A small RNA system ensures accurate homologous pairing and unpaired silencing of meiotic chromosomes

Abstract

During meiosis, chromosomes with homologous partners undergo synaptonemal complex (SC)‐mediated pairing, while the remaining unpaired chromosomes are heterochromatinized through unpaired silencing. Mechanisms underlying homolog recognition during SC formation are still unclear. Here, we show that the Caenorhabditis elegans Argonaute proteins, CSR‐1 and its paralog CSR‐2, interacting with 22G‐RNAs, are required for synaptonemal complex formation with accurate homology. CSR‐1 in nuclei and meiotic cohesin, constituting the SC lateral elements, were associated with nonsimple DNA repeats, including minisatellites and transposons, and weakly associated with coding genes. CSR‐1‐associated CeRep55 minisatellites were expressing 22G‐RNAs and long noncoding (lnc) RNAs that colocalized with synaptonemal complexes on paired chromosomes and with cohesin regions of unpaired chromosomes. CeRep55 multilocus deletions reduced the efficiencies of homologous pairing and unpaired silencing, which were supported by the csr‐1 activity. Moreover, CSR‐1 and CSR‐2 were required for proper heterochromatinization of unpaired chromosomes. These findings suggest that CSR‐1 and CSR‐2 play crucial roles in homology recognition, achieving accurate SC formation between chromosome pairs and condensing unpaired chromosomes by targeting repeat‐derived lncRNAs.

Caenorhabditis elegans Argonaute proteins CSR‐1 and CSR‐2 mediate both synaptonemal complex formation and condensation of unpaired chromosomes by targeting repeat‐derived lncRNAs.

Introduction

Meiosis is a specialized mode of cell division that produces haploid gametes. During the pachytene stage of meiosis I, homologous chromosomes pair up, forming a tetrad held together by the synaptonemal complex (SC; Moses, 1956). This homolog pairing allows crossover recombination, and its spatial and temporal arrangement ensures precise homologous chromosome separation at the end of meiosis I. Homolog pairing initiation depends on species‐specific mechanisms involving interactions between DNA termini generated by double‐strand breaks (DSBs) and between pericentric heterochromatin, centromeres, telomeres, and/or pairing center (PC) sequences (reviewed in Rog & Dernburg, 2013).

In the nematode Caenorhabditis elegans, HIM‐8 or other ZIM proteins with zinc fingers bind to the PCs at the chromosome ends and act as the starting points of homolog pairing (Phillips et al, 2005, 2009). Subsequently, during pachytene, the SC establishes the pairing of other regions of the homologous chromosomes. In budding yeast, mice, and Arabidopsis, DSBs are important for synapse formation, probably because the single‐stranded regions generated by DSBs recognize identical regions on the other homologous chromosome. In C. elegans and Drosophila, DSBs are not necessary for synapse formation (Dernburg et al, 1998; McKim et al, 1998), implying the presence of an unclarified, DNA‐independent mechanism for recognizing homologous chromosomes during SC formation.

Not all chromosomes undergo meiotic pairing. In male mammals, most regions of the X and Y chromosomes do not undergo pairing. In C. elegans, while the hermaphrodite has an XX genotype, the male has an X0 genotype in which the monosomic X chromosome is unpaired during meiosis. Free‐type duplications and artificially formed extrachromosomal transgenes also remain unpaired. These unpaired DNAs are silenced at the transcriptional level during meiosis in Neurospora, C. elegans, and mice (Aramayo & Metzenberg, 1996; Bean et al, 2004; Turner et al, 2005) by a process called meiotic silencing of unpaired DNA (unpaired silencing). Unpaired silencing regulates the expression of sex chromosome genes, likely inactivates unpaired monosomic autosomes to suppress the production of aneuploid gametes missing autosomes and is hypothesized to defend against foreign DNA, including transposons and DNA viruses (reviewed in Kelly & Aramayo, 2007). In Neurospora and C. elegans, Argonaute proteins and RNA‐dependent RNA polymerases (RdRP) are involved in the mechanisms of both unpaired silencing and RNA interference (RNAi) (Shiu et al, 2001; Maine et al, 2005; She et al, 2009; Hammond et al, 2013).

The exogenous (exo‐)RNAi pathway plays a role in defense against RNA viruses (reviewed in Ding & Voinnet, 2007). In exo‐RNAi in C. elegans (Fire et al, 1998), Dicer with RNase III activity produces primary small interfering RNAs (siRNAs) with monophosphorylated 5′‐ends (5′P). Subsequently, RdRP complexes synthesize secondary siRNAs with triphosphorylated 5′‐ends (5′PPP). Secondary siRNAs involved in exo‐RNAi (Pak & Fire, 2007; Sijen et al, 2007) correspond to 22G‐RNAs of endogenous (endo‐)RNAi (Gu et al, 2009). The core component of RdRP is RRF‐1 or EGO‐1 (Sijen et al, 2001; Maine et al, 2005), which is structurally different from the viral RdRP. In vitro RNAi analyses have shown that 5′PPP siRNAs interact with the Argonaute protein CSR‐1 and induce Slicer activity that cleaves target cytoplasmic mRNAs (Aoki et al, 2007).

In C. elegans, some RNAi mutants, including ego‐1 and csr‐1, exhibit abnormal heterochromatinization in unpaired silencing of meiotic chromosomes (Maine et al, 2005; She et al, 2009). Unpaired silencing is a nuclear event, but the chromatin proteins and RNAs that cooperate with RNAi factors to accomplish unpaired chromosome heterochromatinization are relatively unknown.

In this study, we found that the csr‐1 mutants exhibited defects not only in unpaired silencing but also in homologous pairing. We then investigated the relationship between SC formation on paired chromosomes, unpaired chromosome heterochromatinization, and an Argonaute–small RNA pathway in C. elegans. We found that in nuclei, CSR‐1 and meiotic cohesin containing COH‐3/4 were associated with nonsimple repeats expressing 22G‐RNAs and lncRNAs, of which lncRNAs partially colocalized with the SC. CSR‐1 and its paralog were required for SC formation between chromosome pairs with accurate homology and were also involved in the condensation of unpaired chromosomes. The unpaired chromosomes eventually formed heterochromatin rich in the chromodomain protein CEC‐5. Our study suggests that the small RNA system is involved in homology recognition between chromosomes, excluding the PC regions. We also discuss pleiotropic roles of CSR‐1.

Results

Argonaute protein CSR‐1 paralog has slicer activity

We analyzed the Argonaute proteins involved in both RNAi and unpaired silencing in C. elegans (Fig 1A). Our previous in vitro analyses showed that 5′PPP siRNAs mimicking RdRP products induce prominent Slicer activity that cleaves target RNAs far more effectively than 5′P siRNAs. The primary factor responsible for 5′PPP siRNA‐induced RNA cleavage is CSR‐1 (Aoki et al, 2007). Further analysis revealed that 5′P single‐stranded siRNAs could induce minor Slicer activity in lysates from the wild‐type (WT) strain, but not from the C04F12.1(tm1637) mutant (Fig EV1A). C04F12.1 is an Argonaute protein with a D‐D‐H/D motif necessary for RNA cleavage (Fig EV1B) and is the closest paralog of CSR‐1 (Yigit et al, 2006). Hence, we designate C04F12.1 as CSR‐2 in this study. A recombinant CSR‐2 protein showed RNA cleavage activity in conjunction with 5′P and 5′PPP siRNAs (Fig 1B), raising questions as to which small RNAs partner with CSR‐2 in vivo.

Figure 1. Nature of two Slicer‐type Argonaute proteins and current knowledge on the regulation of meiotic chromosomes by RNAi.

1. Schematic diagrams showing homologous pairing and unpaired silencing of chromosomes during meiosis I. Meiosis is divided into meiosis I—consisting of interphase, leptotene, zygotene, pachytene, diplotene, and diakinesis—and meiosis II.
2. CSR‐2 (C04F12.1) is an Argonaute protein that has RNA cleavage activity in conjunction with small RNAs. We reacted 6 × His::MBP::CSR‐2 fusion proteins (final 6 ng/μl) with a cap‐radiolabeled mRNA and the non‐, mono‐, and tri‐phosphorylated (5′OH, 5′P, and 5′PPP) siRNAs. Reaction products were resolved by 6% sequencing gels.
3. Similar to CSR‐1, the majority of small RNAs interacting with CSR‐2 in vivo have 5′PPP ends. The RNAs in the immunopurified Argonaute complexes were treated with or without Vaccinia virus capping enzyme followed by an alkaline phosphatase, and finally labeled with [γ‐³²P]ATP and a polynucleotide kinase. RNA products were resolved by 15% sequencing gels.
4. The proportion of small RNA species detected in CSR‐1 and CSR‐2 complexes from adult hermaphrodites. Small RNAs recovered from FLAG‐tagged Argonaute complexes and input cell lysates were amplified by RT‐PCR and were analyzed by a parallel sequencer. The average of the technical duplicates is presented. Recovery rates (IP/lysate) were categorized into three grades and are presented in colors. The relative frequency of sense products in the small RNAs mapped to coding genes is shown at the bottom of the panel.
5. Male fertility test. The males of the WT and Argonaute mutants were mated to unc‐5(e152) hermaphrodites. For each strain, 10 sets of matings were arranged as indicated by n. Mobile males in F1 progeny were counted as the male cross progeny. The thick lines in the box plot indicate the median values. The length of the box represents the interquartile range (IQR) between the 25^th and 75^th percentiles. The ends of the whiskers represent the farthest data points within an interval of 1.5 times the IQR. P‐value was determined by Mann–Whitney U‐test.
6. Semirandom mutagenesis of the second K‐rich region in the csr‐1 gene, and the genetic screen for csr‐1 mutants exhibiting the RNAi‐deficient phenotype. Progeny of worms mutagenized by genome editing with a degenerate oligonucleotide were exposed to dsRNA targeting an essential gene sca‐1.
7. Three csr‐1 hypomorphic mutants showed a diminished RNAi response to exo‐dsRNA. The WT, the csr‐1(fj150) mutant, and several strains were injected with dsRNA (1.25 μg/μL) targeting the maternal gene pos‐1, essential for embryogenesis. The incidence of lethal embryos in the F1 progeny was examined. The number of F1 progeny counted is indicated by n. Error bars represent 95% confidence intervals (CIs). See also Appendix Fig S1A.
8. Structure and mutations of csr‐1. Exon regions encoding the PAZ and Piwi domains are shown in dark gray.

Source data are available online for this figure.

Figure EV1. Additional information on the Argonaute proteins CSR‐1 and CSR‐2, and the small RNAs in their complexes.

1. Cell lysates from a csr‐2(tm1637) mutant lacked the minor Slicer activity for cleaving target RNAs. Lysates from the WT strain and the csr‐2 mutant were reacted with a cap‐radiolabeled mRNA and the following siRNAs: nonphosphorylated (5′OH), monophosphorylated (5′P), and triphosphorylated (5′PPP). The reaction products were analyzed on a 6% sequencing gel. We also showed that the recombinant CSR‐2 protein had RNA cleavage activity in conjunction with 5′P and 5′PPP siRNAs (see Fig 1B). This may indicate that the strong Slicer activity induced by 5′PPP siRNAs in cell lysates is a mixture of activities, which derive mainly from CSR‐1 and partly from CSR‐2.
2. The catalytic domains of CSR‐1, CSR‐2 (C04F12.1), and human (Hu) AGO2 proteins were aligned based on peptide sequences. The shaded residues are the D‐D‐H/D motif important for RNase activity in CSR‐1 (Yigit et al, 2006; Aoki et al, 2007; Gerson‐Gurwitz et al, 2016), CSR‐2, and HuAGO2 (Liu et al, 2004). Less conserved peptide sequences, which connect between highly conserved peptide sequences, are denoted as “.+”.
3. The small RNA population in the CSR‐1 complex partially overlapped with that in the CSR‐2 complex. The percentages on the bars were calculated from raw reads. The mean content of C or A in the middle region (from 4 to 18 bases) of the small RNAs was calculated from non‐redundant species. Adult (#1 and #2) stands for merged data of duplicates.
4. Relative nucleotide frequencies of small RNAs interacting with CSR‐1 and CSR‐2 are shown as sequence logos. Raw indicates the nucleotide frequencies calculated from raw reads, which connote the multiplicity of expressed RNAs. The analysis of raw reads detected guanine at the 5′ ends in 82% of the CSR‐1‐interacting small RNAs and 96% of the CSR‐2‐interacting small RNAs. Non‐redundant sequences indicate the nucleotide frequencies calculated from species information of RNAs, which were classified based on 18‐nt sequences from their 5′ ends. The mean content of C or A in the middle region of the small RNAs is shown on the right side.
5. The mRNA expression ratio between an unpaired X chromosome versus a pair of autosomes was abnormal in X0 males of the csr‐2 mutant and the csr‐1 mutant. mRNA expression of mes‐1 on the X chromosome and pgl‐3 on autosome V, both expressed in the germline, was analyzed by semi‐quantitative RT‐PCR. Total RNA of adult worms used for RT‐PCR were prepared from XX hermaphrodites (herm.) and X0 males of the WT strain, XX males of the tra‐2 mutant, and X0 males of the csr‐1 mutant, the csr‐2 mutant, the csr‐2; csr‐1 double mutant, and the MAGO12 duodecuple mutant. tra‐2(q276) is a sex determination mutant, and MAGO12 is an RNAi‐deficient mutant strain. PCR products were resolved by 8.5% native polyacrylamide gels followed by staining with a sensitive DNA‐binding dye (GelRed; Biotium). This result is consistent with a model in which CSR‐1 and CSR‐2 condense the monosomic X chromosome and also prevent irregular contact of the X chromosome with the autosomes.
6. The percentage of sense products in CSR‐1‐ and CSR‐2‐interacting small RNAs were examined in the whole exons and introns as well as in the ChIP peak regions for CSR‐1, SMC‐1, and COH‐3/4 located within exons and introns. The ChIP regions common (∩) to the duplicates were used. For sense small RNAs in the ChIP regions for CSR‐1, the relative frequency of RNAs with 5′‐G is shown (‡). Many of the antisense small RNAs were likely to originate in the cytoplasm.
7. ChIP intensity proportion of exon subsets in all exons and of intron subsets in all introns. Subsets expressing sense small RNAs in the CSR‐1 or CSR‐2 complex were defined using the data in this study. Subsets expressing antisense capped nuclear RNAs were prepared using the data from Jänes et al (2018a). Subsets expressing poly(A) RNAs in the germline were prepared using the data from Wang et al (2009). The proportional lengths of the exon and intron subsets relative to the exon and intron supersets are presented for comparison.
8. Genome browser visualization of read densities of small RNAs interacting with CSR‐1 and capped nuclear RNAs, and ChIP signals for CSR‐1, SMC‐1, and COH‐3/4 in the vit‐6 gene region.

Source data are available online for this figure.

We constructed single‐copy transgenic worms expressing FLAG‐tagged CSR‐2 and FLAG‐tagged CSR‐1. The FLAG‐tagged Argonaute complexes were immunopurified and used for identifying the Argonaute‐interacting small RNAs. CSR‐1 interacts with 22G‐RNAs, which are RdRP products bearing 5′PPP, frequently starting with guanine (Gu et al, 2009), and lacking 2′‐O‐methylation (Montgomery et al, 2012). Our assay, using a virus‐capping enzyme that reacts with 5′PPP but not 5′P RNAs, demonstrated that most small RNAs present in the CSR‐2 complex had 5′PPP (Fig 1C), suggesting that the majority of the endogenous small RNAs interacting with CSR‐2 were RdRP products. Considering that some 5′P small RNAs can be generated from 5′PPP small RNAs by polyphosphatases in C. elegans (Chaves et al, 2021), a very small part of the CSR‐2‐interacting small RNAs could still be 5′P small RNAs.

We next examined the sequences of the small RNAs interacting with the Slicer‐type Argonaute proteins using RT‐PCR and high‐throughput sequencing. Based on approximate discrimination by 18‐nucleotide (nt)‐sequences from their 5′ ends, 33% of the CSR‐1‐interacting small RNAs overlapped with 78% of the CSR‐2‐interacting small RNAs (Fig EV1C). The middle region of the CSR‐1‐interacting small RNAs showed a mean cytosine content higher than that of the CSR‐2‐interacting small RNAs (Fig EV1C and D). We classified the derivations of small RNAs based on their genomic features (Fig 1D). The most abundant population in the CSR‐1‐ and CSR‐2‐interacting small RNAs was a class of 22G‐RNAs mapped to protein‐coding exons, and this class of 22G‐RNAs frequently showed complementarity to the mRNAs, suggesting that many of those were synthesized on cytoplasmic mRNAs. However, of the small RNAs interacting with CSR‐1 and CSR‐2, around 5% of exon‐derived small RNAs and around one‐third of intron‐derived small RNAs showed sense orientation (Fig 1D, bottom). These data raise the possibility that antisense transcripts from coding genes in nuclei might also be targeted by CSR‐1 and CSR‐2, because the antisense transcripts are unlikely to be transported to the cytoplasm. A unique population of small RNAs interacting with CSR‐1 and CSR‐2 was the X‐cluster 22G‐RNAs, which are abundantly expressed from a cluster (several kb in size) at the center of the X chromosome (Ambros et al, 2003). Moreover, 2.2% of the CSR‐1‐interacting small RNAs and 1.9% of the CSR‐2‐interacting small RNAs were 22G‐RNAs derived from repeat sequences (Fig 1D). Compared with the recovery rates (IP/lysate) of miRNAs into the CSR‐1 and CSR‐2 complexes, the recovery rates of 22G‐RNAs mapped to protein‐coding exons were over 72‐fold more efficient, and those of repeat‐derived 22G‐RNAs were over 21‐fold more efficient. These observations suggest that CSR‐1 and CSR‐2 not only target coding gene transcripts but also may target subpopulations of repeat‐derived RNAs.

New mutant csr‐1 and csr‐2 strains

csr‐2 mutants can be bred homozygously and show a subtly diminished RNAi response to exo‐dsRNA (Yigit et al, 2006). The csr‐1 null mutant is sterile; however, some csr‐1 homozygotes occasionally lay a few dead eggs with defects in chromosome segregation (Claycomb et al, 2009). In addition, the csr‐1 mutant exhibits a considerably diminished response against exo‐dsRNA (Yigit et al, 2006). We created new mutant strains for further analysis of csr‐1 and csr‐2.

First, a csr‐2(tm1637); csr‐1(fj54) double‐null mutant was constructed. To examine male fertility, we counted F1 cross‐progeny produced by mating the males of the WT strain and the Argonaute mutants with csr‐1(+) hermaphrodites (Fig 1E). In comparison with the mean number of F1 progeny from the WT male, the F1 number from the csr‐1 single‐mutant male was at the 40% level, and the F1 number from the csr‐2; csr‐1 double‐mutant male was at the 21% level. The csr‐2; csr‐1 double mutant seemed partially defective in the formation of functional sperm.

We made several new csr‐1 mutants by genome editing (Fig 1F–H). The csr‐1(fj70) mutant was created so that CSR‐1 had a D769L substitution in the catalytic D‐D‐H motif for RNA cleavage. The csr‐1(fj126) mutant was created by the insertion of a nuclear export signal (NES). The hermaphrodites of both csr‐1(fj70) and csr‐1(fj126) mutants were sterile. Considering that lysine (K) residues are often involved in intra‐ and intermolecular interactions, we mutated two regions encoding K‐rich peptides in the csr‐1 gene. The hermaphrodite of the csr‐1(fj67) mutant, which had a 60‐bp deletion in the first K‐rich region, was sterile. In addition, we mutagenized worms by cleaving the sequence for the second K‐rich region between the PAZ and Piwi domains and repairing it with a degenerate oligonucleotide (Fig 1F). The progeny of mutagenized worms were subjected to RNAi by exposure to dsRNA targeting an essential gene. While most worms were growth‐arrested by RNAi, the csr‐1(fj150) strain, exhibiting the high incidence of males (Him) phenotype, was isolated as a viable RNAi‐deficient (Rde) mutant (Figs 1G and 8A, Appendix Fig S1A). Independent mutant isolates with similar changes in the second K‐rich region, csr‐1(fj160) and csr‐1(fj162), also exhibited the Rde and Him phenotypes.

Figure 8. Effect of CeRep55 multilocus deletions on homologous pairing and unpaired silencing, and models for the relationship between homologous pairing, unpaired silencing, and endo‐RNAi during meiosis.

Pachytene nuclei were analyzed.

• A
  The Him phenotypes of the csr‐1(fj150) and smc‐1(e879) mutants were enhanced by the CeRep55_X quadruple (Q: fj115, fj85, fj123, and fj120) deletions. The incidence of males in F1 progeny from self‐fertilized hermaphrodites was examined. The number of F1 progeny counted is indicated by n. Error bars represent 95% CIs.
• B
  Multicolor FISH using the 5S rDNA and LGX‐R probes in the csr‐1(fj150) mutant and the csr‐1(fj150); CeRep55_X(Q) mutant. Scale bar: 2 μm.
• C
  The nuclei of the csr‐1(fj150); CeRep55_X(Q) mutant showed an X‐chromosome pairing defect more frequently than those of the csr‐1(fj150) single mutant. Nuclei showing a pair of FISH signals ≥ 0.9 μm apart were interpreted to have unpaired chromosomes. The number of nuclei counted is indicated by n. Error bars represent 95% CIs. P‐value (two‐sided) was determined by two‐proportion z‐test.
• D
  Nuclei of male testes carrying an unpaired X chromosome were immunostained with an antibody against histone H3K9me2. H3K9me2‐positive chromosomes showing an extended state (arrow) were observed in the males of the CeRep55_X(Q) deletion mutant more frequently than in WT males. Scale bar (yellow): 2.5 μm.
• E
  In male pachytene nuclei, the relative frequency of extended chromosomes in H3K9me2‐positive chromosomes was examined. H3K9me2‐positive chromosomes ≥ 2.5 μm were interpreted to be in an extended state. The number of nuclei counted is indicated by n. Error bars represent 95% CIs. P‐value (two‐sided) was determined by two‐proportion z‐test.
• F–H
  Models for the heterochromatinization of unpaired chromosomes (F), homology recognition during SC formation between chromosome pairs (G), and the multifaceted roles of CSR‐1 in the cytoplasm, germ granules, and nuclei (H).

Source data are available online for this figure.

Both CSR‐1 and CSR‐2 are involved in unpaired silencing

During meiosis in C. elegans, unpaired DNA, such as the X chromosome in males, forms facultative heterochromatin that is rich in histone H3 lysine 9 dimethylation (H3K9me2) (Bean et al, 2004), as confirmed by simultaneous staining with fluorescence in situ hybridization (FISH) and immunofluorescence (Fig 2A). The csr‐1 null mutant exhibits an abnormality in unpaired silencing of meiotic chromosomes (She et al, 2009). We checked whether the csr‐2 mutant also shows any abnormality in unpaired silencing.

Figure 2. Two Slicer‐type Argonaute proteins are involved in meiotic silencing of unpaired chromosomes in a partially overlapping manner.

Pachytene nuclei were analyzed.

1. Nuclei of the WT strain were stained by the immuno‐FISH method using an anti‐histone H3K9me2 antibody and a probe against the right region of the X chromosome. Scale bar: 2 μm.
2. Shapes of H3K9me2‐positive chromosomes in male testes of the WT and the csr‐2 mutant. The relative frequency of extended chromosomes in H3K9me2‐positive chromosomes was examined. H3K9me2‐positive chromosomes ≥ 2.5 μm were interpreted to be in an extended state. The number of nuclei counted is indicated by n. Error bars represent 95% CIs.
3. Immunofluorescence images of histone H3K9me2 at the pachytene region of male testes in the WT and the Argonaute mutants. DNA was counterstained with DAPI. H3K9me2 signals were detected in the Argonaute mutants, but the H3K9me2 signal distributions were abnormal. In males of the csr‐2 mutant and the csr‐1 mutant, some chromosomes deposited with H3K9me2 signals were not condensed properly, as indicated by arrows. Scale bar: 2.5 μm.
4. Results of digital segmentation of H3K9me2 signals in nuclei of X0 males. The numbers and volumes of H3K9me2 segments are summarized. The volume distributions of the largest H3K9me2 segments are classified into three classes and presented in stacked bar charts. The number of nuclei counted is indicated by n. The error bars represent 95% CIs for the incidence of H3K9me2 segments larger than 1.6 μm³. The raw data are plotted in Appendix Fig S2A.
5. Immunofluorescence images of H3K9me2 and SYP‐1 in male nuclei of the WT and the Argonaute mutants. High‐resolution fluorescence images, 3D reconstruction images (3D CG), and maximum intensity projection (MIP) images of segmented H3K9me2 signals are presented. In the male nuclei of the csr‐1 mutant and the csr‐2; csr‐1 double mutant, some H3K9me2 segments showed abnormally large volumes (narrow arrow) and/or overlapped partially with SYP‐1 signals (arrowhead). In the double mutant, DAPI signals were often detected in an aggregated state (dashed line). Scale bar: 2.5 μm.
6. Immunofluorescence images of H3K9me2 and SYP‐1 in hermaphrodites. Tiny weak H3K9me2 signals were occasionally detected in the csr‐2; csr‐1 double mutant. Scale bar: 2.5 μm.

Source data are available online for this figure.

In pachytene nuclei of X0 male testes, heterochromatin stained by an anti‐H3K9me2 antibody is detected as a condensed structure corresponding to an unpaired X chromosome in WT, but some chromosomes with H3K9me2 signals were not appropriately condensed in the csr‐2(tm1637) mutant (Fig 2B and C).

Next, male testes of the WT and the Argonaute mutants were double‐stained with antibodies against H3K9me2 and SYP‐1, and the H3K9me2 signals were analyzed by digital segmentation (Fig 2D and E, and Appendix Fig S2A). SYP‐1 is a component of the SC central region (MacQueen et al, 2002). Similar to the previous study of a csr‐1 mutant (She et al, 2009), a large H3K9me2 segment and several extra H3K9me2 segments were detected per pachytene nucleus in the males of the csr‐1(fj54) mutant and the csr‐2(tm1637); csr‐1(fj54) double mutant. The large H3K9me2 segments in the WT and the csr‐2 mutant were smaller than 1.6 μm³ in volume, but 17% of the largest segments in the csr‐1 mutant and 23% of those in the csr‐2; csr‐1 double mutant were detected with volumes of greater than 1.6 μm³. The abnormally large H3K9me2 segment in the csr‐2; csr‐1 double‐mutant males might indicate that the heterochromatin formed on the monosomic X chromosome often spread to autosomes by irregular contact. The extra H3K9me2 segments in the double‐mutant males were generally detected at the aggregated chromosomal regions (Fig 2E, dashed line), implying that extra unpaired regions were often generated by a chain collision of chromosomes, which started with the monosomic X chromosome. In addition, we found that the csr‐2; csr‐1 double mutant was partially defective in the distinctive positioning of H3K9me2 and SYP‐1. All H3K9me2 segments in the WT males were detected on chromosomes lacking SYP‐1, while 29% of H3K9me2 segments in the csr‐2; csr‐1 double‐mutant males partially overlapped with SYP‐1 signals (Fig 2D).

The H3K9me2 distribution was examined also in the pachytene nuclei of XX worms lacking monosomic chromosomes. The H3K9me2 signal was not prominent in the hermaphrodites of the WT and the csr‐1 mutant, but tiny weak H3K9me2 signals were occasionally detected at the deformed chromosomal region in the csr‐2; csr‐1 mutant hermaphrodites (Fig 2F).

CSR‐1 and CSR‐2 may function to maintain physical distances between a monosomic chromosome and the chromosome pairs by participating in chromosome alignment and may also function to condense the monosomic chromosome during meiosis. Consistently, abnormalities in the mRNA expression ratio between an unpaired X chromosome versus a pair of autosomes were observed in X0 males of both the csr‐1 mutant and the csr‐2 mutant by semiquantitative RT‐PCR analysis (Fig EV1E).

Chromodomain protein CEC‐5 is enriched on unpaired chromosomes

To identify the components of heterochromatin formed on unpaired DNA, we immunopurified the heterochromatin complex using an antibody against H3K9me2 from a population of him‐8 mutants rich in X0 males. A protein with a chromodomain called CEC‐5 (C. elegans chromodomain protein) was identified in the H3K9me2‐rich complex using mass spectrometry (Fig EV2A). H3K9me2 was mutually detected in the immunoprecipitate with anti‐CEC‐5 antibodies by Western blot (Fig EV2B). The chromodomain is a conserved protein motif that recognizes H3K9 dimethylation (me2) or trimethylation (me3) (Bannister et al, 2001; Lachner et al, 2001), which are typically found in heterochromatin.

Figure EV2. Detection of the interaction between histone H3K9me2 and CEC‐5 by immunoprecipitation, and the regional annotation of the ChIP‐seq peaks for histone H3K9me2 and CEC‐5.

1. The chromatin fraction of the him‐8(tm611) mutant was subjected to immunoprecipitation with an anti‐histone H3K9me2 antibody or normal IgG. Proteins in the immunoprecipitates were analyzed using an LC/MS–MS spectrometer.
2. Histone H3K9me2 was co‐immunoprecipitated with CEC‐5 by anti‐CEC‐5 antibodies. Immunoprecipitates using antibodies against CEC‐5 or FLAG (control) were analyzed with SDS‐PAGE and Western blot.
3. Molecular phylogenetic tree of proteins with a chromodomain in C. elegans. A neighbor‐joining phylogenetic analysis was performed with the full‐length peptide sequence of CEC‐5 and with the chromodomain peptide sequences and its surrounding regions in other proteins. CEC‐5, CEC‐4, and CEC‐8 do not have a clear chromo shadow domain, but their C‐terminal regions still show a low degree of similarity to the chromo shadow domains in HPL‐1 and HPL‐2. Among the chromodomain proteins in C. elegans, HPL‐1 and HPL‐2 are phylogenetically closest to HP1 in Drosophila (James & Elgin, 1986) and Swi6 in fission yeast (Lorentz et al, 1994). HPL‐1 and HPL‐2 are important for fertility in C. elegans (Couteau et al, 2002).
4. Structures and mutations of the cec‐5, cec‐4, and cec‐8 genes. The cec‐4 and cec‐5 genes are located next to each other on autosome IV. Exon regions encoding the chromodomain are shown in dark gray. The locations of deletion mutations are indicated by black lines. The regions of recombinant peptides used for in vitro pull‐down assays are indicated by green lines.
5. ChIP intensity for H3K9me2 in major chromosomal regions. The experiments were performed with the WT, him‐8(tm611), and met‐2(ok2307) set‐25(n5021) strains. H3K9me2 is known to be enriched in repeat sequences (Gerstein et al, 2010; reviewed in Ahringer & Gasser, 2018).
6. ChIP intensity for CEC‐5 in major chromosomal regions. The amount of CEC‐5 on chromosomes was reduced in the met‐2 set‐25 mutant.
7. ChIP intensity for H3K9me2 in the typical repeats (CeRep55, Helitron2, LONGPAL3, CELE45, and Tc1) and the X cluster.
8. For H3K9me2 and CEC‐5, the accumulated length of ChIP peaks was compared between the X chromosome and the autosomes. The difference between the WT and the him‐8 mutant is likely to reflect the enrichment of H3K9me2 and CEC‐5 on unpaired X chromosomes in the him‐8 population.

Data information: (E–G) The data from technical duplicates (#1 and #2) are individually presented, and the peak intensities are presented with base counts that were scaled against 1 Mb of input DNA by the spike‐in normalization. Error bars represent 95% CIs for ChIP intensities in each library, which were estimated by down‐converting the base counts for the peak intensities to comparable read numbers.

Source data are available online for this figure.

CEC‐5 is phylogenetically similar to CEC‐4 and CEC‐8 of the chromodomain protein family in C. elegans (Fig EV2C). According to previous studies, a cec‐4 single mutant is deficient in perinuclear anchoring of repeat‐rich heterochromatin in somatic cells, and CEC‐4 binds to H3K9me2 and H3K9me3 (Gonzalez‐Sandoval et al, 2015; Cabianca et al, 2019). Our in vitro binding assays showed that the recombinant chromodomains of CEC‐5 and CEC‐8 bound weakly to the unmethylated H3 peptide, strongly to the H3K9me3 peptide, and subtly more strongly to the H3K9me2 peptide (Fig 3A). In C. elegans, H3K9me3 correlates with constitutive heterochromatin, and histone H3K9me2 correlates with facultative heterochromatin formed by unpaired silencing (Bean et al, 2004; Bessler et al, 2010).

Figure 3. Heterochromatinization of the unpaired X chromosome in the male germline is impaired in some mutants corresponding to a new class of chromodomain proteins and components of a meiotic cohesin complex.

Immunofluorescence experiments were performed on the pachytene nuclei of X0 males.

1. The affinities of the recombinant chromodomain peptides of CEC‐5 and CEC‐8 for the N‐terminus of histone H3 were assayed by in vitro pull‐down experiments. CEC‐4 served as a control.
2. CEC‐5 and H3K9me2 signals in male pachytene nuclei of the WT and the met‐2 set‐25 double mutant. CEC‐5 was drastically enriched on the H3K9me2‐positive unpaired X chromosome in the WT males. Scale bar: 2 μm.
3. H3K9me2 signals at the pachytene region in the males of cec mutants and cohesin mutants. Worms were cultured at 20°C, except for the smc‐1(e870) mutant that was cultured at 23°C. The intensity of the H3K9me2 signal in the him‐8 mutant male was similar to that in the WT male. Scale bar: 5 μm.
4. The nuclear intensities of H3K9me2 signals were measured from wide‐field (WF) images. The image quantification was performed as described in Fig  EV3B.
5. The nuclear intensities of H3K9me2 and H3K9me3 signals were measured from deconvolved (DV) images. The H3K9me2 signals in the major strains were double‐checked using deconvolution microscopy. P‐values were determined by Mann–Whitney U‐test.

Data information: (D, E) The thick lines in the box plot indicate the median values. The length of the box represents the IQR covering the middle 50% of the data points. The ends of the whiskers represent the farthest data points within 1.5 times the IQR. The number of nuclei counted is indicated by n.

Source data are available online for this figure.

We examined the localization of CEC‐5 in pachytene nuclei of X0 males and a strain carrying a free‐type duplication sDp2(I;f) via immunofluorescence. In the germline, the CEC‐5 staining signal was detected abundantly in the H3K9me2 staining regions corresponding to the unpaired X chromosome and sDp2, less abundantly in the other paired chromosomes, and hardly at the nonchromosomal region (Fig 3B and Appendix Fig S3B). CEC‐5 was also detected in somatic nuclei (Appendix Fig S3E). In the pachytene nuclei of a met‐2 set‐25 double mutant lacking K9 di‐ and trimethylation of histone H3 (Towbin et al, 2012), we detected CEC‐5 on chromosomes and ectopically in nonchromosomal regions (Figs 3B and EV3C).

Figure EV3. H3K9me3 signals correlating with constitutive heterochromatin were almost normally detected in a strain carrying the cec‐8; cec‐4 cec‐5 mutations.

1. H3K9me3 signals at the pachytene region in the males of the WT, him‐8 mutant, cec‐8; cec‐4 cec‐5 him‐8 mutant, and coh‐3/4 double mutant. For quantified intensities, see also Fig 3E. Scale bar: 5 μm.
2. Strategy for quantification of fluorescent intensities in nuclei. As an example of the quantification, the image data from the WT in panel A were processed and displayed at the same scale.
3. CEC‐5 and H3K9me3 signals in hermaphrodite pachytene nuclei of the WT and the met‐2 set‐25 double mutant. Scale bar: 2 μm.

Source data are available online for this figure.

To determine whether cec‐5 and the two similar genes are involved in meiotic silencing of unpaired X chromosomes in males, we performed immunofluorescence staining and image quantification for histone H3K9me2 in single mutants for cec‐4, cec‐5 and cec‐8, a cec‐4 cec‐5 double mutant, and a cec‐8; cec‐4 cec‐5 triple mutant (Figs 3C–E and EV2D, and Appendix Figs S2B and S3A). Assays for the double and triple mutations were performed in the presence of a him‐8 mutation to produce X0 males without mating because the cec‐4 cec‐5 double mutant and the cec‐8; cec‐4 cec‐5 triple mutant exhibited partial sterility, reduced mating rates, and no significant defects in chromosome segregation. The cec‐5(tm6207), cec‐4(ok3124), and cec‐8(fj63) single mutants showed intense, almost normal H3K9me2 signals in the nuclei of male testes. However, compared with the H3K9me2 signals thought to correspond with unpaired X chromosomes in the him‐8(e1489) single mutant, the H3K9me2 signals in male testes were considerably diminished in the cec‐4(ok3124) cec‐5(fj61) him‐8(e1489) mutant and further diminished in the cec‐8(fj63); cec‐4(ok3124) cec‐5(fj61) him‐8(e1489) mutant.

It should be noted that levels of H3K9me3 correlating with constitutive heterochromatin were similar between the males of the him‐8 single mutant and the cec‐8; cec‐4 cec‐5 him‐8 mutant (Figs 3E and EV3A and B).

These data suggest that the chromodomain proteins CEC‐5, CEC‐4, and CEC‐8 function redundantly in unpaired silencing, most likely as heterochromatin components.

A form of meiotic cohesin is required for unpaired silencing

We next examined some chromatin protein mutants to identify additional factors required for unpaired silencing. We found that the H3K9me2 signals were decreased in the pachytene nuclei of him‐1(e879) mutant males cultured above 22°C (Fig 3C–E). Furthermore, the him‐1 mutant showed activated transposition of DNA transposons (Table 1), which has been observed in some RNAi mutants (Ketting et al, 1999; Tabara et al, 1999). The him‐1 mutant population has a high incidence of males owing to a chromosome segregation abnormality (Hodgkin et al, 1979). The him‐1 gene encodes SMC‐1, a constitutive component of cohesin, and e879 is a temperature‐sensitive mutation located in an ATPase domain of SMC‐1 (Chan et al, 2003). Hereafter, we use smc‐1 to denote him‐1.

Table 1.

Tc1 transposon mobilization is de‐silenced in the smc‐1(e879) mutant.

Genotype	unc‐22(st136::Tc1) reversion frequency
	%	(95% CI)	Sample size
Wild‐type	0	(0–0.07)	5,431
csr‐2(tm1637)	0	(0–0.07)	5,624
rde‐3(ne298) [aka mut‐2]	0.41	(0.24–0.65)	4,153
smc‐1(e879) [aka him‐1]	1.24	(0.90–1.67)	3,539
smc‐3(t2553)	0	(0–0.14)	2,713

The unc‐22(st136::Tc1) mutation, which is a transposon insertion into a muscle gene, was introduced into a background of cohesin or RNAi mutants. Tc1 transposon mobilization was monitored as the recovery of movement, which resulted from the repair of the unc‐22 gene by Tc1 excision. The rde‐3(ne298) mutant shows a Him phenotype as well as an RNAi‐deficient phenotype. him‐1(e879) is a temperature‐sensitive mutation (Hodgkin et al, 1979) located at a SMC‐1 ATPase domain that interacts with kleisins (Chan et al, 2003). smc‐3(t2553) is a different temperature‐sensitive mutation located outside the ATPase domains (Baudrimont et al, 2011). This transposon excision assay is not feasible for sterile mutants such as the coh‐3/4 double mutant and the csr‐1 null mutant. CI, confidence interval.

Cohesins are involved in sister‐chromatid cohesion after DNA replication, chromatin boundary formation, and chromatin structure control (Michaelis et al, 1997; reviewed in Nasmyth & Haering, 2009). The binding of different kleisins to the ATPase domains of SMC‐1 and SMC‐3 heterodimers forms functionally differentiated cohesin complexes. Kleisins identified to date in C. elegans include constitutively expressed SCC‐1, somatic COH‐1, meiosis‐specific REC‐8 (Pasierbek et al, 2001) to facilitate sister‐chromatid cohesion, and meiosis‐specific COH‐3/4 (Severson et al, 2009; Severson & Meyer, 2014). Several regulators of mammalian cohesin have RNA‐binding domains associated with lncRNAs (Cho et al, 2018; Hansen et al, 2019; Saldaña‐Meyer et al, 2019).

We analyzed C. elegans kleisin mutants and found that pachytene nuclei of the coh‐4(tm1857) coh‐3(gk112) double‐mutant male did not show a clear facultative histone H3K9me2 signal (Fig 3C–E and Appendix Fig S2B). In contrast, the testes of the rec‐8(ok978) mutant male showed an almost normal level of H3K9me2 signals (Appendix Fig S2C). The histone H3K9me3 signal correlating with constitutive heterochromatin was almost normally observed in the smc‐1 mutant and the coh‐3/4 double mutant (Figs 3E and EV3A). Based on these data, we suggest that a meiosis‐specific cohesin complex containing SMC‐1 and COH‐3/4 plays an important role in the facultative heterochromatinization of unpaired chromosomes.

COH‐3/4 is present on unpaired and paired chromosomes

To examine the presence of COH‐3/4 kleisin(s) on unpaired chromosomes, we raised antibodies against COH‐3/4 and then immunostained pachytene nuclei of X0 males and hermaphrodites carrying the free‐type duplication sDp2(I;f). The H3K9me2 staining signal overlapped widely with DAPI staining in the unpaired X chromosome and sDp2 (Fig 4A and Appendix Fig S3C). In contrast, the COH‐3/4 staining signal was detected in the central portion of the unpaired X chromosome and sDp2. For comparison, we examined the localization of SYP‐1 and found that the signal was absent on the unpaired X chromosome (Fig 2E) and was very faint on sDp2 (Appendix Fig S3D).

Figure 4. Localization patterns of COH‐3/4 and SYP‐1 during meiosis and their abnormality in the csr‐2; csr‐1 Argonaute double mutant and the ego‐1 mutant.

Pachytene nuclei were analyzed by immunofluorescence. Scale bar: 2 μm.

1. The meiosis‐specific kleisin(s) COH‐3/4 was detected on paired chromosomes and the heterochromatinized unpaired X chromosome. The testes of WT males were immunostained with anti‐COH‐3/4 antibodies and an anti‐H3K9me2 antibody.
2. Immunofluorescence images of COH‐3/4 and SYP‐1 on paired chromosomes in the gonads of WT hermaphrodites. COH‐3/4 kleisin(s) was detected laterally to the string‐like signal of SYP‐1, which corresponds to the central region of the SC.
3. SC formation in the WT and multiple mutants were analyzed by immunostaining against SYP‐1 and COH‐3/4. The SC formed string‐like structures in the csr‐2; csr‐1 double mutant and the ego‐1 mutant, but the superstructure of the SC was often abnormally branched.
4. To analyze the shape of the SC, hermaphrodite nuclei were stained with antibodies against SYP‐1 and HIM‐8. Their images were segmented and then converted to 3D reconstruction images.
5. The incidence of branching states in SYP‐1 signals adjacent to HIM‐8 was examined in the WT and mutant hermaphrodites. The number of nuclei counted is indicated by n. Error bars represent 95% CIs.

Source data are available online for this figure.

The meiotic cohesin containing COH‐3/4 in C. elegans (Köhler et al, 2017), and some forms of meiotic cohesin in Drosophila (Manheim & McKim, 2003) and mice (Herrán et al, 2011; Ishiguro et al, 2011; Lee & Hirano, 2011), are known components of the axial elements, which correspond to lateral elements of the SC (reviewed in Page & Hawley, 2004). The SC is composed of layers of strings, with two lateral elements sandwiching a central region. In C. elegans, the central region consists of a coiled‐coil protein called SYP‐1 (MacQueen et al, 2002) and other SYP proteins. COH‐3/4 have been detected as parallel lines of staining on paired chromosomes by a super‐resolution microscope (Köhler et al, 2017). Apart from unpaired silencing, we investigated the role of COH‐3/4 in homologous pairing.

First, we compared the distributions of COH‐3/4 and SYP‐1 in the hermaphrodites of several strains (Figs 4B and C, and EV4A). In the WT nuclei, the COH‐3/4 signals were often observed on both sides of the SYP‐1 signals. Meanwhile, the nuclei of the coh‐4(tm1857) coh‐3(gk112) double mutant showed severely fragmented SYP‐1 signal localization, as reported previously (Severson et al, 2009). The nuclei of the syp‐1(me17) mutant showed a wavy and less uniform COH‐3/4 signal localization.

Figure EV4. Additional immunofluorescence images of SYP‐1, COH‐3/4, and HIM‐8 in mutant strains, and the immunofluorescence images of epitope‐tagged CSR‐2 and EGO‐1.

Images of hermaphrodite pachytene nuclei are presented, except for an image on the right side of panel C that presents the male nuclei. DNA was counterstained with DAPI.

1. SYP‐1 and COH‐3/4 distribution in the nuclei of the WT strain, csr‐2(tm1637) mutant, csr‐1(fj54) mutant, and csr‐1(fj126) mutant. In the csr‐1 mutants, the string‐like superstructure of the SC was often branched, and some meiotic chromosomes were abnormally aggregated. Scale bar: 2 μm. See also Fig 4C.
2. The coh‐3/4 mutant and the csr‐2; csr‐1 double mutant showed an almost normal distribution of HIM‐8. Scale bar: 2 μm.
3. Pachytene regions in the WT strain and the csr‐1 null mutant were stained using anti‐CSR‐1 antibodies. The csr‐1(fj54) mutant served as background control. n, intranuclear region; c, cytoplasm. Scale bar: 5 μm.
4. Distributions of epitope‐tagged CSR‐2 and EGO‐1 in the pachytene regions. The 2 × HA::CSR‐2 was expressed from a single‐copy transgene under the rpl‐21 promoter. The PA::EGO‐1 was expressed from the edited endogenous ego‐1 loci. Scale bar: 5 μm.
5. SYP‐1 and HIM‐8 distributions in nuclei of the csr‐1(fj150) mutant and the csr‐1(fj150); CeRep55_X(Q) mutant are shown with 3D reconstruction images. Scale bar: 2 μm.

Source data are available online for this figure.

Next, we assessed the localization of HIM‐8, which is essential for accurate initiation of X‐chromosome homologous pairing. In pachytene nuclei of WT hermaphrodites, the HIM‐8 staining signal bound to the PC was detected as a single dot or a pair of closely adjacent dots. Likewise, the HIM‐8 signal was detected as a single dot in nuclei of coh‐3/4 double‐mutant hermaphrodites (Fig EV4B), suggesting that homolog pairing at the PC was normal.

The csr‐2; csr‐1 double mutant also shows a defect in homologous pairing

The meiotic cohesin containing COH‐3/4 is involved in unpaired silencing and homolog pairing, while the Argonaute proteins CSR‐1 and CSR‐2 are involved in RNAi and unpaired silencing. Thus, we next examined whether the csr‐2; csr‐1 double mutant and an RdRP mutant also showed defects in homolog pairing. Normal staining for HIM‐8 was detected as a single dot in pachytene nuclei of the csr‐2; csr‐1 double mutant (Fig EV4B), suggesting that the cohesion of sister chromatids was mostly normal, at least for the X chromosome. The SC components SYP‐1 and COH‐3/4 were also detected as string‐like structures, but we noticed that the superstructure of the SC was often abnormally branched in nuclei of the csr‐2(tm1637); csr‐1(fj54) double mutant, ego‐1(om58) mutant, and several csr‐1 mutants (Figs 4C and EV4A). Likewise, DAPI signals showed that some meiotic chromosomes were partially aggregated in these mutants.

The gonads of the Argonaute mutants were stained with antibodies against SYP‐1 and the X‐chromosome binding protein HIM‐8, and then the superstructure of the SC was analyzed by computer graphics (CG) reconstruction of 3D microscopy images (Fig 4D). As for SYP‐1 signals adjacent to HIM‐8 signals, none of the SYP‐1 signals were branched in the WT, whereas the SYP‐1 signals were branched in 56% of the csr‐1 null‐mutant nuclei and 64% of the csr‐2; csr‐1 double‐mutant nuclei (Fig 4E).

Using chromosome FISH, we further examined the defects in homolog pairing in mutants of the kleisin, Argonaute, and RdRP genes (Fig 5A–D). The 5S rRNA locus (rDNA) of autosome V and a fosmid derived from the right region of the X chromosome (X‐R) were used as the probes (Fig 5B). Both probes detect regions distant from the PCs. In WT gonads, the 5S rDNA FISH signal was detected as a single dot or a pair of closely adjacent dots per pachytene nucleus because homologous chromosomes are well‐paired. In the coh‐4(tm1857) coh‐3(gk112) double mutant, 69% of the nuclei had two distinct 5S rDNA dots (≥ 0.9 μm apart), indicating that the 5S rDNA loci in the homologous V chromosomes were frequently unpaired. Interestingly, 24% of nuclei in the csr‐1(fj54) mutant had distinct 5S rDNA dots, while the csr‐2(tm1637) mutant did not show a significant defect in pairing of 5S rDNA loci. Notably, in the csr‐2(tm1637); csr‐1(fj54) double mutant, 43% of nuclei had two distinct 5S rDNA dots (Fig 5C, green bars). Thus, the csr‐2 mutation enhanced the pairing defect caused by the csr‐1 mutation. Similarly, 26% of nuclei in the ego‐1 mutant had separated 5S rDNA loci. Severe defects in X‐chromosome pairing were also observed in the coh‐3/4 double mutant and the csr‐2; csr‐1 double mutant (Fig 5C, magenta bars).

Figure 5. The coh‐3/4 double mutant and the csr‐2; csr‐1 double mutant show severe homolog pairing defects.

Pachytene nuclei in hermaphrodites were analyzed.

1. Multicolor FISH for the 5S rRNA region (rDNA, green) of autosome V and the right region of the X chromosome (LGX‐R, magenta) in nuclei of the WT strain, coh‐3/4 double mutant, csr‐1 mutants, csr‐2; csr‐1 double mutant, and ego‐1 mutant. Arrows indicate occasional colocalizations of the 5S rDNA and X‐R loci detected in the mutant strains. Scale bar: 2 μm.
2. Locations of pairing centers (Phillips et al, 2005, 2009) and probes used for FISH.
3. Statistical data for unpaired states of homologous chromosomes examined by FISH using the 5S rDNA and LGX‐R probes. Nuclei showing a pair of FISH signals ≥ 0.9 μm apart were interpreted to have unpaired chromosomes. The number of nuclei counted is indicated by n. Error bars represent 95% CIs. P‐values (two‐sided) were determined by two‐proportion z‐test.
4. Simultaneous staining of the 5S rRNA region (rDNA, green) by FISH, and SYP‐1 (magenta) by immunofluorescence. Scale bar: 2 μm.
5. csr‐1(fj67) and csr‐1(fj126) in‐frame mutations reduced the abundance of CSR‐1 at the chromosomal regions in pachytene nuclei. The WT strain and the csr‐1 mutants were stained by anti‐CSR‐1 antibodies. Scale bar: 5 μm.

Source data are available online for this figure.

We then analyzed the distribution of CSR‐1 in the pachytene region of the gonads in the WT strain and several csr‐1 mutants using immunofluorescence (Figs 5E and EV4C). In the WT, the CSR‐1 signal was detected strongly at germ granules outside the nuclei, and moderately at the nuclei and other cytoplasmic regions, as reported previously (Claycomb et al, 2009). In the csr‐1(fj67) mutant caused by deletion of the first K‐rich region, CSR‐1 aggregated abnormally at the nonchromosomal region in the nuclei and did not show enrichment at the germ granules. In the csr‐1(fj126) mutant caused by NES insertion, the nuclear abundance of CSR‐1 was substantially reduced (Fig 5E). In FISH analysis, a considerable percentage of nuclei exhibited two distinct 5S rDNA dots in the csr‐1(fj67) and csr‐1(fj126) mutants (19% and 12%, respectively) (Fig 5C). These findings suggest that CSR‐1 in the chromosomal region is important for the integrity of meiotic chromosomes. In the csr‐1(fj70) mutant with substitution of a catalytic residue in the Piwi domain, 15% of nuclei had two distinct 5S rDNA dots.

In addition, we analyzed the distributions of CSR‐2 and EGO‐1. Epitope‐tagged CSR‐2 and EGO‐1 were detected at the nuclei and the cytoplasm without showing strong enrichment at the germ granules (Fig EV4D), suggesting that germ granules might be less important sites in chromosome control by CSR‐2 and EGO‐1.

Next, two different chromosomes were simultaneously analyzed using multicolor FISH (Fig 5A). In the pachytene nuclei, the 5S rDNA loci of chromosome V and the X‐R loci of the X chromosome were detected separately in the WT. However, these loci were occasionally detected as almost‐overlapping dots (<0.4 μm apart) on the same DAPI‐stained structure in the kleisin double mutant (3/164 nuclei) and the Argonaute double mutant (3/161 nuclei) (Fig 5A, arrows), indicating that chromosomes V and X in some nuclei are occasionally paired without accurate location information in these double‐mutant strains. The colocalization of different loci was not observed in the WT (0/211 nuclei).

These findings suggest that the Argonaute proteins CSR‐1 and CSR‐2 support the formation of an SC between homologous chromosomes with accurate homology through a PC‐independent mechanism.

CSR‐1 and COH‐3/4 are associated with nonsimple repeats of chromosomes and weakly with coding genes

In the regulation of meiotic chromosomes, CSR‐1 and the cohesin containing COH‐3/4 appeared to be involved in both homologous pairing and unpaired silencing, whereas H3K9me2 and CEC‐5 were associated with unpaired silencing. To compare the chromosomal distributions of these protein factors, chromatin immunoprecipitation followed by high‐throughput DNA sequencing (ChIP‐seq) was conducted with the crude chromatin fractions from adult populations (Figs 6 and EV5). In the presence of calibrator chromatin, we performed ChIP reactions of CSR‐1, SMC‐1, COH‐3/4, H3K9me2, CEC‐5, and mock controls from WT. The ChIP reactions for H3K9me2 and CEC‐5 were also performed from the him‐8 mutant rich in X0 males and the met‐2 set‐25 mutant. Using the theory of Model‐based Analysis of ChIP‐Seq data (MACS) (Zhang et al, 2008) with our idea to minimize the peak calling background (Fig EV5A), the peaks of the ChIP reads were calculated.

Figure 6. ChIP‐seq experiments showing that CSR‐1 and COH‐3/4 are associated with non‐simple repeats expressing small RNAs.

• A, B
  ChIP‐seq data for CSR‐1, SMC‐1, COH‐3/4, and histone H3K9me2 were compared with subpopulations in CSR‐1‐interacting small RNAs. Adult worms were used for experiments. ChIP signals on autosome III (A) and the X chromosome (B) from the WT hermaphrodites were visualized by aggregation plots with a bin size of 20 kbp. H3K9me2 ChIP signals derived from the him‐8 mutant population rich in X0 males are shown in the lower part. The common ChIP regions computed from the technical duplicates were used for panels A and B. In the adult body, about half of the nuclei correspond to germ cells. Most H3K9me2 ChIP signals from the WT hermaphrodites were likely to originate in somatic cells because staining signals of H3K9me2 in adult hermaphrodites were detected abundantly in somatic cells and very weakly in the germline. Most meiosis‐specific COH‐3/4 ChIP signals were thought to originate in the germline. For comparison, aggregation plots were also constructed from the following sub‐populations of CSR‐1‐interacting small RNAs: small RNAs mapped in sense orientation to coding genes (exons and introns), and repeat‐derived small RNAs. The small RNA data merged from the technical duplicates were used for panels A and B. Densities of protein‐coding exons, all non‐simple repeats, and five major repeats on the reference genome are attached. The CeRep55 sequences were partially degenerate, and no clear ChIP signal for CSR‐1 was detected in CeRep55 cluster #4 on the X chromosome in the ce10 reference genome.
• C
  ChIP intensities over genomic features. The proportional length of genomic features in the reference genome are presented for comparison. The tRNA_etc stands for tRNA, snRNA, snlRNA, and snoRNA. The dashed line indicates the proportion of repeats relative to the reference genome.
• D–F
  ChIP intensities for CSR‐1 (D), SMC‐1 (E), and COH‐3/4 (F). Using the information from the spike‐in calibrator chromatin, normalization across multiple ChIP‐seq data was carried out as described in Fig  EV5A. Mock indicates ChIP reactions with an anti‐FLAG antibody. Repeats (> 10 nt) stand for all non‐simple repeats. Regions of CeRep55, Helitron2, LONGPAL3, CELE45, and Tc1 served as major repeats. The ChIP signal for CSR‐1 was weak in the X cluster. The interaction between CSR‐1 and X‐cluster 22G‐RNAs may not be a nuclear event. The peak intensities are presented with base counts that were scaled against 1 Mb of input DNA by the spike‐in normalization. The data from technical duplicates (#1 and #2) are individually presented. Error bars represent 95% CIs for ChIP intensities in each library, which were estimated by down‐converting the base counts for the peak intensities to comparable read numbers.
• G
  Frequencies of small RNAs expressed from the major non‐simple repeats were counted in the CSR‐1 and CSR‐2 complexes and the input cell lysates. The small RNA reads obtained by the experiments in Fig 1D were analyzed. The data from technical duplicates (#1 and #2) are individually presented. The y‐axis indicates reads per million (rpm). Error bars represent 95% CIs. The red border indicates the expected value of non‐specific recovery, which was predicted with the recovery rate of miRNAs.
• H
  ChIP intensity proportion of exon subsets relative to all exons. Within the exon regions, ChIP signals for CSR‐1 were detected with a bias towards the exons expressing sense small RNAs in the CSR‐1 or CSR‐2 complex and the exons expressing antisense capped nuclear RNAs. Six patterns of exon subsets were prepared based on RNA expression features. See also Fig  EV1G.

Source data are available online for this figure.

Figure EV5. Distributions of ChIP‐seq signals and the ChIP‐seq data processing strategy using spike‐in calibrators.

1. The ChIP‐seq experiments with paired‐end sequencing were performed with the spike‐in normalization strategy (Egan et al, 2016). The peak calling on the ChIP‐seq data was performed based on the MACS theory (Zhang et al, 2008) with several modifications. Abs stand for antibodies.
2. ChIP signal intensities for CSR‐1 were compared between duplicate libraries using genome‐wide scatterplot. Correlation between the two datasets was assessed with the Spearman's rank correlation coefficient (SRCC). In addition, the ChIP signals corresponding to the five major repeats were plotted.
3. ChIP intensity proportion of repeat subsets relative to all repeats. Within the repeat regions, ChIP signals for CSR‐1, SMC‐1, and COH‐3/4 were detected with a bias towards both repeats with homologous small RNAs in the CSR‐1 or CSR‐2 complex and repeats with homologous capped nuclear RNAs. Four patterns of repeat subsets were prepared based on the RNA expression features.
4. Frequencies of the major 22G‐RNAs from CeRep55 were counted in the CSR‐1 and CSR‐2 complexes and the input cell lysates. The sequences of RNA species (i, ii, and iii) are shown in Fig 7A. The data from technical duplicates (#1 and #2) are individually presented. The y‐axis indicates reads per million (rpm). The total number of small RNA reads mapped to C. elegans sequences is indicated by n. Error bars represent the 95% CI for RNA abundances in each library.

Source data are available online for this figure.

We observed partial similarities among the chromosomal distributions of CSR‐1, SMC‐1, and COH‐3 with aggregation plots of ChIP‐seq signals (Fig 6A and B). ChIP signals for H3K9me2 also overlapped partially with those for CSR‐1 and the cohesin components, but fluctuations in H3K9me2 ChIP intensity every 20 kbp were milder than those in the ChIP intensities for CSR‐1 and the cohesin components, probably indicating that H3K9me2 was more broadly distributed. ChIP signals for CSR‐1, SMC‐1, COH‐3/4, and H3K9me2 were rich in the distal regions of autosomes and were also detected on X chromosomes without the distal preference. Compared with the chromosomal distribution of H3K9me2 ChIP signals in the WT hermaphrodites, H3K9me2 ChIP peaks were more frequently detected on the X chromosome in the him‐8(tm611) male‐rich population. This difference in H3K9me2 signals (Figs 6B and EV2H) was consistent with meiotic silencing of unpaired X chromosomes.

We examined ChIP intensities over genomic features (Fig 6C). Compared with the proportion of repeats (Fig 6C, dashed line) relative to the reference genome, ChIP intensities for CSR‐1, SMC‐1, COH‐3/4, H3K9me2, and CEC‐5 in repeats relative to the whole genome were more intensely detected, suggesting that these chromatin complexes were present densely in repeats. In particular, ChIP signals for CEC‐5 were detected principally in repeats and much less in coding exons, probably because CEC‐5 differs from other factors in terms of its association with H3K9me3 as well as H3K9me2. Meanwhile, ChIP intensities for CSR‐1, SMC‐1, COH‐3/4, and H3K9me2 were also observed in coding exons at proportions similar to or somewhat lower than the proportion of exons in the reference, suggesting the temporal presence of these chromatin complexes even in the exon regions.

The repeat‐associated ChIP peaks for CSR‐1, SMC‐1, COH‐3/4, H3K9me2, and CEC‐5 were mainly located in the nonsimple repeats longer than 10 nt (Figs 6D–F and EV2E–G). Among the nonsimple repeats, the ChIP signals were quite abundantly detected at the minisatellite tandem repeat CeRep55 and the non‐LTR DNA transposon Helitron2 (Figs 6D–F, EV2G, and EV5B). The tandem location may be beneficial to increase the concentration of interacting protein per locus. Some Helitrons carry minisatellite repeats internally (Kapitonov & Jurka, 2001). Weak ChIP signals were also detected at LTR DNA transposons such as LONGPAL3 and Tc1 and at short interspersed nuclear elements (SINEs) such as CELE45.

We focused on CSR‐1 and investigated its relationship with repeat‐derived small RNAs. The above five repeats showed small RNA expression. Compared with the frequency in the input lysates, the CeRep55 small RNAs were more frequently detected in both the CSR‐1 and CSR‐2 complexes, and the CELE45 small RNAs were more frequently detected in the CSR‐1 complex (Fig 6G and Appendix Fig S1B). The small RNA populations expressed from the other three repeats were counted less frequently in this simple comparison. However, compared with the recovery rate of miRNAs (Fig 6G, red border), the small RNA populations of Helitron2, LONGPAL3, and Tc1 were more efficiently recovered into the CSR‐1 complex, implying that CSR‐1 selectively interacts with a sub‐population of small RNAs generated from multiple regions within these transposons. We observed a higher frequency of some major RNA sequences derived from Helitron2, LONGPAL3, and Tc1 in the CSR‐1 complex compared with the input lysate (Appendix Fig S1B). For a genome‐wide perspective, we selected repeats expressing small RNAs and examined the CSR‐1 ChIP intensities in the repeat subsets. The repeat subset with homologous small RNAs present in the CSR‐1 or CSR‐2 complex was responsible for 79% of the CSR‐1 ChIP intensities in repeats, while the same repeat subset accounted for only 63% of the accumulative length of repeats (Fig EV5C).

In addition, we investigated what kinds of exons tended to have ChIP signals for CSR‐1, SMC‐1, and COH‐3/4. Of the exon‐derived small RNAs interacting with CSR‐1, 7.5% were located at the ChIP regions for CSR‐1, showing sense orientation along exons with a frequency of 8.3% (Fig EV1F). We focused on exceptional exons that expressed sense small RNAs interacting with CSR‐1 or CSR‐2, and then compared them with the ChIP data, because we predicted that the exon‐derived sense small RNAs might be located mainly in nuclei. The exon subset expressing sense small RNAs present in the CSR‐1 or CSR‐2 complex was responsible for 23% of the CSR‐1 ChIP intensities, 30% of the SMC‐1 ChIP intensities, and 35% of the COH‐3/4 ChIP intensities in exons, while the same exon subset accounted for 14% of the accumulative length of exons (Figs 6H and EV1G, magenta bars). In C. elegans, more than 1,000 promoters produce long antisense capped transcripts from protein‐coding genes (Jänes et al, 2018a). We next compared the exons expressing the antisense capped nuclear RNAs with the ChIP data. The ChIP signals for CSR‐1, SMC‐1, and COH‐3/4 were also detected with a bias towards the exons expressing antisense capped nuclear RNAs. Meanwhile, the ChIP intensities for the three proteins appeared not to be biased toward introns expressing CSR‐1‐interacting sense small RNAs, possibly because the ChIP signals were less abundant in introns. CSR‐1 and cohesin might be associated preferentially with chromosomal exon regions that produce long antisense transcripts, which serve as templates for synthesizing sense small RNAs in nuclei. We looked for typical cases and found that the vit‐6 gene expressed exon‐derived sense small RNAs, antisense capped nuclear RNAs, and sense mRNAs, and moreover had ChIP signals for CSR‐1, SMC‐1, and COH‐3/4 (Fig EV1H).

Motif analysis revealed that CeRep55 repeat sequences were the most prevalent motifs found in common among the ChIP peaks for CSR‐1, SMC‐1, COH‐3/4, and H3K9me2 (Fig 7A).

Figure 7. Chromosome‐associated lncRNAs are expressed from CeRep55 repeats that are common targets of CSR‐1, SMC‐1, COH‐3/4, and histone H3K9me2.

1. Sequences of the minisatellite repeat CeRep55 were found as common motifs, frequently present at regions where the ChIP‐seq peaks for CSR‐1, SMC‐1, COH‐3/4, and H3K9me2 were colocalized. Several sequences of 22G‐RNAs interacting with CSR‐1 and CSR‐2 were expressed from the CeRep55 repeats. F and R indicate the direction of probes used for northern and in situ hybridizations.
2. Four major clusters of CeRep55 tandem repeats are present on the X chromosome. A quadruple‐deletion strain lacking these four clusters was generated.
3. Small RNAs expressed from CeRep55 repeats were detected by northern hybridization. Total RNAs from adult worms were electrophoresed on a 10% sequencing gel.
4. lncRNAs expressed from CeRep55 repeats were detected by northern hybridization. RNAs from nuclei of mixed‐stage worms were electrophoresed on a 1% denaturing agarose gel.
5. CeRep55 lncRNAs are associated with chromosomes. Pachytene nuclei were stained by RNA‐directed in situ hybridization with the F probe against CeRep55 and by immunofluorescence with anti‐SYP‐1 antibodies. Scale bar: 2 μm.
6. The CeRep55 lncRNA signals partially colocalized with those of COH‐3/4 in hermaphrodites. Pachytene nuclei in the WT strain and the csr‐2; csr‐1 double mutant were stained. Scale bar: 2 μm.
7. The number and shape of the CeRep55 lncRNA signals in pachytene nuclei were different between the WT strain and the csr‐2; csr‐1 double mutant. The hybridization signals in 20 nuclei were analyzed in the hermaphrodites. Data presented as mean ± SD.
8. The CeRep55 lncRNA signal was also detected at a region of the unpaired X chromosome where the CEC‐5 and COH‐3/4 signals overlapped. The pachytene chromosome showing the strongest CEC‐5 enrichment was considered to be the unpaired X chromosome in males. The RNA‐directed in situ hybridizations were performed with stringent conditions that detected only long RNAs. Scale bar: 2 μm.

Source data are available online for this figure.

CeRep55‐repeat‐derived lncRNAs partially colocalize with the SC

Based on small RNA abundance, we further analyzed CeRep55 as a prevalent target, but not the sole target, of CSR‐1 and cohesin. CeRep55 is a class of minisatellite sequences consisting of a 27‐nt tandem repeat that is present on all chromosomes (Fig 7A and B). The entire genome has over 1,750 copies of CeRep55. Each chromosome has an average of six clusters of CeRep55 tandem repeats. At least three sequences of 22G‐RNAs that interact with CSR‐1 and CSR‐2 were expressed from the CeRep55 repeats (Figs 7A and EV5D).

We analyzed the RNAs derived from CeRep55 repeats by northern hybridization using single‐stranded probes. In the total RNA blots, forward (F) direction CeRep55 probes did not detect any signals, but reverse (R) direction probes detected small molecules corresponding to 22G‐RNAs (Fig 7C). Conversely, in nuclear RNA blots, the F probe against CeRep55 detected long transcripts with sizes ranging from 0.3 to 10 kb (Fig 7D). The lncRNAs detected with the F probe were 3.4‐fold more abundant than those detected with the R probe. This strand‐biased expression of small RNAs and lncRNAs from CeRep55 repeats implies that RdRPs such as EGO‐1 synthesize 22G‐RNAs on the lncRNAs expressed from the CeRep55 clusters.

We analyzed the expression patterns of CeRep55 lncRNAs by RNA‐directed in situ hybridization. In chromogenic detection, the F (but not the R) probe against CeRep55 strongly stained the chromosomes (Appendix Fig S4A–D). The signals for CeRep55 lncRNAs were detected in both germ and somatic cells.

Using RNA‐directed FISH (Fig 7E and F, and Appendix Fig S4E–G), we compared the localization of CeRep55 lncRNAs to those of proteins involved in homolog pairing. In pachytene nuclei of WT hermaphrodites, FISH with the F probe detected CeRep55 lncRNAs with about six regions of strong signals and one or two regions of weak signals per nucleus, all of which were associated with chromosomes (Fig 7E and F). Among these multiple regions, an average of 4.5 regions of CeRep55 lncRNA signals were observed as dumbbell‐like shapes, each of which were bridged between the COH‐3/4 signal regions of paired chromosomes (Fig 7G). Meanwhile, in the csr‐2; csr‐1 double mutant, many CeRep55 lncRNA signals were detected near COH‐3/4 signals, but the number of CeRep55 lncRNA signal regions in the double mutant was about two regions greater than in the WT. We noticed that the number of CeRep55 lncRNA signal regions with a dumbbell‐like shape in the csr‐2; csr‐1 double mutant decreased to less than half of that in the WT (Fig 7G). These results might imply that the csr‐2; csr‐1 double mutant is deficient in the locating reaction by which CeRep55 lncRNA counterparts find each other between a pair of homologous chromosomes.

We further examined the localization of CeRep55 lncRNAs, COH‐3/4, and the chromodomain protein CEC‐5 on unpaired X chromosomes in WT males. The CeRep55 lncRNA signal was detected in a region where the CEC‐5 and COH‐3/4 signals overlapped with each other on the presumed unpaired X chromosome (Fig 7H).

CeRep55 repeats are involved in the efficient homologous pairing and the condensation of unpaired X chromosomes

To examine the function of CeRep55, which is a typical target of CSR‐1 and cohesin, we made a mutant strain lacking four major CeRep55 clusters in the X chromosome (CeRep55_X) (Fig 7B) and analyzed its meiotic phenotype with and without an alternative genetic background. In C. elegans, meiotic defects often result in a high incidence of the male phenotype, which reflects nondisjunction of the X chromosome. The CeRep55_X quadruple‐deletion mutant was fertile and did not exhibit a clear Him phenotype, but the Him phenotypes of the csr‐1(fj150) and smc‐1(e879) mutants were enhanced by the CeRep55_X quadruple deletions (Fig 8A). Next, we examined the superstructure of the SC by staining SYP‐1 and HIM‐8. The shapes of the SYP‐1 signals were almost normal in the hermaphrodite nuclei of the CeRep55_X quadruple‐deletion mutant. However, for the SYP‐1 signals adjacent to HIM‐8, the abnormal branching of SYP‐1 signals in the csr‐1(fj150) mutant were enhanced by the CeRep55_X quadruple deletions (Figs 4E and EV4E). Moreover, the pairing states of the X chromosome and autosome V were examined by FISH. The quintuple mutant of csr‐1(fj150) in combination with the CeRep55_X quadruple deletions showed pairing defects of X chromosomes more frequently than the csr‐1(fj150) mutant or the CeRep55_X quadruple‐deletion mutant (Fig 8B and C).

Using immunofluorescence against histone H3K9me2, we also examined the condensation of unpaired X chromosomes in male testes of the CeRep55_X quadruple‐deletion mutant. H3K9me2‐positive chromosomes showing an extended state were observed in males of the CeRep55 quadruple‐deletion mutant about 3.1‐fold more frequently than in the WT males (Fig 8D and E).

Discussion

Pleiotropic roles of CSR‐1 in gene silencing and gene expression maintenance at the cytoplasm and nucleus

It has been debated whether the Argonaute protein CSR‐1 functions in gene silencing or gene expression maintenance. CSR‐1 is considered to silence target mRNAs in the cytoplasm (Aoki et al, 2007; Updike & Strome, 2009). Several genetic studies (Rocheleau et al, 2008; Seth et al, 2013) have signified that csr‐1(+) activity is required for maintaining the expression of some genes. CSR‐1 appears to have additional functions in the nuclei (Claycomb et al, 2009; She et al, 2009). Here, we discuss the pleiotropy of CSR‐1 using several models.

The first pleiotropic role of CSR‐1 is mRNA silencing in the cytoplasm. We suggest that the cytoplasmic exo‐RNAi pathway involving Dicer, host‐encoded RdRPs, and CSR‐1 with high sequence specificity may function as an impermanent adaptive defense to silence non‐self RNAs. The csr‐1 null mutant (Yigit et al, 2006) and several csr‐1 hypomorphic mutants (Fig 1G) exhibit diminished responses against exo‐dsRNA. CSR‐1 interacts with RdRP‐type siRNAs and shows Slicer activity that cleaves target mRNAs (Aoki et al, 2007). A paralog of CSR‐1 also showed Slicer activity (Fig 1B). Similarly, the endo‐RNAi pathway involving RdRPs and CSR‐1 seems to downregulate RdRP‐high‐sensitive self mRNAs (Campbell & Updike, 2015; Gerson‐Gurwitz et al, 2016; Quarato et al, 2021).

The second pleiotropic role of CSR‐1 is gene expression maintenance, possibly associated with germ granules and piRNAs. piRNA‐mediated surveillance, involving a predefined set of piRNAs, PRG‐1, and PRG‐2 (Batista et al, 2008), whose complexes are thought to permit mismatch binding, may function as an innate defense against nonself RNAs. Depletion of CSR‐1 results in an increase of piRNAs on mRNAs normally targeted by CSR‐1 (Shen et al, 2018). The csr‐1 null mutations disrupt the perinuclear localization of germ granules (Claycomb et al, 2009; Updike & Strome, 2009), and PRG‐1 is detected mainly in germ granules (Batista et al, 2008). As a possible model consistent with the previous study (Seth et al, 2013), we hypothesize that CSR‐1 could protect some self mRNAs, which are partially sensitive to piRNAs, from excessive piRNA‐mediated silencing by ensuring the perinuclear formation of germ granules containing the piRNA machinery and then restricting the reaction field of the piRNA‐mediated silencing. Perinuclear germ granules are inferred to protect transcripts from runaway silencing by compartmentalizing some components of the small RNA pathways (reviewed in Ouyang & Seydoux, 2022).

The third pleiotropic role of CSR‐1 is chromosome alignment that supports unpaired silencing and homologous pairing in meiotic nuclei, as introduced by this study. In csr‐2; csr‐1 double‐mutant males carrying monosomic X chromosomes, H3K9me2 signals in the pachytene nuclei were detected as aberrant multiple segments and sometimes as abnormally large segments at aggregated chromosomal regions (Fig 2D and E). The inaccurate pairing of chromosomes was also detected in csr‐2; csr‐1 double‐mutant hermaphrodites. By ensuring the accuracy of homolog pairing to prevent nonhomologous chromosome association, CSR‐1 and CSR‐2 may function to maintain physical distances between a monosomic X chromosome and the autosome pairs and then restrict heterochromatinization only to unpaired X chromosomes (Fig 8F–H). The abnormal alignment of chromosomes induced by the csr‐1 mutation might obstruct the meiotic expression of some genetic loci, at least in males. The piRNA pathway is not important in this chromosome alignment because heterochromatinization of unpaired chromosomes (She et al, 2009) and chromosome segregation (Batista et al, 2008) during meiosis are properly observed in prg‐1 and prg‐2 mutants.

New factors involved in unpaired silencing of meiotic chromosomes

Unpaired silencing of meiotic chromosomes has two aspects: chromosome condensation and H3K9me2 deposition. In our study, male pachytene nuclei of the coh‐3/4 meiotic kleisin(s) double mutant did not show a clear H3K9me2 staining signal, indicating a deficiency in meiotic silencing of the unpaired X chromosome. COH‐3/4 was detected in the central portion of unpaired chromosomes. Unpaired chromosome silencing may be initiated via the cohesin containing COH‐3/4 as a core for heterochromatinization. The partial colocalization of heterochromatin and cohesin is unsurprising because cohesin is enriched at the pericentromeric heterochromatin in mitotic cells in several organisms (Nonaka et al, 2002).

We identified the chromodomain protein CEC‐5 as a component of heterochromatin formed by unpaired silencing. Similar to CEC‐4 (Gonzalez‐Sandoval et al, 2015), CEC‐5 and CEC‐8 bound to H3K9me2 and subtly less strongly to H3K9me3. In male pachytene nuclei, the H3K9me2 staining signals correlating with facultative heterochromatinization were notably reduced by cec‐8; cec‐4 cec‐5 triple mutation, whereas the H3K9me3 signals correlating with constitutive heterochromatin were not affected by the triple mutation, implying further functional redundancy with additional chromodomain proteins in the maintenance of constitutive heterochromatin.

In C. elegans, MET‐2 and SET‐25 are the methyltransferases responsible for H3K9me2 and H3K9me3 (Bessler et al, 2010; Checchi & Engebrecht, 2011; Towbin et al, 2012). We thus propose a model where unpaired chromosomes receive the histone H3K9me2 modification via the methyltransferases and then turn heterochromatic by increased binding to CEC‐5 and similar proteins. Meanwhile, SYP‐1, a major component of the SC central region, was almost absent on unpaired chromosomes. Unpaired silencing, involving H3K9me2 modification and Argonaute‐mediated condensation, may exclude the SC central region components from unpaired chromosomes.

This study focused on the mechanism of unpaired silencing at the chromosome level. Extrachromosomal transgenes are also silenced during meiosis in C. elegans, whereas protein‐coding loci heterozygous for short deletions can usually express their mRNAs. The question of how short gene‐level unpaired states are handled by the unpaired silencing machinery remains a question for the future.

Involvement of the Argonaute‐mediated 22G‐RNA pathway in homology recognition of meiotic chromosomes

It remains unclear how the homologous sequences of chromosomes, excluding the PC regions, are recognized during SC formation. We observed severe defects in homolog pairing in the ego‐1 mutant (corresponding to an RdRP) and the csr‐2; csr‐1 double mutant (corresponding to the Argonaute proteins). CSR‐1 and CSR‐2 are dispensable during chromosomal association of SC components because SYP‐1 and COH‐3/4 were detected on chromosomes in the double mutant. The COH‐3/4 interactors HIM‐3, HTP‐1/2, and HTP‐3 (Kim et al, 2014) were also detected on chromosomes in the double mutant (Appendix Fig S4H). Instead, CSR‐1 and CSR‐2 are necessary for SC formation with accurate homology.

We showed that CSR‐1 and COH‐3/4 are associated with nonsimple repeats of chromosomes. Small RNA expression was observed from some nonsimple repeats (minisatellites and transposons) that were targets of CSR‐1 and COH‐3/4. The expression of chromosome‐associated lncRNAs was observed from CeRep55 minisatellite repeats, whose 22G‐RNAs interacted with CSR‐1 and CSR‐2. The CeRep55 lncRNAs localized to zones contacting each other within the SC on paired chromosomes, and the distribution of these zones was perturbed in the csr‐2; csr‐1 double mutant. The inaccurate pairing of chromosomes was observed in a csr‐1 mutant that had substitution of a catalytic residue for Slicer activity.

Taking our findings into consideration, we propose the following model for homolog recognition during meiosis. First, EGO‐1 synthesizes 22G‐RNAs on lncRNAs expressed from some nonsimple repeats, including CeRep55, which are the binding sites of COH‐3/4‐containing cohesin. Subsequently, the lncRNAs are cleaved by the Slicer activity of CSR‐1 and CSR‐2 and activated. In homologous pairing (Fig 8G and H), the cleaved lncRNA counterparts recognize each other between a pair of homologous chromosomes, possibly via Argonaute proteins with the 22G‐RNAs. This interchromosomal reaction, like a zipper, would support the deposition of SC central‐region components between homologous chromosomes with accurate homology. This zipper model requires a starting point using different machinery because the repeat‐derived lncRNAs correspond to the zipper teeth. In fact, C. elegans uses pairing centers and DNA‐binding proteins to initiate chromosome pairing (Phillips et al, 2005, 2009). In the case of a monosomic chromosome (Fig 8F and H) where the interchromosomal recognition of lncRNAs fails, CSR‐1 and CSR‐2 may reluctantly recognize homologous lncRNA copies intrachromosomally and accelerate condensation of the unpaired monosomic chromosome. Indeed, an unpaired X chromosome lacking the major CeRep55 clusters showed defects in the condensation process during unpaired silencing.

While our ChIP‐seq analyses pointed out the association of CSR‐1 and COH‐3/4 with non‐simple repeats, we observed ChIP signals for CSR‐1 and COH‐3/4 even in coding gene regions at lower density. The ChIP signals for CSR‐1 and COH‐3/4 were detected with a bias towards the exons expressing sense small RNAs present in the CSR‐1 complex and the exons expressing antisense capped nuclear RNAs, the latter of which were found in Jänes et al (2018a). A portion of exon‐derived sense small RNAs might be synthesized by RdRPs on exon‐derived long antisense transcripts, which can be capped or noncapped. CSR‐1 and meiotic cohesin might be associated with exceptional exons that produce long antisense transcripts, and this association could also contribute weakly to homology recognition in meiotic pairing.

An opposite model could be that junk lncRNAs generating nonspecific pairing reactions are eliminated by CSR‐1 and CSR‐2. This looks unlikely because deletions of the major CeRep55 clusters on the X chromosome enhanced the defect in X‐chromosome pairing in the csr‐1(fj150) mutant. If the absence of CeRep55 lncRNAs was important for accurate pairing, deletions of the CeRep55 clusters would not enhance the pairing defect in the csr‐1 mutant.

Over four decades ago, histochemical studies on meiotic chromosomes by electron microscopy showed that SC lateral elements in house snails, a monocot plant (Esponda & Stockert, 1971), and rats (Vázquez‐Nin & Echeverría, 1976) were rich in RNAs. In addition, the presence of RNAs on unpaired X chromosomes or their adjacent dense structures was reported in the testes of male crickets (Wolstenholme & Meyer, 1966). However, the RNA sequences present in the SC remain unidentified. Our findings on the SC‐associated lncRNAs, originating from repeats, provide new insights into these previous observations. In mice, homolog pairing is facilitated by DSBs, and also involves DSB‐independent homology recognition (Boateng et al, 2013) that requires a meiotic kleisin, RAD21L, in the SC lateral elements (Ishiguro et al, 2014). Chromosome regions associated with the meiotic kleisin might contain non‐DNA molecules involved in homology recognition, even in mice. In fission yeast, which undergoes homologous pairing without SC formation, several lncRNAs, including sme2, have recently been reported to promote robust pairing (Ding et al, 2012, 2019). Our study suggests a new model where lncRNAs and small RNAs are involved in homolog recognition during SC formation in C. elegans, a multicellular organism with holocentric chromosomes. An interesting question is whether lncRNA‐mediated homolog recognition is also involved in SC formation in other multicellular organisms with monocentric chromosomes.

Materials and Methods

Worm strains and culture conditions

N2 was used as the WT C. elegans strain. Strains and alleles are described in the Appendix Materials and Methods. The tm alleles were isolated by the Mitani group (National BioResource Project) (Gengyo‐Ando & Mitani, 2000) and the ok alleles were isolated by the Barstead group (C. elegans Deletion Mutant Consortium, 2012). The fj alleles, ranging from 67 to 163, were generated using the CRISPR/Cas9 system.

Worms were ordinarily cultured by feeding E. coli on NGM plates. For large‐scale cultivation, the worms were fed a mixture of E. coli and plant‐based infant formula, or E. coli only.

Antibodies

Polyclonal antibodies against CSR‐1, SMC‐1, and COH‐3/4 were raised in rabbits, and polyclonal antibodies against COH‐3/4, SYP‐1, and CEC‐5 were raised in guinea pigs. All antibodies were affinity purified with antigen proteins. The mouse monoclonal antibodies used to detect histone H3K9me2 were MABI0307 (MBL) for immunofluorescence and ab1220 (Abcam) for immunoprecipitation and Western blot. The rabbit monoclonal antibody used to detect histone H3K9me3 was EPR16601 (Abcam). Polyclonal antibodies against HIM‐3, HTP‐1/2, and HTP‐3 were gifts from A. F. Dernburg. Rabbit polyclonal antibodies against Drosophila histone H2Av (61686; Active Motif) were used for the spike‐in normalization in ChIP‐seq experiments.

Immunofluorescence and in situ hybridization

Worms were fixed with mixtures of methanol and formalin, subjected to epitope retrieval treatments, and then processed for immunofluorescence, chromosome FISH, and RNA‐directed in situ hybridization, similar to the method described in a previous report (Tabara et al, 1996). Detailed protocols are described in the Appendix Materials and Methods. Chromosome FISH was performed with the following DNA probes prepared by nick translation: 5S rDNA labeled with digoxigenin (DIG) and fosmid WRM0636aF12 derived from the right region of the X chromosome labeled with biotin. RNA in situ hybridization was performed using RNA probes. The probes were synthesized from a plasmid carrying about 23 tandem repeats of CeRep55 sequences by in vitro transcription with a nucleotide mix containing DIG‐ or biotin‐UTP, followed by fragmentation into about 250 nt in length.

Microscopy

Fluorescent images were captured in 3D on the following microscopes: a super‐resolution structured illumination microscope (SR‐SIM; Elyra S.1; Zeiss), Figs 2A and C, 3C, 4B and C, 5A and D, 7E and F, and EV4B–D, and Appendix Fig S3A–D; and a deconvolution microscope (DeltaVision; GE), Figs 2E and F, 3B, 4A and D, 5E, 7H, 8B and D, EV3A–C, and EV4A, and Appendix Fig S2B. The images for Figs 2E and 4A were visualized by high‐resolution 3D imaging, which took multiple shots per Z‐slice. Single‐plane images are displayed in these figures.

Analysis of 3D microscopic images

Immunofluorescence images of nuclei counterstained with DAPI were analyzed mainly with Fiji (ImageJ2) and its plugins.

The spatial distribution of histone H3K9me2 in nuclei was analyzed as follows. The high‐resolution 3D image of H3K9me2 and SYP‐1 signals was split into single‐channel images, and the fluorescence signals were segmented by Trainable Weka segmentation 3D (Arganda‐Carreras et al, 2017) in Fiji. The volumes and centroid 3D coordinates of the H3K9me2 segments were measured by 3D Object Counter (Bolte & Cordelières, 2006) in Fiji.

The immunofluorescence intensities of H3K9me2 and H3K9me3 in nuclei were quantitatively analyzed as described in Fig EV3B. Working images to quantify nuclear‐specific immunofluorescence signals were prepared by subtracting the intensity of the cytoplasmic background from the raw images. The DAPI signals were smoothed by Gaussian blur and Median 3D‐filters in CLIJ2 (Haase et al, 2020). The partially blurred DAPI signals were segmented by 3D Iterative Thresholding (Gul‐Mohammed et al, 2014) in 3D ImageJ Suite. The nuclear segments were numbered by 3D Object Counter. Immunofluorescence signals of H3K9me2 or H3K9me3 positioning in a segmented nuclear object were quantified by 3D Intensity Measure in 3D ImageJ Suite (Ollion et al, 2013).

The SYP‐1 and HIM‐8 staining signals were computer‐graphically reconstructed to analyze the superstructure of the SC. The fluorescence images were segmented by Trainable Weka segmentation 3D and were converted to 3D mesh data in OBJ format. The mesh data were imported into Blender (3D graphics software) and were visualized in 3D.

Analyses of small RNAs

Cell lysates were prepared from adult or mixed‐stage populations of worms carrying 2×FLAG::csr‐1 and HA_2×FLAG::C04F12.1 transgenes by homogenization and brief sonication to solubilize chromatin. C04F12.1 was designated csr‐2 herein, but it is also called vsra‐1. The Argonaute complexes were immunoprecipitated with anti‐FLAG M2 antibody beads (Sigma). Small RNAs recovered from the immunoprecipitates were ligated with adapters as described in our previous study (Aoki et al, 2007), amplified by RT‐PCR, and then sequenced with Illumina sequencers. The sequence data of small RNAs from adult populations were used for the figures.

Small RNA FASTQ data were aligned to reference sequences by either of the following three mapping strategies. For overall analyses (Figs 1D and EV1C and D), small RNA reads were mapped using BWA (Li & Durbin, 2009), first against the dataset comprising multiple classes of RNAs (exon transcripts, miRNAs, etc.) and consensus sequences of repeats, followed by second mapping against introns, third mapping against individual collections of all repeats, and fourth mapping against genome sequences. For analysis of the major repeats (Figs 6G and EV5D, and Appendix Fig S1B), the small RNA reads were mapped against the individual collections of five repeat sequences and miRNAs using BWA. For analyzing the orientation of small RNAs located in the exon regions of ChIP peaks (Figs 6H and EV1F–H), the small RNA reads were mapped to genome sequences in one step using BBmap (Bushnell, 2014). For comparison, capped nuclear RNA reads (Jänes et al, 2018a; Data ref: Jänes et al, 2018b) were mapped to genome sequences using BBmap. The number and orientation of small RNAs mapped to introns were examined, omitting reads matched to miRNAs, piRNAs, and known transcription units within introns. As for the strand orientation, the mapping results using BWA were adopted in Fig 1D, and the mapping results using BBmap were adopted in Figs 6H and EV1F–H.

Chromatin immunoprecipitation

Chromatin immunoprecipitations were performed with the crude chromatin fractions from adult worm populations. The worms were treated in a 1% formaldehyde buffered solution with brief homogenization, followed by several washes with a Tris‐based buffer. To obtain crude chromatin fractions, the formalin‐treated worms were mixed with four volumes of high‐salt lysis buffer, followed by sonication in the presence of 0.4 M LiCl and 0.03 M NaCl. The details are described in the Appendix Materials and Methods.

The first set of ChIP‐seq analyses was executed on the following chromatin complexes: (i) CSR‐1, SMC‐1, COH‐3/4, and H3K9me2 in N2, the strain KR16 carrying sDp2, and the him‐8(tm611) mutant. This set of ChIP libraries was sequenced with a SOLiD sequencer. The single‐end ChIP‐seq reads in CSFASTA format were mapped to the C. elegans genome sequence using SHRiMP2 (David et al, 2011).

The second and third sets of ChIP experiments were executed with the spike‐in normalization strategy (Egan et al, 2016) by adding a heterogenous calibrator chromatin. C. elegans chromatin mixed with a small amount of Drosophila chromatin was subjected to ChIP reactions with antibodies against target proteins in C. elegans and histone H2Av in Drosophila. The target chromatin complexes were as follows: (ii) CSR‐1, SMC‐1, COH‐3/4, and a mock in N2; and (iii) H3K9me2 and CEC‐5 in N2, the him‐8 mutant, and the met‐2 set‐25 double mutant. These sets of ChIP libraries were sequenced with an Illumina sequencer. The ChIP‐seq data were processed as described in Fig EV5A. The paired‐end ChIP‐seq reads in FASTQ format were mapped to the combined reference genome of C. elegans (ce10) and Drosophila (dm6) using BBmap. The peak calling on the ChIP‐seq data was performed based on the MACS theory (Zhang et al, 2008), except that the mapped reads were piled up using BEDTools (Quinlan & Hall, 2010). We combined the mock ChIP data with the control lambda of the input data and then used it as the control for peak calling by MACS2. The ChIP intensities were counted using BEDOPS (Neph et al, 2012) and BEDTools, and then scale‐normalized by referring to the calibrator information. These ChIP‐seq data with paired‐end reads were used for the figures.

In addition, a chromatin complex possessing H3K9me2 was immunoprecipitated from the chromatin fraction of the him‐8 mutant, and its protein components were analyzed using an LC/MS–MS spectrometer.

Charts and statistics

Statistical analyses were performed using the R software. Error bars in bar charts represent 95% confidence intervals, except that error bars in Fig 7G represent standard deviations.

Author contributions

Hiroaki Tabara: Conceptualization; resources; formal analysis; funding acquisition; investigation; visualization; methodology; writing – original draft; project administration; writing – review and editing. Shohei Mitani: Conceptualization; resources; investigation; writing – review and editing. Megumi Mochizuki: Investigation. Yuji Kohara: Conceptualization; resources; funding acquisition; investigation; writing – review and editing. Kyosuke Nagata: Conceptualization; resources; funding acquisition; writing – review and editing.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View and Appendix

PDF+

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 7

Source Data for Figure 8

The EMBO Journal (2023) 42: e105002

Data availability

The datasets generated by the Illumina and SOLiD sequencers are available in the DDBJ Sequence Read Archive (DRA, https://www.ddbj.nig.ac.jp/dra/index‐e.html).

Small RNA‐seq data of adults (duplicate sets): DRA013274 (https://ddbj.nig.ac.jp/resource/sra‐submission/DRA013274).

Small RNA‐seq data of mixed‐stage populations (a single set): DRA004203 (https://ddbj.nig.ac.jp/resource/sra‐submission/DRA004203).

ChIP‐seq data of adults (duplicate sets, Illumina): DRA013275 (https://ddbj.nig.ac.jp/resource/sra‐submission/DRA013275).

ChIP‐seq data of adults (a single set, SOLiD): DRA004202 (https://ddbj.nig.ac.jp/resource/sra‐submission/DRA004202).

The mapping and peak data obtained from the ChIP‐seq datasets are available in the DDBJ Genomic Expression Archive (GEA, https://www.ddbj.nig.ac.jp/gea/index‐e.html).

ChIP peak data for DRA013275: E‐GEAD‐510 (https://ddbj.nig.ac.jp/public/ddbj_database/gea/experiment/E‐GEAD‐000/E‐GEAD‐510).

ChIP peak data for DRA004202: E‐GEAD‐340 (https://ddbj.nig.ac.jp/public/ddbj_database/gea/experiment/E‐GEAD‐000/E‐GEAD‐340).

The cDNA sequences of cec‐5, cec‐4, and cec‐8 are available in the DDBJ/EMBL/GenBank databases under accession numbers LC127218, LC127219, and LC128409, respectively.