Epigenetic inheritance of diet-induced and sperm-borne mitochondrial RNAs

Abstract

Spermatozoa harbour a complex and environment-sensitive pool of small non-coding RNAs (sncRNAs)¹, which influences offspring development and adult phenotypes^1-7. Whether spermatozoa in the epididymis are directly susceptible to environmental cues is not fully understood⁸. Here we used two distinct paradigms of preconception acute high-fat diet to dissect epididymal versus testicular contributions to the sperm sncRNA pool and offspring health. We show that epididymal spermatozoa, but not developing germ cells, are sensitive to the environment and identify mitochondrial tRNAs (mt-tRNAs) and their fragments (mt-tsRNAs) as sperm-borne factors. In humans, mt-tsRNAs in spermatozoa correlate with body mass index, and paternal overweight at conception doubles offspring obesity risk and compromises metabolic health. Sperm sncRNA sequencing of mice mutant for genes involved in mitochondrial function, and metabolic phenotyping of their wild-type offspring, suggest that the upregulation of mt-tsRNAs is downstream of mitochondrial dysfunction. Single-embryo transcriptomics of genetically hybrid two-cell embryos demonstrated sperm-to-oocyte transfer of mt-tRNAs at fertilization and suggested their involvement in the control of early-embryo transcription. Our study supports the importance of paternal health at conception for offspring metabolism, shows that mt-tRNAs are diet-induced and sperm-borne and demonstrates, in a physiological setting, father-to-offspring transfer of sperm mitochondrial RNAs at fertilization.

Fig. 1. Paternal overweight at conception is important for offspring metabolism in mice and humans.

a, Experimental design. CD, chow diet. b, Glucose tolerance of unexposed male offspring (F₁) of HFD-exposed bucks. Top: AUC_ipGTT. Bottom: frequency distribution analysis to identify tolerant and intolerant animals. n = 60 male mice across 4 cohorts with 5 litters each and 3 males per litter (eLFD and eHFD bucks); n = 10–15 (sLFD and sHFD bucks) including 1 cohort with 5 litters and 3 males per litter. c,d, Glucose tolerance (c; mean ± s.e.m.) and insulin sensitivity (d; mean ± s.e.m.) of male offspring (F₁) of HFD-exposed bucks. Data represent a re-phenotyping of the animals in c carried out 8 weeks after the first phenotyping. Significance calculated by a two-way (c; n in graph) or one-way (d; n as in c) analysis of variance (ANOVA; ***P < 10⁻⁴). ITT, insulin tolerance test. e,f, PCA plot (e) and functional enrichment analysis (KEGG (Kyoto Encyclopedia of Genes and Genomes); f) of peripheral tissue RNA-seq data from HFDt and HFDi F₁ animals. eWAT, epididymal white adipose tissue; metab., metabolism; AA, amino acid; TH, thyroid hormone; OXPHOS, oxidative phosphorylation; FA, fatty acid; TCA, tricarboxylic acid. g, Left: overlap between genes differentially expressed in tissues from HFDi mice and their human orthologues associated with childhood obesity. Right: functional enrichment analysis (KEGG) of the overlapping genes (n = 693) pre-classified as protective and risk genes for childhood obesity on the basis of the β-score for BMI-SDS. DEGs, differentially expressed genes; Hs, Homo sapiens; FDR, false discovery rate; NSCLC, non-small cell lung cancer; CML, chronic myeloid leukaemia; NAFLD, non-alcoholic fatty liver disease. h, Scatter plot of children’s body weight trajectories as a function of paternal BMI at conception in families with mothers who were lean (red line; r = 0.2611; P value < 10⁻⁴) or overweight (blue line; r = 0.3467; P value < 10⁻⁴) at conception. Significant association calculated by linear regression analysis. i, Insulin sensitivity, measured as ISI_Matsuda (top) or homeostatic model assessment for insulin resistance (HOMA-IR; bottom) indices in children as a function of parental weight status at conception. n lean–lean = 106; overweight–lean = 184; lean–overweight = 114; overweight–overweight = 415. Data represented as mean ± s.e.m. Significance calculated by two-way ANOVA (details in the graph). Source Data

Fig. 2. mt-tsRNAs are sperm-borne sensors of male health.

a, Distribution of sncRNA biotypes in cauda spermatozoa from LFD- and HFD-fed mice. n = 3 samples per diet. lincRNA, long intergenic non-coding RNA; miRNA, microRNA. b, Biotype-specific differential expression analysis (P_adj < 0.05; log₂[fold change (FC)] > |1|) of cauda spermatozoa sncRNAs. c, Histogram (bars at a bin centre of 0.5 with the overlapping non-linear fit curve) showing the distribution of the log₂[FC(HFD versus LFD)] in nuclear- (n) and mitochondrial- (mt) derived tsRNAs (left) and rsRNAs (right). d, Volcano plot representation of differentially expressed mt-tsRNAs (differential expression calculated by EdgeR; significance defined as P_adj < 0.05). e, Fragmentation pattern of mt-tRNAs in cauda spermatozoa from LFD- and HFD-fed mice. Significance tested by a two-tailed t-test, HFD versus LFD (mean ± s.e.m.; n = 3; *P < 0.05). f, Pearson-based correlation analysis of individual sncRNA biotype expression in human sperm with BMI (two-tailed P value with 95% confidence interval). g, Differentially expressed tsRNAs in human sperm from lean and overweight donors calculated by DESeq2-based continuous differential expression analysis (n = 18 donors). VST, variance-stabilizing transformation. h, Heat map representation of mature mt-tRNA levels in cauda spermatozoa from LFD- and HFD-fed mice. i, Relative abundance of the indicated sncRNA biotypes in epididymosomes (Epi) or spermatozoa (Sp) isolated from the caput, corpus and cauda epididymis. Data reanalysed from refs. ^,. j, Uniform manifold approximation and projection (UMAP) representation of mtDNA transcription during spermatogenesis from testis single-cell RNA-seq data.

Fig. 3. Sperm mt-tRNA sequences are transferred to the oocytes at fertilization.

a, Experimental design to generate and analyse single hybrid early two-cell embryos generated through IVF with BL6 sperm (from LFD- or HFD-fed mice) and ST oocytes. b,c, PCA plot representation of mitochondrial transcriptomes in single female (b) and male (c) embryos. Insets represent the variance at the two main principal components (whiskers are the 5th and 95th percentile of the distribution; experimental and analytical details in the Methods). d,e, Density plot (d) and heat map (e) representation of the quantified heteroplasmy at the 416 SNPs mapped between BL6 and ST mitochondrial genomes. f, Relative heteroplasmy enrichment of the highlighted mt-tRNAs against LFD embryos. g, Biotype-specific heteroplasmy enrichment (over LFD) in male and female HFD embryos (HFD_B shown on the x axis; data represented as mean ± s.e.m.).

Fig. 4. Double-edge connection between mitochondrial metabolism and paternal epigenetic inheritance.

a, MA plot representation of differentially expressed genes in HFD_A versus LFD embryos. b, Gene Ontology-based functional enrichment analysis of differentially expressed genes in HFD_A versus LFD embryos (see Methods for details). c,d, Heat map representation of the relative expression (HFD versus LFD) of OXPHOS genes in early two-cell embryos (c) and in a publicly available dataset of single-embryo RNA-seq analysis of mouse pre-implantation development (d). TE_blast, trophectoderm blastocyst; ICM_blast, inner cell mass blastocyst. e, PCA-based representation of the developmental timing in HFD and LFD embryos laid over the analysis of mouse pre-implantation development of ref. . f, RNA-seq-based quantification of the expression of genes important for mitochondrial function across adult tissues and germ cells of male mice exposed to 2 weeks of HFD as well as early embryos and adult tissues from unexposed male offspring of HFD-fed fathers. 2CEs, two-cell embryos; TF, transcription factor; ROS, reactive oxygen species; gastro, gastrocnemius. g, Pearson-based correlation matrix of IMPC-derived metabolic phenotypes in WT offspring of parents heterozygous (het.) for IMPC-selected genes (see Methods for details). h, Bar plot representation of the relative (WT versus control) glucose intolerance (measured as AUC_ipGTT) in WT offspring of fathers heterozygous for genes important for mitochondrial structure and function. Black arrows indicate genes for which cryopreserved heterozygous sperm samples were analysed. i, Distribution of sncRNA biotypes in cauda spermatozoa from the indicated mutant mice (n = 10 mice per gene). j, Heat map representation of the relative abundance of 5′ n-tsRNAs and 5′ mt-tsRNAs in mutant spermatozoa. Control, LFD and HFD samples are cryopreserved and resequenced to serve as reference and technical controls.

Extended Data Fig. 1. (Extended Data Fig. 1 - related to Fig. 1) - Acute HFD treatment does not affect spermatogenesis.

(a) Representative H&E (top) and TRA98 (bottom) staining in testes from LFD (left) or HFD (right) fed mice. N = 6-7 mice/group independently stained and analysed (b) Quantification of the testis tubule diameter (average of the horizontal and vertical diameters) from H&E-stained testes (N = 60–70 seminiferous tubules – 10/mouse/group – significance calculated with Mann-Whitney test to compare ranks). (c) Motility of cauda spermatozoa from LFD and HFD-fed mice (N = 7 – significance calculated by two-tailed t-test). (d-e) First cleavage (d) and positive IVF (% of fertilised oocytes that complete pre-implantation development - e) rates in embryos generated via IVF with sperm from LFD or HFD-fed mice (N = 4 IVF each done with a pool of spermatozoa from 10 mice – significance calculated by two-tailed t-test). (f) PCA representation of the RNA-seq analysis of Round Spermatids (RS) isolated from testes of mice fed with LFD or HFD (f). (g-h) Scatter plot (g) and heatmap (h) representation of differentially expressed genes in round spermatids. (i-j) UMAP representation (i) and cluster annotation (j - based on publicly available datasets - GSE112393) of single-cell RNA-seq analysis of testes from LFD and HFD-fed mice (n = 3). (k) Representative marker genes for the different germ cell populations. Source Data

Extended Data Fig. 2. (Extended Data Fig. 2 - related to Fig. 1) - Paternal overweight at conception is important for offspring metabolism in mice.

(a-b) Body weight (a-left), body composition (a-right) and glucose tolerance (b) in mice exposed to two weeks of LFD or HFD feeding. (mean ± SEM – significance calculated by two-tailed t-test (a; N = 18/diet – p-value = 0.0003 BW, and <10⁻⁴ Fat mass) or two-way ANOVA (b; N = 10-11/diet – p-value < 10⁻⁴)) (c-d) Effect of four weeks recovery on chow diet on diet-induced body weight (c-left), body composition (c-right) and glucose tolerance (d). N = 8/diet (mean ± SEM – significance calculated by two-tailed t-test (c – p-value = 0.03 BW, and 0.008 Fat mass) or two-way ANOVA (d – p-value = 0.006)). (e-f) Body weight (e) and Fat mass (f) development in male offspring of LFD and HFD-exposed bucks and bucks allowed to recover for four weeks on chow diet (N = 60 male mice across 4 cohorts with 5 litters each and 3 males/litter (eLFD and eHFD); N = 10–15 (sLFD and sHFD) including one cohort with 5 litters and 3 males/litter – significance calculated with Mann-Whitney test to compare ranks for weight and fat-mass gain). (g) Frequency of HFDi offspring across four different F0-F1 cohorts. (h-j) Body weight (h), Fat mass (i) and glucose tolerance (j) in female offspring of LFD and HFD-exposed bucks and bucks allowed to recover for four weeks on chow diet. (N = 60 female mice across 4 cohorts with 5 litters each and 3 female/litter (eLFD and eHFD); N = 10–15 (sLFD and sHFD) including one cohort with 5 litters and 2-3 females/litter – significance calculated with Mann-Whitney test to compare ranks for weight and fat-mass gain). (k) Differentially Expressed Genes (male HFDi vs HFDt) in peripheral tissues. (l) Childhood obesity risk (top) and protective (bottom) genes differentially expressed in HFDi tissues. Source Data

Extended Data Fig. 3. (Extended Data Fig. 3 - related to Fig. 1) - Paternal overweight at conception is important for offspring metabolism in humans.

(a-b) Control linear regression analysis (r² and p-value in the respective panels) for associations between (a) maternal and paternal BMIs at conception and (b) maternal and offspring BMI (dotted lines indicate the 95% confidence intervals of the regression line). (c) Frequency of overweight and obesity in offspring stratified by paternal BMI groups in families with lean mothers (analysis detailed in the methods section; results in the figure). (d-e) Control linear regression analysis (r² and p-value in the respective panels) for associations between ISI-MATSUDA (d) and HOMA-IR (e) insulin sensitivity indices and paternal BMI at conception (dotted lines indicate the 95% confidence intervals of the regression line).

Extended Data Fig. 4. (Extended Data Fig. 4 - related to Fig. 2) - SncRNA-sequencing analysis of mouse cauda spermatozoa and purified round spermatids.

(a-b) Distribution of sncRNA biotypes in cauda spermatozoa from LFD and HFD-fed mice after recovery on chow diet (a - N = 3 samples/diet) and in purified round spermatids from LFD and HFD-fed mice (b - N = 3 samples/diet). (c) Differential expression analysis of cauda spermatozoa sncRNA sequencing from LFD and HFD-fed mice. (d) Volcano plot representation of DE n-tsRNAs. (e) Fragmentation pattern of n-tRNAs in cauda spermatozoa from LFD and HFD-fed mice. Significance tested by a two-tail t-test HFDvsLFD (p-value = 0.0017, N = 3). (f) Violin plots illustrating the downregulation of tRF-Glu^CTC-5 (left) and tRF-Gly^GCC-5 (right) in spermatozoa from HFD-fed mice. Significance tested by a two-tailed Mann-Whitney test to compare ranks (p-value < 10⁻⁴). (g) Volcano plot representation of DE mt-rsRNAs. (h) Biotype-specific differential expression analysis of cauda spermatozoa sncRNAs from LFD and HFD-fed mice after recovery on chow diet. (i) Bar plot showing the relative abundance of mt-tsRNAs in epididymosomes or spermatozoa from caput, corpus and cauda epididymis (N = 3)^,. (j) Relative abundance of the indicated sncRNA biotypes in cauda spermatozoa, its associated cytoplasmic droplets (CD) and CD-depleted cauda spermatozoa (pur. Sperm). Data re-analysed from Wang H. et al. . (k) Enrichment of the indicated small RNAs in cauda spermatozoa over its associated cytoplasmic droplets (N = 2). (l) Biotype-specific differential expression analysis of round spermatids sncRNAs from LFD and HFD-fed mice.

Extended Data Fig. 5. (Extended Data Fig. 5 - related to Fig. 2) - SncRNA-sequencing analysis of human spermatozoa.

(a) Distribution of sncRNA biotypes in human spermatozoa purified from ejaculates. (up, left) biotypes distribution in single donors stratified per BMI; (up, right) average biotypes distribution in lean (BMI < 24) and owt (BMI > 24) donors. (down, left) relative abundance of n- and mt-tsRNAs in single donors stratified per BMI; (down, right) average relative abundance of n- and mt-tsRNAs in lean and owt donors. (b-d) Pearson-based co-correlation matrix of sncRNA biotypes and BMI (b) and scatter plot representation for mt-tsRNAs (c) and n-tsRNAs (d) Exact p-values for c and d indicated in the figure. (e) Biotype-specific differential expression analysis of human sperm sncRNAs. (f) Heatmap representation of the Continuous Differential Expression analysis results for n- and mt-tsRNAs. (g) Dot plot showing the upregulation of n-tRNA-Gly^GCC_5_end in spermatozoa from owt donors.

Extended Data Fig. 6. (Extended Data Fig. 6 - related to Fig. 3) - Male embryos sired by HFD-fed bucks cluster into two independent clusters characterized by differential expression of mt-tRNAs.

(a) PCA plot representation of nuclear transcriptomes in single male and female hybrid embryos. (b-c) Hierarchical (b) and Seurat-based (c) clustering of female embryos sired by HFD-fed fathers. Both methods identify one single cluster in female embryos. (d-e) Hierarchical (d) and Seurat-based (e) clustering of male embryos sired by HFD-fed fathers. Both methods identify two independent clusters in male embryos defined by differential expression of mt-tRNAs. (f) Heatmap representation of mt-tRNAs (left) and mt-genes (right) expression in the different embryo populations. Differential expression calculated with the DESeq2 algorithm. * = adj.p value < 0.05 (see methods for details).

Extended Data Fig. 7. (Extended Data Fig. 7 - related to Fig. 3) - Sperm mt-RNAs are transferred to the oocytes at fertilization.

(a-b) Average heteroplasmy in male (a) and female (b) early 2CE. (Whiskers are the 10^th–90^th percentile of the data distribution. N = 373 SNPs with quantifiable heteroplasmy. Significance calculated by two-tailed paired t-test. **** = p value < 10⁻⁴). (c) Density plot (top) and heatmap (bottom) representation of the quantified heteroplasmy in female HFD and LFD embryos at the 416 SNPs mapped between BL6 and ST mitochondrial genomes. mt-tRNAs are highlighted. For comparison between male and female embryos, the heatmap from Fig. 3e is also shown.

Extended Data Fig. 8. (Extended Data Fig. 8 - related to Fig. 4) - Double-edge connection between mitochondrial metabolism and paternal epigenetic inheritance.

(a-c) MA plot representation of differentially expressed genes in HFD_A vs HFD_B (a), HFD_B vs LFD (b) and HFD vs LFD Female (c) embryos. (d) Heatmap representation of the genes used for the PCA plot in Fig. 4e. (e) Scheme of the working hypothesis by which exposure of male mice to two weeks of HFD induces mild mitochondrial dysfunction, which leads to a compensatory up-regulation of mt-DNA transcription in spermatozoa, transfer of mt-tRNAs from sperm-to-oocytes at fertilisation and reprogramming of offspring glucose metabolism. (f) Heatmap representation of the IMPC metabolic phenotypes analysed in the WT offspring of IMPC-selected genes mutants with indication of offspring gender and parental effects (Het_x_Wt = paternal effect). (g-i) Heatmap representation of the tsRNA expression (g) and the relative abundance of 3’ (h) and CCA (i) fragments in mutant spermatozoa.