Male obesity causes adipose mitochondrial dysfunction in F₁ mouse progeny via a let-7-DICER axis

Abstract

Male obesity negative affects gametic function and offspring metabolism. We here describe that (F₀) obesity and weight loss in male mice reversibly alter metabolism and impair adipose mitochondrial function. These metabolic aberrations are transmitted to male offsprings (F₁), which display reduced mitochondrial gene expression. Mechanistically, we identify microRNAs let-7d/e as epigenetic mediators induced in obese F₀ sperm and in F₀/F₁ adipose tissue, where they silence the miRNA processor DICER1 and impair mitochondrial activity. Microinjecting let-7d/e into lean zygotes phenocopies the paternal obesity phenotype, inducing glucose intolerance and mitochondrial gene suppression in sired offspring. Single-cell RNA sequencing of blastomeres reveals that let-7d/e impair oxidative metabolism in early embryos. Furthermore, lifestyle-induced weight loss in males with obesity downregulates human HSA-LET-7D/E in semen, indicating a conserved role for let-7 in transmission of metabolic health. These findings demonstrate that microRNA let-7 in sperm reprograms offspring metabolism by modulating mitochondrial function during early development.

Paternal obesity impacts offspring health, though the underlying mechanisms remain poorly understood. Here, the authors show that male obesity drives adipose mitochondrial dysfunction in F₁ mouse progeny via a let-7-DICER axis, identifying a pathway for intergenerational metabolic inheritance

Introduction

Genome-wide association studies have demonstrated that genetic (i.e., DNA) variants only explain 20% of obesity heritability¹. Given pandemic rises in obesity in evolutionarily relatively short timescales, it was proposed that non-genetic or 'epi-genetic' processes, that are spurred by obesogenic lifestyles and other changes in environmental exposures, might increase obesity susceptibility and contribute to disease incidences. This form of epigenetic inheritance, if convincingly demonstrated², is particularly intriguing for cases of paternal programming where affected fathers contribute lifestyle-affected sperm to embryos (F₁ generation, so-called “intergenerational” inheritance) or even in situations that arise as a consequence of F₁ phenotypes but in F₂ and beyond (so-called “transgenerational” inheritance). Epidemiological support for the existence of hereditary, paternally-acquired traits comes from historical records demonstrating that patterns of ancestral food supply affect cardiometabolic traits for two generations^3,4, showing that individuals can carry increased risks of cardiometabolic dysfunction if their parents or even grandparents were obese or malnourished⁵. In line with these correlative observations in humans, mechanistic studies in isogenic (i.e., genetically identical) mouse strains like C57BL/6 demonstrated that paternal obesity impairs metabolism in offsprings (i.e., F₁ generation and beyond) and that low-protein⁶ or high-fat diet feeding^7–9 and stress-related cues like depression-like behavior¹⁰, fear-conditioning¹¹ and psychological traumata¹² impart highly specific epigenetic behavioral and metabolic effects in F₁ offsprings. Recently, beneficial types of epigenetic programming in adipose tissue were demonstrated where offsprings of mice exposed to cold temperatures display increased metabolically favorable brown adipose tissue¹³, suggesting an unexpectedly plasticity of the molecular and cellular mechanisms underlying hereditary epigenetic responses.

The molecular and cellular basis for epigenetic heredity remains largely enigmatic: Although changes in epigenetic marks like methylation of genomic DNA, histone post-translational modifications and alterations in small noncoding RNAs (sncRNAs) correlate with hereditary processes in male gametes¹⁴. Although lifestyle/diet-induced changes to DNA methylation and histones modifications were reported, the evidence for causatively implicating these in eliciting paternal response in progeny is rare for higher mammalian species like rodents and humans. Other emerging instigators of intergenerational response are found amongst the different types of gametic sncRNAs such as microRNAs (miRNAs): miRNAs in gametes were initially identified and functionally dissected in nematodes and fruit flies^15–18, thus species where sncRNAs-evoked gametic gene silencing processes are well-understood¹⁹. In contrast to DNA-linked marks, sncRNAs are mobile in nature and subject to microvesicular trafficking between cells and organs and secreted by somatic cell types in testes^20,21. Thus, due to their high abundance of miRNAs in sperm, their dynamic changes following diet exposures^22,23, their exchange between sperm and oocytes²³ sperm-born sncRNAs have the unique potential for exchanging information about past lifestyle choices to zygotes and F₁ progeny. Consistent with this, metabolic trait/disease-associated miRNAs can affect intergenerational phenotypes and phenocopy F₀ lifestyles if microinjected^24,25 or removed from zygotes²⁶. Beyond controlling gametic epigenomes, miRNAs are also essential for germ cell development processes such as spermatogenesis²⁷ and embryogenesis²⁸, and they are exchanged between zygotes and other cell types^{12,25,29–31}. Taken together, it is plausible to hypothesize that obesity-evoked changes in sperm miRNAs alter zygotic miRNA pools and embryonic development, instigating epigenetic phenotypic traits and somatic gene expression in offsprings. As diverse paternal lifestyle exposures ranging from inflammation¹⁵, nutrition^29,31 to stress^12,30 alter gametic miRNA repertoires and F₁ phenotypes in non-overlapping ways, the intriguing possibility exists that environmental experiences are encoded in individual gametic sncRNA signatures and that these sperm sncRNAs have the potential to transmit complex lifestyle exposures, given the high bandwidth of RNA-mediated effects³².

Adipose tissue is a multifunctional endocrine organ that coordinates tissue-level (autocrine, juxtacrine, paracrine) and systemic (endocrine) responses by release of adipokines, lipids, inflammatory factors, and, recently, miRNA-containing microvesicles (exosomes)^33–35. It is believed that adipose tissue can safely store ingested nutrients as lipids, although chronic calorie excess causes insulin resistance³⁶, aberrant gene expression in adipocytes³⁷, vascular remodeling and immune cell infiltration³³. Mitochondria in adipose tissue exert important functions for tissue/energy homeostasis, particularly during nutrient deprivation/excess³⁵: Whereas brown adipocytes in contain a high numbers of mitochondria to generate heat by dissipation of the electron transport chain (ETC) proton gradients³⁸, inguinal (beige, iWAT) and epididymal (white, eWAT) adipose tissue protects from calorie overload by removing nutrients from circulation and storage as energetic, indolent macromolecules such as triglycerides. Mitochondrial function in WAT is impaired in obesity³⁹, and reduced mitochondrial function in fat causes metabolic imbalance^40–43. Mitochondrial DNA content, oxidative phosphorylation (OxPhos), abundance and composition of ETC complexes are all functionally impaired in obesity and morphological aberrations in mitochondrial structures are found in obesity^39,43 and aging⁴⁴. Conversely, treatment with the AMP-activated protein kinase (AMPK) agonist Metformin improves mitochondrial function in adipose tissue, as do exercise or calorie restriction, and all these interventions correlate with metabolic improvement⁴⁵. Recently, miRNAs were identified as important regulators of mitochondrial formation and function. Known mitochondrial microRNAs or ‘mito-miRs' include miR-143, miR-181a, miR-378 that instigate, whereas miR-532 impairs mitochondrial function, mitochondrial translocation and mt-RNA homeostasis^46–48. We and others showed that removing miRNAs from adipocytes by conditionally ablating Dicer1, a central component of the conserved miRNA processing machinery, impairs adipogenesis, triggering lipodystrophy and metabolic dysfunction in mice^49–51.

Collectively, we report that obesity in male mice is linked to reversible miRNA, mRNA and proteins alterations in adipose mitochondrial gene regulation that match those in adipose tissue of lean isogenic F₁ progeny. Using integrative analyses of adipose, liver and sperm miRNA:mRNA gene networks, we identify several let-7 isoforms, including let-7d/e, as being induced by obesity across cell types and generations. Mechanistically, we implicate let-7d/e driven suppression of DICER1 and the concomitant declines in mitochondrial function in adipocytes in inherited mouse phenotypes. This suggess sperm-born let-7d/e as an intergenerational contributing factor for mitochondrial dysfunction and glucose impairments in the F₁ offspring generation.

Results

Paternal obesity elicits intergenerational impairments in adipose mitochondrial function

Whilst intergenerational effects of paternal (F₀) obesity on offspring (F₁) phenotypes and metabolic health have been reported, the specific organ-level manifestations of this dysfunction in progeny, and the precise somatic and germ cell-intrinsic molecular mechanisms in obese mice and their descendants, remain surprisingly elusive. To investigate this using isogenic rodents (Mus musculus), we used diet-induced obese (DIO) and weight-regressed C57BL/6 J male mice, and fed 6-week-old animals with 1) low-fat diet (LF) for 18 weeks (F₀ LF), 2) high-fat diet (HF) for 18 weeks (F₀ HF) or 3) HF for 9 weeks, followed by 9 weeks of LF-driven weight loss (F₀ HF-LF, Fig. 1a) to allow at least three cycles of spermatogenesis. For each F₀ group, we mated eight randomly selected mice with unexposed lean virgin, female C57BL/6 J mice, and exposed both male and female F₁ progeny of each group, i.e., F₁ (F₀ LF, n = 44 pups), F₁ (F₀ HF, n = 52 pups) and F₁ (F₀ HF-LF, n = 38 pups) mice to LF after weaning to test for metabolic and metabolic sequel of F₀ diet exposure and removal. We did not additionally challenge F₁ mice with HF given that effects of HF and epigenetic F₀ effects were reported as cumulative with regards to F₁ phenotypes⁵² (Fig. 1b). Noteworthy, we detected no alterations in fecundity, male-to-female pup ratio and litters sizes (Fig. S1a, b), suggesting that F₀ obesity did not affect development and embryonic implantation rates nor indirectly affected maternal provisioning after birth due to different litter sizes, two potential confounders.

Fig. 1. Paternal obesity elicits heritable impairments in glucose metabolism and adipose mitochondrial function.

a, b Schematic of the intergenerational study design showing paternal (F0) obesity/regression models and F₁ offspring (a) alongside multi-omics workflows for eWAT, liver, sperm, and serum (b). Figure created in BioRender (https://BioRender.com/dbu6k78) c–e Metabolic phenotyping of F0 and F₁ male mice. c Body weights (F₀: LF n = 6, HF n = 13 [15–24 weeks of age n = 6], HF-LF n = 7; F1: LF n = 18, HF n = 25, HF-LF n = 21). d Blood glucose during the intraperitoneal glucose tolerance test (F₀: LF n = 6, HF n = 6, HF-LF n = 7; F₁: LF, HF, HF-LF n = 10 each). e Blood glucose during intraperitoneal insulin tolerance test (F₀: LF n = 6, HF n = 6, HF-LF n = 7; F₁: LF n = 9, HF n = 10, HF-LF n = 10). f Relative eWAT weights (F₀: LF n = 6, HF n = 6, HF-LF n = 7; F₁: LF n = 10, HF n = 10, HF-LF n = 7). Box plots show the Q1, the median, the Q3 and min/max with individual replicates. g Serum leptin measured by ELISA (F₀: LF n = 6, HF n = 6, HF-LF n = 7; F₁: LF, HF, HF-LF n = 8 each). h Fractional area size distribution of eWAT adipocytes (F₀ and F₁ groups: n = 6 each). i, j Representative haematoxylin and eosin staining of eWAT from i F₀ and j F₁ groups. Scale bar, 100 μm. k, l Visualization of high-confidence STRING interaction networks of proteins (top) and GOBP enrichment (bottom) of downregulated proteins in eWAT for (k) F₀ (LF vs HF, both n = 6) and l F₁ (LF vs HF, both n = 6). Data are presented as mean ± standard error with individual values. Two-tailed Student’s t test (c–e, h) or one-way ANOVA followed by Tukey’s multiple corrections (f, g). P values are indicated in the panel or represented as letters (a, p < 0.05, F₀ LF versus F₀ HF or F₁ (F₀ LF) versus F₁ (F₀ HF); b p < 0.05, F₀ HF versus F₀ HF-LF or F₁ (F₀ HF) versus F₁ (F₀ HF-LF); c p < 0.05, F₀ LF versus F₀ HF-LF or F₁ (F₀ LF) versus F₁ (F₀ HF-LF)).

When phenotyping F₀ and male F₁ animals for metabolic traits commonly associated with obesity and type 2 diabetes (T2D), we found that HF increased and ensuing HF-LF feeding reduced body weight to levels seen in F₀ LF mice, whilst F₁ progeny were indistinguishable in terms of body weight (Fig. 1c). Intriguingly, male obesity impaired glucose tolerance (Fig. 1d, Fig. S2a) and insulin sensitivity (Fig. 1e) not only in F₀ HF mice, but also in their F₁ (F₀ HF) offsprings. Whilst circulating F₁ insulin (Fig. S2b) was unchanged, we found that obesity increased iWAT (Fig. S2c) and eWAT (Fig. 1f) masses and elevated serum levels of leptin, an adipokine reflecting adipocyte size and numbers, in F₀ and F₁ mice (Fig. 1g). Adipocyte size determination using haematoxylin/eosine (H/E) staining of formalin-fixated, paraffine-embedded adipose sections demonstrated comparable shifts towards increased eWAT adipocyte sizes in F₀ and F₁ (Fig. 1h–j), suggesting that paternal obesity might promote lipid accrual/lipolysis as well as metabolic processes governing adipocyte health by similar molecular mechanisms in both generations and via this mechanism contribute to glucose intolerance in F₀ and male F₁. When phenotyping female offsprings of lean, obese and weight-regressed F₀ fathers, we observed no changes in F₁ weight (Fig. S2d), mild glucose intolerance (Fig. S2e, f) and only trends towards changes in insulin sensitivity (Fig. S2g). In support of overall mild impairments in female offsprings, random-fed glycemia (Fig. S2h) and organ weights (Fig. S2i) were also unchanged in female progeny. Thus, male obesity and weight loss reversibly affect glucose tolerance in F₁, with metabolic effects predominantly observed in male progeny.

For comprehensive insights into molecular processes affected by paternal obesity and weight loss, and those governing intergenerational effects, we performed mass spectrometry to determine differentially abundant proteins in F₀ and F₁ eWAT. We observed 980/431 proteins increased in F₀ HF compared to F₀ LF mice and in F₁ (F₀ HF) compared to F₁ (F₀ LF) mice. Conversely, 632/305 proteins were less abundant comparing F₀ HF to F₀ LF mice (Fig. S3a) and F₁ (F₀ HF) compared to F₁ (F₀ LF) mice (Fig. S3b). Using the STRING database (https://string-db.org/) followed by Gene Ontology (GO) analysis of overrepresented biological processes (GOBP), we observed that F₀ HF mice (Fig. 1k) and F₁ (F₀ HF) progeny (Fig. 1l) shared the same downregulated, mitochondria-associated GOBPs such as Oxphos, adenosine triphosphate (ATP) synthesis, and regulation of ETC subunits. Examples of proteins similarly repressed in F₀ HF and F₁ (F₀ HF) eWAT included components of ETC Complex I like NADH dehydrogenase 1-alpha subcomplex subunit 8 (NDUFA8) and NADH dehydrogenase iron-sulfur protein 6 (NDUFS6), of Complex IV like Cytochrome C Oxidase Subunit 6B1 (COX6B1), of mitochondrial membrane-intrinsic lipid carriers such as carnitine palmitoyltransferase 1 A (CPT1A) and of mitochondrial protein translocation like translocase of outer mitochondrial membrane 20 (TOMM20). In contrast, upregulated GOBP terms in F₀ HF mice (Fig. S3a) and their F₁ (F₀ HF) progeny (Fig. S3b) included unrelated GOBP terms like protein translation, Golgi-mediated transport and secretion, as well as protein localization. Importantly, abundance of metabolite- (Fig. S4a–c) and lipid-based (Fig. S4d, e) biomarkers that can reflect systemic mitochondrial dysfunction, for instance tricarboxylic cycle acid (TCA) intermediates like citrate, malate, fumarate, alpha-ketoglutarate and citrate, were reduced by paternal obesity and restored by weight loss (Fig. S4f) and mitochondrial lipid carriers like (acetylated) carnitine (Fig. S4g) were reduced. Furthermore, saturated, short-chain C:16:0/C:18:0 ceramides, i.e., lipid species reported to disrupt mitochondrial function and elicit mitochondrial ER stress in obese adipocytes^53,54 were induced, whilst mito-protective long-chain C18:1/C24:1 ceramides were reduced in F₀ HF mice (Fig. S4h). Thus, male obesity downregulates proteins associated with mitochondrial processes in obese F₀ and F₁ offsprings that correlate with reduced levels of mitochondrial function and lipid catabolism, likely reflecting a system-wide hereditary form of mitochondrial dysfunction, also in adipose tissue.

Paternal obesity negatively affects the expression of mitochondria-associated miRNA-mRNA networks

We next turned our attention to in detail study of gene-regulatory processes at the messenger and miRNA level that might underlie the observed repression of mitochondria-associated proteins in F₀ and F₁ adipose tissue, and to this end, performed mRNA-seq of eWAT and liver of F₀ and F₁ groups and performed miRNA and mRNA differential expression analysis, specifically focussing on mitochondrial processes. Comparing F₀ HF to F₀ LF eWAT using adjusted p values (pV_adj) ≤ 0.05, we detected yielded 3265 upregulated and 2890 downregulated genes (Fig. 2a). When comparing F₁ (F₀ HF) to F₁ (F₀ LF) mice, we detected no significantly regulated genes when accounting for multiple testing (pV_adj). Yet when specifically interrogating differential expression to reflect the more variable transcriptional effects in F₁ mice, we found 766 upregulated and 893 downregulated genes using unadjusted p Values (pV) ≤ 0.05 (Fig. 2b). When performing gene set enrichment analysis (GSEA) of affected GOBP we found upregulated processes linked to cell proliferation and immune activation that likely reflect adipocyte hyperplasia and inflammation, thus cellular processes commonly associated with obesity, and GOBP associated with germ cell and cilial function in F₁. Importantly, we confirmed our proteomic findings and detected overrepresented GOBP related to lipid catabolism, ATP synthesis, OxPhos and ETC downregulated in adipose tissue of F₀ HF mice (Fig. 2c) and their F₁ offsprings (Fig. 2d) and, when interrogating expression of mRNAs encoding mitochondrial ETC I-V, we confirmed the repression of nuclear-expressed mitochondrial constituents at the mRNA level (Fig. 2a, b). Importantly, aberrant mitochondrial effects were to a large extent reversible by weight loss, as comparing F₀ HF to F₀ HF-LF eWAT yielded 2687 upregulated and 2237 downregulated genes using pV_adj (Fig. S5a), whilst comparing F₁ (F₀ HF) to F₁ (F₀ HF-LF) eWAT yielded 547 upregulated and 732 downregulated genes without adjusting for multiple testing (Fig. S5b), and GOBP restored by weight loss also fell into genes linked to mitochondrial function (Fig. S5c, d).

Fig. 2. Paternal obesity dysregulates miRNA:mRNA gene networks spurring intergenerational mitochondrial dysfunction.

a, b Volcano plot of significantly up- (gray) and downregulated (white) mRNAs in eWAT from a F₀ HF versus F₀ LF with 3,265 upregulated and 2890 downregulated mRNAs and b F₁ (F₀ HF) versus F₁ (F₀ LF) with 766 up- and 893 downregulated mRNAs. Plots show log10 transformed fold-changes (fc) and log10-transformed a adjusted or b non-adjusted p values of mRNA changes. Mitochondrial complex I-V genes are annotated in the plot. c, d Gene set enrichment analysis and gene ontology enrichment of biological processes (GOBP) of differentially expressed mRNAs from (a, b). GOBP linked to mitochondrial respiration, ATP synthesis and Complex I assembly are marked with blue bars. e, f Volcano plot of significantly regulated miRNAs in eWAT from e F₀ HF versus F₀ LF with 154 up- and 193 downregulated miRNAs and f F₁ (F₀ HF) versus F₁ (F₀ LF) with 63 up- and 87 downregulated miRNAs. Plots depict log10 transformed fc and log10-transformed p values of expression changes. Mitochondria-activating miRNAs are annotated in pink. g, h STRING interaction networks of g miR-378/miR-143 and h miR-532-5p targets predicted by AGO2-CLIP (left) and network subset for eWAT proteins downregulated when comparing F₀ LF (n = 6) versus F₀ HF (n = 6) male C57BL/6 J mice. Submodules containing mitochondrial proteins are marked by a red square, and proteins marked by red dots are AGO2-CLIP predicted let-7 targets and downregulated in F0 HF versus F₀ LF eWAT. Differentially expressed mRNAs and miRNAs were determined using two-tailed tests with P values adjusted for multiple comparisons. NES normalized enrichment score, FDR false discovery rate.

Given that miRNAs are important (rheostatic) components in post-transcriptional gene network and have previously been linked to mitochondrial maturation and function⁵⁵, albeit not in adipocytes, we next wondered whether concerted downregulation of mitochondrial programs in obese eWAT might (partially) be instigated by altered expression of mito-miRs¹‘ like let-7², miR-378³, miR-181⁴, miR-143⁵ that functionally support mitochondrial formation and oxidative metabolism. Thus, when conducting small RNA (sRNA)-seq to determine differential miRNA expression, we found that F₀ obesity repressed mito-miRs that positively affect mitochondrial biogenesis and function, for instance miR-143-3p⁵⁶, miR-181a-2-3p^57,58, miR-182⁵⁹, miR-200⁶⁰ and miR-378a-3p⁶¹ in eWAT of F₀ HF (Fig. 2e) and F₁ (F₀ HF, Fig. 2f) mice, whereas mitochondria-inactivating miR-532⁶² was induced in F₀ HF (Fig. S6) and F₁ (F₀ HF, Fig. S6) eWAT, suggesting that miRNA dysregulation contributed to mitochondrial impediments observed in obese F₀ eWAT. Integration of miRNA-mRNA NGS datasets with adipocyte-specific miRNA-mRNA-Argonaute 2 interactomes using AGO-HITS-CLIP⁶ revealed that miR-378 and miR-143, despite being predicted, did not coincide with downregulation of their mitochondrial protein targets (Fig. 2g), whilst miR-532-5p was both predicted to target and observed to coincide with mitochondrial proteins repressed in F₀ obesity, arguing for roles of miR-532, but not miR-143/378 in driving mitochondria dysfunction in F₀ HF eWAT (Fig. 2h). In line with body weight loss-driven mRNA transcriptional responses (Fig. S6a, b), miRNA alterations were highly dynamic, and weight loss restored expression of mitochondria-promoting in F₀ HF (Fig. S5e) and F₁ (F₀ HF, Fig. S5f) mice. Thus, F₀ obesity causes reversible repression of mitochondrial and lipid catabolic genes, and miRNA-mRNA network analysis implicates obesity-regulated mito-miRs in this process.

The specificity of obesity-induced heritable epigenetic mitochondrial alterations to adipose tissue

We were in the following interested to address how specific the observed heritability of mitochondrial dysfunction was as a cellular and metabolic process. Further, we aim to address whether this impairment in mitochondrial gene expression was confined to adipose tissue. Thus, we next performed mRNA-seq to assess the heredity of other obesity-associated processes, such as chronic, low-grade inflammation (meta-inflammation). Focussing on expression of gene markers of classically (M1) and alternatively activated (M2) monocytes/macrophages, cell types that are linked to meta-inflammation-driven adipose tissue dysfunction³³, we found that, as expected, mRNAs encoding M1 markers such as interleukin 1b (IL1b), monocyte chemoattractant protein-1 (Mcp1/Ccl2) and tumor necrosis factor alpha (TNF); but also M2 markers like arginase 1 (Arg1), mannose receptor C-Type 1 (Mrc1), C-Type lectin domain containing 10 A (Clec10a) and resistin like beta (Retnlb) were induced in F₀ HF, but not F₁ eWAT (Fig. S6e, f), emphasizing that a meta-inflammatory tone is not passed on from F₀ HF mice.

Given the dysfunction in a specialized cellular process like mitochondrial respiration in eWAT, we next aimed to juxtapose adipose tissue with transcriptional effects in liver, an organ essential for glucose and lipid homeostasis: Analysis of mRNAseq data using multidimensional scaling (MDS) plots showed that F₀ LF segregated from F₀ HF and F₀ HF-LF, arguing for mostly irreversible effects of obesity in F₀ liver (Fig. S7a) and modest effects of F₀ obesity on F₁ liver transcriptomes (Fig. S7b). When comparing F₀ HF to F₀ LF livers, we found 2,474 up- and 2,612 downregulated genes with pV_adj ≤ 0.05 (Fig. S7c), whereas comparing F₁ (F₀ HF) to F₁ (F₀ LF) without adjusting for multiple testing (pV ≤0.05) yielded 716up- and 546 downregulated genes (Fig. S7d). GSEA⁶³ of differentially regulated mRNAs identified lipid metabolism and organelle formation as being induced, whereas repressed terms included lipoprotein trafficking, endoplasmic reticulum and ribosome, yet no transcriptional evidence for altered mitochondrial processes in the liver of obese mice and their F₁ progeny (Fig. S7e, f), emphasising the specificity of mitochondrial alterations in obese fat.

The non-hereditary effects of paternal obesity were supported by histological evidence, and pathologist assessment of H/E (morphology), Sirius Red (collagen/fibrosis) and anti-CD45 (leukocytes) staining revealed important hallmarks of metabolic dysfunction-associated steatosis liver disease (MASLD) like micro-/macro-vesicular steatosis, fibrosis and locular inflammation in F₀ HF mice, but not lean F₀ or F₁ descendants (Fig. S8a–h). Targeted analysis of expression changes in genes linked to hepatic lipid metabolism and meta-inflammation, such as adhesion G protein-coupled receptor E1 (Adgre, F4/80), apolipoprotein A4 (Apoa4), apolipoprotein E (Apoe), low density lipoprotein receptor (Ldlr), peroxisome proliferator-activated receptor gamma (Pparg), cluster of differentiation 36 (Cd36) and sterol regulatory element binding transcription factor 1 (Srebf/Srebp1), confirmed that induction was confined to F₀ HF livers (Fig. S8i). Furthermore, mitochondria-associated genes that were activated in obesity, presumably counteracting lipid accumulation in MASLD, for instance carnithine palmitoyltransferase 1 A (Cpt1a), carnithine O-acetyltransferase (Crat), acyl-CoA-oxidase (Acox1) and acetyl-CoA-acetyltransferase (Acat1) were induced in livers from F₀ HF, but not F₁, mice. Thus, combinatorial analyses of transcriptional and histological changes argue against prominent effects of paternal obesity onto F₁ gene expression and mitochondrial processes in hepatic cell types.

Paternal obesity causes qualitatively similar alterations in eWAT and sperm of obese male mice

Spermic sncRNA were reported to be subjected to dietary cues such as high-fat and sugar diets in mice and humans²²; and it is therefore conceivable that F₀ HF feeding and weight loss alter spermic sncRNA levels that can then, following oocyte fertilization, be passed on to fertilized oocytes (zygotes) as shown recently²³. By changing sRNA composition in zygotes, sncRNAs thus have the ability to post-transcriptionally affect decision points in embryogenesis, thereby contributing to the observed glucose intolerance and mitochondrial protein repression observed in F₁ offsprings sired by F₀ HF bucks. A particular aspect of sperm transcriptomes is the complex profile of sncRNAs involving miRNAs, ribosomal RNAs (rRNAs), PIWI-interacting RNAs (piRNAs), (mitochondrial) transfer RNAs ((mt)-tRNAs) and (mt)-tRNA fragments²⁶ that, as described by Tomar et al.²³, are affected by mitochondrial dysfunction in obesity and, particularly relevant in this setting, are physically exchanged between sperm and oocytes during fertilization, a scenario also conceivable for other types of mitochondria-associated sncRNAs, for example mito-miRs. To map the effects of paternal obesity and weight loss on spermic sncRNAs and to gain insights into sRNA-driven metabolic adaptations in F₀ HF sperm, we next isolated motile spermatozoa using swim-up assays and charted differential spermic sncRNA responses as a consequence of body weight gain and losses and found sperm harbored complex sncRNA pools of thousands of (mt)-tRNAs, piRNAs, rRNAs and miRNAs (Fig. 3a–c). In addition to effects elicited by obesity and/or LF feeding at the single sncRNA level, we observed that (mt)-tRNAs and miRNAs as an entire RNA biotype were repressed by HF feeding. As a prior study from Tomar et al.²³ had shown effects on mt-RNA in sperm after HF feeding, this was investigated further: Mt-tsRNA fragments were found for nine mt-tRNAs (mt-Tl1, mt-Tq, mt-Tm, mt-Ts1, mt-Tr, mt-Th, mt-Ts2, mt-Te, mt-Tp), where mean cpm above 10 was observed in mt-Tm, mt-Ts1, mt-Ts2 and mt-Th (Fig. S9a), but only 5’ fragments of mt-Ts-1 were significantly changed in sperm from F₀ HF and F₀ HF-LF diet (Fig. b). As miRNAs are abundant in somatic cell types and sperm, harbor the possibility of horizontal transfer between them, and exert gene-regulatory functions within both cellular compartments, we turned our attention to miRNA regulation in sperm. We found that F₀ obesity and weight loss caused a coordinated and inverse (i.e., downregulation and upregulation, respectively) effect on miRNome regulation (Fig. 3d, e), a finding previously observed by our labs⁵¹ and others^49,50,64 for obese adipose tissues; in the latter case, at least partially due to reduced abundance of DICER1. Intrigued by the repression of miRNAs in two functionally distinct, yet anatomically adjacent tissues and cell types like eWAT and sperm, we asked if differential miRNA responses were qualitative similar in both organ which, if present, would support the possibility of physical miRNA exchange between eWAT adipocytes and sperm as shown for other testicular cell types and sperm²¹: Indeed, upon comparison of lean, obese and weight-regressed sperm and eWAT miRNAs, we found highly congruent responses across both compartments with activating mito-miRNAs like miR-378 and miR-143 being downregulated, whilst inactivating mito-miRs of the let-7 family upregulated in obese F₀ eWAT and sperm (Fig. 3f). Importantly, when broadly mapping the regulation of let-7 family isoforms, we found mild, but synchronous inductions of let-7 in F₀ eWAT and sperm and F₁ eWAT (Fig. 3g–i). As increases of let-7c were reported in intergenerational rat studies by us⁸ and others⁶⁵ and because let-7 impairs embryo implantation and glucose homeostasis when overexpressed in mice^66,67, we focussed on let-7 isoforms for functional follow-up studies in germ cells and adipocytes. Thus, paternal obesity triggers similar alterations in adipose tissue and sperm miRNAs and let-7 isoforms are consistently induced across cell types and generations.

Fig. 3. Obesity reversibly affects sperm sncRNomes and induces let-7 in eWAT and sperm.

a Size distribution on nucleotide length of biotypes present in F₀ HF-LF sperm. The data presented is the mean cpm per diet. b Size distribution on nucleotide length of biotypes present in F₀ HF sperm. The data presented is the mean cpm per diet. c Size distribution on nucleotide length of biotypes present in F₀ LF sperm. The data presented is the mean cpm per diet. d Fold change of F₀ HF vs F₀ LF sncRNA in sperm. Each point represents one sequence. e Fold change of F₀ HF-LF vs F₀ HF sncRNA in sperm. Each point represents one sequence. Dark blue = mitochondrial tRNA, dark green=piRNA, green=ribosomal RNA, light green =long non-coding RNA, yellow=mitochondrial ribosomal RNA, orange = miRNA, dark orange=protein coding, red = miscellaneous RNA, dark red = tRNA, gray=unannotated. f Heatmap depicting scaled abundances of miRNAs in F₀ LF, F₀ HF and F₀ HF-LF (n = 6/group) eWAT and sperm as determined by sRNA-seq. g–i Abundance of miR-143-3p, let-7d-5p and in F₀ LF (n = 6), F₀ HF (n = 6) and F₀ HF-LF (n = 6) eWAT and sperm. One-way ANOVA followed by Tukey’s multiple comparisons (g–i) were used for statistical analysis. Box plots overlaid with individual data points representing biological replicates. The box indicates the 25th (Q1), the median, and the 75th (Q3) percentiles, while the whiskers extend to the minimum and maximum values. Data are presented as mean ± standard error with individual values shown for n ≤ 10. P values are indicated in the panel.

Zygotic delivery of let-7 causes glucose intolerance and mitochondrial impairment in sired mice

Prompted by profound and reversible effects of male obesity and weight loss on sperm miRNomes and let-7 in particular, the observation that environmentally and artificially induced (e.g., elicited by miRNA microinjection) alteration in zygote miRNA (pools) trigger abnormalities in progeny^25,68, and because embryonic trajectories require maturation of pre- into mature miRNAs⁶⁹, we next tested the functional relevance of let-7 gain-of-function in one-cell embryos for F₁ phenotypes. We focused on let-7d-5p and let-7e-5p given that these exhibit the most concordant increases amongst let-7 family members in F₀ and F₁ eWAT and sperm (Fig. 3g–i), and because let-7 has been connected to embryonic cell fate decisions^70,71. To test if zygotic delivery of let-7 corresponding at levels seen in few obese spermatozoa triggers phenotypical and metabolic changes in mice sired from let-7 microinjected embryos (F_MI generation) and whether these would resemble mice sired by F₀ HF fathers, we injected physiological amounts (50 fmol, i.e., let-7 found in 5-10 spermatozoa) of cel-mir-67 mimic as non-mammalian (C.elegans) control, and mimics for let-7d-5p and let-7e-5p by microinjection into zygotes conceived from lean males and females C57BL/6 J mice using standard protocols²⁹ (Fig. 4a). We injected 126/138/106 zygotes with cel-mir-67, let-7d-5p and let-7e-5p, but observed no discernible difference in implantation success, litter sizes and number of pregnancies in F_MI mice (Fig. S10), arguing against roles of let-7 for zygote viability and implantation. Next, we analyzed at least 19 pups per group from 3-4 pregnancies but observed no overt differences in postnatal body weight trajectories (Fig. 4b). Despite similar weight, we detected impaired glucose tolerance (Fig. 4c, d), insulin resistance (Fig. 4e) and elevated random-fed glycemia (Fig. 4f) in mice sired by let-7 injected F_MI embryos, whilst eWAT and liver weights were indistinguishable between F_MI groups (Fig. 4g). Importantly, when performing mRNA-seq in eWAT of F_MI mice, despite mild changes, we found that let-7 downregulated mitochondrial processes in F_MI mice (Fig. 4h). Thus, zygotic delivery of let-7 at levels corresponding to few F₀ HF sperm cells, causes glucose intolerance and aberrant mitochondrial gene regulation in mice derived from these embryos.

Fig. 4. Zygotic delivery of let-7 causes glucose intolerance in sired F_MI mice and impairs embryonic mitochondria.

a Illustration of experimental workflow for zygotic miRNA injection and phenotypic characterization of sired offsprings (F_MI mice). b Body weights of male F_MI mice sired from cel-mir-67 (n = 9), let-7d-5p (n = 14) and let-7e-5p (n = 14) injected zygotes. c, d Blood glucose (c) and glucose AUC (d) during intraperitoneal glucose tolerance test in male F_MI mice sired from cel-mir-67cel-mir-67 (n = 7), let-7d-5p (n = 10) and let-7e-5p (n = 10) injected zygotes. e Blood glucose during intraperitoneal insulin tolerance test in male F_MI mice sired from cel-mir-67 (n = 7), let-7d-5p (n = 9) and let-7e-5p (n = 11) injected zygotes. f Random-fed blood glucose at indicated in male F_MI mice sired from cel-mir-67 (n = 6), let-7d-5p (n = 6) and let-7e-5p (n = 6) injected zygotes. g Relative weights of indicated organs from male F_MI mice sired from cel-mir-67 (n = 7), let-7d-5p (n = 6) and let-7e-5p (n = 6) injected zygotes. h GSEA-GOBP analysis of eWAT of male F_MI mice sired from cel-mir-67 (n = 5), let-7d-5p (n = 6) and let-7e-5p (n = 6) treated zygotes. GOBP linked to mitochondrial respiration, ATP synthesis and Complex I assembly are marked with blue bars. i Illustration of workflow to obtain single 8-cell stage blastomeres derived from zygotes injected with cel-mir-67, let-7d-5p and let-7e-5p subjected to smart-seq2 single-cell sequencing. Figure created in BioRender (https://BioRender.com/dbu6k78). j Multidimensional scaling plots of single-cell RNAseq profiles in 8-cell stage blastomeres obtained from cel-mir-67 (n = 22 blastomeres derived from n = 3 zygotes), let-7d-5p (n = 19 blastomere cells from n = 3 zygotes) and let-7e-5p (n = 20 blastomeres derived from n = 4 zygotes). k Over-representation analysis (ORA) of differential mRNA programs in F_MI blastomeres. l, m Correlation plot and regression analysis between differential BMI loss (∆BMI) and expression changes (∆Ct) in human semen (l) LET-7D-5P and (m) LET-7E-5P. Two-tailed Student’s t test (c, e) per timepoint, one-way ANOVA followed by Tukey’s multiple comparisons (d–g) or Pearson’s regression analysis (l, m) were used for statistical analysis. Significance in Pearson’s regression analysis was determined by two-tailed tests. ORA was determined using one-tailed tests with P values adjusted for multiple comparisons. Box plots show the Q1, the median, the Q3 and min/max with individual replicates. Data are presented as mean ± standard error with individual values. P-Values are indicated in the panel or represented as letters (a, p < 0.05, cel-mir-67 versus let-7d-5p, b p < 0.05, cel-mir-67 versus let-7e-5p).

Obesity-associated let-7 rewires mitochondrial metabolism during embryogenesis

As transfer of miRNA perturbation in sperm to oocytes can have deleterious consequences and give rise to incompletely developed embryos^20,72,73, and as zygotes and blastocysts have the capacity to differentiate into virtually all cell types, (early) embryogenesis represents an important window of developmental plasticity that might affect F_MI phenotypes if perturbed by miRNA delivery, despite the physiological levels of miRNA delivery and overt absence of embryonic lethality. Nonetheless, it is conceivable that alterations in zygotic miRNAs still entail far-reaching consequences in sired mice due to a combination of altered developmental, epigenetic and metabolic processes⁷³. Based on the specific mitochondrial alterations, but absence of increased let-7 levels in F_MI eWAT, we next hypothesized that let-7d-5p and let-7e-5p could carry out important roles in cell fate decisions during embryogenesis, which will indirectly contribute to the glucose intolerance in F_MI pups via altered embryonic development. To test this experimentally, we repeated zygotic let-7 injections but allowed miRNA-injected embryos to divide thrice until the 8-cell blastomere stage. We dissociated embryos into single blastomeres and performed single-cell RNA-Seq (scRNA-seq) using Smart-seq2⁷⁴ (Fig. 4i). Remarkably, we found that after three cell divisions, miRNA-injected blastomeres were still transcriptionally distinct, reflecting the profound transcriptional and developmental consequences of inducing let-7 in zygotes (Fig. 4j). When performing GSEA of let-7-dependent gene sets, we corroborated that let-7d-5p repressed terms linked to mitochondrial processes like ATP synthesis, TCA cycle and ETC (Fig. 4k). Intergenerational programming in isogenic rodents has been described before, yet it is elusive if these effects are translatable to species with highly different reproductive apparatuses, genetic backgrounds, and timescales of obesity development like humans. Although epidemiological evidence^4,75 suggests heritable effects of overfeeding and malnutrition in humans (e.g., the famous ‘Dutch Hunger Winter’⁷⁶), causal evidence for epigenetic inheritance is hard to separate from cultural and ecological inheritance in humans². As it is becoming appreciated that nutritional and metabolic cues shape human sncRNA expression²², we tested effects of weight change on semen let-7 in a small cohort of n = 15 men with obesity (BMI, 39.49 ± 1.80) undergoing voluntary weight loss⁷⁷ and quantified hsa-let-7 using TaqMan qPCR. Intriguingly, we found that hsa-let-7d-5p and hsa-let-7e-5p followed the trends observed for weight regulation, suggesting similar correlations between metabolic improvement and let-7 reductions in human semen (Fig. 4l, m). Thus, obesity-associated let-7d/e-5p affects murine embryogenesis by impairing mitochondrial gene regulation and similar processes might be occurring following weight loss in humans with obesity.

Adipocyte let-7 impairs mitochondrial function via post-transcriptional silencing of DICER1

Beyond F₀ HF sperm (Fig. 3g–i), we found let-7d-5p and let-7e-5p upregulated in F₀/F₁ eWAT, and AGO2-CLIP interactome analyses confirmed that let-7 gene targets are enriched for genes encoding mitochondrial proteins, that are downregulated by obesity in F₀/F₁ eWAT (Fig. 5a). For this, and to add to their previously ascribed role in fertilized oocytes, we aimed to dissect the transcriptional and functional consequences of elevated let-7 in adipocytes, with a specific focus on mitochondrial gene regulation as the highly sequence-related let-7i can impair commitment of mitochondria-rich adipocytes⁷⁸. Yet, whether roles for let-7 in mitochondrial formation and function exist in mature adipocytes is unknown. To test this, we delivered by reverse lipofection mimetics for let-7d-5p and let-7e-5p and cel-mir-67 into stromal vascular fraction-derived mature epididymal white adipocytes (1°eWA) to mimic obesity-associated increased in let-7 in eWAT. When interrogating the effects of let-7 overexpression, we observed that let-7d-5p elicited more pronounced effects (1768 upregulated, 1387 downregulated genes, Fig. 5b) than let-7e-5p (557 upregulated, 893 downregulated genes, Fig. 5c) compared to cel-mir-67, yet both let-7 isoforms downregulated processes related to lipid metabolism, Insulin-AKT (Protein Kinase B) signaling and carbohydrate metabolism (Fig. S11). Although not overrepresented at the GOBP level, we found rate-limiting genes in glycolysis, TCA, and mitochondrial lipid import like Carnithine Palmitoyltransferase 2 (Cpt2), Carnithine O-Acetyltransferase (Crat), Hexokinase 2 (Hk2), Isocitrate Dehydrogenase 2 (Idh2), Pyruvate Carboxylase (Pcx), Pyruvate Dehydrogenase Kinase 2 (Pdk2) and mitochondrial ATP-magnesium/phosphate antiporter Solute Carrier Family 25 Member 24 (Slc25a24, Fig. 5b, c; in blue) repressed upon let-7 delivery. Importantly, we also detected pivotal genes in miRNA maturation, such as Ribonuclease III DICER1 (Dicer1) and Argonaute RISC Catalytic Complex 2 (Ago2, Fig. 5b, c; in red). Next, we experimentally validated these predicted let-7 targets by dual luciferase reporter assays (DLRA)-based miRNA-mRNA 3’UTR interaction studies in HEK293 cells for genes linked to mitochondrial lipid and ATP uptake (Slc25a24, Cpt2), glycolysis (Hk2, Pdk2), TCA activity (Sdha, Fig. 5d) and miRNA processing (Dicer1, Ago2).

Fig. 5. Adipocyte let-7 blunts mitochondrial function via a let-7-DICER1 axis.

STRING interaction network of downregulated AGO2-CLIP-predicted let-7 targets in eWAT of F₀ HF (n = 6) versus F₀ LF (n = 6) male C57BL/6 J mice. Red squares mark mitochondrial submodules; red dots indicate let-7 targets. (b, c) Volcano plots showing significantly up- (gray) and downregulated (white) genes 1°eWA transfected with (b) let-7d-5p or (c) let-7e-5p versus cel-mir-67. Blue dots: mitochondrial genes; red dots: miRNA maturation genes. d, e Relative luciferase activity in HEK293 cells co-transfected with cel-mir-67 or let-7 mimics and constructs harboring d 3′-UTRs or exons containing let-7 miRNA responsive elements (MRE) of indicated genes or e wild-type and mutated let-7 MREs in Dicer1 transfected with 100 nM of cel-mir-67, let-7d-5p and let-7d-5p mimics. (n = 4 independent biological replicates). f–h Western blot analysis of 1°eWA transfected with cel-mir-67, let-7d-5p and let-7d-5p mimics performed using the indicated antibodies detecting (f) mitochondrial and (g) DICER1 proteins. HSC70 served as a loading control. h Densitometric quantification of (f, g) representing n = 3 biological replicates. i Western blot (top) and quantitative RT-PCR (bottom) and densitometric quantification of DICER1 in 1°eWA transfected using two independent locked nucleic acid (LNA) inhibitors targeting Dicer1 mRNA. Data represent n = 3 biological replicates. j, k Oxygen consumption rates (j, OCR) and extracellular acidification rates (k, ECAR) in 1°eWA transfected with negative control LNA and Dicer1 LNAs and stimulated with oligomycin (O), FCCP (F) and Antimycin A plus Rotenone (A/R). Experiments are representative of n = 3 independent biological replicates. l, m Western blot analysis (l) and densitometric quantification (m) of mitochondrial and DICER1 proteins in eWAT from F₀ LF, HF, and HF-LF mice (all n = 4). HSP90 served as loading control. n Proposed mechanism connecting obesity and elevated let-7 in adipocytes to post-transcriptional silencing of DICER1 and impaired oxidative phosphorylation and respiration. Statistical analysis used two-tailed Student’s t test (j, k) or one-way ANOVA followed by Tukey’s multiple comparisons (d, e, h, i). Differentially expressed mRNAs were determined using two-tailed tests with P values adjusted for multiple comparisons. Data are presented as mean ± standard error with individual values. P values are indicated in the panel or represented as letters (a p < 0.05, cel-mir-67 versus let-7d-5p; b p < 0.05, cel-mir-67 versus let-7e-5p).

Dicer1 3’UTRs and exons harbored several independent let-7 seeds of which, after mutation of putative let-7 microRNA responsive elements (MREs), two from exonic regions became refractory to let-7 repression, demonstrating direct negative effects of let-7 on DICER1 translation (Fig. 5e). We strived to validate let-7 targets identified in HEK293 cells in adipocytes, and delivered let-7 and control mimics to 1°eWA and confirmed that let-7 also repressed DICER1 in white adipocytes, yet caused only modest reductions in mitochondria-/TCA-/glycolysis-associated proteins upon overexpression (Fig. 5f–h), prompting us to hypothesize that the chronic let-7 induction seen in obese adipocytes might repress mitochondrial function rather via sustained DICER1 loss than by direct let-7 effects on mitochondrial function; and that Dicer1 inactivation might be a better proxy for mimicking the prolonged upregulation in let-7d-5p and let-7e-5p in obesity than acute overexpression of the miRNA.

To address whether Dicer1 silencing, because of sustained let-7 upregulation, is sufficient to phenocopy in vitro the obesity-evoked impediments in mitochondrial function, we designed two independent Dicer1-targeting locked nucleic acid (LNA) inhibitors that repressed Dicer1 at the mRNA and protein level, yet with varying efficiencies (Fig. 5i) and interrogated mitochondrial respiration using Seahorse. We found that Dicer1 knockdown reduced oxygen consumption rates (OCR) as proxy for mitochondrial respiration (Fig. 5j) and extracellular acidification rates (ECAR) as proxy for glycolytic rates in 1°eWA (Fig. 5k). We further validated that not only DICER1 but also other mitochondrial proteins like CPT2, IDH2 were downregulated and partially restored in F₀ HF eWAT (Fig. 5l, m) supporting our hypothesis that obesity impairs mitochondrial respiration and gene regulation, at least partially, by post-transcriptional silencing of DICER1 via let-7 (Fig. 5m). Thus, obesity-associated let-7 impairs mitochondrial oxidative respiration by silencing of DICER1 in mature adipocytes.

Discussion

High transcriptional plasticity of mitochondrial dysfunction in adipose tissue

In our study, we found that hepatic expression changes inflicted by HFD feeding were largely irreversible, whereas adipose tissue miRNA/mRNA transcriptomes and translatomes, and particularly mitochondrial OxPhos and TCA gene sets, reverted to a state ante. Similar studies have investigated the effects of reversible (i.e., diet- and exercise-evoked weight gain and loss) on hepatic and adipose transcriptomes: Gonzalez-Francesa et al. observed⁴² that sustained diet-induced weight loss correlated with appreciable adipose transcriptional plasticity and mitochondrial gene expression, where short-term, exercise induced did not. Thus, durations of weight loss and differences in control diet exposure might reflect the importance of nutritional and exposure factors during weight loss interventions on tissue responses in obesity. Of note, expression of mitogenic transcription factor Nuclear Respiratory Factor 1 (Nrf1⁷⁹) and of transcriptional co-regulator Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (Ppargc1a) were repressed in F₀ HF eWAT (Fig. 2a). As Brandao et al.⁸⁰ reported that genetic ablation of DICER1 in adipocytes repressed PPARGC1A/NRF1, we can conceive a scenario where reduced DICER1 (e.g., due to increased silencing by let-7d/e-5p) drives and/or synergizes with reduced Ppargc1a/Nrf1 levels to impair mitochondrial biogenesis in hypercaloric adipocytes. While our findings demonstrate a robust paternal transmission of mitochondrial dysfunction to the F₁ generation, it is important to distinguish between intergenerational and transgenerational inheritance. Given that the let-7-DICER axis shifts were observed in the F₁ adipose tissue following direct paternal obesity exposure, this represents a bona fide intergenerational effect. Whether these epigenetic signatures can bypass reprogramming in subsequent germlines to persist into the F₂ generation or even beyond remains to be determined, though current evidence suggests such mechanisms may be attenuated in later generations.”

Rewiring of embryonal development by gametic miRNAs

In C. elegans and D. melanogaster, germline gene silencing by sncRNA is well-understood and similar processes are suggested for mammals¹⁹. miRNAs are not only critical for sperm cell formation²⁷ and embryogenesis²⁸, but can also be transferred from sperm to zygotes, and it is plausible that oocyte-sperm fusion alters zygotic miRNA pools, affects embryonic development, and instructs phenotypic traits and somatic gene expression in offspring, as observed in our intergenerational paradigm. In contrast to DNA-intrinsic epigenetic modifications like DNA methylation (DNAme) and histone post-translational modifications, miRNAs diffuse between and across cellular compartments, are highly abundant in most cells, and, crucially, can be exchanged between cells through extravesicular transport³². As mature sperm is transcriptionally inert, and no mechanism exists for how sperm senses endocrine information and translates this into differential miRNA levels, sperm relies on ‘decoding somatic cell types’ that possess these sentinel features and are capable of instructing sperm by transfer of miRNAs and other sncRNAs like mitochondrial tRNA fragments that are exchanged between sperm and oocyte at fertilization²³. Furthermore, epididymal cells, somatic epithelial cells of the extragonadal male reproductive tracts, secrete epididysomes, microvesicles that fuse with nascent sperm⁸¹, and as we observed several miRNAs concordantly regulated in F₀ adipose tissue and sperm, horizontal miRNA transfer from adipocytes to sperm, potentially via epididysomes is a plausible scenario that still lacks experimental proof in our paradigm. Finally, altering miRNA levels like miR-199a-5p in zygotes by microinjection is sufficient to reprogram embryonal metabolism⁸², highlighting that miRNA alterations can rewire embryonal development and metabolism, which are still compatible with fetal development.

Methods

Mice housing

All animal studies received ethical approval and were approved by the Ministry of Environment and Agriculture Denmark (Miljø- og Fødevarestyrelsen), under license 2023-15-0201-01544. All mice were housed under a 12-h light/12-h dark cycle in a temperature-controlled (22 °C) and humidity-controlled facility that had ad libitum access to diets and drinking water. 4-week-old C57BL/6 J male mice (Janvier Labs) were fed with a chow diet (NIH-31, Zeigler Brothers Inc., 8% calories from fat) for 2 weeks of acclimatization and then were fed with either a low-fat diet (LF, D12450J, 10% kcal from fat, Research Diet) or a high-fat diet (HF, D12492, 60% kcal from fat, Research Diet) for 9 weeks. After 9 weeks of feeding, half of the HF-fed mice were randomly assigned to the diet regression group, which was given LFD. LFD is carefully chosen to match the food source (animal vs plant) and the micronutrient composition of HFD.

Mouse husbandry

All male virgin mice were maintained on the assigned diet until 21 weeks of age and mated with 10-week-old C57BL/6 J female virgin mice (Janvier Labs). 8-week-old C57BL/6 J female mice were fed with a chow diet for 2 weeks of acclimatization and then were fed with LF during breeding and lactation. Mating was set up for 5 days in a female’s cage, and vaginal plugs were checked every morning. All mice, including the F₀ HF group, were maintained on LF during breeding to remove confounding effects from maternal diets. Males were sent from breeding cages to the original cage to be fed with the assigned diet for an additional 2 weeks and then sacrificed. The offspring from all groups of founders consumed only LF. Representative male offspring from six litters per group were included in all analyses.

Intraperitoneal glucose and insulin tolerance tests

Mice were fasted for 6 h for glucose-tolerance tests. Fasting glucose levels were measured, and then mice were intraperitoneally injected with 2 g/kg body weight of 20% glucose solution (B. Braun Medical A/S). For intraperitoneal insulin-tolerance tests, random-fed glucose levels were measured, and mice were intraperitoneally injected with 0.75 IU/kg body weight of 100 IU stock insulin (Novo Nordisk). Blood glucose levels were measured at 15, 30, 60, 90, and 120 min after injecting glucose or insulin using a CONTOUR®NEXT glucometer (Bayer). For ipGTT/ipITT, we omitted mice with strongly delayed rises in glucose and no discernible drop in blood glucose, respectively, from analyses, as this was the consequence of mal injection outside the peritoneum. For area under the curve (AUC) calculations, we have normalized serum [Glu] values to [Glu, T₀] of each animal to account for intra-animal variability in glucose responses and relative glucose responses.

Tissue collection

Mice were sacrificed by carbon dioxide euthanasia. Livers, eWAT, iWAT and BAT were weighed, snap-frozen in liquid nitrogen, and stored in −80 °C. Pieces from liver and eWAT were fixed in 4% paraformaldehyde (PFA) for histology analysis. For serum collection, blood collected by the cardiac puncture was centrifuged at 2000 × g for 15 min. The supernatant was collected and stored in −80 °C.

Sperm collection

The prospective intervention study complies with relevant ethical regulations and was conducted between May 2020 and April 2022 at the Fertility Clinic, Zealand University Hospital, Denmark. The study was approved by the Danish National Committee on Health Research Ethics (SJ-769), the Danish Data Protection Agency (REG-048-2019) and registered on the Clinical Trial website (www.ClinicalTrials.gov identifier NCT04721938). The primary outcomes have been published⁷⁷. Briefly, mature sperm cells were collected from the cauda epididymis. Sperm cells are allowed to swim out from the cauda epididymis and incubate in phosphate-buffered saline (PBS) at 37 °C for 15 min. Supernatant containing sperms was passed through a 40-µm cell strainer and incubated in a somatic cell lysis buffer (0.5% Triton-X and 0.1% SDS) for 40 min on ice. Sperms were centrifuged at 600 × g for 5 min and pellets were washed with PBS and pelted again at 600 × g for 5 min. The supernatant was discarded, and the pellets were snap-frozen in liquid nitrogen and stored at −80 °C.

Zygotic miRNA microinjection and embryo implantation

Ovulation of prepubescent (4-week-old) C57BL/6 J females was induced by intraperitoneal (IP) injection of PMSG, 5 IU/female, followed by IP injection of hCG, 5 IU/female, 47 h later. After the second injection, females were set in breeding with C57BL/6 J stud males. The next morning, they were monitored for the formation of a vaginal plug (indicating mating had occurred during the night). Females were euthanized and oviducts were dissected to harvest the cumuli containing zygotes and unfertilized oocytes. Cumuli were disaggregated by incubation in a 10 mg/ml solution of Hyaluronidase (Sigma-Aldrich ref. H4272) in M2 medium (Sigma-Aldrich ref. M7167) at RT for 10 minutes. Zygotes were sorted out, assessed by the presence of the second polar body and cultured in KSOM medium (Merck Millipore ref. MR-106-D). Oocytes were discarded. Each miRNA was diluted at 10 ng/ul in microinjection buffer (5 mM Tris, 0.1 mM EDTA, pH 7.4) as a stock solution and kept at −80 °C. The day of microinjection, 1.5 ng/ul solutions were prepared from the stock, and injected in the pronuclei of the zygotes following standard microinjection procedures. miRIDIAN mimics were ordered from Dharmacon: Inert miRNA control is based on C.elegans cel-mir-67, which is not expressed in mouse or human cells (5‘-UCACAACCUCCUAGAAAGAGUAGA-3‘, cat. no. CN-001000-01-05), mmu-let-7d-5p (cat. no. C-310486-07-0002) and mmu-let-7e-5p (cat. no. C-310507-07-0005). Injected zygotes were cultured in KSOM until the microinjection session was completed and then cryopreserved by the “Slow Freezing” Method (Manipulating the Mouse Embryo, a Laboratory Manual, fourth edition, Chapter 16, Richard Behringer et al. Cold Spring Harbor Laboratory Press, 2014) and transported to SDU at −150 °C for embryo transfer and recovery.

The frozen microinjected embryos were thawed at SDU and cultured in KSOM (Merck, MR-121-D) for 40 minutes (Manipulating the Mouse Embryo, a Laboratory Manual, fourth edition, Chapter 16, Richard Behringer et al. Cold Spring Harbor Laboratory Press, 2014). Viable embryos with intact zona pellucida were selected for further transfer. These embryos were then washed in 10 drops of M2 and surgically transferred to the oviducts of 0.5 dpc pseudo-pregnant females. Pseudopregnancy was induced in the females by mating them with a vasectomized male. Prior to implantation, the pseudo pregnant females were anesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) through intraperitoneal injection, and the surgical embryo transfer procedure was performed following the protocol described by Behringer et al. (Manipulating the Mouse Embryo, a Laboratory Manual, fourth edition, Chapter 6, Richard Behringer Cold Spring Harbor Laboratory Press, 2014). Postoperative analgesic treatment with carprofen (5 mg/kg) was administered subcutaneously for 48 hours.

8-cell-morula blastomere disaggregation

miRNAs were injected into the embryo at the zygote stage as described above. After injection, embryos were incubated in KSOM (KSOM-Embryomax advanced embryo medium, Merck Millipore #MD-101-D) for 48-56 hours at 37 °C, in 5% C0₂. The culture medium was refreshed daily. In the afternoon of the third day of culture, when the embryos had reached the 8-cell morula stage, they were incubated at RT in Tyrod’s solution (Sigma-Aldrich T1788) for up to 30 seconds until the zona pellucida was completely dissolved. Afterwards, the naked morulae were washed through three drops of M2 medium (Sigma-Aldrich M7167) and finally incubated in KSOM without Ca²⁺ and Mg²⁺ at RT for 20 minutes. Then, embryos were moved to a new drop of fresh M2 medium and there, pipetted up and down repeatedly to facilitate the blastomere disaggregation. When single blastomeres were obtained, they were individually transferred to single wells of a 96-well single-cell-sequencing plate from VWR (Cat. No. 82006-636) with 2μl of RNAse inhibitor mix (containing 0.12 μl Triton-X-100 10%, 0.1 μl RNAse inhibitor (Takara (Cat. No. 2313B, 40 U/μl), 1.78μl MilliQ water (RNAse-free)). Plates were then frozen at −80 °C until the sequencing was performed.

Lipidomics and metabolomics

Serum extraction

100 µL of serum was mixed with 1000 µL of a 2:1 chloroform/methanol solution and 200 µL of water. Four blank samples, each containing 100 µL of water, were processed in parallel. The mixtures were shaken at 1000 rpm for 30 minutes at 4 °C, followed by centrifugation at 16,000 × g for 10 minutes at 4 °C to separate the phases. The aqueous phase was re-extracted with 350 µL of an 86:14:1 chloroform/methanol/water solution. The organic phase from this re-extraction was combined with the corresponding organic phase from the initial extraction. The aqueous fraction was dried using SpeedVac, while the organic phase was dried under a stream of nitrogen. For subsequent analysis, the dried aqueous fractions were reconstituted in 30 µL of 1% formic acid for metabolomics. The organic phase was reconstituted in 30 µL of lipidomic solvent A (5:1:4 isopropanol/methanol/water with 5 mM ammonium acetate and 0.1% acetic acid). A pooled quality control sample was prepared for each analytical setup by combining 5 µL from each sample, excluding blanks.

Lipidomic analysis

For lipidomics analysis, 2.5 µL of each sample was injected into an Agilent 1290 Infinity UPLC system (Agilent Technologies, Santa Clara, CA, USA) equipped with a Zorbax Eclipse Plus C18 guard column (2.1 × 5 mm, 1.8 µm) and an analytical column (2.1 × 150 mm, 1.8 µm), maintained at 50 °C. Analytes were eluted at a flow rate of 400 µL/min using eluent A (water with 5 mM ammonium acetate and 0.1% acetic acid) and eluent B (99:1 isopropanol/water with 5 mM ammonium acetate and 0.1% acetic acid). The gradient was as follows: 0% B from 0 to 1 min, 0–25% B from 1 to 1.5 min, 25–95% B from 1.5 to 12 min, 95% B from 12 to 14 min, and 95-0% B from 14 to 15 min, followed by a 3-minute equilibration. For metabolomics analysis, 1.5 µL of each sample was injected into the same UPLC-MS system, but with a ZORBAX Eclipse Plus C18 guard column (2.1 × 50 mm, 1.8 µm) and an analytical column (2.1 × 150 mm, 1.8 µm) maintained at 40 °C. Eluent A was 0.1% formic acid, and eluent B was 0.1% formic acid in acetonitrile. The gradient was as follows: 3% B from 0 to 1.5 min, 3–40% B from 1.5 to 4.5 min, 40–95% B from 4.5 to 7.5 min, 9 5% B from 7.5 to 10.1 min, and 95–3% B from 10.1 to 10.5 min, followed by a 3.5-minute equilibration. Both analyses were performed on a 6530B quadrupole time-of-flight mass spectrometer (Agilent Technologies, Santa Clara, CA, USA) for mass spectrometric detection in positive and negative ion modes. MS1 mode settings included a scanning range of 50–1700 m/z for lipidomic and 10-1050 m/z for metabolomics, with 2 scans/sec. Internal calibration was performed using Hexakis (1H,1H,3H-tetrafluoropropoxy) phosphazene delivered through a second needle in the ion source via an isocratic pump running at 20 µL/min. An iterative data-dependent MS2 mode was used for fragmentation in both analyses, with the following settings: collision energy at 40 V (lipidomics) and 20-40 V (metabolomics), precursor threshold at 5000 counts, and active exclusion after 2 spectra, with re-inclusion after 0.5 min.

Data processing and normalization

All data was converted. mzML format using ProteoWizard⁸³. Lipidomics data were annotated in MS-Dial (v. 4.9) against the incorporated lipid-blast⁸⁴ with an MS1/MS2 mass tolerance of 0.005/0.01 Da and a minimum identification score of 70%. The annotated compounds were exported to PCDL manager B.08.00 (Agilent Technologies) to create a database for area extraction in Profinder 10.0 (Agilent Technologies). Metabolomics data were processed in MzMine (v2.59) utilizing modules such as ADAP chromatogram builder and deconvolution. Compounds were annotated at Metabolomics Standards Initiative (MSI) levels 3 using the libraries of National Institute of Standards and Technology 2017 (NIST17) and MassBank of North America (MoNa). Features with signals less than 5 times those in blanks or missing in more than 20% of QC samples were removed, and signals were QC corrected using statTarget⁸⁵.

Data analysis

For metabolomics pathway enrichment analysis (KEGG pathways of Mus musculus), comparing F₀ LF and F₀ HF serum (parameters 'Scatter plot', 'Hypergeometric test', 'relative-betweenness Centrality', 'Use all compounds in the selected pathway library') of differentially abundant polar metabolites (pV ≤ 0.1) were loaded into Metaboanalyst 6.0. For lipidomics and metabolomics, principal component analysis and heatmap representation of an indicated number of differentially abundant lipids and metabolites were generated using 'Statistical Feature' (one-factor option), Sample normalization = 'Normalization by sum', Data transformation = 'Log transformation (base 10)' and Data scaling = 'Mean Centring'.

Serum insulin and leptin ELISA

Serum insulin and leptin levels were measured by mouse insulin ELISA kit (Mercodia, Uppsala, Sweden) and mouse leptin ELISA kit (Crystal Chem, IL, USA) according to manufactures’ protocol.

Histology and staining

Liver and adipose tissues fixed with 4% of PFA were used for histology analyses. Formalin-fixed, paraffin-embedded tissues were cut into 3 μm sections. Sections were stained by haematoxylin and eosin (H&E) and Sirius Red by standard procedures at the Department of Pathology at Odense University Hospital, Odense, Denmark. Immunohistochemistry for CD45 was performed on Ventana BenchMark Ultra (Roche) with ready-to-use CD45 (LCA) (2B11 & PD7/26) Mouse Monoclonal Antibody.

A liver pathologist (SD) assigned steatosis, lobular inflammation, ballooning, fibrosis stage, and portal inflammation. Steatosis was graded as S0 ( < 5% hepatocytes with large fat vacuoles), S1 (5–33%), S2 ( > 33–66%), and S3 ( > 66% hepatocytes with large fat vacuoles)⁸⁶. Lobular inflammation (0-3) and hepatocellular ballooning (0-2) were assessed according to the NAS-CRN Activity Score by Kleiner et al.⁸⁶. Also, portal inflammation was assessed, using a 4-tiered score by Kleine et al. Liver fibrosis stage was assessed according to NAS-CRN for fibrosis; F-0, no fibrosis; F-1, perisinusoidal or periportal fibrosis; F-2, perisinusoidal and portal/periportal fibrosis; F-3, bridging fibrosis; F-4, cirrhosis.

Measurement of adipocyte volumes

The volume weighted mean volume of adipocytes was determined from the H&E sections with newCAST stereology software (Visiopharm), autodisector; measurements on systematic, randomly selected fields from 6 mice of all F₀ and F₁ groups were performed on at least 75 adipocyte profiles in each tissue.

Proteomics from epididymal white adipose tissue

Deep frozen fat pads (200 mg of epididymal adipose tissue) were powdered with a mortar-pestle and subjected to homogenization with ceramic beads (Percellys, 5500 rpm, 2 × 10 seconds, 30 seconds break) in the buffer containing 50 mM HEPES pH 7.4, 1% Triton-X-100, 100 mM NaF, 10 mM sodium orthovanadate, 10 mM EDTA, 0.2% SDS, 100 mM NaCl. Tissue homogenates were further processed to remove the nucleic acid traces (Diagenode SA; 10 min, cycle 30/30 seconds, at 4 °C). Floating fat and tissue debris were removed from protein extracts by centrifugation at 20,000 × g. 150 mg of proteins were precipitated with 4x volume of ice-cold acetone and washed with ice-cold acetone twice. Air-dry pellets were resuspended in 8 M Urea/50 mM triethylammonium bicarbonate (TAEB) buffer supplemented with a protease inhibitor cocktail (Sigma). 50 mg of the samples were incubated with 5 mM dithiothreitol at 25 °C for 60 min, followed by treatment with 40 mM chloroacetamide for 30 min (room temperature, light-protected). Lys-C protease (enzyme: substrate ratio of 1:75) was added for 4 hours at 25 °C. The TAEB-diluted sample (2 M Urea final) was supplemented with trypsin (1:75 ratio) followed by overnight digestion at 25 °C. Acidified peptides (1% formic acid final) were purified on the SDR-RP (C18) multi-stop-and-go-tip (Stage Tip)⁸⁷.

LC MS/MS label-free quantitative proteomic analysis from epididymal white adipose tissue

All samples were analyzed by the Proteomics Facility on a Q Exactive Plus Orbitrap mass spectrometer coupled to an EASY nLC 1000 (Thermo Scientific). Peptides were loaded with solvent A (0.1% formic acid in water) onto an in-house packed analytical column (50 cm–75 µm I.D., filled with 2.7 µm Poroshell EC120 C18, Agilent) and separated with 150 min gradients. Peptides were chromatographically separated at a constant flow rate of 250 nL/min using the following gradient: 4–6% solvent B (0.1% formic acid in 80 % acetonitrile) within 5.0 min, 6–23% solvent B within 120.0 min, 23–54% solvent B within 7.0 min, 54–85% solvent B within 6.0 min, followed by washing and column equilibration. The mass spectrometer was operated in data-dependent acquisition mode. The MS1 survey scan was acquired from 300–1750 m/z at a resolution of 70,000. The top 10 most abundant peptides were isolated within a 2.1 Th window and subjected to HCD fragmentation at a normalized collision energy of 27%. The AGC target was set to 5E5 charges, allowing a maximum injection time of 60 ms. Product ions were detected in the Orbitrap at a resolution of 17,500. Precursors were dynamically excluded for 20.0 s. All mass spectrometric raw data were processed with MaxQuant⁸⁸ (version 1.5.3.8, using default parameters and searched against the canonical murine Uniprot reference proteome (UP589, downloaded on 28/08/2020). Label-free quantification option was enabled, as was the match-between runs option between replicates. Follow-up analysis was done in Perseus⁸⁹ 1.6.15. Hits from the decoy database, the contaminant list, and those only identified by modified peptides were removed. Afterwards, results were filtered for data completeness in replicate groups, and LFQ values were imputed using sigma downshift with standard settings. Finally, FDR-controlled T-tests between sample groups were performed with s0 = 0.2. Proteins were annotated on GOBP, GOCC, KEGG and Pfam terms. Afterwards, all one-way ANOVA (s = 0.2) significant candidates were Z-scored and hierarchically clustered using the Euclidean distance model. Resulting clusters were merged onto the whole dataset, and a Fisher's Exact Test was calculated on cluster level with the whole dataset as background.

RNA isolation from mouse tissue and primary adipocytes

Total RNAs from tissues, sperm, and primary adipocytes were isolated using the TRI reagent (Sigma) followed by clean-up with RPE buffer (Qiagen, Germany). The quality of RNA was validated by the Agilent RNA 6000 Nano-Kit in Agilent 2100 Bioanalyzer according to the manufacturer’s protocol. (Agilent Technologies, Waldbronn, Germany).

RNA sequencing

Small RNA sequencing: (1) For sRNA-seq from F₀ sperm, library preparation and sRNA-seq were performed at Cologne Centre for Genomics, Germany. 50 ng of total RNA was used for library preparation, followed by size selection (Small RNA-Seq Library Prep Kit, Lexogen). The library was sequenced in 1 × 50-bp single-end reads on NovaSeq6000 serie no. A00316. (2) Small RNA sequencing from F₀ and F₁ eWAT was carried out after quality control, 125 ng of total RNA was used for library preparation (Nextflex library kit), followed by sequencing by 1 × 50-bp single-end reads on Illumina Novaseq 6000. mRNA sequencing: mRNA sequencing was performed in-house. After quality control, 1 µg of total RNA was used for library preparation (NEBNext RNA prep kit) followed by 2 × 50 bp paired-end sequencing on Illumina Novaseq 6000.

RNAseq data analysis

Small RNA sequencing of spermatozoa

SncRNA-seq analysis of murine sperm was performed with Seqpac ver. 1.2.0, where sncRNA biotypes were annotated with 1 mismatch against Ensembl protein and non-coding RNA from GRCm39, miRbase, piRNA from piRBase ver. 3.0, tRNA from tRNAscan-SE, rRNA from RNAcentral and mitochondria from GRCm39:MT. Biotypes were assessed based on the following hierarchy: miRNAs, mitochondrial rRNAs, rRNAs, mitochondrial protein-coding mRNAs, mitochondrial tRNAs, tRNAs, piRNAs, nuclear protein-coding mRNAs, lncRNAs, and miscRNAs. Baseline filtering of one read per sample, and a size of 15–75 nt long was performed. Normalization was performed with counts per million (cpm).

Small RNA sequencing of somatic (adipose) tissue

Adapter sequences were trimmed from raw miRNA-seq reads using Cutadapt. The processed reads were mapped to the reference genome and aligned using the Bowtie algorithm within the miRDeep2 pipeline⁹⁰. Normalization was performed using the Trimmed Mean of M-values (TMM) method to account for sequencing depth and composition bias. Subsequent data analysis, including differential expression analysis, was conducted in R⁹¹ using respective Bioconductor packages⁹².

mRNA sequencing

Sequencing reads were aligned to the mouse reference genome from Gencode (vM25)⁹³ using STAR aligner (v2.7.9a)⁹⁴, and gene features were counted by using featureCountsv (2.0.3)⁹⁵. The genes with low expression were filtered out using filterByExpr function with default settings from edgeR (v4.0.16)⁹⁶ R package. Differential expression analyses were performed using edgeR’s voomLmFit⁹⁷ function with sample quality weights (sample.weights = TRUE) and linear model “0 + group”, whereas group represents the combination of all experimental groups of interest. The experimental conditions and linear models used for each dataset are summarized below. Genes with adjusted p values below 0.05 were considered as differentially expressed. Gene enrichment analyses were performed by clusterProfiler (v4.10.1)⁹⁸ R package using both over-representation analysis (ORA) and gene-set enrichment analysis (GSEA) using differentially expressed genes and log-fold-change (logFC) ranked genes (n_permutation = 1,000,000), respectively. Gene sets were retrieved from Gene Ontology (GO.db v3.18.0) and Reactome (reactome.db, v1.86.2) databases.

Smart-seq2 mRNA sequencing of murine blastomeres

Custom-made Smart-seq2 protocol was implemented as before⁷⁴. Briefly, after thawing the plate, 1 µL of dNTP mix (10 mM), 0.1 µL of oligo-dT (100 µM, /5Biosg/AAGCAGTGGTATCAACGCAGAGTAC(T)30VN-3′) and 0.9 µl nuclease-free water were added to each well, and the lysis protocol was run when the samples were incubated at 72 °C for 3 min in a thermocycler. Samples were then put on ice for at least 1 min. Next, 5.60 µL of reverse-transcription mix containing 0.25 µL of RNase I, 2 µL of 5× first-strand buffer, 2 µL of betaine (5 M), 0.06 µL of MgCl₂ (1 M), 0.5 µL of DTT (100 mM), 0.5 µL of SuperScript III (200 U/µL), and 0.29 µl of nuclease-free water was added to each well and reverse transcription was carried out as follows: 42 °C–90:00, 42 °C hold, 42 °C–12:20, 10 cycles (50 °C–2:00, 42 °C–2:00), 39 °C–12:00, 70 °C–15:00, 4 °C hold. After the first 90 min of RT, 0.4 µL of 12,5 µM TSO (0.83 µM final concentration) was added to each reaction at room temperature. The plate was resealed, centrifuged for 1 min at 700 × g, and placed back on the thermocycler to resume cycling from 42 °C–12:20. After reverse transcription, 15 µL of cDNA enrichment PCR reaction mix containing 12.5 µL of 2× KAPA HiFi HotStart ReadyMix, 0.25 µL of ISPCR primer (10 µM, /5Biosg/AAGCAGTGGTATCAACGCAGAGT-3′), and 2.25 µL of nuclease-free water was added to each well and cDNA enrichment PCR was performed as follows for 21 cycles: 98 °C–3:00, 7 cycles (98 °C–0:20, 60 °C–4:00, and 72 °C–6:00), 7 cycles (98 °C–0:20, 64 °C–0:30, and 72 °C–6:00), and 7 cycles (98 °C–0:20, 67 °C–0:30, and 72 °C–7:00), 72 °C–10:00, 4 °C hold. The cDNA purification was performed using AMPure XP beads at a ratio of 1:1 (Beckmann). For elution, Qiagen elution buffer (EB) was used. cDNA concentrations were measured using the Qubit 3.0 fluorometer according to the manufacturer’s protocol, and all cDNA samples were analyzed on Agilent Tape station using the High Sensitivity DNA assay.

Dual-indexed Illumina Nextera XT sequencing libraries were prepared using 25% of the recommended reaction volumes of Nextera XT components. All Nextera XT libraries were prepared using 125 pg of input cDNA for tagmentation, and were subsequently enriched for 12 PCR cycles, and purified using AMPure XP beads (Beckmann) at a sample:bead ratio of 0.6:1. Library concentrations were measured using Qubit 3.0 fluorometer, and the fragments were analyzed on the Agilent Tape station using the High Sensitivity DNA assay. All libraries were individually normalized and diluted to a concentration of the pooled library to 5.4 ng/µL with an average size ~515 bp (16 nM). The libraries were sequenced on Illumina NextSeq500 in a single-end 75-bp format (FC-404-2005, Illumina). Single-nuclei libraries were sequenced at an average depth of 3 million reads.

Male participants, sperm samples and analysis

The intervention study was approved by the Danish National Committee on Health Research Ethics (SJ-769), the Danish Data Protection Agency (REG-048- 2019) and performed at the Fertility Clinic, Zealand University Hospital, Denmark⁷⁷. The semen samples were collected from patients living with obesity ( ≥  30 kg/m2) via masturbation following 2 days of abstinence. The sperm analyses were performed in the IVF laboratory in an authorized tissue establishment. The other semen samples were diluted with MHM-C (Fujifilm) and frozen immediately for further analysis. The semen samples were collected via masturbation following two days of abstinence, and the samples were analyzed a maximum of two to three hours after production. After liquefaction, ejaculate volume and sperm concentration (total, motile and progressive) were assessed using a microscope (Leica DM750). Sperm concentration was determined with a CellVision counting chamber (CellVision Technology, Heerhugowaard, the Netherlands). The sperm analyses were performed in the IVF laboratory in an authorized tissue establishment.

Study design and approvals

This prospective intervention study was conducted between May 2020 and April 2022 at the Fertility Clinic, Zealand University Hospital, Denmark. The primary outcomes have been published⁷⁷. The study was approved by the Danish National Committee on Health Research Ethics (SJ-769), the Danish Data Protection Agency (REG-048- 2019) and registered on the Clinical Trial website (www.ClinicalTrials.gov identifier NCT04721938). All participants provided written informed consent.

Primary adipocyte culture and transfection

Epididymal white adipose tissues (eWAT) from 6- to 8-week-old male C57BL/6 J mice were removed, minced, and digested with pre-warmed collagenase media (500 U/mL collagenase type 2 (LS004183, Worthington) and 3.75 U/mL DNase I (10104159001, Sigma). Isolated cells were seeded in 6-well plates and grown in DMEM/Ham’s F12 medium (21331, Vendor: Gibco), supplemented with 0.1% Biotin/D-Pantothenate (33 mM/17 mM), 1% penicillin-streptomycin, and 20% FBS. Until reaching 80% confluency, primary cells were induced by induction medium with 1 µM rosiglitazone, 850 nM insulin, 1 µM dexamethasone, 250 µM 3-isobutyl-1-methyl-xanthin (IBMX) and 10% FBS. Subsequently, the medium was changed every other day by a differentiation medium containing 1 µM rosiglitazone, 850 nM insulin and 10% FBS. 100 nM of Negative Control and mmu-let-7-d/e-5p mimics or LNA oligos targeting Dicer1 were reversely transfected using Lipofectamine RNAiMAX (Thermo Fisher Scientific) at day 4 after induction. Completely differentiated cells were harvested at 7 days after induction for further analysis.

Plasmid design and dual luciferase assay

For the generation of reporter constructs, the specific region of murine Ago2, Dicer-1, Hif1an, Hk2, Cpt2, Crat, Pcx, Idh2, Sdha, Sdhaf3, Pdk2, Slc25a22 and Slc25a24 containing either wild-type (wt) or mutated (mut) let-7 binding sites was cloned behind the translation stop codon of the firefly-luciferase gene in pmirGLO Dual Luciferase vector using custom DNA oligos provided by IDT. Renilla luciferase coding in the same plasmid was used as a control reporter for normalization purposes. To report the miRNA activity in mammalian cells, HEK 293 cells were reversely transfected in 96-well plates with 60 ng DNA using the TransIT transfection reagent (mirusbio) for 3 days and forwardly transfected with let-7d/e-5p mimics for 2 days. HEK-293 cells were cultured in DMEM medium (21013, Gibco) supplemented with 5.6 mM L-glutamine, 1% penicillin-streptomycin, and 10% FBS. After 72-h vector transfection and 48-h miRNA mimics transfection, DLRAs (Promega) were performed. The relative activities of firefly luciferase, indicating the interaction between miRNA and the specific region of targeted mRNA were normalized by the activities of Renilla luciferase.

TaqMan assay and general qPCR

To quantify the levels of mature miRNAs, TaqMan MicroRNA Reverse-Transcription Kit (4366596, Thermo Fisher), TaqMan MicroRNA Assays and TaqMan Universal Master Mix II (4440042, Thermo Fisher) were applied according to manufacturer’s protocols. The level of mature hsa-miR-16 was used as an internal control in human samples. The relative expression of mature miRNA was calculated using a comparative method (2- ∂∂CT) according to the ABI Relative Quantification Method.

Determination of OCRs and measurement of glycolytic activity

Primary adipocytes from epididymal SVF (1°eWA) were seeded into Agilent Seahorse XFe96 Bioanalyzer microplates. Induction and differentiation were performed as described above. After four days of differentiation, 16,000 per well mature adipocytes were re-seeded and transfected with LNA oligos targeting Dicer1 in DMEM/Ham’s F-12 medium plus 10% Fetal Bovine Serum, 1% P/S, 0.1% Biotin and 0.1% Pantothenic acid (Growth medium) at 37 °C and 5% CO₂. Seahorse measurement was conducted at day 7 from differentiation. For each seahorse plate, the corresponding calibration plate was prepared 24 hours prior to experiments using 200 µl XF Seahorse Calibrant Agilent per well. Corresponding media were prepared before the experiment and consisted of Assay Medium supplemented with 25 mM Glucose, 1 mM Glutamine, 2 mM Sodium Pyruvate, set to pH = 7.4 and filtered sterile. The calibration plate possessed a cartridge having 4 pockets per well. Shortly before the measurement pocket A was filled with 20 µl 20 µM Oligomycin, pocket B with 22 µl 20 µM FCCP and pocket C with 25 µl 5 µM Antimycin A and Rotenone. One hour before measurements, plates were washed with 1× PBS and changed to 180 µl Assay Medium in a non-CO₂ incubator at 37 °C. Calibration was started using calibration plates, measuring O₂ and pH LED Value/emission/Initial reference Delta for each well. All measurements started with measuring basal values, followed by injection of Oligomycin, FCCP, and Rotenone plus Antimycin A (MitoStressKit). Measurement parameters were: Mix 3 min, wait 0 min, measure 3 min with each reagent’s effect assessed within three consecutive measurement cycles with a total duration of 18 min per reagent injection.

Protein extraction and western blot analysis

After being washed once by PBS, the treated cells at harvest time were lysed by RIPA buffer supplemented with cOmplete and PhosSTOP and quantified by Pierce BCA Protein Assay Kit (23227, Thermo Fisher). Western blot analyses were carried out according to standard protocols with primary antibodies against COXIV (Cell Signaling Technologies #4850, 1:11,000 dilution), VDAC (Cell Signaling Technologies #4661, 1:11,000 dilution), HK2 (Cell Signaling Technologies #2897, 1:1,1000 dilution), SLC25A24 (Abcam Ab221120, 1:1000 dilution), CPT2 (Abcam Ab181114, 1:1,1000 dilution), CRAT (AP Proteintech #15170, 1:1000) and IDH2 (Cell Signaling Technologies #60322, 1:1,1000 dilution). Relative protein quantification was normalized by HSC70 (Santa Cruz sc-7298, 1:1000 dilution) or HSP90 (Cell Signaling Technologies #4877, 1:1000 dilution).

Access to reagents

All requests for data and reagent sharing should be directed to the corresponding author (janwilhelmkornfeld@bmb.sdu.dk).

Statistical analysis

Data were shown as mean ± standard error. One-way ANOVA and two-way ANOVA followed by Tukey’s multiple comparisons test were used for statistical analysis in in vivo experiments. Significant differences between groups were noted in the figure legends, respectively.

Reporting summary

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

Author contributions

C.H., J.H.P., J.M.G., K.K., M.A.M., M.R.F., R.B., and J.W.K. contributed to the conceptualization of the research idea and study design. C.H., J.H.P., A.A., N.S., H.T., R.V., K.K.D.L., S.I., P.K., H.T., T.M., S.Z., P.M.M.R., R.V., J.H., B.D.M.L., Y.T.C., W.C.H., Y.P.B., Y.F.J., A.L.R., N.C.L.J., K.S., A.T., J.W.L., E.K.H., I.E.J., S.H.K., R.L.B., J.M.G., K.K. performed experiments and data analysis. C.H., J.H.P., A.A., S.I., H.T., B.D.M.L., K.S., A.T., S.D., S.H.K., K.K., N.J.F., M.A.M., M.R.F., A.Ö., R.B., and J.W.K. contributed to discussions and performed training. C.H., J.H.P., J.M.G., K.K., M.A.M., R.B. and J.W.K. conceived experiments. C.H. and J.W.K. wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Next-generation sequencing: (1) Illumina sRNA-seq and mRNA-seq datasets from F₀ and F₁ liver and eWAT (2) Illumina sRNA-seq from F₀ sperm (3) Illumina mRNA-seq from offsprings sired from miRNA-injected zygotes (F_m /F_MI) eWAT (4) Illumina mRNA-seq from 1° eWA overexpressing cel-mir-67 (negative control), let-7d-5p and let-7e-5p mimics, and (5) Smart-Seq2 single-cell mRNA-seq data from miRNA-injected 8-cell embryos are available as Gene Expression Omnibus (GEO) SuperSeries GSE280278. Proteomics: The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE⁹⁹ partner repository with the dataset identifier PXD042800. Metabolomics & lipidomics:All metabolomic and lipidomic datasets are available at the Metabolomics Workbench¹⁰⁰ data repository as project ID PR002238. All data supporting the findings described in this manuscript are available in the article and in the Supplementary Information and from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-69686-5.