worms were enriched for genotype ΔrDNA/73-rDNA (diploid rDNA number = 73) and old worms for unc-54/73-rDNA (diploid rDNA number = 202) (χ²  P < 0.0001; Fisher's exact P < 0.0001). c) Trial 2 (experiment ID 062320, Supplementary Table 4). n = 140 F1s, of which 26 were categorized as “young” and 36 as "old.” Young worms were enriched for genotype ΔrDNA/73-rDNA (diploid rDNA number = 73) and old worms for unc-54/73-rDNA (diploid rDNA number = 202) (χ²  P = 0.001; Fisher's exact P = 0.0007).

We repeated this assay using strains with alternatively sourced reduced rDNA copy number alleles (Supplementary Fig. 8a). One strain (SEA302) was generated by introgression into an N2 background from wild isolate JU775, possessing a 45S rDNA of approximately 81 copies (81-rDNA) per haploid genome (Fig. 1b). A third strain (SEA296) was obtained through a spontaneous copy number reduction event in a transgenic strain of wild isolate MY1; rDNA copy number was reduced to ∼64 copies (64-rDNA) per haploid genome (Fig. 1b, see Methods). We set up crosses of each of these 2 additional sources of low-copy rDNA arrays as described above. Progeny phenotypes of crosses involving the 81-rDNA allele trended in the same direction as observed for 73-rDNA, with developmentally young worms enriched for ΔrDNA/81-rDNA (diploid rDNA number = 81) and old for unc-54/81-rDNA (diploid rDNA number = 210), but did not reach significance (Fisher's exact = 0.06 and 0.1 for 2 independent replicates) (Supplementary Fig. 8, b and c; Supplementary Table 4). Crosses using the 64-rDNA allele reproduced the 73-rDNA result with high significance (Fisher's exact P < 0.0001) (Supplementary Fig. 8d; Supplementary Table 4). Together, these results indicate that there is an rDNA copy number at which fertility and post-embryonic tissue development is accomplished, but which is insufficient for fully wild-type developmental timing.

A developmental timing defect is rescued by a hemizygous high copy number rDNA allele

If the developmental timing phenotype described above is truly a consequence of reduced rDNA copy number, we would predict that supplying higher rDNA copy number in the crossed-in allele would rescue the phenotype. We thus generated a near isogenic line with high rDNA copy number by introgressing into the N2 background one of the largest observed naturally occurring rDNA alleles among C. elegans wild isolates: an array with copy number ∼417 and a diploid copy number of ∼834 (Thompson et al. 2013; Morton et al. 2020) (Fig. 6a; Supplementary 1c). As N2 worms natively have ∼200 copies per diploid genome, putting a >400-copy array over an rDNA deletion should more than restore the total cellular complement of wild-type rDNA. We set up crosses in parallel between unc-54/ΔrDNA hermaphrodites and males of either 73-rDNA or 417-rDNA homozygous genotypes (Fig. 6b). Mated hermaphrodites from these crosses were moved to individual plates and blinded for paternal genotype. F1 worms were scored for stage to sort into developmentally oldest and youngest subsets and then subsequently genotyped via F2 progeny, as described. In 2 independent experiments, the young subset of the 73-rDNA parental cross were significantly enriched for reduced rDNA copy number (ΔrDNA/73-rDNA, diploid rDNA number = 73) and the old subset for wild-type rDNA copy number (unc-54/73-rDNA, diploid rDNA number = 202) (Fisher's exact P = 0.006 and 0.003); whereas progeny of the cross to 417-rDNA (resulting in diploid rDNA numbers of 546 or 417) showed no such association (Fisher's exact both P > 0.5) (Fig. 6, c and d).

Fig. 6.

A high copy number rDNA array restores developmental timing. a) The available alleles are diagramed (not to scale). b) Hermaphrodite PD2621 (unc-54/ΔrDNA) worms were mated to either SEA305 (73-rDNA/73-rDNA) or SEA300 (417-rDNA/417-rDNA) males. After 2 days, mated hermaphrodites were blinded for mate genotype and allowed to pulse lay embryos for 6 h. F1 progeny were developmentally scored as described and the developmentally youngest and oldest were identified (Supplementary Tables 4 and 5). Three days after the stage scoring, the F2 progeny were used to retroactively assign the genotype of each F1. c) Trial one of the experiment diagramed in B (experiment ID 062920, see Supplementary Tables 3 and 4). For the test of 73-rDNA (left), n = 106 F1s, of which 45 were categorized as “young” and 33 as “old.” Genotypes were determined after categorization. Young worms were enriched for genotype ΔrDNA/73-rDNA (diploid rDNA number = 73) and old worms for unc-54/73-rDNA (diploid rDNA number = 202) (X²  P = 0.01; Fisher's exact P = 0.006). For the test of 417-rDNA (right), n = 178 F1s, of which 70 were categorized as “young” and 39 as “old.” Association between genotype and developmental category was not observed (X²  P = 0.9; Fisher's exact P = 0.8). d) A replicate experiment of C is presented (experiment ID 080520). For the test of 73-rDNA, n = 116 F1s, of which 49 were categorized as “young” and 6 “old.” Association between genotype and developmental category was observed (X²  P = 0.009; Fisher's exact P = 0.003). For the test of 417-rDNA, n = 127 F1s, of which 31 were categorized as “young” and 18 “old.” (X²  P > 0.5; Fisher's exact P > 0.5). Outcross was determined by presence of males in the F1 population (Supplementary Table 3).

In an additional trial, we set up crosses in parallel with 4 separate paternal genotypes: SEA305 (homozygous 73-rDNA), SEA302 (homozygous 81-rDNA), SEA51 (homozygous 133-rDNA with a linked GFP transgene, mIs13), and SEA300 (homozygous 417-rDNA) (Supplementary Fig. 9a). These crosses were performed as described above, with unc-54/ΔrDNA P0 hermaphrodites singled after mating and each individual blinded for paternal genotype (Supplementary Fig. 9b). This experiment produced the expected results of both 73- and 81-copy rDNA arrays associating with developmental timing skew: the reduced-rDNA genotype was overrepresented in young worms and under-represented in old worms (Fisher's exact P = 0.001 for SEA305 and 0.0002 for SEA302) (Supplementary Fig. 9, c and e; Supplementary Table 4). The cross involving 417-rDNA also reproduced earlier results of no association (Fisher's exact P > 0.5) (Supplementary Fig. 9d; Supplementary Table 4). The cross involving an rDNA array of 133 copies also produced no association (Fisher's exact P > 0.5), suggesting that ∼133 rDNA copies total is sufficient to rescue the developmental timing defect of moderately reduced rDNA copy number. We thus conclude that the developmental timing phenotype observed in worms with low-copy rDNA alleles is indeed the result of too few rDNA copies and is not produced by alleles bearing higher copy numbers. This result also suggests that sufficient rDNA copies for standard developmental timing may be provided on a single array and need not be present on both chromosomes.

Discussion

In humans, rDNA copy number variation has emerged as a factor to consider in human health, diagnostics, and association studies. Loss of rDNA copies has been observed in some cancers but not others (Salim et al. 2017; Wang and Lemos 2017; Xu et al. 2017; Udugama et al. 2018; Hosgood et al. 2019; Feng et al. 2020; Valori et al. 2020) and in some reports examining aging tissues (Johnson and Strehler 1972; Gaubatz and Cutler 1978; Zafiropoulos et al. 2005; Watada et al. 2020). rDNA copy number variation in humans and flies has also been associated with altered genome-wide gene expression (Paredes et al. 2011; Gibbons et al. 2014). We present here investigation into the phenotypic consequences of varying copy number in an animal model. We observe tissue-specific phenotypes, suggesting different minimal rDNA requirements among tissues. We conclude that ∼11 copies of 45S rDNA—∼5% of the amount found in a wild-type diploid—are sufficient to complete many aspects of development to adulthood. However, this development happens over an extended length of time and with variable penetrance. Furthermore, certain tissues fail to develop or develop improperly, most notably the vulva and germline. More moderate, but still substantial, rDNA copy number reduction to ∼22 copies, equating to about 10% of the total amount of 45S rDNA found in wild type, was sufficient for fertility. However, these worms displayed severe developmental delay, as did worms with ∼44 total rDNA copies, with less severe delay. Finally, reduction to ∼73 copies of rDNA, about one-third the rDNA dose of wild type, rescues morphological phenotypes but confers a very mild defect in developmental timing, which is eliminated by supplying excess rDNA copies. Collectively, our data suggest that rDNA sufficiency for development and fertility is not a threshold trait, but rather a continuous relationship between increasing rDNA copy number and ability to develop, at least for copy numbers below the natural range.

Diseases of ribosomal components or production (ribosomopathies) are surprisingly tissue-specific in their presentation (Freed et al. 2010; Yelick and Trainor 2015). Ribosomopathies may involve craniofacial abnormalities, osteopenia, brain developmental defects, liver malfunction, or many other symptoms (Yelick and Trainor 2015). A C. elegans ribosomopathy mutant model with impaired rDNA transcription shows delayed gonadogenesis, a higher rate of germline apoptosis, and reduced fertility (Lee et al. 2014). We do not have direct evidence of reduced ribosome biogenesis in our individual worms with ∼11, ∼22, or ∼73 total 45S rDNA copies, due to the technical difficulty of such quantification on small numbers of heterozygous worms. However, insufficient rRNA transcription and thus impaired ribosome production is a plausible mechanism by which the observed phenotypes could arise. The amount of ribosome production supported by ∼11 copies of 45S rDNA in our heterozygous worms is enough to allow worms to overcome the stage transition from L1 to L2, presumably facilitated by the maternally supplied ribosomes present through L1 (Cenik et al. 2019). C. elegans heterozygotes with ∼11 rDNA copies exhibit severe defects past the L2 stage, however. Our results suggest that tissue types that require substantial post-embryonic cell division (Sulston and Horvitz 1977) may suffer the most under these rDNA-restricted conditions.

There is additional evidence suggesting germ cell development and proliferation may be processes particularly sensitive to rDNA abundance. One of the few established examples of tissue-specific rDNA copy number requirements is oogenesis; high amplification of rDNA copies has been observed in oogenesis of many cold-blooded animals (Gall 1968; Spring et al. 1996; Ortega-Recalde et al. 2019; Davidian et al. 2021). Mouse oocytes, too, upregulate rRNA production, although without concomitant increases in rDNA copy number (Iuchi and Green 1999; Tian et al. 2001). Germline stem cell proliferation in L3 and L4 C. elegans is similarly dependent on sufficient ribosome production (Kudron and Reinke 2008; Waters and Reinke 2011). Thus, the germline represents one important tissue type with specific rDNA requirements. We observed that the functional outcome of hermaphrodite worms possessing few copies of the 45S rDNA was 100% sterility. It remains undetermined if this reflects a failure of oogenesis or other processes, such as spermatogenesis. The sperm viability of ΔrDNA/eDf24 males was not tested. The increase in egg holding/internal hatching in ∼22 total copy worms (Supplementary Fig. 5) may or may not be further indication of tissue-specific abnormalities, as it could be attributable to several different causes, including vulval defects or starvation state. While the 22-rDNA worms had a plentiful food supply, it is possible that a metabolic or feeding defect could mimic starvation, which can cause embryo retention (Seidel and Kimble 2011).

The extreme variability of phenotypes in ∼11-copy rDNA worms was particularly noticeable. Only about half of these worms developed to a stage displaying hallmarks of adulthood. The observed developmental and phenotypic variability in these worms might well reflect small individual-specific changes in the rDNA allele copy number (e.g. expansion to 12 rDNA copies from 11); unfortunately, current methods do not allow fine-resolution quantification of rDNA copy number in individual sterile worms. Instead, we investigated worms in the ∼22 and ∼44 copy number range with extreme phenotypes, in which we indeed observed some variation in actual copy number (Supplementary Fig. 7). Although we cannot claim that this copy number variation explains the phenotypic variability, our results are consistent with a model in which at least some of the phenotypic variability present in individuals with short rDNA arrays is due to copy number fluctuation. In Drosophila, inherited short rDNA arrays can recover to wild-type copy number levels over a single generation (Lu et al. 2018). We did not observe this rapid recovery in C. elegans, but the existence of copy number variation in conjunction with a fitness advantage of higher copy number would imply that population-level recovery would happen given enough variation and generational time. To truly understand the degree of mutability in rDNA copy number and how it relates to phenotype will require copy number assessment on a far larger and finer scale than presented here, an endeavor challenged by the substantial difficulty of detecting single-digit copy number differences.

Variable penetrance of development of 11-rDNA worms might equally be due to variation in maternal deposition of ribosomes; however, if that were the case, we might have expected more variation to have been observed in the L1 arrest phenotype with total absence of rDNA (Cenik et al. 2019). Among worms that developed, variation also existed in the severity of phenotypic abnormalities (Supplementary Figs. 2 and 3). It similarly remains possible that reduction in rDNA copy number created a sensitized environment for haploinsufficiencies (due to the eDf24 deleted region) to become partially penetrant, which would have very interesting implications for the epistatic role of rDNA copy number. In the absence of underlying genetic or maternal variation, the phenotypic variation might reflect ways in which the needs and thresholds of ribosome biogenesis respond to micro-environmental variation or other stochastic factors. As discussed, disease presentation of ribosomopathies is noted to show considerable variability (Yelick and Trainor 2015); perhaps cellular function operating at the edge of ribosome debilitation is particularly sensitive to metabolic or environmental fluctuation.

The embryonic lethality of deficiency allele eDf24 has been attributed to its truncated rDNA, to the extent that the lethal designation let-209 is associated as an alternative gene name of the 18S rRNA (Albertson 1984). However, our results suggest that the rDNA was misclassified as this lethal allele. We have shown that the eDf24 reduction in rDNA copy number alone does not result in embryonic lethality, either through insufficient rDNA copies or through toxicity of a few copies being worse than none (Cenik et al. 2019). We sequenced the boundary of eDf24 (Supplementary Fig. 1), allowing us to propose alternative identities of let-209. pole-1, at genomic location 15005603 on chromosome I, is a homolog of DNA polymerase epsilon with an annotated embryonic lethal RNAi phenotype (Fraser et al. 2000; Sönnichsen et al. 2005); this has also been proposed as the causal gene of eDf24 lethality by others in unpublished work (Vijayaratnam 2000).

Isolates of classic model organisms that have been recently obtained from the wild have wide variation in their strain-specific rDNA copy number, observed in flies (Ritossa and Scala 1969; Mohan and Ritossa 1970; Lyckegaard and Clark 1989), yeast (Peter et al. 2018; Morton et al. 2020; Sharma et al. 2021), and C. elegans (Thompson et al. 2013). The lowest verified haploid 45S rDNA copy number among C. elegans wild isolates is ∼70 copies [out of a sampling of 40 wild isolates (Thompson et al. 2013) of which the lowest and highest were verified with CHEF gel (Morton et al. 2020)]. Natively, of course, this allele occurs in diploid worms, supplying ∼140 total copies. It is perhaps not a coincidence that this haploid number appears to be nearing the threshold of detrimental fitness effects. When present over a deletion allele, ∼73 rDNA copies allows superficially wild-type anatomy and development, but subtly perturbs developmental timing (Figs. 4 and 5). Further study is needed to determine if persistent subtle fitness defects occur with copy numbers ranged between 70 and 140 total rDNA copies, and if so what form such defects take. Furthermore, high rDNA copy numbers, particularly above the natural range, may also have the potential to influence phenotype, a question that warrants further investigation.

Our study supports the conclusions that the minimal rDNA copy number for a multicellular animal to complete development and reproduce is smaller than even the lowest naturally occurring total copy number, and substantially smaller than the mean copy number among strains (Ritossa and Scala 1969; Thompson et al. 2013). Similar observations about the minimum viable copy number have been made in the unicellular model S. cerevisiae (Kaback and Halvorson 1977; French et al. 2003; Kobayashi 2006). Why then, does the rDNA exist in such an excess of copies? One possibility is a situational need for exceptional rRNA production. In frogs and some reptiles, rDNA is amplified to presumably meet heightened ribosome production needs during oogenesis (Gall 1968; Macgregor 1972; Spring et al. 1996; Davidian et al. 2021). It is conceivable, for instance, that environmental conditions outside the lab might exist that require rDNA copy number to be maintained in such high copy numbers. There is substantial evidence, however, that rDNA copy number reflects properties independent from ribosome biogenesis. Above a certain threshold, rDNA copy number does not correlate with rRNA abundance (Mohan and Ritossa 1970; Buescher et al. 1984; French et al. 2003; Paredes and Maggert 2009b; Li et al. 2018), and copy number has been implicated in several other cellular processes. rDNA copy number deletion alleles in flies affect heterochromatic silencing and position effect variegation elsewhere in the genome (Paredes and Maggert 2009b). In yeast, replication and cell cycle regulation can be impacted through rDNA copy number change (Kwan et al. 2021), as can genome stability in yeast and plants (Kobayashi 2008, 2014; Picart-Picolo et al. 2020; Kwan et al. 2021). Yeast can even use rDNA arrays to apparently functionally replace telomeres (Begnis et al. 2018). The roles of the ribosomal DNA locus in cell biology are thus myriad and have yet to be fully explored.

Thus far, investigation into the consequences of copy number variation has been hampered by technological limitations on accurate quantification at this locus (McStay 2016; Morton et al. 2020), and for organisms with large rDNA unit sizes such as humans, precise copy number calls remain elusive (Miga et al. 2020; Hall et al. 2021; Nurk et al. 2021). Through the use of the model organism C. elegans, we have expanded our knowledge of the lower bounds of rDNA copy number sufficiency, observing tissue-specific defects and substantial phenotypic variability as a result of inadequate rDNA copy number. Further investigation is warranted to explore the extent to which phenotype is influenced by minor variation in rDNA copy number and to characterize the complex operation and functional consequences of variation in this genomic feature.

Data availability

Strains (Supplementary Table 6) are available upon request. Supplementary File S2 includes read depth analysis code; read depth files are available at figshare: https://doi.org/10.25386/genetics.22197457. Also available through figshare are alignment and VCF files for analyzed NILs. FASTQ files for sequenced NILs and CB3740 are available through SRA (PRJNA746007). Supplemental material available at GENETICS online.

Funding

This work was supported by funding from NIGMS (grants 5R01GM122088 and 1R35GM139532 to CQ), NHGRI (grant 1RM1HG010461 to CQ), and NIA (National Institute on Aging) (F31 AG063450 to ANH).