Selfish conflict underlies RNA-mediated parent-of-origin effects – Nature

Maintenance of worm strains

Nematodes were grown on modified nematode growth medium (NGM) plates with 1% agar/0.7% agarose to prevent C. tropicalis burrowing. Experiments were conducted at either 25 °C (C. tropicalis) or 20 °C (C. elegans). csr-1(+/−) strains were cultured on 6-cm NGM plates supplemented with 500 μl of G418 (25 mg ml⁻¹) for selecting heterozygous null individuals. Supplementary Table 2 lists all study strains, some of which were provided by the Caenorhabditis Genetics Centre, funded by the NIH Office of Research Infrastructure Programs (P40 OD010440).

Phenotyping and genotyping of crosses

For crosses, 4–5 L4 hermaphrodites were mated with 30–40 males in a 12-well plate with modified NGM. After 2 days, 10 L4 F₁ progeny were transferred to separate plates, genotyped by PCR, and at least 10 embryos per F₁ hermaphrodite were singled into 6-cm NGM plates. Each F2 individual was visually inspected daily for up to 7 days, classified for developmental stage, and any phenotypic abnormalities. Embryonic lethality, arrested development, and delayed reproduction were assessed. Sterility was noted for adults not producing progeny. After 7 days, worms were lysed and genotyped. A list of primers used for genotyping can be found in Supplementary Table 3. Crosses involving csr-1(−); slow-1/grow-1 hermaphrodites vs EG6180 males or injected hermaphrodites vs NIL males were selected based on a pmyo-2::mScarlet reporter.

Generation of C. tropicalis transgenic lines

For CRISPR–Cas gene editing, we adapted previous protocols⁵³. In brief, 250 ng µl⁻¹ Cas9 or Cas12a proteins were incubated with 200 ng µl⁻¹ CRISPR RNA (crRNA) and 333 ng µl⁻¹ trans-activating crRNA (tracrRNA) before adding 2.5 ng µl⁻¹ co-injection marker plasmid (pCFJ90-mScarlet-I). For HDR, donor oligos (IDT) or biotinylated and melted PCR products were added at a final concentration of 200 ng µl⁻¹ or 100 ng µl⁻¹, respectively. Following injections into young hermaphrodites, mScarlet-positive F1 were singled, and their offspring screened by PCR and Sanger sequencing to detect successful editing. To clone the mScarlet::SLOW-1 donor, we added ~300-bp homology arms amplified from QX2345 genomic DNA to mScarlet-I (from pMS050) in pBluescript via Gibson assembly. Because csr-1 is essential for viability in C. elegans, we first devised a strategy to stably propagate a csr-1 heterozygous line in the absence of classical genetic balancers. To do so, we used CRISPR–Cas9 to introduce a premature stop mutation in the endogenous csr-1 locus followed by a neoR cassette, which confers resistance to the G418 antibiotic (Extended Data Fig. 6d). For the csr-1::neoR donor, we first replaced the C. elegans rps-27 promoter and unc-54 3′ UTR in pCFJ910 with 500 bp upstream and 250 bp downstream of the C. tropicalis rps-20 gene. This rps-20::neoR cassette was then flanked with ~550-bp homology arms amplified from EG6180 worms and inserted into pBluescript. Correct targeting introduces a stop codon after residue L337 of CSR-1 followed by a ubiquitously expressed neomycin resistance. We propagated the mutant line in plates containing G418 and thus actively selecting for heterozygous csr-1(−) null individuals. Upon drug removal, most homozygous csr-1(−) individuals derived from heterozygous mothers developed into adulthood but were either sterile or laid mostly dead embryos. However, a small fraction of null mutants was partially fertile and homozygous csr-1(−) lines could be stably propagated for multiple generations despite extensive embryonic lethality in the population (Extended Data Fig. 6d). All gRNAs and HDR templates are available on Supplementary Tables 4 and 5.

In vitro RNA transcription and injection

The slow-1 cDNA was cloned into pGEM-T Easy (Promega, A1360), with a 5′ T7 RNA polymerase site and the start codon mutated RNA-only transcription (ATG>TTG). The plasmid was digested with NotI to release the insert (NEB, R0189), which was subsequently purified by gel-extraction and used as template for RNA synthesis. RNA was prepared using the HiScribe T7 Quick High Yield kit (NEB, E2050) with the following modifications: addition of 3 µl of 10 mM DTT and 1 µl of RNaseOUT (Thermo, 10777019). After overnight transcription, the reaction was diluted, treated with RNase-free DNase I (NEB, M0303S), bead-purified (Vienna Biocenter MBS 5001111, High Performance RNA Bead Isolation), quantified (Thermo, Q32852), and stored at −80 °C. Injections were repeated twice using independently transcribed RNA at concentrations: 150 nM and 400 nM yielding identical results.

Reciprocal crosses with the mScarlet::slow-1 reporter line

To assess SLOW-1 expression in F₁ progeny from reciprocal crosses between mScarlet::SLOW-1 NIL and EG6180 strains, we conducted 2 sets of crosses: (1) SLOW-1::mScarlet dpy (INK461) hermaphrodites to EG6180 males for maternal inheritance; and (2) EG6180 dpy (QX2355) hermaphrodites to mScarlet::SLOW-1 NIL males (INK459) for paternal inheritance. Wild-type young adult F₁ progeny were immobilized in NemaGel on a glass slide and imaged using an Axio Imager.Z2 (Carl Zeiss) widefield microscope with a Hamamatsu Orca Flash 4 camera, (excitation 545/30 nm filter). The analysis was performed in FIJI, by tracing the germline in the DIC channel and measuring mean fluorescence, including gut autofluorescence.

Sequencing and genome assembly of EG6180

We extracted high molecular weight genomic DNA using the Masterpure Complete DNA and RNA purification kit (tissue sample protocol, Lucigen). We prepared 8 kb, 20 kb and unfragmented sequencing libraries using the 1D Ligation Sequencing Kit (Oxford Nanopore SQK-LSK109). The 8 kb fragmentation was done using g-TUBE (Covaris). Library was loaded on a MinION MK1B device (Oxford Nanopore). Read calling was done using MinKNOW software. We performed a hybrid assembly, incorporating Illumina sequencing reads of EG6180 with some modifications as detailed below⁹. We used assembled Illumina reads to correct raw Nanopore reads, which were assembled using Flye Assembler⁵⁴. The preliminary assembly included 119 contigs in 107 scaffolds (Scaffold N50 was 1,489,504 bp). We derived synteny blocks between the provisional assembly and our chromosome-level NIC203 assembly using Sibelia⁵⁵ and used the synteny blocks to scaffold the contigs to chromosome level using Ragout⁵⁶.

Identification of C. tropicalis Argonaute proteins and piRNA pathway effectors

We annotated functional domains in C. tropicalis NIC203 using Interproscan 5 as part of our previous NIC203 genome assembly⁹. We identified Argonaute proteins with PFAM domains, including Piwi (PF02171), PAZ (PF02170), N-terminal domain of Argonaute (PF16486), Argonaute linker 1 (PF08699), Mid domain of Argonaute (PF16487) and Argonaute linker 2 (PF16488) domains. We excluded a protein with low molecular weight (41 kDa) as unlikely to be an Argonaute and the orthologue of C. elegans Dicer that represented an outgroup to the rest of the proteins. After aligning those sequences to C. elegans Argonautes identified in a previous study⁵⁷ using Clustal Omega we conducted phylogenetic analysis using iqtree2 (ref. ⁵⁸), with 1,000 replicates of the approximate likelihood-ratio test (–alrt 1000) and 1,000 boostraps (-b 1000). iqtree2 carries out an initial model selection step, and a substitution model with the general Q matrix, empirical codon frequencies, a proportion of invariable sites and a free rate heterogeneity (Q.pfam+F + I + R4) was selected. Additional orthologues of C. elegans piRNA effector genes were identified through reciprocal blastp searches, synteny conservation, and gene trees from Wormbase Parasite⁵⁹. C. elegans mut-16, rrf-1, and simr-1 have 1:1 orthologues in C. tropicalis. The evolutionary history of SET proteins is complex due to their propensity to gain and lose paralogues within Caenorhabditis. The gene annotated gene as C. tropicalis set-25, is the closest among six paralogues in its genome. Thus, the absence of a phenotype in the mutant may be attributed to genetic redundancy. The gene annotated as C. tropicalis set-32 is a close orthologue of two C. elegans genes: set-21 and set-32. The SET domains of C.tr-SET-32 and C.el-SET-32 are ~48% identical at the protein level. Additionally, using Alphafold2 (ref. ⁶⁰) we found that these two proteins have high structural similarity (root mean square deviation = 0.962) and using the predicted structure of C.tr-SET-32 as a query retrieved C.el-SET-32 as the top hit in C. elegans (Foldseek)⁶¹.

Transgenerational silencing of slow-1/grow-1

In the transgenerational inheritance experiments, EG6180 hermaphrodites were crossed to NIL (QX2345) males. F₁ individuals were genotyped after laying embryos to distinguish between self-progeny from cross-progeny. F₂ embryos from cross-progeny mothers were singled, allowed to lay eggs and genotyped. F₃ homozygous carriers for slow-1/grow-1 propagated for multiple generations and mated to EG6180 males. The slow-1/grow-1 TA activity was assessed by determining the proportion of delayed EG/EG non-carriers.

Single molecule in situ hybridization

Stellaris FISH Probes targeting slow-1, slow-2 and pgl-1 were designed using the Stellaris RNA FISH Probe Designer (Biosearch Technologies). The probes were labelled with Quasar 570, CAL Fluor Red 610 or Quasar 670, respectively (Biosearch Technologies). The protocol was adapted from Raj et al.⁶² and described in ref. ⁹. For imaging, an Axio Imager.Z2 (Carl Zeiss) widefield microscope with a Hamamatsu Orca Flash 4 camera and a 63×/1.4 plan-apochromat Oil DIC objective was used. Filters used were: DAPI excitation 406/15 nm, emission 457/50 nm and Quasar 570 excitation 545/30 nm, emission 610/75 nm. z-stack images with 40 slices (step size 0.2 µm) were acquired. Image analysis was performed with the FIJI plugin RS-FISH⁶³ with parameters set at Sigma 1.44, and threshold 0.0062.

RNA extraction and RNA-seq

Total RNA was extracted from approximately 100 young adult hermaphrodites and F₁ progeny, with the later using recessive mutations to visually discriminate cross-progeny from self-progeny. Reciprocal crosses were set up between parental strains for maternal or paternal inheritance of slow-1/grow-1 by mating INK531 hermaphrodites (uncoordinated worms in NIC203 background) to EG6180 males and QX2355 hermaphrodites (dumpy worms in EG6180 background) to NIC203 males and selecting phenotypically wild-type progeny for RNA extraction. Reciprocal crosses between NIL and EG6180 strains were performed analogously (INK255 hermaphrodites (dumpy worms in NIL background) to EG6180 males and QX2355 hermaphrodites (dumpy worms in EG6180 background) to QX2345 NIL males). Total RNA was extracted following a modified version of the protocol in⁶⁴ including multiple M9 washes, TRizol and chloroform incubation, phase-separation, isopropanol precipitation and resuspension in RNase-free water. Samples with RNA integrity number (RIN) > 8 were used for library preparation using the NEBNext Poly(A) kit and sequenced on NextSeq2000 P2 SR100 or NovaSeq S1 PE100 at the Vienna Biocenter NGS facility. To reduce reference bias, raw reads were aligned to a concatenated NIC203 + EG6180 genome/transcriptome assembly using STAR and bcbio-nextgen (https://github.com/bcbio/bcbio-nextgen). Transcript quantification and normalization were performed with tximport and Deseq2 (ref. ⁶⁵). We used Deseq2 to fit a model for the normalized counts using the strain identity of the mother and sequencing batch (Nextseq vs NovaSeq libraries) as fixed effects and compared the model to a null model that included only batch using a likelihood-ratio test. Despite identifying an outlier in the slow-1/grow-1 paternal inheritance samples (Fig. 1d), no obvious difference between the outlier and the other samples in terms of RNA quality and mRNA-seq quality control were identified. However, since each library was derived from an independent genetic cross, we cannot discard a human error, and therefore decided that it would be best practice to keep the outlier in the final analysis.

RT–qPCR

RNA was extracted from adult worms (50 males or 100 hermaphrodites per biological replicate) using TRIzol-chloroform extraction, followed by Dnase I digestion⁶⁶ and then RNA concentrations were measured using the Qubit High-Sensitivity RNA fluorescence kit (Thermo). cDNA was prepared with SuperScript III reverse transcriptase (Thermo) using random hexamers. Intron-spanning primers were validated with standard curves from QX2345 cDNA to ensure amplification efficiency and an r² value above 0.95. The following primers were used: FW-slow-1-mRNA: 5′-GAGCTACCGGAACTGGATAAAG-3′, RV-slow-1-mRNA: 5′-CAGAGTTCTCGGAAGTCTCCTC-3′, FW-slow-1-pre-mRNA: 5′-CGGACTGGATGAAACATTTAGC-3′, RV-slow-1-pre-mRNA: 5′-GAGCGGTGTTGACctgaatc-3′, FW-cdc-42: 5′-CGATTAAATGTGTCGTCGTAGG-3′, and RV-cdc-42: 5′-ACCGATCGTAATCTTCTTGTCC-3′. All samples had at least 3 biological replicates. We used the ∆∆C_t method to calculate relative fold change and chose cdc-42 as a housekeeping gene^67,68. Cdc-42 expression showed a low coefficient of variation in our RNA-seq datasets suggesting its validity as a housekeeping gene. All RT–qPCR reactions were prepared with the Luna Universal qPCR and RT–qPCR kit (NEB) and run with an annealing temperature of 58 °C. All biological replicates were run in technical quadruplicate and any reactions with abnormal amplification curves or melting temperatures were omitted before analysis (distinct from reactions for which we observed no amplification, which were not omitted). Representative samples from each condition were Sanger sequenced. We confirmed the absence of genomic DNA contamination in RNA samples by performing PCRs with gDNA-specific primers using the RNA as template and observed no amplification after 40 cycles. RT–qPCR indicated specific amplification of slow-1 in both hermaphrodites and males. However, the higher C_t values for males (34.27 versus 28.31 on average) and greater variability (s.d. of 1.55 versus 0.65 in the NIL) suggest much lower expression levels in males. This variability hinders a reliable estimate of abundance and assessment of the parent-of-origin effect in males.

Small RNA library preparation and sequencing

We isolated sRNAs, using the TraPR protocol⁶⁹. In brief, frozen worm pellets (2,000 worms per parental line) were supplemented with 350 µl lysis buffer, (20 mM HEPES-KOH, pH 7.9, 10% (v/v) glycerol, 1.5 mM MgCl₂, 0.2 mM EDTA, 1 mM DTT, 0.1% v/v Triton X-100). Samples were mechanically disintegrated and subjected to 4 freeze–thaw cycles in liquid nitrogen. The resulting lysates were cleared by centrifugation and the sRNA fraction was isolated using the TraPR Small RNA Isolation Kit (135.24, LEXOGEN). Isolated sRNA was treated with RppH (M0356S, BioLabs), to ensure 5′ monophosphate-independent capturing of small RNAs⁷⁰, following purification with Agencourt RNA Clean XP magnetic beads (BECKMAN COULTER). The sRNA was ligated to a 32-nt 3′ adapter with unique barcodes (sRBC, Supplementary Table 6, IDT) using truncated T4 RNA ligase 2 (M0373L, NEB). The resulting RNA was run on 12% SequaGel–UreaGel (National Diagnostics) and purified with ZR small-RNA PAGE Recovery Kit (R1070, ZYMO RESEARCH). The 37-nt-long 5′ adapter was ligated to the sRNAs using T4 RNA ligase (M0204S, NEB). The resulting RNA was cleaned up (R1015, ZYMO RESEARCH), reverse-transcribed, and PCR amplified. The cDNA fragments (160–190 nt) were extracted and gel purified (D4008, ZYMO RESEARCH). Small RNA Libraries were sequenced in triplicates on a NovaSeq S1 SR100 mode (Illumina) at the Vienna Biocenter NGS facility. All sequencing libraries generated for this project are listed in Supplementary Table 7.

sRNA immunoprecipitation

To study piRNA binding preferences of PRG-1.1 and PRG-1.2, we performed sRNA immunoprecipitation of N-terminally Flag-tagged PRG-1.1 (INK775) and PRG-1.2 (INK735) followed by sRNA-seq. For each of the 3 biological replicates (50,000 worms each), 18 worm plates (9 cm) were bleached to synchronize the population. Young adults were collected, frozen at −70 °C, thawed and washed with RIP buffer (50 mM Hepes pH 7.2, 150 mM NaCl, 0.01% NP-40). For lysis, RIP buffer and Benzonase were added and sonicated in a Diagenode Bioruptor followed by cleaning via centrifugation. For immunoprecipitation, 200 µl of Anti-Flag M2 Magnetic Beads (Millipore) were used (4 °C, overnight). The bound proteins were eluted in 500 µl 0.1 M GlycinHCl pH 2.7 for 5 min at room temperature. And transferred into a vial with 50 µl 1 M Tris-HCl pH 8. The proteins were digested with Proteinase K (0.7 mg ml⁻¹), and denatured proteins were removed by centrifugation following proteinase K inactivation. Samples were stored at −70 °C until library preparation.

Small RNA analysis

Sequencing adapters were trimmed from 5′ and 3′ ends using Cutadapt v1.18 (ref. ⁷¹). Extracted 21U and 22G reads aligned to the genome using hisat2 v2.1 (ref. ⁷²). For 22 G, only reads mapped to the coding sequences were analysed; for 21U, reads mapped to coding sequences, tRNAs and rRNAs were excluded using seqkit v0.13 and samtools v1.10. 22 G reads were quantified using featureCounts (Rsubread, R), normalized by the total number of 22 G per replicate, and visualized using the Gviz R package⁶². Candidate 21U-RNAs were identified based on perfect mapping and abundance criteria (>0.1 ppm). A custom script quantified 21U-RNAs and reads were normalized to miRNAs predicted based on homology to C. elegans miRNAs. To identify potential 21U-RNAs slow-1 candidates we used known targeting rules in C. elegans and binding energies. First, putative binding sites and energies for all 21U-RNAs against slow-1 mRNA were predicted with RNAduplex (ViennaRNA Package v2.0.58)⁶³, of which five best duplexes for every piRNA were taken. Candidate piRNAs without bubbles during binding and no more than 4 mismatches outside the seed region were extracted and ranked by binding energy (Supplementary Data 1). The second candidate list was generated considering the overall level of binding continuity by using Nucleotide blast v2.2.26 in blastn-short mode. Only 21U-RNAs with no mismatches or gaps in the seed region were selected for further analysis. Finally, we ranked 21U-RNAs by the total length of the ungapped alignment to slow-1 (Supplementary Data 1).

Chromatin immunoprecipitation

For chromatin immunoprecipitation, we collected an F₄ population of homozygous carriers for the repressed slow-1 allele after paternal inheritance, which was highly enriched in s22G-RNA complementary to slow-1 (Fig. 3i,j). First, we crossed EG6180 hermaphrodites to NIL males. The F₂ were genotyped to identify repressed slow-1/grow-1 (NIC/NIC) worms which were expanded for two generations (F₄) and collected as young adults. Each ChIP sample represents an independent genetic cross. Worms (200 µl) were collected, washed and incubated to minimize bacterial content and frozen in liquid nitrogen. For ChIP, we used the protocol described⁶⁴. Shortly the frozen worm pellet was pulverized by grinding in mortar with liquid nitrogen and the powder was crosslinked in 1 ml ice-cold RIPA buffer supplemented with 2% formaldehyde to crosslink (10 min, 4 °C). After quenching by addition of 100 µl 1 M Tris-HCl (pH 7.5), the sample was sonicated using Covaris for 600 s to achieve chromatin fragments of 200–500 bp. Fifty microlitres of the lysate was saved as an input fraction. Chromatin was immunoprecipitated using anti-H3K9me3 antibody (Ab8898, Abcam). The immunoprecipitation product was incubated with Protein A Dynabeads (Thermofisher scientific) and washed with LiCl. The immunoprecipitation product was eluted from beads and DNA was purified using ChIP DNA Clean and Concentrator kit (Zymo Research). Input control fractions were treated similarly to immunoprecipitation samples. DNA libraries were prepared with NEBNext Ultra II DNA Library Prep Kit (Illumina), deduplicated using bbmap v38.26, aligned using bwa mem v0.7.17 (ref. ⁶⁵), and normalized by the number of reads that mapped to the genome with samtools v1.10 (ref. ⁷³). Peaks were called by macs2 v2.2.5 with –broad and –mfold 1 50 options⁷⁴. Quality control plots were made using deeptools v3.3.1 (ref. ⁷⁵). H3K9me3 signal was calculated as read counts per genomic position in the ChIP sample normalized by counts in the corresponding input sample using bedtools v2.27 (ref. ⁷⁶) and custom R (v4.3) script.

Immunohistochemistry

Gravid nematodes were washed from plates, and embryos were extracted using bleach solution. The embryo suspension was applied to prepared poly-l-lysine slides (Sigma-Aldrich, P8920), and immersed into liquid nitrogen, fixed in ice-cold methanol (10 min) followed by acetone (10 min), and rehydrated in descending ethanol concentrations (95%, 70%, 50% and 30% ethanol). Fixed embryos were blocked in 3% BSA (VWR Life Science, 422351 S), followed by incubation with anti-Flag M2 primary antibody (Sigma-Aldrich, F3165, diluted 1:3,000). After washing, a secondary antibody Alexa Fluor A568 (ThermoFisher Scientific, A-11031, diluted 1:3,000) was applied, followed by additional washes. The final wash contained DAPI (Merck, D9542, 5 ng ml⁻¹). Processed embryos were mounted with Fluoroshield (Sigma-Aldrich, F6182) and imaged at Axio Imager 2 (ZEISS).

Fluorescence intensity quantification

Twenty-four-bit raw images were analysed in Fiji (v1.53r)⁷⁷. Embryos were selected by freehand tool and the same selection mask was used to capture background fluorescence intensity for each embryo. To compare fluorescence intensities between strains we used corrected total cell fluorescence (CTCF) parameter (CTCF = integrated density − (area of selected cell × mean fluorescence of background readings)). At least 23 embryos were used for quantification.

Worm protein lysate preparation and western blot

Gravid adult worms were collected, washed, and flash-frozen in the liquid nitrogen. Worm pellets were resuspended in ice-cold lysis buffer (30 mM HEPES pH 7.4, 100 mM KCl, 2 mM MgCl2, 0.05% IGEPAL, 10% glycerol and 1 tablet of protease inhibitors (Roche, 11836153001)) and lysed by sonication in Bioruptor (UCD-200, Diagenode) followed by centrifugation to obtain the supernatant. After protein quantification by Bradford assay (Thermo Scientific, 23238), samples were diluted, resuspended in SDS loading buffer, and loaded onto NuPAGE gels (Invitrogen). Samples were transferred to 0.45 µm PVDF membrane (Thermo Scientific, 88518) and blocked with 4% non-fat milk in TBS-T. Membranes were incubated with anti-Flag M2 (mouse, 1:2,000, Sigma-Aldrich, F3165) or anti-actin (rabbit, 1:3,000, Abcam, ab13772) primary antibody overnight followed by incubation with HRP-conjugated anti-mouse (1:10,000, Invitrogen, G-21040) or anti-rabbit (1:10,000, Jackson Immuno, 111-035-045) secondary antibody. Detection was performed using ECL reagent (Cytiva, RPN2106) and imaged with ChemiDoc MP (Bio-Rad). Membranes were stripped before reprobing (Thermo Scientific, 21059).

Live imaging of mScarlet::SLOW-1

Approximately 20 gravid adults were dissected in M9 medium under a stereo microscope. Embryos were transferred to individual wells in a Thermo Scientific Nunc MicroWell 384-Well Optical-Bottom Plate (Thermo Scientific). Embryos were imaged using an Olympus spinning disk confocal based on an Olympus IX3 Series (IX83) inverted microscope, equipped with a dual-camera Yokogawa W1 spinning disk (Yokogawa Electric Corporation) and two ORCA-Flash 4.0 V3 Digital CMOS cameras (Hamamatsu). Each field was imaged using a 40×/0.75 NA (air) objective, 16 z-sections at 2 µm and conditions were as follows: bright-field (100% power 30 ms) 568 nm, (100% power, 500 ms). Image acquisition was performed using CellSense software (Olympus). Image processing and montages were created using Fiji and embryoCropUI⁷⁸.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.