Pseudouridine guides germline small RNA transport and epigenetic inheritance

Abstract

Developmental epigenetic modifications in plants and animals are mostly reset during gamete formation but some are inherited from the germline. Small RNAs guide these epigenetic modifications but how inherited small RNAs are distinguished in plants and animals is unknown. Pseudouridine (Ψ) is the most abundant RNA modification but has not been explored in small RNAs. Here, we develop assays to detect Ψ in short RNA sequences, demonstrating its presence in mouse and Arabidopsis microRNAs. Germline small RNAs, namely epigenetically activated small interfering RNAs (easiRNAs) in Arabidopsis pollen and Piwi-interacting RNAs in mouse testes, are enriched for Ψ. In pollen, pseudouridylated easiRNAs are transported to sperm cells from the vegetative nucleus, and PAUSED/HEN5 (PSD), the plant homolog of Exportin-t, interacts genetically with Ψ and is required for this transport. We further show that Exportin-t is required for the triploid block: small RNA dosage-dependent seed lethality that is epigenetically inherited from pollen. Thus, Ψ has a conserved role in marking inherited small RNAs in the germline.

Fig. 1. microRNAs are pseudouridylated in plants and mammals.

a, BiFC confirming interaction of AGO3 with the DSKC1 cofactor NHP2 in Arabidopsis (Methods). Split yellow fluorescence protein (YFP) was fused to the C and N termini or both N termini of AGO3 and NHP2. Reconstituted YFP signifies interaction. p35S::RFP (red fluorescence protein) acted as a transformation control. b, Volcano plot showing proteins copurified with GFP-tagged NHP2 compared to IP performed in WT plants. NHP and AGO family proteins are depicted in orange (n = 3 biological replicates; statistical significance was calculated using edgeR (Methods); P value was adjusted by Benjamini–Hochberg correction). c, Structure of uridine and Ψ, enzymes involved in catalysis and methods of detection with specific antibodies and chemical modification with CMC (Methods and Extended Data Fig. 2). d, Volcano plot of miRNAs enriched by Ψ-IP compared to unbound fractions from Arabidopsis flower buds. Blue dots, significantly enriched or unbound miRNAs (adjusted P < 0.01); orange, both Ψ-IP enriched and depleted from libraries following CMC treatment (adjusted P < 0.01); dark red, not significant. Statistical significance was calculated using DESeq2; P value was adjusted by Benjamini and Hochberg method (n = 5 biological replicates). e, Venn diagram showing overlap of miRNAs detected in flower buds by each technique. Predicted Ψ sites in miRNAs detected by all three techniques. f, miRNAs enriched by Ψ-IP in WT and dkc1 mutants in Arabidopsis leaves. Error bars indicate the s.e.m. of log₂(fold change) estimated by DESeq2; n = 3 WT and n = 4 dkc1 biological replicates. g, Metaplot of modification frequency of miRNAs at each site based on proximity to the 5′ or 3′ end determined by CMC/Mn²⁺ sequencing. h, Northern blots of three synthetic 21-mer oligoribonucleotides (sequence: UGACACAGGACUACGGACGUAU) either unpseudouridylated or pseudouridylated at position 10 or 21, treated (+) or mock-treated (−) with CMC and probed with a matching DIG-labeled probe (representative image of three independent experiments). i, Flower bud small RNA treated (+), mock-treated (−) or untreated (no mock treatment) with CMC and probed with miR159b to reveal mobility shifts (representative image of four independent experiments). Source data

Fig. 2. Pollen siRNAs are highly pseudouridylated and loaded into AGO proteins in sperm cells.

a, Heat maps of individual miRNA Ψ enrichment in flower buds and pollen detected by Ψ-IP (log₂ IP/unbound; n = 5 biological replicates) and by CMC-mediated depletion from small RNA libraries (log₂ CMC⁺/CMC⁻; n = 2 treatment replicates) and assessed by DESeq2. b, Heat maps of siRNA Ψ enrichment from individual TE families in flower buds and pollen detected by Ψ-IP and CMC depletion and assessed as in a. Small RNAs from most TE families were enriched for Ψ in pollen. c, Small RNAs were sequenced from sperm cells isolated from pollen grains by FACS, followed by RIP-seq with antibodies to AGO proteins. SC, sperm cell. d, GFP::AGO5 fusion protein expressed from AGO5 promoter (pAGO5) in mature pollen is found in the perinuclear space around sperm cells. DAPI was used to stain the VN. Reproduced with permission from Borges et al.. e, AGO9::GFP fusion protein expressed from the AGO9 promoter (pAGO9) in mature pollen. f–i, Association of small RNA pseudouridylation in pollen, determined by Ψ-IP (f,h) and CMC depletion (g,i) with sperm cell abundance by FACS, for both miRNAs (f,g) and siRNAs from TE families (h,i)^,. j, Bar chart showing number of preferred TE families for each AGO protein split by siRNA size (20–22 and 23–25 nt). Each TE family was assessed for enrichment in AGO1, AGO5 and AGO9 IP as a function of two size classes (20–22 and 23–25 nt), with the highest enrichment signaling a preference for a certain AGO. k, Heat maps showing siRNA Ψ enrichment from individual TE families split by AGO preference and size class. Source data

Fig. 3. Transport of pseudouridylated siRNAs into sperm cells depends on PAUSED (Exportin-t).

a, Ψ enrichment of miRNAs in WT and psd-13 pollen. A modest reduction in Ψ enrichment of miRNAs was observed in psd pollen (P = 0.0019, determined by two-tailed paired t-test; n = 3 biological replicates; 64 miRNAs are plotted). b, Ψ enrichment of siRNAs was determined by Ψ-IP and small RNA-seq in WT and psd-13 mutant pollen and plotted by size class (20–22 nt or 23–25 nt) for individual TEs belonging to each group, namely, miR845 targets, Pol IV and non-Pol IV targets and TEs whose siRNAs were derived from sperm cells (SC) or from the VN in ref. . P values show differences between group means, determined by one-way ANOVA with Bonferroni correction for multiple comparisons (P = 0.002, 0.015, <0.0001, 0.234, 0.5964, 0.5153, <0.0001, 0.0009, 0.0002 and 0.0049; n = 3 biological replicates). Numbers above violins indicate the number of individual TEs plotted (including replicates). c, siRNA abundance in SC was determined by sequencing small RNA from FACS-sorted WT, psd-13 and hst-6 SC. VN-derived 20–22-nt siRNAs were depleted from psd SC but enriched in hst SC relative to WT, while VN-derived 23–25-nt siRNAs and SC-derived siRNAs were at levels similar to WT. Numbers on bars indicate the number of individual TEs plotted (including replicates); error bars indicate the s.e.m. d,e, Abundance of highly prevalent siRNA from 30 Gypsy and 2 Copia TE families in psd and hst sperm relative to WT for 20–22 nt (d) and 23–25 nt (e) size classes. f, Abundance of highly prevalent miRNAs (reads per million mapped reads (RPM) > 100) in psd and hst sperm relative to WT. miRNAs with terminal Ψ are indicated with a black bar; miRNAs presented in h are shown in bold. g, Abundance of miRNAs with and without terminal Ψ in psd and hst mutants relative to WT in sperm cells sorted by FACS. miRNAs without terminal Ψ were lost in hst, but not in psd sperm cells (P < 0.0001, 0.6571 and <0.0001, determined by two-way ANOVA; n = 3 biological replicates). Numbers above violins indicate the number of individual miRNAs plotted. h, Ψ enrichment of miRNAs in WT and psd-13 pollen for those miRNAs dependent on PSD for pseudouridylation (P = 0.02, 0.02, 0.021, 0.037 and 0.039, determined by two-sided Student’s t-test; n = 3 biological replicates). Error bars indicate the estimated s.e.m. of log₂(fold change) calculated by DESeq2. In a,b,g,h, *P < 0.05, **P < 0.01 and ***P < 0.001. Source data

Fig. 4. Ψ and Exportin-t mediate epigenetic inheritance and the triploid block.

a, Mean Ψ enrichment (log₂ Ψ-IP/unbound) of 20–22-nt siRNAs and 23–25-nt siRNAs surrounding imprinted genes (−2.5 kb upstream of start; CDS; +2.5 kb downstream of stop) targeted by DME in WT (blue) and psd-13 (orange) pollen (P = <0.0001, 0.4528, <0.0001, 0.0002, 0.0428 and 0.0003, determined by two-way ANOVA with Bonferroni correction for multiple comparisons; n = 3 biological replicates). Error bars indicate the s.e.m. CDS, coding sequence. b,c, Metaplots of sperm cell abundance of 20–22-nt siRNAs (red) and 23–25-nt siRNAs (black) surrounding imprinted genes targeted by DME in WT (b) and psd-13 (c) pollen. d–g, Images of representative seeds from crosses between WT Col-0 female and osd1 diploid pollen (d), double-mutant osd1;psd-13^−/− diploid pollen (e), double-mutant osd1;psd-13^+/− diploid pollen (f) and double-mutant osd1;nrpd1a-3 diploid pollen (g). NRPD1a encodes the large subunit of Pol IV. h, Frequency of aborted seeds from the same crosses. i, Model for transport of small RNA during pollen development. As the pollen matures, 21–22-nt easiRNAs and 24-nt siRNAs are produced by Pol IV in the VN and pseudouridylated along with some miRNAs. Pseudouridylated miRNAs and easiRNAs are exported by PSD (Exportin-t) into sperm cells, where they are loaded onto AGO1, AGO5 and AGO9, while unmodified miRNAs are transported by HST (Exportin-5) and loaded onto AGO5 and AGO1. Pseudouridylated sperm cell easiRNAs mediate dose-dependent lethality (the triploid block) by targeting MEGs in the seed. j, mCherry::AGO9 fusion protein expressed from the AGO9 promoter (pAGO9) in developing pollen. DAPI was used to stain nucleic acids (blue). AGO9 is present in the VN at the bicellular stage but is enriched in the sperm cells in mature pollen. k, AGO9 IP from WT and psd-13 pollen shows increased binding of VN-derived 20–22-nt siRNAs and decreased binding of 23–25-nt sperm-cell-derived siRNAs to AGO9 in psd-13. Numbers on bars indicate the number of TEs analyzed; error bars indicate the s.e.m. P = <0.0001, 0.99, 0.99 and <0.0001, determined by one-way ANOVA with Bonferroni correction for multiple comparisons. In a,k, *P < 0.05, **P < 0.01 and ***P < 0.001. NS, not significant. Source data

Extended Data Fig. 1. miRNA from plants and mammals are pseudouridylated.

(a) Immunoprecipitation with anti-Ψ antibody enriches for Ψ-containing RNAs produced through in vitro transcription (IVT) as determined by mass spectrometry (n = 3 biological replicates, error bars ± SD). (b) Immuno-dot blot of 100 ng total input RNA or 100 ng RNA immunoprecipitated with anti-Ψ antibody or control immunoglobulin G (IgG). 100 ng of 100% Ψ IVT RNA was used as control. (c) Volcano plot of miRNAs enriched by small RNA immunoprecipitation using a Ψ-specific antibody (Ψ-IP) in mouse fibroblast NIH/3T3 cells compared to input fractions. Significantly enriched and depleted miRNAs are highlighted in blue (padj<0.01; DESeq2, Benjamini and Hochberg correction for multiple comparisons; n = 3 biological replicates). (d) Validation of enriched miRNAs using qRT-PCR. (e) Ψ-enrichment of miRNAs affected by PUS1 knockdown (KD) 3 days and 6 days after PUS1 shRNA transfection of NIH/3T3 cells. (f) Arabidopsis AGO proteins and Dyskerin subunit NHP2 were fused at the N-terminus to N- and C- terminally split YFP, respectively. Positive interaction is shown by YFP signal, with 35S::RFP as an expression control. (g) Ψ-enrichment of precursor pri-miRNAs in Arabidopsis flower buds plotted against pri-miRNAs bound to NHP2 (blue) and detected by RIP-seq in 2 or more replicates (orange). (h) Violin plot showing significant difference between NHP2-enriched and non-enriched pri-miRNA in Ψ enrichment of the mature miRNA by Ψ -IP (Log₂ IP/IgG; numbers below violins indicate number of miRNA analyzed;one-way ANOVA; **p < 0.01). Source data

Extended Data Fig. 2. Strategies for robust detection of Ψs in small RNAs.

(a) Three strategies to detect Ψ in small RNAs and their predicted effect on pseudouridylated sequences. Details are in Methods (b) Known sites of pseudouridylation in tRNAs mediated by PUS7 (targets UNUAR) and PUS10 (targets Ψ-55). Ψ-55 is 22 nt ‘upstream’ of the 3ʹ end of -CCA tailed tRNA fragments. (c) Heatmap showing individual tRNA fragments (tRFs) with and without the UNUAR motif in wild-type and pus7−/− libraries treated with CMC. (d) 3′CCA tRFs ≥22 nt contain canonical Ψ-55 and are enriched by anti-Ψ antibody and depleted by CMC treatment, whereas shorter fragments do not contain this nucleoside and are not enriched/depleted. Ψ-55 is ablated in mutants of PUS10. (e) Deletions in 3ʹ CCA tRFs based on position in CMC/Mn²⁺- and mock-treated libraries. (f, g) Metaplots showing coverage depletion/enrichment in CMC/IP libraries at rRNA pseudouridylation sites (f) compared to mismatches from Mn2 + /CMC+ libraries (g). Average across 42 predicted pseudouridylated sites in 18S and 25S rRNA and from 17 predicted sites in 18S rRNA (CMC mismatch/Mock mismatch). (h, i) Enrichment/depletion/mismatches across 18S rRNA in CMC/IP (h) and CMC/Mn²⁺ (i) libraries. Predicted Ψs are marked with dotted lines. Source data

Extended Data Fig. 3. Arabidopsis mutants in Dyskerin/NAP57.

(a) Generation of dkc1-2 transgenic plants. nap57/dkc1 (SALK_031065 T-DNA insertion) was complemented with the pABI3:AtNAP57 transgene (b) AtNAP57 transcript level in dkc1-2 transgenic lines analyzed by qPCR (n = 3 biological replicates; error bars show mean and SD) (c) dkc1-2 transgenic plants show severe developmental defects including dwarfism, abnormal leaf shape, deformed and infertile flowers. (d) Enrichment of miRNAs by Ψ-IP in wild-type and dkc1-3 leaves based on DESeq2 analysis. Boxes show median, upper/lower quartiles and the Tukey min/max with outliers represented as dots, X marks the mean value. N = 3 biological replicates, 44 miRNAs with L2FC > 1, p-adj<0.01 are plotted; significance calculated using Mann-Whitney test, p = 0.003. Source data

Extended Data Fig. 4. Mapping of miRNA pseudouridylation sites in Arabidopsis by CMC labeling and reverse transcription in the presence of Mn²⁺ (CMC/Mn²⁺).

(a, b) Bar chart showing proportions of 5ʹ and 3ʹ nucleotides in Ψ-IP-enriched (a) and CMC-depleted (b) siRNAs from inflorescence compared to background (non-structural) small RNAs. IP-enrichment and CMC-depletion show biases towards 5ʹ and 3ʹ U. (c) Sites of predicted Ψ (insertions and deletions) in precursor pri-miR159b PCR amplicons (upper panels) and mature miR159b (lower panel; high-throughput sequencing) in seedlings, leaves and flower buds of wild-type and hyl1-2 microprocessor mutants. pri-miRNA amplicons (upper) and mature small RNA sequences (lower) had insertions and deletions at predicted sites. Number of clones with the respective amplicon is shown in brackets. Source data

Extended Data Fig. 5. Pseudouridine in miRNAs and TE-derived siRNAs in pollen from WT and paused (psd-13)/exportin-t mutants.

(a) Metaplots of read coverage at 42 known pseudouridylated sites in 18S and 25S rRNA, relative to untreated controls, in CMC-treated WT (magenta) and Ψ-IP WT (blue) and psd-13 (orange) small RNA sequencing libraries from pollen. (b) Read coverage from the same libraries at 17 rRNA pseudouridylation sites in 18S rRNA. (c, d) metaplot (c) and browser (d) analysis of predicted sites in rRNA from CMC/Mn²⁺ analysis relative to mock treated controls. (e, f) Metaplots (e) and read coverage (f) relative to untreated controls in WT (blue) and psd-13 (orange) Ψ-IP small RNA sequencing libraries from inflorescence. Known positions of Ψ in b, d and f are marked with dotted lines. (g) Ψ-enrichment of individual miRNAs detected by Ψ-IP and small RNA sequencing in WT and psd mutant inflorescence (L2FC calculated by DESeq2 with estimate of SE; n = 5 and 2 biological replicates for WT and psd, respectively; miR403 p = 0.01027, two-tailed students t-test). (h) Ψ-enrichment of miRNAs overall in WT and psd mutant inflorescence (two-way ANOVA, p = 0.2747; n = 153 miRNAs) (i) Violin plot of Ψ-enrichment of siRNAs from individual TEs based on IP-enrichment (Log₂ IP/Unbound) in flower buds (inflorescence) and pollen from wild-type and psd mutants. Significant Ψ enrichment was observed in pollen, but not in inflorescence, and depended on PSD (p < 0.0001; p = 0.9181, one-way ANOVA; numbers under violins indicate number of TEs analyzed). Asterisks in g and i indicate significance (p < 0.05 (*); p < 0.01 (**); p < 0.001 (***)). Source data

Extended Data Fig. 6. Mutation of psd and pus7 is synthetically lethal in Arabidopsis.

(a) Alexander staining of pollen from Col-0, pus7, psd-13, psd-13−/−; pus7 + /−, and psd-13 + /−; pus7−/− parents. psd-13 + /; pus7−/− had semisterile pollen abortion (loss of Alexander staining, arrows). (b) Mature seeds from each parent. Aborted seeds in psd-13are highlighted with arrows. (c) Quantification of approximately 500 seeds from each genotype revealed psd-13−/−; pus7 + /− aborts seeds at a rate of ~25% (* indicates no difference from expected 3:1 normal:aborted ratio, p > 0.1 Chi-square test). Double homozygous mutants could not be obtained (n = 150 plants). Source data

Extended Data Fig. 7. Characterization of pseudouridylation of siRNA in psd-13 mutants.

(a,b) Ψ-enrichment of PolIV-dependent and -independent (a) or VN- and SC-derived (b) TE-siRNAs in wild-type and psd-13 pollen for each size class (average enrichment of TE-siRNAs from each classification is shown +/− s.e.m. for each biological replicate; numbers below bars indicate number of TEs analyzed). (c) siRNA abundance was determined by sequencing small RNA from FACS sorted wild-type, psd-13 and hst-6 sperm cells. SC abundance of PolIV-dependent and non-PolIV-dependent TE-siRNAs was assessed in each size class (average enrichment of TE-siRNAs relative to wild-type from each TE classification is shown +/− s.e.m.; numbers below bars indicate number of TEs analyzed). (d-i) Metaplots of 20-22nt (red) and 23-25nt (black) siRNA abundance across PolIV-TEs from biological replicates of wild-type (a,d), psd-13 (b,e) and hst-6 (c,f) sperm cells sorted by FACS. 250 bp up- and down-stream regions as well as 500 bp from each end of the TE are shown. (j-l) Scatterplots of small RNA sequencing data from two biological replicates mapped to TE families in wild-type (j), psd-13 (k) and hst-6 (l) sperm cells isolated by FACS. Simple linear regression used to determine R-squared value. (m) Ψ-enrichment of siRNAs from TEs surrounding maternally expressed imprinted genes (MEGs, n = 42) and paternally expressed imprinted genes (PEGs, n = 9) in wild type (blue) and psd-13 (orange) pollen. * = p < 0.05, two-sided paired t-test. (n) Abundance of 20-22 nt and 23-25 nt siRNA surrounding imprinted genes in wild-type (blue) and psd-13 (orange) sperm cells. siRNA abundance in sperm cells matching MEGs and PEGs was lower in psd-13 mutants for 20-22nt siRNA but relatively higher for 23-25nt siRNA. * = p < 0.05; ** p < 0.01, *** = p < 0.001, paired t-test with Bonferroni correction for multiple comparisons; numbers below boxes indicate individual TEs analyzed. Boxes indicate median, interquartile range and min/max (y-axis clipped to visualize median). (o) Levels of miR845a and miR845b were unchanged or higher in psd-13 mutant pollen (Bars show average RPM; error bars show SD of biological replicates, n = 3). Source data

Extended Data Fig. 8. Pseudouridine in piRNAs and tRNA fragments (tRFs) from mouse testis.

(a, b) Ψ detection by Ψ-IP (a) and CMC-depletion (b) of small RNA from testes of 8 weeks-old male mice. Dots correspond to tRF counts per tRNA isoacceptor. Differential expression (fold change) and statistical significance were calculated using the DESeq2 package. Statistically significant (p-adj. <0.01, log2 fold change > |1|) enriched or depleted tRFs are in black, 3’-tRFs are colored blue; grey dots mark tRFs without significant enrichment/depletion. (c) Volcano plot of read counts per piRNA sequence in testes (8 weeks old males) by CMC-depletion and sequencing. Black: significant fold changes (p-adj. <0.01, log2 fold change > |1|) as per DESeq2 analysis; grey: not significant; blue: significantly enriched or depleted piRNAs overlapping LTR and LINE transposon sequences. (d) Volcano plot of read counts per piRNA sequence by Ψ-IP. Black: significant fold changes (p-adj. <0.01, log2 fold change > |1|) as per DESeq2 analysis; grey: not significant; blue: significantly enriched or depleted piRNAs overlapping LTR and LINE transposon sequences. (e,f) Volcano plot of read counts per piRNA cluster using Ψ-IP (e) or CMC-depletion (f) and DESeq2 for statistical analysis. (g,h) Size distribution of 3’-tRFs (g) and piRNA (h) sequences (without structural RNAs) using CMC-depletion and sequencing. (i, j) Size distribution of 3’-tRFs (i) and piRNA (j) sequences (without structural RNAs) using Ψ-IP and sequencing. RPM: reads per million mapped reads. Source data