| INVESTIGATION

Transgenerational Epigenetic Inheritance Is Negatively Regulated by the HERI-1 Chromodomain

Roberto Perales,* Daniel Pagano,* Gang Wan,* Brandon D. Fields,*,† Arneet L. Saltzman,‡ and Scott G. Kennedy*,1 *Department of Genetics, Harvard Medical School, Boston, Massachusetts 02115, †Laboratory of Genetics, University of Wisconsin–Madison, Wisconsin 53706, and ‡Department of Cell and Systems Biology, University of Toronto, Ontario M5S 3G5, Canada ORCID ID: 0000-0002-3964-9734 (R.P.)

ABSTRACT Transgenerational epigenetic inheritance (TEI) is the inheritance of epigenetic information for two or more generations. In most cases, TEI is limited to a small number of generations (two to three). The short-term nature of TEI could be set by innate biochemical limitations to TEI or by genetically encoded systems that actively limit TEI. In , double-stranded RNA (dsRNA)-mediated silencing [RNAi (RNA interference)] can be inherited (termed RNAi inheritance or RNA-directed TEI). To identify systems that might actively limit RNA-directed TEI, we conducted a forward genetic screen for factors whose mutation enhanced RNAi inheritance. This screen identified the gene heritable enhancer of RNAi (heri-1), whose mutation causes RNAi inheritance to last longer (. 20 generations) than normal. heri-1 encodes a protein with a chromodomain, and a kinase homology domain that is expressed in germ cells and localizes to nuclei. In C. elegans, a nuclear branch of the RNAi pathway [termed the nuclear RNAi or NRDE (nuclear RNA defective) pathway] promotes RNAi inheritance. We find that heri-1(2) animals have defects in spermatogenesis that are suppressible by mutations in the nuclear RNAi (Ago) HRDE-1, suggesting that HERI-1 might normally act in sperm progenitor cells to limit nuclear RNAi and/or RNAi inheritance. Consistent with this idea, we find that the NRDE nuclear RNAi pathway is hyperresponsive to experimental RNAi treatments in heri-1 mutant animals. Interestingly, HERI-1 binds to targeted by RNAi, suggesting that HERI-1 may have a direct role in limiting nuclear RNAi and, therefore, RNAi inheritance. Finally, the recruitment of HERI-1 to chromatin depends upon the same factors that drive cotranscriptional gene silencing, suggesting that the generational perdurance of RNAi inheritance in C. elegans may be set by competing pro- and antisilencing outputs of the nuclear RNAi machinery.

KEYWORDS transgenerational epigenetic inheritance; RNAi; chromatin; small

HE inheritance of epigenetic information for two or more et al. 2012), and the inheritance of acquired traits in mice Tgenerations is referred to as transgenerational epigenetic (Carone et al. 2010; Walker and Gore 2011; Radford et al. inheritance (TEI) (Heard and Martienssen 2014). Many ex- 2012; Castel and Martienssen 2013; Padmanabhan et al. amples of TEI have now been documented including, but 2013; Somer and Thummel 2014; Holoch and Moazed 2015; not limited to, in plants (Arteaga-Vazquez and Martienssen and Moazed 2015; Rankin 2015). In many cases, Chandler 2010), protein-based inheritance in yeast (Shorter TEI is limited to a small number of generations (e.g.,twoto and Lindquist 2005), double-stranded RNA (dsRNA)-mediated three) (D’Urso and Brickner 2014; Heard and Martienssen gene silencing [RNA interference (RNAi)] in Caenorhabditis 2014). In other cases, such as -interacting RNAs (piRNA)- elegans (Vastenhouw et al. 2006; Ashe et al. 2012; Buckley mediated gene silencing in C. elegans and paramutation in plants, TEI can be long-lasting (. 10 generations). Molecular

Copyright © 2018 by the Genetics Society of America mechanisms that set the generational duration of TEI are doi: https://doi.org/10.1534/genetics.118.301456 largely a mystery. Manuscript received August 2, 2018; accepted for publication October 23, 2018; Recently, small noncoding RNAs have emerged as impor- published Early Online November 2, 2018. Supplemental material available at Figshare: https://doi.org/10.25386/genetics. tant mediators of epigenetic inheritance in . For 7229294. example, in plants, the RNA-dependent RNA Polymerase 1Corresponding author: Department of Genetics, Harvard Medical School, 77 Ave. Louis Pasteur, New Research Bldg. 266, Boston, MA 02115. E-mail: kennedy@ (RdRP) mop1 produces small interfering (si)RNAs thought genetics.med.harvard.edu to mediate paramutation (Alleman et al. 2006). In the yeast

Genetics, Vol. 210, 1287–1299 December 2018 1287 Schizosaccharomyces pombe, siRNAs help direct and maintain H3K27me3), as well as inhibition of RNAP II during the elon- stable, and in some cases heritable, heterochromatic states at gation step of via an unknown mechanism pericentromeres (Volpe et al. 2002; Martienssen et al. 2005; (Guang et al. 2008, 2010; Burkhart et al. 2011; Buckley et al. Ragunathan et al. 2015). In Drosophila, maternally inherited 2012; Mao et al. 2015). While it is not yet clear why nuclear piRNAs direct heritable silencing of transposable elements RNAi is needed for RNAi inheritance, it is known that nuclear (Le Thomas et al. 2014). Finally, in mice, the effects of stress RNAi is needed for transgenerational siRNA expression, sug- and metabolic disease are reported to pass from parent to gesting that cotranscriptional gene silencing and cytoplasmic progeny, and and short tRNA fragments have siRNA amplification systems may be connected in some way. been implicated in mediating this inheritance (Carone et al. If RdRP enzymes can amplify small RNA populations each 2010; Gapp et al. 2014; Sharma et al. 2016). Thus, small generation during RNAi inheritance, then why does RNAi regulatory RNAs are good candidates for being the informa- inheritance not last forever? There could be fundamental tional vectors of TEI in eukaryotes (termed RNA-directed biochemical limitations that prevent RNAi inheritance from TEI). Small noncoding RNAs are also linked to TEI in lasting forever, or C. elegans might possess systems that ac- C. elegans. For instance, piRNAs can trigger transgenerational tively prevent long-term inheritance. Recently,loss-of-function gene silencing via a process termed RNA-induced epigenetic mutations in met-2,whichisoneoftwoC. elegans H3K9 silencing (Ashe et al. 2012; Shirayama et al. 2012). Addition- responsible for depositing the majority ally, dsRNA-mediated gene silencing (RNAi) is heritable in of H3K9me2 found in C. elegans, were shown to cause RNA- C. elegans; the progeny of animals treated with dsRNAs retain directed TEI to last longer than normal (Andersen and Horvitz the ability to silence RNAi-targeted genes for many genera- 2007; Checchi and Engebrecht 2011; Towbin et al. 2012; tions, even after the removal of dsRNA triggers (termed RNAi Lev et al. 2017). The reasons why loss of MET-2 enhances inheritance) (Vastenhouw et al. 2006; Alcazar et al. 2008). RNAi inheritance are not known, but may involve global al- During RNAi inheritance in C. elegans, dsRNAs are processed terations in that indirectly promote RNAi by into “primary” siRNAs, which are bound by Argo- inheritance (Lev et al. 2017). To further our understanding naute (AGO) to regulate complementary cellular of why RNAi inheritance does not last forever, we conducted RNAs (Meister 2013). AGO-bound primary siRNAs recruit, a forward genetic screen to identify additional factors that by an unknown mechanism, RdRPs to homologous mRNA limit the duration of RNAi inheritance. Our screen identified templates to produce amplified pools of “secondary siRNAs” the gene-heritable enhancer of RNAi 1 (heri-1). We find that (Sijen et al. 2001). Repeated RdRP-based siRNA amplifica- HERI-1 inhibits nuclear RNAi-based cotranscriptional gene tion in germ cells each generation is likely responsible for silencing and that this ability is likely the reason why loss driving RNAi inheritance in C. elegans (Ashe et al. 2012; of HERI-1 promotes RNAi inheritance. Interestingly, HERI-1 Buckley et al. 2012; Sapetschnig et al. 2015). is physically recruited to genes undergoing nuclear RNAi, Forward genetic screens have identified factors that are suggesting that HERI-1 may be a direct and dedicated regu- required for RNAi inheritance in C. elegans (Ashe et al. 2012; lator of nuclear RNAi, and therefore TEI. Surprisingly, the Buckley et al. 2012; Spracklin et al. 2017; Wan et al. 2018). recruitment of this negative regulator of nuclear RNAi to The factors fall into two general categories. The first category genes is itself dependent upon nuclear RNAi. To explain these includes factors that localize to cytoplasmic liquid-like con- results, we propose that nuclear RNAi has both pro- and densates such as the P granule, the Z granule, or Mutator foci. antisilencing functions, and that the generational perdurance These factors likely promote RNAi inheritance by acting of RNAi inheritance is set by the relative contributions of with RdRPs to amplify siRNA populations each generation these two opposing outputs. (Spracklin et al. 2017; Wan et al. 2018). The second category of factors are members of a nucleus-specific branch of the Materials and Methods RNAi pathway [the nuclear RNAi or NRDE (nuclear RNA Strains defective) pathway] (Ashe et al. 2012; Buckley et al. 2012; Shirayama et al. 2012; Spracklin et al. 2017; Wan et al. 2018). For a complete list of strains used in this study, see Supple- According to current models of nuclear RNAi, AGOs bind and mental Material, Table S1. All strains were maintained using escort siRNAs to nuclei, where these ribonucleoprotein com- standard laboratory conditions. plexes locate RNA Polymerase II (RNAP II)-dependent nascent heri screen transcripts based on complementarity to trigger cotranscrip- tional gene silencing (termed nuclear RNAi) (Guang et al. Larval stage 4 (L4) worms carrying both the oma-1(zu405) 2008, 2010; Buckley et al. 2012). HRDE-1 and NRDE-3 are allele and the pie-1::gfp::h2b transgene were mutagenized two tissue-specific nuclear AGOs that drive nuclear RNAi in with ethyl methanesulfonate (EMS) at a concentration of germ cells and somatic cells, respectively (Guang et al. 2008; 47 mM for 4 hr at room temperature. Control worms were Ashe et al. 2012; Buckley et al. 2012; Shirayama et al. 2012). not exposed to EMS. Mutagenized and control worms were The nuclear AGOs recruit downstream nuclear RNAi effec- washed 33 with M9 buffer, and allowed to recover for 2 hr at tors (NRDE-1/2/4) to genomic sites of RNAi to direct 20° on NGM plates seeded with OP50. F2 progeny of muta- post-translational modifications (PTMs) (e.g., H3K9me3 and genized and control worms were isolated by hypochlorite

1288 R. Perales et al. treatment, and exposed to oma-1 and gfp RNAi simulta- neously on 10-cm RNAi plates at 20°. In total, 90,000 muta- genized were screened in 74 independent pools. In the following generation, embryos were isolated with hypo- chlorite treatment and fed OP50 at 20°. We repeated this process until the control worms failed to inherit oma-1 and gfp silencing (F7). In the F9, we isolated mutants that con- tinued to silence gfp and oma-1. Five generations later, lines that continued to silence reporter genes were kept for further study. To identify candidate heri genes, we used whole- sequencing coupled with the bioinformatics tool CloudMap as describedinMinevichet al. (2012). In short, genomic DNA was purified from control and mutant strains. Next-generation se- quencing libraries were prepared and sequenced using the Illu- mina HiSeq platform. Each genome was sequenced using paired-end sequencing at 203 coverage. CloudMap software was used to identify, map, and compare SNP’s between heri mutant strains. Genes that contained three (or greater) unique SNPs were kept as potential heri genes.

RNAi inheritance assays For gfp inheritance assays: embryos were collected by hypo- chlorite treatment and placed on either control (HTT115) or gfp RNAi-expressing at 20°. The F1 embryos were collected again by hypochlorite treatment and fed OP50. When animals reached gravid adult stage, they were scored for germline gfp expression in groups of 50 worms per bi- ological replicate. For Figure 2, C and D, this process was repeated for 10 and 23 generations, respectively. For oma-1 in- heritance assays in Figure 1B, embryos were collected by hypo- chlorite treatment, and placed in either control (HT115) or oma-1 RNAi-expressing bacteria at 20°. The F1 embryos were Figure 1 A genetic screen to identify heritable enhancers of RNAi (heri) collected again by hypochlorite treatment and fed OP50.For genes. pie-1::h2b::gfp and oma-1(zu405ts) can be silenced heritably by RNAi (Vastenhouw et al. 2006; Alcazar et al. 2008). We treated each biological replicate and for each generation, we singled out pie-1::h2b::gfp;oma-1(zu405ts) animals with the mutagen EMS. Control six individuals at the L4 stage of development, and 3 days later animals were left untreated. After two generations, animals were fed we counted the number of hatched embryos as well as arrested bacteria expressing double-stranded RNA targeting gfp and oma-1 embryos. The percent viability (oma-1 RNAi inheritance) of simultaneously. Embryos from these animals were isolated and shifted ° each strain was calculated as the number of embryos that to a non-RNAi (OP50) food source and grown at 20 , which is the non- permissive temperature for oma-1 (zu405ts). Embryos were isolated for hatched within 24 hr over the total number of laid eggs. an additional five generations until control animals stopped inheriting gfp and oma-1(zu405ts) silencing (gfp ON and embryonic arrest, F7). Muta- Construction of HERI-1-tagged strains using clustered genized pools were allowed to grow for two more generations and regularly interspaced short palindromic repeats potential Heri strains (animals were gfp OFF and alive, indicating that For construction of the C-terminal heri-1::3xflag strain, we pie-1::h2b::gfp and oma-1(zu405ts) were still being silenced) were sin- gled. After five additional generations, those populations still inheriting used the Co-CRISPR (clustered regularly interspaced short gene silencing were kept for further study. heri genes were identified by a palindromic repeats) strategy described in Arribere et al. combination of whole-genome sequencing and bioinformatic analysis (2014). In short, candidate guide RNAs (gRNAs) were calcu- (see Materials and Methods and Figure S1). Rare animals within these lated using the CRIPSR design tool at Massachusetts Institute populations had lost gene silencing at pie-1::h2b::gfp and oma-1- of Technology (.mit.edu). We chose the gRNA sequence (zu405ts). These animals were isolated to establish mutant lines that “ ” 9 9 expressed reporter genes (termed reset ). EMS, ethyl methanesulfate; 5 -TCATGCGAAACGAGAGAAAG-3 because its protospacer RNAi, RNA interference; RT, room temperature. adjacent motif (PAM) sequence of GTGG is located four nucle- otides upstream of heri-1’s termination codon, and it had low off-target effect predictions. As a repair template, we used a cutting. Germlines of gravid hermaphrodites were injected with single-stranded DNA oligo [4 nM ultramer HPLC purified from the following injection mix: gRNA (20 ng/ml), repair template Integrated DNA Technologies (IDT)], which contained 50 bp of (20 ng/ml), and Co-CRISPR markers pDD162 (50 ng/ml) homology on each side of heri-1’s termination codon, and it had (#47549; Addgene), unc-58 gRNA (20 ng/ml), and AF-JA-76 the PAM sequence mutated to GTTG to prevent repeated Cas9 (20 ng/ml) all in 13 taq buffer (New England Biolabs, Beverly,

HERI-1 Limits TEI 1289 MA). Uncoordinated (Unc) progeny of injected worms were mixtures were centrifuged for 30 sec at 3000 rpm and super- singled, allowed to self-fertilize, lay progeny, and then were natants were transferred to new tubes. For H3K9me3 genotyped for the presence of the 3xflag insertion at the 39– chromatin immunoprecipitation (ChIP), we used 2 mgof end of heri-1.FortheconstructionoftheC-terminalheri-1::- anti-H3K9me3 antibody (07-523; Millipore) and for HERI- gfp::3xflag strain and the heri-1::chromomutant::gfp::3xflag,we 1::3xFLAG ChIP we used 3 mg of M2 anti-flag antibody followed the CRISPR/Cas9 homologous recombination protocol (F1804; Sigma [Sigma Chemical], St. Louis, MO). Antibodies described in Dickinson et al. (2015). Details of this approach can were added and incubated overnight at 4°, while rotating. be found at http://wormcas9hr.weebly.com/. Repair template The following day, 20 ml of protein A (for H3K9me3) and had 500 bp of homology to heri-1 and contained a self-excis- protein G (for Flag) were added to the extracts and allowed ing cassette (SEC) with three features: a hygromycin resistance to bind for 2 hr at 4°, while rotating. Beads were washed gene, a Roller (Rol) marker gene, and a heat shock-inducible Cre twice with FA lysis buffer, twice with FA lysis buffer + recombinase gene. The SEC is flanked by LoxP sites. Gravid 500 mM NaCl (50 mM Tris/HCl pH 7.5, 1 mM EDTA, 1% hermaphrodites were injected with an injection mix containing Triton X-100, 0.1% sodium deoxycholate, and 500 mM the following: 50 ng/ml of pDD162 plasmid containing the Cas9 NaCl), once with LiCl buffer (0.25 M LiCl, 1% NP-40, 1% and sgRNA (#47549; Addgene), 10 ng/ml of repair template, sodium deoxycholate, 1 mM EDTA, and 10 mM Tris/HCl 10 ng/mlpGH8(Prab-3::mCherry neuronal co-injection marker; pH 8.0), and twice with TE buffer (1 mM EDTA and 10 mM #19359; Addgene), 5 ng/ml pCFJ104 (Pmyo-3::mCherry body Tris/HCl pH 8.0). All washes were performed in 1-ml vol- wall muscle co-injection marker; #19328; Addgene), and umes with 1 min rotation at room temperature, and beads 2.5 ng/mlpCFJ90(Pmyo-2::mCherry pharyngeal co-injection were concentrated by centrifugation for 30 sec at 3000 rpm. marker; #19327; Addgene). Injected worms (three per plate) Immunoprecipitated chromatin was eluted by incubating were incubated at 25° for 2–3 days. Then, hygromycin was beads/ChIPs in 500 ml of 0.1 M sodium carbonate and 1% added directly to the plates at a concentration of 250 mg/ml. SDS for 30 min, while rotating at room temperature. Beads Plates were incubated at 25° for 4 days. Animals that were were concentrated by centrifugation for 30 sec at 3000 rpm, Hygro-resistant, Rol,andlackedtheredfluorescent markers and the supernatant was collected and transferred to a new were isolated. To excise SEC, larval stage one (L1) animals were tube. To reverse cross-linking, each sample received 30 mlof heat-shocked at 34° for 4–5hrandshiftedto20° for 3–5days. 5 M NaCl and was incubated at 65° overnight. The follow- Non-Rol animals were isolated and genotyped by PCR and ing day, ChIP DNA was purified using a PCR cleanup kit Sanger sequencing for the gfp::3xflag insertion. (QIAGEN, Valencia, CA) and concentrated into 35 mlof water. Enrichment of immunoprecipitated DNA was quan- Chromatin immunoprecipitation and qPCR tified using the Bio-Rad 23 SYBR Green Master mix, with We isolated embryos by hypochlorite treatment and fed larval 2 ml of ChIP DNA per reaction and primers at a final con- animals HT115 bacteria expressing oma-1 dsRNA. The fol- centration of 250 nM. For a list of primer sequences, see lowing generation, embryos were again collected by hypo- Table S2. chlorite treatment and were fed OP50 (non-RNAi bacteria). Stacked oocyte scoring and rescue Gravid hermaphrodites were collected and 100 ml pellet of packed worms were washed 23 with 1 ml of M9 buffer, and heri-1 mutant or control animals were crossed to N2 males, then flash frozen on liquid nitrogen in a 1.5-ml tube. The and maintained as heterozygous for one to two generations. pellet was cross-linked by incubation in 1 ml of M9 buffer Homozygous heri-1 mutant animals were isolated and with 2% formaldehyde, and allowed to rotate for 30 min. The stacked oocytes were quantified in multiple siblings the next cross-linking reaction was quenched by addition of 54 mlof generation. This strategy was adopted to minimize the 2.5 M glycine and samples were then rotated for an addi- chance that transgenerational epigenetic effects could accrue tional 5 min. Animals were then washed 23 with 3 ml of in heri-1 lines prior to analysis. To score stacked oocyte de- M9 buffer and resuspended in 0.5 ml of FA lysis buffer fects, L4 animals were placed in OP50 food at 20° for 24 hr, (50 mM Tris/HCl pH 7.5, 1 mM EDTA, 1% Triton X-100, then placed in M9 buffer with 0.1% sodium azide and 0.1% sodium deoxycholate, and 150 mM NaCl) supple- mounted onto 2% agarose pads. For each biological replicate, mented with complete Mini protease inhibitors (Roche). An- 50 worms were scored. To ask if mating with wild-type males imals were sonicated in a Qsonica Q800R sonicator for could rescue stacked oocyte defects, we singled L4 stage heri- 25 min with a cycle of 30 sec on, 30 sec off, and 70% output. 1(2); pie-1::gfp::h2b animals and allowed these animals Insoluble debris was cleared by a quick spin in a microcen- grow at 20° for 24 hr. Animals with stacked oocytes in both trifuge for 2 min at 5000 rpm at 4°. Extracts were transferred gonad arms were identified with a dissecting microscope and to new tubes and centrifuged at 15,000 rpm for 10 min at 4°. split into two groups. Group #1 worms were not crossed to Protein concentrations were calculated by Bradford assay wild-type males (no cross). Group #2 worms were mated (Bio-Rad, Hercules, CA). Equal amounts of protein extracts with three wild-type males (cross). Approximately 48-hr were precleared using 15 ml of a 50:50 slurry of unblocked later, males were removed and animals were allowed to grow protein A or G agarose beads (Millipore, Bedford, MA) in FA at 20° for another 24 hr. The number of progeny for each lysis buffer for 15 min at 4°, while rotating. Extract/bead worm in each group was counted.

1290 R. Perales et al. Immunofluorescence and microscopy for six to nine generations (Vastenhouw et al. 2006; Buckley et al. 2012). oma-1(zu405ts) is a temperature-sensitive (ts) For HERI-1::3xFLAG imaging, gonads were dissected and gain-of-function (gf) lethal allele of oma-1 (Lin 2003). oma-1 fixed as follows: 20 gravid hermaphrodites where picked RNAi silences oma-1(gf) and, therefore, OMA-1(GF)-medi- onto a 20 ml drop of M9 in a coverslip. Gonads were dissected ated lethality for four to seven generations (Alcazar et al. using a 25G ⅝ needle by cutting worms in half near the vulva. 2008; Buckley et al. 2012). We EMS mutagenized gfp::h2b; The coverslip containing dissected gonads was placed on a oma-1(gf) animals and, two generations after mutagenesis, Gold Color Frost Plus slide (9951GLPLUS; Thermo Fisher exposed animals to gfp and oma-1 dsRNAs. In parallel, we Scientific) and immediately placed on a block of dry ice for propagated nonmutagenized control gfp::h2b; oma-1(gf) an- 10 min. Using a razor blade, the coverslip was popped off the imals similarly treated with gfp and oma-1 dsRNAs. Animals sample then quickly placed in 220° methanol for 10 min. The were propagated in the absence of further RNAi until non- slide was allowed to air dry for 1–2 min at room temperature. mutagenized control animals failed to inherit oma-1 or gfp Excess methanol was removed with a kimwipe. Samples were silencing (approximately equal to seven generations). Strains blocked with 500 ml of 0.5% BSA in 13 PBS at room temper- that exhibited silencing for an additional seven generations ature. The M2 anti-flag antibody (F1804; Sigma) was diluted were isolated and kept for further study. From 90,000 1:100 in 0.5% BSA in 13 PBS and 50 ml was added to the mutagenized genomes, we isolated 20 strains fulfilling these sample. Primary incubation was overnight at 4° in a humid- criteria. We refer to the genes mutated in these strains as the ified chamber. Slides were then washed 43 with 150 mlof heritable enhancer of RNAi (heri) genes. 0.5% BSA in 13 PBS with 6–7 min for each wash. An Alexa Fluor 488 goat anti-mouse secondary antibody (A10667; Mo- heri-1 encodes a chromodomain protein with homology lecular Probes, Eugene, OR) was diluted 1:250 in 0.5% BSA to Ser/Thr protein kinases in 13 PBS and 50 ml were added to the sample. Secondary To assign identity to the heri genes, we used whole-genome incubation was for 2 hr at room temperature. Slides were sequencing coupled with custom scripts and the bioinfor- again washed 43 with 150 ml of 0.5% BSA in 13 PBS with matic software CloudMap (Minevich et al. 2012; Spracklin 6–7 min each wash. After the final wash, 15 ml of Vectashield et al. 2017) (Figure S1). This approach identified four inde- with DAPI (H-1200; Vector Laboratories, Burlingame, CA) pendent mutations within the C. elegans gene cec-9/c29h12.5 was added to the slide and a new coverslip was carefully (Figure 2A). For reasons outlined below, we refer to cec-9/ placed over the sample. Most images were taken using a c29h12.5 as heri-1. From strains harboring mutations in heri-1, wide-field Zeiss Axio Observer.Z1 microscope ([Carl Zeiss], we isolated rare individuals in which gfp and oma-1 silenc- Thornwood, NY) using a Plan-Apochromat 633/1.40 Oil DIC ing had been lost, and we used these “reset” strains to ask if M27 objective and an ORCA-Flash 4.0 CMOS camera. Images these lines were indeed enhanced for RNAi inheritance. This in Figure 3 and Figure S5 were generated using a Nikon analysis showed that RNAi inheritance after oma-1 or gfp (Garden City, NY) Eclipse Ti microscope equipped with a RNAi was enhanced in all four heri-1 mutant strains (Figure W1 Yokogawa Spinning disk, with a 50 mm pinhole disk 2, B and C). To confirm the molecular identity of heri-1,we and an Andor Zyla 4.2 Plus sCMOS monochrome camera tested an independently isolated allele of heri-1 (gk961392) with a 603/1.4 Plan Apo Oil objective. for RNAi inheritance. Note that gk961392 likely represents a Data availability loss-of-function allele of heri-1,asgk961392 is a deletion that removes conserved domains of HERI-1 (see below) and alters Strains, plasmids, and whole-genome sequencing data are the reading frame (Figure 2A). gk961392 animals exhibited available upon request. The authors affirm that all data an enhanced RNAi inheritance phenotype (Figure 2D and necessary for confirming the conclusions of the article are Figure S2). We conclude that cec-9/c29h12.5 is heri-1 and present within the article, figures, and tables. Supplemental that one function of HERI-1 is to limit the generational per- data has been uploaded to Figshare, including WGS sequenc- durance of RNAi inheritance. ing data for each of the four heri-1 alleles identified by WGS. Table S1 contains the strains used in this study. The HERI-1 is expressed in germ cell nuclei Supplemental Figures document contains Figures S1–S10. Supplemental material available at Figshare: https://doi. heri-1 encodes a protein with two conserved domains: a chro- org/10.25386/genetics.7229294. modomain and a serine/threonine kinase-like domain (Fig- ure 2A). To understand more about HERI-1, we used CRISPR/Cas9 to introduce a C-terminal gfp::3xflag epitope Results to the 39 terminus of the heri-1 locus (Arribere et al. 2014; Paix et al. 2014; Dickinson et al. 2015; Farboud and Meyer A genetic screen identifies heritable enhancers of RNAi 2015). In heri-1::gfp::3xflag animals, we observed HERI-1:: (heri) genes GFP expression in the nuclei of adult germ cells (Figure 3A). We conducted a genetic screen to identify factors that limit the No HERI-1::GFP signal was observed in somatic tissues in generational perdurance of RNAi inheritance in C. elegans adult animals (R. Perales, unpublished data). GFP fluores- (Figure 1). gfp RNAi silences a gfp::h2b reporter transgene cence was expressed diffusely throughout nuclei and did

HERI-1 Limits TEI 1291 Figure 2 HERI-1 limits transgenerational epigenetic inheritance. (A) Diagram of HERI-1 domain structure. heri-1 alleles identified in our genetic screen and the heri-1(gk961392) deletion allele used in this study are indicated. Asterisk represents premature stop codon. (B) RNAi inheritance assay (see Materials and Methods) showing percent embryonic viability [oma-1(zu405ts) silencing] over generations after oma-1 RNAi feeding in our starting strain (control, YY565), and heri-1 mutants that were identified in the Heri screen and that had been “reset” for reporter gene expression (see Figure 1). Error bars represent SD of the mean. (C) gfp inheritance assay showing the percentage of animals of the indicated genotypes showing gfp silencing in control and reset heri-1 strains over generations after gfp RNAi exposure. (D) Representative images of oocytes in wild-type and heri-1(gk961392) animals that harbor the pie-1::h2b::gfp transgene during the gfp RNAi inheritance assay. Generations after gfp RNAi are indicated. RNAi, RNA interference. not colocalize with mitotic chromosomes (Figure 3A and Fig- that are undergoing heritable silencing (Burton et al. 2011; ure S3). Similar results were seen when CRISPR was used to Ashe et al. 2012; Buckley et al. 2012; Gu et al. 2012; introduce a 3xflag epitope to heri-1 and anti-FLAG immuno- Shirayama et al. 2012). Therefore, inherited siRNAs and/or fluorescence was used to detect HERI-1::FLAG (Figure S3). H3K9me3 may be the informational vectors that drive RNAi RNAi inheritance assays indicated that the CRISPR-modified inheritance. We asked if the enhanced RNAi we see in heri-1 heri-1 locus produced functional HERI-1 protein (Figure S4). mutant animals was associated with an increase in the num- During embryogenesis in C. elegans, four asymmetric cell di- ber of generations during which siRNAs and repressive chro- visions partition germline determinants into the germline matin marks were inherited. To do this, we used H3K9me3 blastomeres termed the P0–P4 cells. During these early stages ChIP and gfp siRNA Taqman probes to measure H3K9me3 de- of embryogenesis, HERI-1::GFP was expressed in both germ- positedonthegfp gene and gfp siRNA levels, respectively, in line and somatic blastomeres (Figure S5 and R. Perales, un- wild-type and heri-1 mutant animals 20 generations after published data). In the P1 blastomere, we detected a low gfp RNAi had been initiated. Both H3K9me3 and siRNAs level of GFP fluorescence in cytoplasmic puncta, whose sub- remained elevated in the F21 progeny of heri-1 mutant an- cellular distribution and morphology were reminiscent of P imals exposed to dsRNA (Figure 4, A and B), indicating that granules (Figure S5). By L1, HERI-1::GFP::3xFLAG was seen the Heri phenotype exhibited by heri-1 mutant animals is exclusively in the primordial germ cells Z2 and Z3, and was likely due to an increased generational perdurance of the no longer seen in somatic tissues (Figure 3B). HERI-1::GFP gene silencing pathways mediating RNAi inheritance in was expressed in germ cell nuclei throughout the remainder wild-type animals. of larval development and no GFP expression could be seen in HERI-1 inhibits nuclear RNAi the soma during larval development (Figure 3B and R. Per- ales, unpublished data). We conclude that HERI-1 is a germ- What pathway(s) might HERI-1 regulate to limit RNAi inher- line-expressed protein that localizes predominantly to nuclei. itance? The C. elegans nuclear RNAi pathway is required for RNAi inheritance; in animals lacking components of the nu- HERI-1 limits transgenerational siRNA and clear RNAi machinery, H3K9me3 is not deposited on chroma- H3K9me3 expression tin, siRNAs are not heritably expressed, and gene silencing is RNAi inheritance in C. elegans is correlated with the heritable not maintained in inheriting generations (Burton et al. 2011; expression of siRNAs, which are antisense to genes undergo- Ashe et al. 2012; Buckley et al. 2012; Gu et al. 2012; ing heritable silencing, as well as with the heritable deposi- Shirayama et al. 2012). Note that the reason why nuclear tion of repressive histone marks (e.g., H3K9me3) on genes RNAi is needed for RNAi inheritance is not yet known. Given

1292 R. Perales et al. Figure 3 HERI-1 is expressed in germ cell nuclei. (A) Fluorescent micrographs of the mitotic and transition zones of an adult germline from an animal expressing HERI- 1::GFP::3XFLAG and mCHERRY::H2B. Note that HERI-1::GFP::3XFLAG does not colocal- ize with chromatin in dividing cells (arrows). Distal tip cell is to the left of image. One nuclei (square box) is magnified in panels shown to the right. (B) Micrograph of a lar- val stage one (L1) and a larval stage two (L2) animal. HERI-1::GFP::3XFLAG is expressed in Z2 and Z3 (arrows) in L1, and in the devel- oping germline of L2 animals (dotted line). In these animals, no HERI-1::GFP signal was observed in the soma. Note that the nongermline fluorescence signal is due to autofluorescence, which is also present in wild-type animals. HERI-1::GFP is expressed throughout the remainder of germline devel- opment and no GFP fluorescence could be observed in somatic cells during this time. that nuclear RNAi mediates RNAi inheritance and HERI-1 is a RNAi and/or RNAi inheritance. To test this idea, we used nuclear protein, we wondered if HERI-1 might limit RNAi HERI-1 ChIP to ask if RNAi would cause HERI-1 to associate inheritance by inhibiting nuclear RNAi. The following lines with the chromatin of genes that we had targeted by RNAi. of evidence indicate that this is indeed the case. First, the Indeed, we found that, after oma-1 RNAi, HERI-1::GFP::FLAG nuclear RNAi machinery directs H3K9me3 deposition at ge- associated with the oma-1 locus, but not the genes flanking nomic loci exhibiting sequence homology to RNAi triggers oma-1 (Figure 5). Similar results were seen when animals (Guang et al. 2010; Gu et al. 2012). Using H3K9me3 ChIP expressing HERI-1::FLAG were treated with oma-1 RNAi (Fig- to quantify H3K9me3 levels before and after RNAi, we found ure S6). We conclude that RNAi can direct HERI-1 to interact that RNAi triggered the deposition of H3K9me3 (4–63)in with chromatin of the gfp gene after exposing animals to gfp heri-1 mutant animals more than what is seen in wild-type dsRNA triggers. The physical recruitment of HERI-1 to a gene animals after RNAi (which itself was 3–43 elevated over targeted by RNAi suggests that HERI-1 may play a direct role non-RNAi-treated animals) (Figure 4C). Second, exposure of in inhibiting nuclear RNAi and, therefore, RNAi inheritance. C. elegans to dsRNA results in the nuclear RNAi-based silenc- Recruitment of HERI-1 to chromatin requires ing of unspliced nascent RNAs (pre-mRNA) exhibiting se- nuclear RNAi quence homology to trigger dsRNAs (Guang et al. 2008, 2010; Burton et al. 2011). Using quantitative RT-PCR to How is HERI-1 recruited to genes targeted by RNAi? heri-1 quantify pre-mRNA levels before and after RNAi, we found encodes a chromodomain. We wondered if nuclear RNAi- that the degree to which RNAi silenced pre-mRNAs was directed chromatin modifications (such as H3K9me3 or greater in heri-1 mutant animals than in wild-type animals H3K27me3) might be responsible for recruiting HERI-1 (via (Figure 4D). Finally, HRDE-1 is a germline Argonaute re- chromatin PTM/chromodomain interactions) to genes after quired for RNAi inheritance in wild-type animals. hrde-1 RNAi. Note that such a scenario would be somewhat surpris- was epistatic to heri-1 for RNAi inheritance, suggesting that ing, as nuclear RNAi normally promotes gene silencing and RNAi inheritance in heri-1 mutant animals depends upon HERI-1 limits this silencing. HRDE-1 is a nuclear AGO that nuclear RNAi (Figure 4E). We conclude that HERI-1 is a drives nuclear RNAi in germ cells (Buckley et al. 2012). We negative regulator of nuclear RNAi. We speculate that the exposed wild-type or hrde-1(2) animals to oma-1 RNAi and inhibition of nuclear RNAi is the mechanism by which used HERI-1 ChIP to quantify HERI-1 chromatin interactions HERI-1 negatively regulates RNAi inheritance. in these animals. The analysis showed that HRDE-1 was re- quired for RNAi to direct HERI-1 to bind chromatin, suggest- RNAi directs HERI-1 to chromatin ing that nuclear RNAi is, indeed, needed for RNAi to recruit Chromodomains can interact with post-translationally mod- HERI-1 to chromatin (Figure 5). The putative H3K9 methyl- ified such as H3K9me3 and H3K27me3 (Eissenberg transferase SET-32/HRDE-3 is, at least in some cases, 2012). Nuclear RNAi in C. elegans directs H3K9me3 and required for RNAi inheritance and contributes to RNAi- H3K27me3 (Guang et al. 2010; Buckley et al. 2012; Gu directed H3K9 methylation following RNAi (Ashe et al. et al. 2012; Mao et al. 2015). These observations hint at the 2012; Kalinava et al. 2017; Lev et al. 2017; Spracklin et al. possibility that HERI-1, which possesses a chromodomain, 2017). We found that SET-32/HRDE-3 was required for might be physically recruited to genes undergoing nuclear oma-1 RNAi to recruit HERI-1 to the oma-1 gene, hinting that

HERI-1 Limits TEI 1293 Figure 4 HERI-1 inhibits nuclear RNAi. (A) H3K9me3 ChIP-qPCR on heri-1(+) heri-1(2) ani- mals. Data are expressed as a ratio of H3K9me3 ChIP signals detected in heri-1(+) over heri-1(2) mutant animals before gfp RNAi (bottom panel) and 20 generations after gfp RNAi (bottom panel). Data are from three biological replicates and error bars are SD of the mean. (B) Total RNA isolated from animals of the indicated genotypes was used for Custom Taqman assays (Burton et al. 2011) to detect gfp siRNAs 20 generations after gfp RNAi treatment. Signal from WT (starting strain) animals is defined as one. Data are from three biological replicates and error bars are SD. (C) H3K9me3 ChIP was conducted on WT, heri-1 (gk961392), and hrde-1(tm1200) animals in the progeny of animals exposed +/2 to oma-1 RNAi. Relative location of primers used for qPCR are indicated. Position along x-axis is not to scale with actual genomic loci (see Figure 5). Data are from three biological replicates and error bars are SD of the mean. Note, a value of one in this assay means that RNAi had no effect on the state of H3K9me3 at this locus. Data show that H3K9me3 is enriched within the oma-1 operon af- ter oma-1 RNAi. oma-1 operon contains three genes; oma-1, spr-2, and c27b76.2. Primer set B and C are in oma-1. Primer set D is in spr-2. (D) qPCR data showing that the oma-1 pre-mRNA is more silenced in heri-1(gk961392) animals than in WT animals after oma-1 RNAi. Error bars are SD of the mean. Note, a value of one would indicate that RNAi had no effect on oma-1 pre-mRNA levels for this experiment. (E) Animals of the indicated geno- types [heri-1(gk961392) and hrde-1(tm1200)] and expressing pie-1::gfp::h2b were exposed to +/2 gfp RNAi. Representative micrographs of 21, 22, and 23 oocytes are shown. Percentages of animals inheriting silencing in the indicated generations are shown. ChIP, chromatin immunoprecipitation; qPCR, quantitative PCR; RNAi, RNA interference; siRNA, small interfering RNA; WT, wild-type.

H3K9me3 may be a component of the chromatin signature gonad arms containing more oocytes than normal (termed that recruits HERI-1 to chromatin (Figure 5) [note that heri-1 a stacked oocyte phenotype) (Figure 6, A and B). Animals chromodomain mutations destabilize HERI-1, making addi- harboring any one of the four other heri-1 mutations also tional obvious tests of this model difficult (see Discussion)]. showed this stacked oocyte phenotype, albeit at a somewhat We conclude that HRDE-1 and SET-32/HRDE-3 are required reducedrate(FigureS7).Thedatashowthatthelossof for the recruitment of the negative regulator of nuclear RNAi HERI-1 causes oocytes to accumulate within C. elegans HERI-1 to genomic sites of RNAi. Because HRDE-1 is required germlines. Stacked oocytes are often seen in C. elegans lack- for nuclear RNAi (and SET-32/HRDE-3 contributes to nu- ing functional sperm (Schedl and Kimble 1988). Two addi- clear RNAi), the data suggest that the nuclear RNAi machin- tional lines of evidence support the idea that heri-1 mutant ery in C. elegans has both pro- and antisilencing outputs. First, animals have dysfunctional sperm. First, heri-1(2) animals the machinery links siRNAs to cotranscriptional gene silenc- with stacked oocytes were sterile (Figure 6C). Second, wild- ing. Second, the machinery places limits on this silencing by type sperm (introduced by mating) rescued the fertility de- recruiting the negative regulator HERI-1 to sites of nuclear fects of heri-1 animals, which had stacked oocytes (Figure RNAi, perhaps by altering chromatin landscapes in ways rec- 6C). We conclude that 20% of animals that lack HERI-1 ognized by the HERI-1 chromodomain. produce dysfunctional sperm or have defects in spermato- genesis. To begin to differentiate these possibilities, we iso- heri-1 mutant animals have defective sperm, which may lated and DAPI-stained animals in which one of two gonad be caused by hyperactive nuclear RNAi arms contained stacked oocytes(Figure6D).Spermatheca While working with heri-1 mutant animals, we noticed that from “normal” germline arms possessed sperm; however, in 20% of heri-1(gk961392) animals had a single or both spermatheca from defective germline arms (with stacked

1294 R. Perales et al. Discussion heri-1 encodes a chromo/kinase domain protein that nega- tively regulates nuclear RNAi and RNAi inheritance. We spec- ulate that it is the ability of HERI-1 to limit nuclear RNAi that allows HERI-1 to limit RNAi inheritance. Interestingly, we find that HERI-1 is recruited to genes targeted by RNAi, sug- gesting that HERI-1 may be a direct regulator of RNAi inher- itance and, therefore, that C. elegans possess genetically encoded systems dedicated to limiting RNA-directed TEI dur- ing normal reproduction. Finally, we find that the NRDE nu- clear RNAi system, which is itself needed for nuclear RNAi, is needed to localize the negative regulator of nuclear RNAi HERI-1 to genes. These data suggest that the nuclear RNAi machinery has two outputs during RNAi inheritance: (1) to promote cotranscriptional gene silencing and (2) to recruit negative regulators of nuclear RNAi to limit the number of generations over which epigenetic information is inherited. Figure 5 RNAi causes HERI-1 to associate with chromatin of genes un- dergoing heritable silencing. HERI-1::GFP::3xFlag animals were exposed How does HERI-1 inhibit nuclear RNAi? to +/2 oma-1 RNAi and HERI-1 ChIP was conducted on progeny. Fold enrichment of HERI-1 on the oma-1 locus, as well as neighboring loci, is HERI-1 has homology to serine/threonine protein kinases, shown and is expressed as a ratio of HERI-1 ChIP signals +/2 oma-1 RNAi. and one of the heri-1 alleles that we identified in our genetic Data are from three biological replicates and error bars are SD of the screen (gg538) alters a conserved aspartate residue within mean. Similar results were obtained when HERI-1::3xFLAG animals were the HERI-1 kinase-like domain (Figure 2A). Thus, the HERI-1 subjected to a similar analysis (Figure S6). A value of one would indicate kinase-like domain is likely important for HERI-1 to inhibit that RNAi had no effect on HERI-1 ChIP. Note that primer set D does not show RNAi-induced HERI-1 binding, but did show an increase in nuclear RNAi and, therefore, RNAi inheritance; however, H3K9me3 in Figure 4C (see Discussion). ChIP, chromatin immunoprecip- HERI-1 is unlikely to be an active protein kinase, as its kinase itation; RNAi, RNA interference; WT, wild-type. domain lacks active site residues required for kinase activity in related protein kinases (Nolen et al. 2004). For instance, oocytes), no sperm could be seen (Figure 6D). The data the GxGxxG, VAIK, and DFG motifs, which contribute to 2+ suggest that heri-1 sperm are defective because either: (1) Mg and ATP binding in canonical protein kinases, are not they never develop, or (2) develop but are dysfunctional conserved in HERI-1 (Figure S8). Additionally, we failed to and, consequently, are swept out of the spermatheca by detect kinase activity associated with recombinant HERI-1 ovulating oocytes. To distinguish these scenarios, we used protein (Figure S9). Thus, we predict that HERI-1 is Normarski microscopy to visualize spermatogenesis in wild- a pseudokinase. Pseudokinases are fairly common in eukary- type and heri-1(2) mutant animals at mid-late fourth larval otes where they often play important roles in diverse aspects developmental stages. In total, 33% of heri-1(2) animals of cellular physiology, such as acting as scaffolds, anchors, or displayed defects in spermatogenesis, as early sperm pre- allosteric regulators of other proteins, including other protein cursor cells are present but did not develop into mature kinases (Eyers and Murphy 2013). Given that HERI-1 is spermatids (Figure 6E). Wild-type animals did not exhibit recruited to sites of nuclear RNAi, we speculate that HERI-1 such defects (Figure 6E). We conclude that some heri-1(2) negatively regulates nuclear RNAi by interacting with and mutant animals have defects in spermatogenesis. We won- allosterically regulating other prosilencing factors at sites of dered if heri-1(2) sperm defects were related to HERI-1’s nuclear RNAi (Figure 7). Potential targets of such regulation role in limiting nuclear RNAi. If so, a hrde-1 mutation would include the nuclear RNAi factors HRDE-1 and NRDE-1/2/4, be expected to suppress heri-1(2) sperm defects, as nuclear as well as the chromatin-modifying enzymes that impart re- RNAi should no longer be hyperactive in animals that lack pressive modifications to chromatin in response to nuclear the AGO that directs nuclear RNAi. To test this idea, we RNAi (Guang et al. 2010; Burkhart et al. 2011; Buckley quantified sperm defects in heri-1(2) and hrde-1(2) sin- et al. 2012) (Figure 7). HERI-1 IP-MS might identify the gle-mutant animals as well as heri-1(2); hrde-1(2) dou- regulatory targets of HERI-1. ble-mutant animals and found that, indeed, hrde-1 was How is HERI-1 recruited to chromatin? epistatic to heri-1 with regard to spermatogenesis defects (Figure6B).Thedataareconsistent with the idea that HERI-1 is recruited to chromatin in response to RNAi (Figure HERI-1 limits nuclear RNAi (likely directed by endogenous 5). Recruitment is dependent upon the nuclear RNAi AGO nuclear siRNAs and HRDE-1) in sperm nuclei, and that this HRDE-1, suggesting that nuclear siRNAs direct HERI-1 recruit- regulation is needed for normal sperm function and/or ment to chromatin (Figure 5). How might nuclear siRNAs re- development. cruit HERI-1 to chromatin? The nuclear RNAi pathway directs

HERI-1 Limits TEI 1295 Figure 6 heri-1 mutant animals have defective sperm. (A) Fluorescent micrograph of WT or heri-1(gk961392) oocytes expressing pie-1p::h2b::gfp.Of the heri-1 mutant animals, 20% show a stacked oocytes in either one or both gonad arms. Asterisk indicates vulva. Thick arrows indicate stacked oocytes. Thin arrow indicates an unfertilized oocyte. (B) Quantification of the indicated oocyte defects in wild-type, heri-1(gk961392), hrde-1(tm1200), and double-mutant animals. Error bars are SD of the mean. (C) Number of progeny from heri-1(gk961392) animals that had stacked oocytes in both gonad arms that were selfed (self) or after crossing to WT (N2) males (cross). n = 13 for self and n = 14 for cross. (D) DAPI staining of a heri-1(gk961392); pie-1::h2b::gfp animals that had stacked oocytes in one gonad arm. A magnification of spermatheca from both gonad arms (indicated by box in “merge” panel) is shown below. Arrow indicates gonad arm with stacked oocytes. Asterisk indicates vulva. (E) Images of WT and heri-1(gk961392) animals undergoing spermatogenesis. Spermatogenesis steps are labeled for WT animals. heri-1(gk961392) fail to complete spermatogenesis. “%” refers to the percent of animals displaying spermatogenesis defects. RNAi, RNA interference; WT, wild-type. the deposition of repressive chromatin marks such as RNAi, indicating that H3K9me3 and HERI-1 binding are dis- H3K9me3 and H3K27me3 on the chromatin of genes under- sociable (Figure 4C and Figure 5, primer D). Therefore, al- going nuclear RNAi (Guang et al. 2010; Burkhart et al. 2011; though H3K9me3 may be necessary for recruiting HERI-1 Gu et al. 2012; Mao et al. 2015). Additionally, HERI-1 has a to chromatin, it is unlikely to be sufficient. Indeed, recent chromodomain, and chromodomains are well known to bind studies show that the role of H3K9 methylation during RNAi post-translationally modified histones, such as H3K9me3 and inheritance is complex, involving multiple H3K9 methyl- H3K27me3 (Eissenberg 2012). Together, these observations transferase enzymes, multiple nuclear RNAi-directed chro- suggest the following model. First, nuclear RNAi directs matin modifications, as well as a locus-specific dependence PTMs such as H3K9me3 on chromatin. Second, these PTMs upon H3K9me3 for inheritance (Ashe et al. 2012; Kalinava interact with the chromodomain of HERI-1 to recruit HERI-1 et al. 2017; Lev et al. 2017; Spracklin et al. 2017). For in- to genes, so that HERI-1 can limit nuclear RNAi. Consistent stance, genetic screens, and candidate gene approaches, with the model, we find that the putative H3K9 methyltrans- show that SET-32/HRDE-3 promotes RNAi-directed H3K9 ferase SET-32/HRDE-3 is required for oma-1 RNAi to direct methylation and RNAi inheritance when gfp reporter trans- HERI-1 to oma-1 chromatin (Figure 5). Note, however, genes are targeted for silencing by dsRNA or piRNAs (Ashe that we detected an increase in H3K9me3, but not an in- et al. 2012; Spracklin et al. 2017). However, in other RNAi crease of HERI-1 binding, on the spr-2 gene after oma-1 inheritance assays, (such as oma-1 RNAi inheritance), set-32/

1296 R. Perales et al. inheritance, it seems likely that biochemical studies seeking to identify the histone PTM code bound by the HERI-1 chro- modomain will be needed to figure out how HERI-1 is recruited to chromatin to regulate nuclear RNAi. Identifying and characterizing additional heri genes (defined by our ge- netic screen) might also be helpful in this regard. Note that any in vivo studies exploring how the HERI-1 chromodomain recruits HERI-1 to chromatin by RNAi will be complicated by the fact that HERI-1 chromodomain mutations destabilize HERI-1 (Figure S10). HERI-1 and sperm function Why would C. elegans need factors that limit nuclear RNAi in germ cells? C. elegans expresses an abundant class of endog- enous siRNAs that are thought to direct nuclear RNAi in nu- clei during the normal course of growth and reproduction (Lee et al. 2006; Guang et al. 2008, 2010; Burkhart et al. 2011; Billi et al. 2014). Some of these endogenous siRNAs are thought to be present in sperm progenitors, where they regulate gene expression programs needed for sperm devel- opment and function (Sijen et al. 2001; Kennedy et al. 2004; Duchaine et al. 2006; Gent et al. 2009, 2010; Pavelec et al. 2009; Conine et al. 2010; Vasale et al. 2010). We find that Figure 7 Model for the role of HERI-1 in limiting RNA inheritance. The heri-1 mutant animals exhibit spermatogenesis defects, AGO HRDE-1 uses siRNAs as guides to recognize and bind pre-mRNAs which are suppressed by loss of the germline nuclear RNAi with sequence homology to siRNAs. HRDE-1 then recruits downstream AGO HRDE-1. The data suggest that one function of TEI- silencing factors, such as the NRDEs (unlabeled teal ovals) and chromatin- modifying enzymes, to deposit H3K9me3 as well as another, currently regulating systems in C. elegans is to limit endogenous unknown (X), signal on chromatin to mark genes undergoing nuclear nuclear RNAi (directed by sperm 22G siRNAs) in sperm RNAi. Note: X could be another chromatin PTM or a chromatin/DNA- progenitors, and that this gene regulation is important for binding protein. X and H3K9me3 cooperate to recruit HERI-1 to chroma- spermatogenesis. Such gene misregulation might occur in tin. HERI-1 is a pseudokinase that may limit RNAi inheritance by binding sperm progenitors at genes normally targeted by nuclear and regulating prosilencing proteins (such as HRDE-1 or NRDEs) that are present on genes undergoing nuclear RNAi/RNAi inheritance. The model 22G endo siRNAs in wild-type animals or, alternatively, at predicts that nuclear RNAi drives cotranscriptional gene silencing by inhib- genes that are not normally subjected to nuclear RNAi. Re- iting RNA Polymerase II (RNAP II) while, at the same time, limiting this garding this latter model, endogenous RNAi (like experimen- silencing by recruiting negative regulators of the process, such as HERI-1, tal RNAi) is driven by 22G siRNAs, which are maintained over to sites of nuclear RNAi/RNAi inheritance. NRDE, nuclear RNA defective; generations by RdRP enzymes (Gent et al. 2009, 2010; RNAi, RNA interference; siRNA, small interfering RNAs; WT, wild-type. Buckley et al. 2012). This self-amplifying feedforward mode of gene regulation could be dangerous if genes that are not hrde-3(2) animals may lack most RNAi-directed H3K9me3 normally silenced by endogenous siRNAs (or are only mar- marks but these animals do not show obvious defects in ginally silenced) were to inappropriately enter states of her- inherited gene silencing (Kalinava et al. 2017). Additionally, itable silencing. Therefore, we speculate that HERI-1 might C. elegans lacking two other H3K9 en- promote germ cell health by preventing and/or reversing zymes, MET-2 and SET-25, lack biochemically detectable lev- such unwanted and runaway heritable gene silencing. The els of H3K9me2/3 in germ cells, and yet these animals do not identification of HERI-1 regulated sperm genes (via HERI-1 show obvious defects in oma-1 RNAi inheritance (Towbin ChIP), and/or loci regulated by nuclear RNAi in wild-type et al. 2012; Kalinava et al. 2017). Even more puzzling, ani- and heri-1 mutant animals (H3K9me3 ChIP or RNA-Seq), will mals lacking SET-25 (H3K9me3) alone show defects in RNAi allow this idea to be tested. inheritance when some gfp reporter transgenes and some endogenous loci are targeted by RNAi (L. Wong, unpublished Acknowledgments data), while animals lacking MET-2 (H3K9me2) alone show a Heri phenotype when a gfp reporter gene is targeted by We thank members of the Kennedy laboratory for helpful RNAi (Lev et al. 2017) [note that these later data hint at discussions of the data and the manuscript; Judith Kimble the tantalizing possibility that H3K9me2 (deposited by for helpful discussions; Paula Montero Llopis of the Micros- MET-2) might be the signal responsible for HERI-1 recruit- copy Resources On the North Quad imaging core at Harvard ment to chromatin]. Given the complex relationship that Medical School for microscope access, training, and helpful exists between nuclear RNAi,chromatinPTMs,andRNAi discussions; and Catherine Musselman at the University of

HERI-1 Limits TEI 1297 Iowa for the purified recombinant histone H3. Some strains Checchi, P. M., and J. Engebrecht, 2011 Caenorhabditis elegans were provided by the Caenorhabditis Genetics Center, which histone methyltransferase MET-2 shields the male X chromo- is funded by the National Institutes of Health (NIH) (P40 some from checkpoint machinery and mediates meiotic sex chromosome inactivation. PLoS Genet. 7: e1002267. https:// OD-010440). Some strains were provided by the Mitani lab- doi.org/10.1371/journal.pgen.1002267 oratory through the National Bio-Resource Project of the Conine, C. C., P. J. Batista, W. Gu, J. M. Claycomb, D. A. Chaves Ministry of Education, Culture, Sports, Science and Technol- et al., 2010 ALG-3 and ALG-4 are required for ogy (MEXT), Japan. R.P. and D.P. were funded by F32 spermatogenesis-specific 26G-RNAs and thermotolerant sperm awards from the NIH. B.D.F. was funded by a National Sci- in Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 107: 3588–3593. https://doi.org/10.1073/pnas.0911685107 ence Foundation Graduate Research Fellowship. This work Dickinson, D. J., A. M. Pani, J. K. Heppert, C. D. Higgins, and B. was supported by the NIH, RO1 GM-088289 (S.G.K.). The Goldstein, 2015 Streamlined genome engineering with a self- authors declare no conflict of interest. excising drug selection cassette. Genetics 200: 1035–1049. https://doi.org/10.1534/genetics.115.178335 Duchaine, T. F., J. A. Wohlschlegel, S. Kennedy, Y. Bei, D. Conte, Jr. et al., 2006 Functional proteomics reveals the biochemi- Literature Cited cal niche of C. elegans DCR-1 in multiple small-RNA-mediated pathways. Cell 124: 343–354. https://doi.org/10.1016/j. Alcazar, R. M., R. Lin, and A. Z. Fire, 2008 Transmission dynamics cell.2005.11.036 of heritable silencing induced by double-stranded RNA in Cae- D’Urso, A., and J. H. Brickner, 2014 Mechanisms of epigenetic norhabditis elegans. Genetics 180: 1275–1288. https://doi.org/ memory. Trends Genet. 30: 230–236. https://doi.org/ 10.1534/genetics.108.089433 10.1016/j.tig.2014.04.004 Alleman, M., L. Sidorenko, K. McGinnis, V. Seshadri, J. E. Dorweiler Eissenberg, J. C., 2012 Structural biology of the chromodomain: et al., 2006 An RNA-dependent RNA polymerase is required form and function. Gene 496: 69–78. https://doi.org/10.1016/ for paramutation in maize. Nature 442: 295–298. https://doi. j.gene.2012.01.003 org/10.1038/nature04884 Eyers, P. A., and J. M. Murphy, 2013 Dawn of the dead: protein Andersen, E. C., and H. R. Horvitz, 2007 Two C. elegans histone pseudokinases signal new adventures in cell biology. Biochem. methyltransferases repress lin-3 EGF transcription to inhibit vul- Soc. Trans. 41: 969–974. https://doi.org/10.1042/BST20130115 val development. Development 134: 2991–2999. https://doi. Farboud, B., and B. J. Meyer, 2015 Dramatic enhancement of org/10.1242/dev.009373 genome editing by CRISPR/Cas9 through improved guide Arribere, J. A., R. T. Bell, B. X. Fu, K. L. Artiles, P. S. Hartman et al., RNA design. Genetics 199: 959–971. https://doi.org/10.1534/ 2014 Efficient marker-free recovery of custom genetic modifi- genetics.115.175166 cations with CRISPR/Cas9 in Caenorhabditis elegans. Genetics Gapp, K., A. Jawaid, P. Sarkies, J. Bohacek, P. Pelczar et al., 198: 837–846. https://doi.org/10.1534/genetics.114.169730 2014 Implication of sperm RNAs in transgenerational inheri- Arteaga-Vazquez, M. A., and V. L. Chandler, 2010 Paramutation tance of the effects of early trauma in mice. Nat. Neurosci. 17: in maize: RNA mediated trans-generational gene silencing. Curr. 667–669. https://doi.org/10.1038/nn.3695 Opin. Genet. Dev. 20: 156–163. https://doi.org/10.1016/j. Gent, J. I., M. Schvarzstein, A. M. Villeneuve, S. G. Gu, V. Jantsch gde.2010.01.008 et al., 2009 A Caenorhabditis elegans RNA-directed RNA po- Ashe, A., A. Sapetschnig, E.-M. Weick, J. Mitchell, M. P. Bagijn lymerase in sperm development and endogenous RNA interfer- et al., 2012 piRNAs can trigger a multigenerational epigenetic ence. Genetics 183: 1297–1314. https://doi.org/10.1534/ memory in the germline of C. elegans. Cell 150: 88–99. https:// genetics.109.109686 doi.org/10.1016/j.cell.2012.06.018 Gent, J. I., A. T. Lamm, D. M. Pavelec, J. M. Maniar, P. Parameswaran Billi, A. C., S. E. J. Fischer, and J. K. Kim, 2014 Endogenous et al., 2010 Distinct phases of siRNA synthesis in an endoge- RNAi pathways in C. elegans (May 7, 2014), WormBook, ed. nous RNAi pathway in C. elegans soma. Mol. Cell 37: 679–689. The C. elegans Research Community, WormBook, doi/10.1895/ https://doi.org/10.1016/j.molcel.2010.01.012 wormbook.1.170.1, http://www.wormbook.org. https://doi.org/ Gu, S. G., J. Pak, S. Guang, J. M. Maniar, S. Kennedy et al., 10.1895/wormbook.1.170.1 2012 Amplification of siRNA in Caenorhabditis elegans gener- Buckley, B. A., K. B. Burkhart, S. G. Gu, G. Spracklin, A. Kershner ates a transgenerational sequence-targeted histone H3 lysine et al., 2012 A nuclear Argonaute promotes multigenerational 9 methylation footprint. Nat. Genet. 44: 157–164. https://doi. epigenetic inheritance and germline immortality. Nature 489: org/10.1038/ng.1039 447–451. https://doi.org/10.1038/nature11352 Guang, S., A. F. Bochner, D. M. Pavelec, K. B. Burkhart, S. Harding Burkhart, K. B., S. Guang, B. A. Buckley, L. Wong, A. F. Bochner et al., 2008 An Argonaute transports siRNAs from the cyto- et al., 2011 A pre-mRNA-associating factor links endogenous plasm to the nucleus. Science 321: 537–541. https://doi.org/ siRNAs to chromatin regulation. PLoS Genet. 7: e1002249. 10.1126/science.1157647 https://doi.org/10.1371/journal.pgen.1002249 Guang, S., A. F. Bochner, K. B. Burkhart, N. Burton, D. M. Pavelec Burton, N. O., K. B. Burkhart, and S. Kennedy, 2011 Nuclear RNAi et al., 2010 Small regulatory RNAs inhibit RNA polymerase II maintains heritable gene silencing in Caenorhabditis elegans. during the elongation phase of transcription. Nature 465: 1097– Proc. Natl. Acad. Sci. USA 108: 19683–19688. https://doi. 1101. https://doi.org/10.1038/nature09095 org/10.1073/pnas.1113310108 Heard, E., and R. A. Martienssen, 2014 Transgenerational epige- Carone, B. R., L. Fauquier, N. Habib, J. M. Shea, C. E. Hart et al., netic inheritance: myths and mechanisms. Cell 157: 95–109. 2010 Paternally induced transgenerational environmental re- https://doi.org/10.1016/j.cell.2014.02.045 programming of metabolic gene expression in mammals. Cell Holoch, D., and D. Moazed, 2015 RNA-mediated epigenetic reg- 143: 1084–1096. https://doi.org/10.1016/j.cell.2010.12.008 ulation of gene expression. Nat. Rev. Genet. 16: 71–84. https:// Castel, S. E., and R. A. Martienssen, 2013 RNA interference in the doi.org/10.1038/nrg3863 nucleus: roles for small RNAs in transcription, Kalinava, N., J. Z. Ni, K. Peterman, E. Chen, and S. G. Gu, and beyond. Nat. Rev. Genet. 14: 100–112. https://doi.org/ 2017 Decoupling the downstream effects of germline nuclear 10.1038/nrg3355 RNAi reveals that H3K9me3 is dispensable for heritable RNAi

1298 R. Perales et al. and the maintenance of endogenous siRNA-mediated transcrip- programming. PLoS Genet. 8: e1002605. https://doi.org/10.1371/ tional silencing in Caenorhabditis elegans. Epigenetics Chroma- journal.pgen.1002605 tin 10: 6. https://doi.org/10.1186/s13072-017-0114-8 Ragunathan, K., G. Jih, and D. Moazed, 2015 Epigenetics. Epige- Kennedy, S., D. Wang, and G. Ruvkun, 2004 A conserved siRNA- netic inheritance uncoupled from sequence-specificrecruitment. degrading RNase negatively regulates RNA interference in Science 348: 1258699. https://doi.org/10.1126/science.1258699 C. elegans. Nature 427: 645–649. https://doi.org/10.1038/ Rankin, C. H., 2015 A review of transgenerational epigenetics for nature02302 RNAi, longevity, germline maintenance and olfactory imprinting Lee, R. C., C. M. Hammell, and V. Ambros, 2006 Interacting en- in Caenorhabditis elegans. J. Exp. Biol. 218: 41–49. https://doi. dogenous and exogenous RNAi pathways in Caenorhabditis ele- org/10.1242/jeb.108340 gans. RNA 12: 589–597. https://doi.org/10.1261/rna.2231506 Sapetschnig, A., P. Sarkies, N. J. Lehrbach, and E. A. Miska, Le Thomas, A., E. Stuwe, S. Li, J. Du, G. Marinov et al., 2015 Tertiary siRNAs mediate paramutation in C. elegans. 2014 Transgenerationally inherited piRNAs trigger piRNA bio- PLoS Genet. 11: e1005078. https://doi.org/10.1371/journal. genesis by changing the chromatin of piRNA clusters and induc- pgen.1005078 ing precursor processing. Genes Dev. 28: 1667–1680. https:// Schedl, T., and J. Kimble, 1988 fog-2, a germ-line-specific sex doi.org/10.1101/gad.245514.114 determination gene required for hermaphrodite spermatogene- Lev, I., U. Seroussi, H. Gingold, R. Bril, S. Anava et al., 2017 MET- sis in Caenorhabditis elegans. Genetics 119: 43–61. 2-dependent H3K9 methylation suppresses transgenerational Sharma, U., C. C. Conine, J. M. Shea, A. Boskovic, A. G. Derr et al., small RNA inheritance. Curr. Biol. 27: 1138–1147. https://doi. 2016 Biogenesis and function of tRNA fragments during sperm org/10.1016/j.cub.2017.03.008 maturation and fertilization in mammals. Science 351: 391– Lin, R., 2003 A gain-of-function mutation in oma-1, a C. elegans 396. https://doi.org/10.1126/science.aad6780 gene required for oocyte maturation, results in delayed degrada- Shirayama, M., M. Seth, H.-C. Lee, W. Gu, T. Ishidate et al., tion of maternal proteins and embryonic lethality. Dev. Biol. 258: 2012 piRNAs initiate an epigenetic memory of nonself RNA 226–239. https://doi.org/10.1016/S0012-1606(03)00119-2 in the C. elegans germline. Cell 150: 65–77. https://doi.org/ Mao, H., C. Zhu, D. Zong, C. Weng, X. Yang et al., 2015 The Nrde 10.1016/j.cell.2012.06.015 pathway mediates small-RNA-directed histone H3 lysine 27 tri- Shorter, J., and S. Lindquist, 2005 as adaptive conduits of methylation in Caenorhabditis elegans. Curr. Biol. 25: 2398– memory and inheritance. Nat. Rev. Genet. 6: 435–450. https:// 2403. https://doi.org/10.1016/j.cub.2015.07.051 doi.org/10.1038/nrg1616 Martienssen, R., and D. Moazed, 2015 RNAi and Sijen, T., J. Fleenor, F. Simmer, K. L. Thijssen, S. Parrish et al., assembly. Cold Spring Harb. Perspect. Biol. 7: a019323. https:// 2001 On the role of RNA amplification in dsRNA-triggered doi.org/10.1101/cshperspect.a019323 gene silencing. Cell 107: 465–476. https://doi.org/10.1016/ Martienssen, R. A., M. Zaratiegui, and D. B. Goto, 2005 RNA in- S0092-8674(01)00576-1 terference and heterochromatin in the fission yeast Schizosac- Somer, R. A., and C. S. Thummel, 2014 Epigenetic inheritance of charomyces pombe. Trends Genet. 21: 450–456. https://doi. metabolic state. Curr. Opin. Genet. Dev. 27: 43–47. https://doi. org/10.1016/j.tig.2005.06.005 org/10.1016/j.gde.2014.03.008 Meister, G., 2013 Argonaute proteins: functional insights and Spracklin, G., B. Fields, G. Wan, D. Vijayendran, A. Wallig et al., emerging roles. Nat. Rev. Genet. 14: 447–459. https://doi. 2017 Identification and characterization of Caenorhabditis el- org/10.1038/nrg3462 egans RNAi inheritance machinery. Genetics. 2017 Jul;206(3): Minevich, G., D. S. Park, D. Blankenberg, R. J. Poole, and O. Hobert, 1403–1416 2012 CloudMap: a cloud-based pipeline for analysis of mutant Towbin, B. D., C. González-Aguilera, R. Sack, D. Gaidatzis, V. Kalck genome sequences. Genetics 192: 1249–1269. https://doi.org/ et al., 2012 Step-wise methylation of histone H3K9 positions 10.1534/genetics.112.144204 heterochromatin at the nuclear periphery. Cell 150: 934–947. Nolen, B., S. Taylor, and G. Ghosh, 2004 Regulation of protein https://doi.org/10.1016/j.cell.2012.06.051 kinases; controlling activity through activation segment confor- Vasale, J. J., W. Gu, C. Thivierge, P. J. Batista, J. M. Claycomb et al., mation. Mol. Cell 15: 661–675. https://doi.org/10.1016/j.mol- 2010 Sequential rounds of RNA-dependent RNA transcription cel.2004.08.024 drive endogenous small-RNA biogenesis in the ERGO-1/Argo- Padmanabhan, N., D. Jia, C. Geary-Joo, X. Wu, A. C. Ferguson- naute pathway. Proc. Natl. Acad. Sci. USA 107: 3582–3587. Smith et al., 2013 Mutation in folate metabolism causes epi- https://doi.org/10.1073/pnas.0911908107 genetic instability and transgenerational effects on develop- Vastenhouw, N. L., K. Brunschwig, K. L. Okihara, F. Müller, M. ment. Cell 155: 81–93. https://doi.org/10.1016/j.cell.2013. Tijsterman et al., 2006 Gene expression: long-term gene silenc- 09.002 ing by RNAi. Nature 442: 882. https://doi.org/10.1038/442882a Paix, A., Y. Wang, H. E. Smith, C. Y. Lee, D. Calidas et al., Volpe, T. A., C. Kidner, I. M. Hall, G. Teng, S. I. S. Grewal et al., 2014 Scalable and versatile genome editing using linear 2002 Regulation of heterochromatic silencing and histone H3 with microhomology to Cas9 Sites in Caenorhabditis elegans. lysine-9 methylation by RNAi. Science 297: 1833–1837. Genetics 198: 1347–1356. https://doi.org/10.1534/genet- https://doi.org/10.1126/science.1074973 ics.114.170423 Walker, D. M., and A. C. Gore, 2011 Transgenerational neuroen- Pavelec, D. M., J. Lachowiec, T. F. Duchaine, H. E. Smith, and S. docrine disruption of reproduction. Nat. Rev. Endocrinol. 7: Kennedy, 2009 Requirement for the ERI/DICER complex in 197–207. https://doi.org/10.1038/nrendo.2010.215 endogenous RNA interference and sperm development in Cae- Wan, G., B. D. Fields, G. Spracklin, A. Shukla, C. M. Phillips et al., norhabditis elegans. Genetics 183: 1283–1295. https://doi.org/ 2018 Spatiotemporal regulation of liquid-like condensates in 10.1534/genetics.109.108134 epigenetic inheritance. Nature 557: 679–683. https://doi.org/ Radford, E. J., E. Isganaitis, J. Jimenez-Chillaron, J. Schroeder, M. 10.1038/s41586-018-0132-0 Molla et al., 2012 An unbiased assessment of the role of im- printed genes in an intergenerational model of developmental Communicating editor: D. Greenstein

HERI-1 Limits TEI 1299