Graphical abstract

   graphic file with name fx1.jpg
   [59]Open in a new tab

Highlights

     * •
       PCOS-sons are often obese and have dyslipidemia
     * •
       miRNAs altered in the serum of PCOS-sons and women with PCOS
       targets PCOS-risk genes
     * •
       Small RNAs present in sperm imply transgenerational transmission of
       phenotype in mice
     * •
       Shared miRNAs between mouse sperm of F[1]–F[3] generations and
       human serum are revealed
     __________________________________________________________________

   Risal et al. found that the sons of women affected by polycystic ovary
   syndrome (PCOS) are frequently obese and dyslipidemic. Male descendants
   of obese or androgen-exposed mothers also exhibit reproductive and
   metabolic problems across generations, mediated by sperm small RNAs
   dysregulation. Common predicted small RNA targets are suggested in
   PCOS-affected mice and PCOS-sons’ serum.

Introduction

   Polycystic ovary syndrome (PCOS) is the leading causes of female
   infertility and is associated with a high degree of comorbidities,
   including type 2 diabetes and psychiatric disorders.[60]^1^,[61]^2 The
   key feature of PCOS is hyperandrogenism,[62]^3 and >50% of women with
   PCOS are obese,[63]^4 which exacerbates their symptoms. Although ∼15%
   of women worldwide suffer from PCOS, management of the syndrome is
   hindered by lack of insight into the origin and underlying mechanisms.
   It is known that PCOS runs in families with both genetic and epigenetic
   contributions, the latter of which are phenotypic changes that do not
   involve alterations in the DNA sequence,[64]^5 and that daughters of
   women with PCOS are five times more likely to be diagnosed with the
   syndrome.[65]^6 Although a distinct phenotype of male offspring related
   to PCOS has not yet been defined, sons born to mothers with PCOS
   (PCOS-sons) display increased body mass index (BMI), insulin
   resistance,[66]^7^,[67]^8 and prepubertal signs of reproductive
   dysfunction with increased antimüllerian (AMH) hormone levels,
   indicating increased Sertoli cell number.[68]^9 However, at adult age,
   there were no differences in circulating AMH, sex steroids, or
   gonadotropins, in addition to sperm production.[69]^9 Moreover,
   brothers of women with PCOS have increased AMH hormone levels, altered
   gonadotrophin, and steroidogenic secretion,[70]^10^,[71]^11 as well as
   a metabolic phenotype with insulin resistance and pancreatic β-cell
   dysfunction, dyslipidemia, and an increased cardiovascular disease
   risk.[72]^12^,[73]^13^,[74]^14^,[75]^15^,[76]^16 Recently, it was shown
   that genetic risk factors for PCOS increase the odds of obesity, type 2
   diabetes, cardiovascular disease, and androgenic alopecia in
   men.[77]^17 Although genetic components are likely involved in a
   male-PCOS phenotype, the clinical observations suggest that maternal
   obesity and PCOS could also affect the development of male fetuses and
   predispose them to reproductive and metabolic disorders in later life,
   as shown in their female siblings.

   Our and others’ recent animal studies show that prenatal androgen or
   AMH exposure predisposes the first-generation (F[1]) female offspring
   to develop PCOS-like traits, and both reproductive and metabolic
   phenotypic alterations are passed on to the F[3] of females, suggesting
   non-genetic transgenerational transmission.[78]^6^,[79]^18 Previous
   studies reveal that F[1] male offspring of both rodents and sheep
   develop an aberrant reproductive and metabolic phenotype due to
   prenatal androgen exposure.[80]^19^,[81]^20^,[82]^21^,[83]^22^,[84]^23
   Whether these phenotypic changes are transmitted further to subsequent
   male generations has not yet been explored. Other studies have
   demonstrated that maternal stress and endocrine disruptors in rodent
   models cause a transgenerational transmission of phenotypes on both
   female and male germline.[85]^24^,[86]^25^,[87]^26^,[88]^27
   Diet-induced obesity from early life in male mice affects sperm with
   altered small non-coding RNAs (sncRNAs), which predisposes their male
   offspring in subsequent generations to obesity, suggesting epigenetic
   inheritance potentially driven by the germline.[89]^28^,[90]^29

   Our study now provides evidence that PCOS-sons have altered lipid
   profiles and are at higher risk to childhood obesity. Several
   differentially expressed (DE) miRNAs found in serum from PCOS-sons are
   overlapped with those identified in serum or follicular fluid of women
   with PCOS, likely to regulate PCOS-risk genes identified by genome-wide
   association studies
   (GWASs).[91]^30^,[92]^31^,[93]^32^,[94]^33^,[95]^34^,[96]^35^,[97]^36
   We then turned to the mouse models and showed that prenatal androgen
   exposure and/or maternal obesity resulted in the transmission of
   reproductive and metabolic traits to F[3] male offspring associated
   with common DEsncRNAs in sperm of F[1], F[2], and F[3] offspring (i.e.,
   transgenerational DEsncRNAs). Moreover, we found that several DEsncRNAs
   in serum from PCOS-sons are shared with transgenerational DEsncRNAs in
   mouse sperm, highlighting the translational relevance of our
   transgenerational mouse studies.

Results

PCOS-sons are more obese together with altered circulating lipid profile

   Our previous findings show that PCOS-sons have increased BMI and
   abnormal glucose and lipid metabolism.[98]^7^,[99]^8 Besides, they have
   increased AMH levels during infancy, childhood, and adulthood as well
   as smaller testicular volume.[100]^8 To follow up these findings in a
   large cohort, we performed a Swedish nationwide register-based cohort
   study to investigate whether PCOS-sons are more often diagnosed with
   obesity ([101]Figure 1A; [102]Table S1). Using the Swedish Medical
   Birth Register and the National Patient Register, a total of 467,275
   sons born in Sweden between July 2006 and December 2015 were included
   and followed from 2 years of age. From them, 9,828 (2.10%) were born to
   a mother diagnosed with PCOS. Of the mothers diagnosed with PCOS, 165
   (1.67%) had at least one prescription of metformin discharged from a
   pharmacy. Obesity diagnosed in children was identified by using the
   International Code of Diseases, v.10 (ICD-10: E66). Overall, an
   increased risk of obesity in sons born to mothers with PCOS (with or
   without use of metformin during pregnancy) was found (adjusted hazard
   ratio [HR] = 1.51 95% confidence interval [1.27–1.79]). A similar risk
   was found in the sub-analysis assessing only women with PCOS and
   without use of metformin. Finally, when stratifying maternal BMI, there
   was an association between maternal PCOS and childhood obesity only in
   the group of women with BMI ≥25 (HR = 1.60 [1.33–1.91]) with no
   association in women with BMI <25 (HR = 1.07 [0.66–1.76])
   ([103]Figure 1A). Of note, only 2.1% of the mothers had PCOS, which is
   much lower than the expected 10%–18% in the general population of
   reproductive age.[104]^37 In a longitudinal case-control study from
   Chile,[105]^7 we showed that already at Tanner II-III and Tanner IV-V
   PCOS-sons have higher circulating cholesterol and low-density
   lipoprotein (LDL) cholesterol compared with control sons
   ([106]Figures 1B and 1C; [107]Table S2). Moreover, sons of mothers with
   maternal obesity have higher BMI and larger waist circumference
   compared with sons of mothers with BMI <25 ([108]Figures 1D–1F). The
   prevalence of children who were overweight and obese was higher in
   those children born to mothers with BMI >25 during pregnancy
   ([109]Figure 1G).

Figure 1.

   [110]Figure 1
   [111]Open in a new tab

   Risk of being obese and altered lipid profile of sons of women with
   PCOS

   (A) Risk of obesity during childhood in sons born to mother with
   polycystic ovary syndrome (PCOS) identified in the Swedish National
   Patient Register and in the Multi-Generational Register (n = 467,275),
   expressed as hazard ratios (HRs) and 95% confidence intervals (CIs).
   The covariates in the adjusted model were maternal age at delivery
   stratified as <25, 25–29, 30–34, and ≥35; maternal BMI stratified as
   <18.5, 18.5–24.9, 25–29.9, and ≥30; parity (multiparous, nulliparous),
   cigarette consumption at enrollment (yes/no); assisted reproduction
   (yes/no); size for gestational age (adequate, small or large); preterm
   birth (yes/no); Apgar <7 at 5 min; cesarean section; diabetes
   (gestational diabetes, diabetes mellitus, or use of metformin during
   pregnancy); and finally sub-analyses of women with a BMI <25 and women
   with a BMI ≥25.

   (B and C) Total cholesterol (B) and low-density lipoprotein (C) in sons
   of women with PCOS at Tanner stages I–V.

   (D–G) BMI of sons of women with PCOS (D), waist circumference of sons
   of women with PCOS (E), body weight of sons of women with PCOS (F), and
   prevalence of obesity of sons of women with PCOS (G) with Z score BMI
   <2 (normal weight) and Z score BMI ≥2 (overweight-obesity) according to
   maternal nutritional state at beginning of pregnancy distributed in BMI
   <25 and ≥25 kg/m^2. Differences were calculated by chi-squared test for
   prevalence values and Student’s t test for BMI and waist diameter.
   Control-S, control son; PCOS-S, PCOS son.

   (H) The numbers of differently expressed (DE) sncRNAs in each biotype
   in whole blood from sons of women with PCOS and control women (n =
   9/group).

   (I) Heatmap of DEsncRNA in whole blood from sons of women with PCOS and
   control women (n = 9/group).

   (J) Chord diagram showing DEmiRNA in whole blood from sons of women
   with PCOS and their target genes. Chords in different colors represent
   Gene Ontology (GO) enrichment. The expression of the DEmiRNAs is shown
   in red (up-regulated) and blue (down-regulated).

sncRNA analysis on serum of PCOS-sons identifies miRNAs targeting loci of
PCOS-risk genes

   Next, we investigated molecular features in serum of PCOS-sons (Chilean
   case control) by sncRNA sequencing (sncRNA-seq). Among DEsncRNAs
   between sons of women with and without PCOS, Piwi-interacting RNAs
   (piRNAs), rRNA-derived small RNAs (rsRNAs), and microRNAs (miRNAs) were
   the major biotypes ([112]Figures 1H and 1I). As circulating miRNAs are
   extensively characterized in gene regulation and as stable biomarkers,
   we first asked whether DEmiRNAs of serum of PCOS-sons are also
   identified in women with PCOS by comparing our data with previously
   profiled serum, granulosa cells, or follicular fluid miRNAs expression.
   We found that 7 out 11 DEmiRNAs in the serum of PCOS-sons were also DE
   in women with PCOS ([113]Data S1): hsa-miR-1299,[114]^38
   hsa-miR-122-5p,[115]^39 hsa-miR-199b-3p,[116]^40^,[117]^41
   hsa-miR-199a-5p,[118]^42 hsa-miR-1307-3p,[119]^43
   hsa-miR-96-5p,[120]^44 and hsa-miR-548ar-3p.[121]^41 Moreover, we
   examined in silico targets of these DEmiRNA and revealed 783 potential
   target genes ([122]Data S2), among which six are reported as PCOS-risk
   genes by GWAS, i.e., AOPEP[123]^30^,[124]^33 (hsa-miR-1299),
   TOX3[125]^32^,[126]^45 (hsa-miR-1299), ERBB4[127]^32 (hsa-miR-199b-3p),
   GABRB1[128]^46 (hsa-miR-548ar-3p), ADGRB3[129]^33 (hsa-miR-96-5p), and
   MYRIP[130]^36 (hsa-miR-96-5p). To understand the function of these
   overlapped miRNAs, we performed Gene Ontology (GO) pathway analyses of
   targeted genes ([131]Figure 1J; [132]Data S3). Among the enriched
   pathways that potentially could contribute to the pathology of PCOS are
   insulin resistance (e.g., FOXO1, MTOR, GYS1, PTEN, RPS6KB1, STAT3,
   CREB5, TRIB3), sex differentiation (e.g., FER, FOXF2, LRP2, PGR, SIRT1,
   LHX9, AGO4), response to hormone (e.g., KLF9, GABRB1, ITGA3, BCAR3,
   CYP7B1), regulation of cellular catabolic process (e.g., ABCA2, ABCD1,
   PIK3CA, DISC1, MAP3K5), and GnRH secretion (e.g., ITPR1, PIK3CA,
   PIK3R1, PLCB4, PIK3R3).

Maternal obesity in F[0] dam causes transgenerational reproductive
dysfunction in male offspring

   We used our previously validated mouse models[133]^6 to investigate
   whether F[1] male offspring that were directly exposed to diet-induced
   maternal obesity, prenatal androgens, or the combination of the two
   exposures could develop reproductive traits in adult males and if such
   traits are passed on to F[2] (direct germline exposure, i.e.,
   intergenerational) and F[3] (transgenerational) male offspring. The
   phenotype of F[0] dams has recently been described in detail.[134]^6 In
   total, four experimental lineages were studied: (1) control diet +
   vehicle (control); (2) control diet + dihydrotestosterone
   (androgenized); (3) high-fat, high-sucrose diet + vehicle (obese); and
   (4) high-fat, high-sucrose diet + dihydrotestosterone (obese and
   androgenized) ([135]Figure 2A). F[1] male offspring were mated with
   unrelated healthy females to generate F[2], and F[2] male offspring
   were mated with unrelated healthy females to generate F[3] and each
   generation were compared with parallel bred controls, which is required
   to study transgenerational inheritance. Phenotypic testing was
   performed between 15 and 22 weeks of age in each generation.

Figure 2.

   [136]Figure 2
   [137]Open in a new tab

   Prenatal androgen and maternal obesity exposure causes
   transgenerational reproductive phenotypes in male offspring

   (A) Schematic illustration of the experimental design. (1) CD+Veh
   (control lineage), (2) CD+DHT (androgenized lineage), (3) HFHS+Veh
   (obese lineage), and (4) HFHS+DHT (obese and androgenized lineage).

   (B) Transgenerational increase in anogenital distance (AGD) in the
   obese and the obese and androgenized lineages. CD+Veh (F[1]: n = 10,
   litters = 5; F[2]: n = 15, litters = 4; F[3]: n = 16, litters = 4);
   CD+DHT (F[1]: n = 10, litters = 5; F[2]: n = 13, litters = 4; F[3]: n =
   23, litters = 4); HFHS+Veh (F[1]: n = 10, litters = 5; F[2]: n = 18,
   litters = 4; F[3]: n = 19, litters = 4); and HFHS+DHT (F[1]: n = 10,
   litters = 7; F[2]: n = 17, litters = 4; F[3]: n = 13, litters = 4).

   (C) Testis weight normalized to body weight in F[1]–F[3] male
   offspring. CD+Veh (F[1]: n = 7, litters = 5; F[2]: n = 10, litters = 4;
   F[3]: n = 8, litters = 4); CD+DHT (F[1]: n = 10, litters = 5; F[2]: n =
   13, litters = 4; F[3]: n = 8, litters = 4); HFHS+Veh (F[1]: n = 10,
   litters = 5; F[2]: n = 12, litters = 4; F[3]: n = 8, litters = 4); and
   HFHS+DHT (F[1]: n = 8, litters = 7; F[2]: n = 16, litters = 4; F[3]:
   n = 8, litters = 4).

   (D) Total sperm counts in F[1]–F[3] male offspring. CD+Veh (F[1]: n =
   10, litters = 5; F[2]: n = 6, litters = 4; F[3]: n = 16, litters = 4);
   CD+DHT (F[1]: n = 10, litters = 5; F[2]: n = 8, litters = 4; F[3]: n =
   12, litters = 4); HFHS+Veh (F[1]: n = 11, litters = 5; F[2]: n = 8,
   litters = 4; F[3]: n = 6, litters = 4); and HFHS+DHT (F[1]: n = 10,
   litters = 7; F[2]: n = 8, litters = 4; F[3]: n = 11, litters = 4).

   (E) Representative transmission electronic microscopy images of
   mitochondrial morphology in pachytene spermatocytes in F[1]–F[3]
   testis. Yellow arrows: condense mitochondria, red arrows: intermediate
   form of mitochondria, and pink arrows: elongated form of mitochondria.

   (F) Quantification of mitochondrial morphology: normal type, condense
   (C); abnormal types, intermediate (I) and elongated (E) forms, in
   pachytene spermatocytes in F[1], F[2], and F[3] male offspring. Scale
   bar: 5 μm. CD+Veh (F[1]: n = 3, litters = 3; F[2]: n = 2, litters = 2;
   F[3]: n = 5, litters = 4); CD+DHT (F[1]: n = 3, litters = 3; F[2]: n =
   5, litters = 4; F[3]: n = 5, litters = 4); HFHS+Veh (F[1]: n = 3,
   litters = 3; F[2]: n = 4, litters = 4; F[3]: n = 5, litters = 4); and
   HFHS+DHT (F[1]: n = 3, litters = 3; F[2]: n = 2, litters = 2; F[3]: n =
   3, litters = 3).

   (G) Representative images of the mitochondrial sheath of sperm from
   testis in F[1], F[2], and F[3] male offspring. Yellow arrows: normal
   mitochondria, and red arrows: abnormal mitochondria. Scale bar: 5 μm.

   Lineage: control: CD+Veh, maternal control diet + vehicle exposure;
   androgenized: CD+DHT, maternal control diet + dihydrotestosterone
   exposure; obese: HFHS+Veh, maternal high-fat, high-sucrose diet +
   vehicle exposure; and the obese and androgenized: HFHS+DHT, maternal
   high-fat, high-sucrose diet + dihydrotestosterone exposure. Comparison
   between the groups was performed using linear mixed-effects models and
   non-repeated measures (ANODE, R package car). Each dot represents one
   offspring mouse. All data are presented as mean ± SEM.

   Anogenital distance, a marker of in utero androgen exposure,[138]^47
   was longer in F[1] and F[3] male offspring in the obese lineage,
   demonstrating a transgenerational effect due to maternal obesity,
   whereas it was longer only in the F[1] male offspring in the
   androgenized lineage ([139]Figure 2B). Notably, the transgenerational
   transmission of anogenital distance was independent of circulating sex
   steroids, as we found neither differences in circulating testosterone,
   dihydrotestosterone, and androstenedione nor in testis AMH
   concentrations in F[1] and F[3] male offspring in any of the lineages
   ([140]Figures S1A–S1D), suggesting that the transgenerational effects
   are caused by initial maternal condition rather than excessive
   circulating androgens in F[1] and F[3] male offspring.

   Moreover, we found lower testis weight in F[1] male offspring in the
   androgenized and obese lineages, respectively, compared with controls,
   although no difference was observed in their respective F[3] males
   ([141]Figure 2C). In contrast, there was a latent effect in F[2] and
   F[3] male offspring in the combined obese and androgenized lineage with
   lower testis weight compared with the obese lineage ([142]Figure 2C).
   In line with the low testis weight, the obese and androgenized lineage
   of F[3] showed a low total sperm count ([143]Figure 2D).

   Maternal obesity and prenatal androgen exposure affect mitochondrial
   morphology of MII oocytes.[144]^6 Accordingly, we analyzed
   mitochondrial morphology in the testis by transmission electron
   microscopy. During spermatogenesis, three different types of cristae
   morphology are present in mitochondria: orthodox type (Sertoli cells,
   spermatogonia, and preleptotene and leptotene spermatocytes),
   intermediate type (zygotene spermatocytes), and condense type
   (pachytene and secondary spermatocytes and early spermatids).[145]^48
   F[1] of the androgenized lineage and F[1] and F[3] male offspring of
   the obese lineage had spermatocytes with more intermediate-type
   (abnormal) mitochondria and a declining number of condensed (normal)
   mitochondria in pachytene spermatocytes and round spermatids
   ([146]Figures 2E and 2F). We further analyzed the mitochondrial sheath
   of sperm and found abnormal crista structures with vacuoles, indicating
   morphological abnormalities in F[1] male offspring in the obese and the
   obese and androgenized lineages ([147]Figure 2G). In F[2] male
   offspring, the aberrant mitochondrial sheath was observed in all three
   lineages ([148]Figure 2G), which was retained in F[3] male offspring as
   a transgenerational effect ([149]Figure 2G). These findings were
   further supported by dysregulated expression of key mitochondrial genes
   in testis of F[1], F[2], and F[3] male offspring, namely Tfam1
   (transcription factor A; mitochondrial), Drp1 (dynamin-related protein
   1), Opa1 (OPA1 mitochondrial dynamin-like GTPase), and Guf1 (GUF1
   homolog; GTPase) ([150]Figure S1E). Despite these mitochondrial
   phenotypes, there are no significant changes in sperm morphology in
   F[1]–F[3] male offspring and no effect on the fecundity of F[1] and
   F[2]adult males ([151]Figures S1F–S1H).

   Collectively, these results suggest that altered reproductive (testis
   and sperm mitochondrial) functions are transmitted across generations
   in male offspring of the obese lineage, and there is a strong and
   belated effect in F[3] male progeny in the combined obese and
   androgenized lineage.

Prenatal androgen exposure and maternal obesity cause transgenerational
metabolic dysfunction in male offspring

   F[1] male offspring in the androgenized and in the androgenized and
   obese lineages, respectively, gained more weight, whereas F[2] in the
   androgenized lineage gained less weight with no difference in F[3] male
   offspring ([152]Figure S1I). Both F[1] and F[3] male offspring in the
   androgenized and in the obese lineages, respectively, had more fat mass
   ([153]Figure 3A). Increased fat mass was observed independent of lean
   mass ([154]Figure S1J). In support of the increased adiposity in F[1]
   and F[3] male offspring in the androgenized and obese lineages, we also
   found enlarged epididymal adipocytes ([155]Figures 3B and 3C). The
   observation of increased fat mass and enlarged adipocytes was further
   supported by impaired glucose metabolism in F[1] and F[3] male
   offspring in the obese lineage as shown by increased area under the
   curve-oral glucose tolerance test (AUC-OGTT) ([156]Figures 3D, 3E, and
   [157]S2A).

Figure 3.

   [158]Figure 3
   [159]Open in a new tab

   Prenatal androgen and maternal obesity exposure causes
   transgenerational metabolic dysfunction in male offspring and
   reproductive dysfunction in male cousins caused by androgen exposure

   (A) Body composition presented as percentage of fat mass normalized to
   body weight (grams). CD+Veh (F[1]: n = 10, litters = 5; F[2]: n = 13,
   litters = 4; F[3]: n = 13, litters = 4); CD+DHT (F[1]: n = 7, litters =
   5; F[2]: n = 13, litters = 4; F[3]: n = 11, litters = 4); HFHS+Veh
   (F[1]: n = 10, litters = 5; F[2]: n = 16, litters = 4; F[3]: n = 13,
   litters = 4); and HFHS+DHT (F[1]: n = 7, litters = 4; F[2]: n = 13,
   litters = 4; F[3]: n = 12, litters = 4).

   (B) Epididymal adipocyte size measurements were made on six sections
   per mouse of F[1]–F[3]. CD+Veh (F[1]: n = 3 litters = 3; F[2]: n = 3,
   litters = 4; F[3]: n = 3, litters = 3); CD+DHT (F[1]: n = 3 litters =
   3; F[2]: n = 3, litters = 4; F[3]: n = 3, litters = 3); HFHS+Veh (F[1]:
   n = 3 litters = 3; F[2]: n = 3, litters = 4; F[3]: n = 3, litters = 3);
   and HFHS+DHT (F[1]: n = 3 litters = 3; F[2]: n = 3, litters = 4; F[3]:
   n = 3, litters = 3).

   (C) Representative images of epididymal adipocytes stained with
   hematoxylin and eosin. Scale bar: 200 μm.

   (D) Blood glucose levels at different time points during oral glucose
   tolerance test (OGTT).

   (E) Glucose area under the curve (AUC) at 0 to 90 min in F[1]–F[3]
   adult male offspring. CD+Veh (F[1]: n = 9 litters = 5; F[2]: n = 13,
   litters = 4; F[3]: n = 8, litters = 4); CD+DHT (F[1]: n = 10, litters =
   5; F[2]: n = 9, litters = 4; F[3]: n = 10, litters = 4); HFHS+Veh
   (F[1]: n = 10, litters = 5; F[2]: n = 16, litters = 4; F[3]: n = 13,
   litters = 4); and HFHS+DHT (F[1]: n = 9, litters = 6; F[2]: n = 15,
   litters = 4; F[3]: n = 6, litters = 4).

   (F) Liver triglyceride (TG) content in F[1]–F[3] male offspring
   normalized to tissue weight (mg/g). CD+Veh (F[1]: n = 6 litters = 5;
   F[2]: n = 10, litters = 4; F[3]: n = 10, litters = 4); CD+DHT (F[1]:
   n = 10, litters = 5; F[2]: n = 9, litters = 4; F[3]: n = 9, litters =
   4); HFHS+Veh (F[1]: n = 10, litters = 5; F[2]: n = 10, litters = 4;
   F[3]: n = 8, litters = 4); and HFHS+DHT (F[1]: n = 8, litters = 6;
   F[2]: n = 10, litters = 4; F[3]: n = 9, litters = 4).

   (G) Representative images of neutral lipid accumulation in the liver
   visualized by oil red O staining. Scale bar: 100 μm.

   (H and I) Respiratory exchange ratio (RER; VCO[2]/VO[2]) (H) and energy
   expenditure (EE) (I) was measured by indirect calorimetry by using the
   TSE system in F[1]–F[3] adult male offspring. CD+Veh (F[1]: n = 8
   litters = 4; F[2]: n = 4, litters = 4; F[3]: n = 4, litters = 4);
   CD+DHT (F[1]: n = 8, litters = 4; F[2]: n = 4, litters = 4; F[3]: n =
   4, litters = 4); HFHS+Veh (F[1]: n = 8, litters = 4; F[2]: n = 4,
   litters = 4; F[3]: n = 4, litters = 4); and HFHS+DHT (F[1]: n = 8,
   litters = 5; F[2]: n = 4, litters = 4; F[3]: n = 4, litters = 4).

   Lineage: CD+Veh, maternal control diet + vehicle exposure; CD+DHT,
   maternal control diet + dihydrotestosterone exposure; HFHS+Veh,
   maternal high-fat, high-sucrose diet + vehicle exposure. Comparison
   between the groups were performed using linear mixed-effects models and
   non-repeated measures (ANODE, R package car) except for (D), where
   linear mixed-effects models with repeated measure was used. #, CD+Veh
   vs. HFHS+Veh; ¤, HFHS+Veh vs. HFHS+DHT. Each dot represents one
   offspring mouse. All data are presented as mean ± SEM.

   Although there was no transgenerational transmission of adiposity and
   impaired glucose homeostasis in the combined obese and androgenized
   lineage, we found higher liver triglycerides content in F[3] male
   offspring ([160]Figure 3F). Furthermore, the accumulation of neutral
   lipids in the liver further corroborated these observations in the
   combined lineage, as well as the transgenerational effects of increased
   fat mass and enlarged epididymal adipocytes in the obese and
   androgenized lineages, respectively ([161]Figure 3G).

   To gain a deeper understanding of the metabolic phenotypes, we used
   indirect calorimetry and found that F[1] male offspring in the
   androgenized and the combined lineages, and F[3] male offspring in all
   lineages, had altered respiratory exchange ratio (RER) (day or night),
   which is an indicator of fuel selection and -utilization and altered
   energy expenditure (EE), indicating dysfunctional energy metabolism
   ([162]Figures 3H, 3I, and [163]S2B–S2D). These results demonstrate a
   transition from carbohydrate to fatty acid consumption, which
   corresponded to increased adiposity in F[1] and F[3] male offspring and
   occurred without a change in food intake and total activity.

   These findings suggest that metabolic dysfunction in F[1] male
   offspring as an effect of maternal obesity or prenatal androgen
   exposure are transmitted across generations, whereas the
   transgenerational effect in the combined obese and androgenized lineage
   is less pronounced.

   Reproductive and metabolic function in F[2] male offspring is less
   affected. The phenomenon that phenotypic changes skip one (or even two)
   generations has previously been observed in the female and male
   germline by us and others.[164]^6^,[165]^49^,[166]^50

sncRNAs in sperm accompany transgenerational transmission of phenotypic
traits

   To dissect the molecular basis of transmission of PCOS-related
   dysfunction, sperm was collected from the cauda epididymis and
   subjected to a swim-up assay to ensure that only motile mature
   spermatozoa, i.e., the pure fraction of ∼10 million sperm was used for
   sncRNA-seq analysis of F[1], F[2], and F[3] male offspring
   ([167]Figure 4A). First, we found that sperm from different lineages in
   F[1], F[2], and F[3] male offspring show various sncRNA biotypes: miRNA
   (22–24 nt), piRNA (20–34 nt), rsRNAs (15–44 nt), and tRNA-derived small
   RNAs (tsRNA; 27–36 nt) ([168]Figure S3A). To reveal the molecular basis
   of phenotypic variation, we further analyzed DEsncRNAs in F[1], F[2],
   and F[3] offspring of the androgenized, the obese, and the obese and
   androgen lineages ([169]Figure S3B).

Figure 4.

   [170]Figure 4
   [171]Open in a new tab

   The transgenerational inheritance pattern of sncRNA in sperm of
   androgenized, obese, and obese and androgenized lineages

   (A) Illustration showing the collection of sperm for sncRNA sequencing.

   (B) The principal-component analysis (PCA) showing different lineages
   in F[1], F[2], and F[3] male offspring.

   (C) Transgenerational expression patterns of F[1], F[2], and F[3] sperm
   differentially expressed miRNA, piRNA, rsRNA, and tsRNA.

   The heatmaps show the log[2] fold change (FC) of overlapped
   differentially expressed sncRNAs in F[1], F[2], and F[3]. CD+Veh (F[1]:
   n = 4, litters = 4; F[2]: n = 4, litters = 4; F[3]: n = 4, litters =
   4); CD+DHT (F[1]: n = 3, litters = 3; F[2]: n = 4, litters = 4; F[3]:
   n = 3, litters = 3); HFHS+Veh (F[1]: n = 4, litters = 4; F[2]: n = 4,
   litters = 4; F[3]: n = 4, litters = 4); and HFHS+DHT (F[1]: n = 4,
   litters = 4; F[2]: n = 4, litters = 4; F[3]: n = 4, litters = 4).

   First, we performed principal-component analysis (PCA) on the sncRNA
   profiles for the three comparisons: (1) control (control diet [CD] +
   vehicle [Veh]) vs. androgenized lineages (CD + dihydrotestosterone
   [DHT]), (2) CD+Veh vs. obese lineages (high fat, high sucrose [HFHS] +
   Veh), and (3) HFHS+Veh vs. obese and androgenized lineages (HFHS+DHT),
   respectively. The obese lineages, with or without androgen exposure,
   show a clear separation of F[1]–F[3] sncRNA profiles from controls,
   with less prominent separation in the androgenized lineage
   ([172]Figure 4B). Next, we defined transgenerational up-regulated and
   down-regulated DEsncRNAs of F[1], F[2], and F[3] male offspring sperm
   and identified 8, 140, and 9 transgenerational DEsncRNAs in the
   androgenized, obese, and combined obese and androgenized lineages,
   respectively ([173]Figure 4C; [174]Data S4). Interestingly, the obese
   lineage has the greatest number and diverse biotypes of DEsncRNAs
   across the three generations, indicating that maternal obesity resulted
   in more transgenerational effectors compared with prenatal androgen
   exposure. This interpretation agreed with phenotypic data that the
   obese lineage showed clear aspects of transgenerational reproductive
   and metabolic dysfunctions, whereas the androgenized lineage only shows
   transgenerational metabolic dysfunctions. This may also explain why
   there is a belated transgenerational effect in the combined obese and
   androgenized lineage.

   Next, we investigated where the transgenerational DEsncRNAs are derived
   from.[175]^51 Distinct from prepachytene piRNAs mainly silencing
   retrotransposons to protect the integrity of the
   genome,[176]^52^,[177]^53 pachytene piRNAs are dominantly transcribed
   from genic and intergenic regions and instruct mRNA degradation during
   late spermrmatogenesis[178]^54 Accordingly, the majority of identified
   transgenerational DEpiRNAs are presented within intergenic and genic
   regions in the obese lineage ([179]Figure 5A). However, only two
   transgenerational DEpiRNA in the androgenized (in the genic and
   intergenic region) and one transgenerational DEpiRNA in the combined
   obese and androgenized lineages (in the repeat region) were annotated,
   respectively. rsRNAs were classified according to their origins as
   mitochondrial 12S and 16S rRNAs; ribosomal 18S, 28S, 45S, 4.5S, and
   5.8S rRNAs; and nuclear 5S rRNAs. The transgenerational rsRNAs in the
   obese lineage derives mainly from 28S, 18S, and 45S rRNA
   ([180]Figure 5B). In the obese and androgenized lineage, only one
   transgenerational DErsRNA was derived from 28S rRNA, with no DErsRNAs
   in the androgenized lineage. tsRNAs were systematically annotated from
   tRNAs according to their coupled amino acids from genome and
   mitochondria. The majority of the transgenerational DEtsRNAs in the
   obese lineage are derived from Glu and Gly tRNA, followed by Val, Gln,
   and mitochondrial tRNAs ([181]Figure 5C). One tRNA can give rise to
   various tsRNAs, whose origins are mainly categorized into 5′-tRNA,
   3′-tRNA, internal-tRNA, and 3′CCA-tRNA. Transgenerational tsRNAs in the
   obese lineage were mainly derived from 5′-tRNA and internal tRNA, with
   a small proportion derived from 3′-tRNA with or without CCA end
   ([182]Figure 5C).

Figure 5.

   [183]Figure 5
   [184]Open in a new tab

   Transgenerational differential expression of sperm piRNA, rsRNAs and
   tsRNAs

   (A–C) Proportion of transgenerational (in sperm of F[1], F[2], and
   F[3]) biogenesis of (A) piRNA, (B) rsRNA, and (C) tsRNA in obese
   lineage.

   (D) The overlapped GO enrichment for DEmiRNA target genes in F[1]–F[3]
   transgenerational androgenized and obese and androgenized lineages.

   (E) The overlapped GO enrichment for DEmiRNA target genes in F[1]–F[3]
   transgenerational obese lineage.

   Representative pathways with log[10] (p value) >2 are presented. Size
   of the bubble represents the number of genes enriched in the pathway,
   and the gray color means the number of genes is larger than 200. Color
   represents -log[10] (p value). CD+Veh (F[1]: n = 4, litters = 4; F[2]:
   n = 4, litters = 4; F[3]: n = 4, litters = 4); CD+DHT (F[1]: n = 3,
   litters = 3; F[2]: n = 4, litters = 4; F[3]: n = 3, litters = 3);
   HFHS+Veh (F[1]: n = 4, litters = 4; F[2]: n = 4, litters = 4; F[3]: n =
   4, litters = 4); and HFHS+DHT (F[1]: n = 4, litters = 4; F[2]: n = 4,
   litters = 4; F[3]: n = 4, litters = 4).

   A family of X-linked miRNAs predominantly express in mammalian sperms
   and were named spermatogenesis-related miRNAs (spermiRs).[185]^55 We
   next investigated whether the spermiRs were transgenerationally
   dysregulated ([186]Data S5). In the androgenized and the obese
   lineages, respectively, five X-linked spermiRs (mmu-miR-465a-5p,
   mmu-miR-743b-3p, mmu-miR-470-5p, mmu-miR-871-3p, mmu-miR-741-3p) were
   up-regulated in F[2] male offspring, and four of these miRNAs (except
   mmu-miR-470-5p) were DE in the obese lineage in
   F[3] generation.[187]^55

   Thereafter we annotated the function of transgenerational DEmiRNA
   target genes in each lineage and each generation using pathway
   enrichment analyses ([188]Data S6A). When overlapping enriched
   biological pathways across the three generations, we found different
   pathways enriched in each respective lineage. In the androgenized
   lineage, enriched biological processes are implicated in morphogenesis
   of epithelium (e.g., Edn1, Ajuba, Cited2, Nrp1, Frs2 [mmu-miR-1a-3p]),
   and in the combined obese and androgenized lineage, the three pathways
   enriched are related to protein phosphorylation (e.g., Epha7
   [mmu-miR-130b-5p], Ptk2b, Rock2, Itch [mmu-miR-6240]) ([189]Figure 5D;
   [190]Data S6B). In the obese lineage, the biological processes are
   mainly enriched in utero embryonic development (e.g., Socs3, Sox6,
   Zbtb18 (mmu-miR-19a-3p), Fgfr1, Hif1a [mmu-miR-6240]); vasculature
   development (e.g., Bmpr2 [mmu-miR-467d-3p], Egr3 [mmu-miR-23b-3p], Lrp2
   [mmu-miR-142a-5p, mmu-miR-148a-3p, mmu-miR-152-3p, mmu-miR-199a-3p, and
   mmu-miR-199b-3p], Adipor2 [mmu-miR-19a-3p, mmu-miR-19b-3p, and
   mmu-miR-218-5p], Slc4a7, Esr1 [miR-19a-3p and mmu-miR-148b-3p], Tsc1
   [mmu-miR-130a-3p, miR-19a-3p, and miR-19b-3p]); and neuron-related
   pathways (e.g., Dlx1 [mmu-miR-19a-3p and mmu-miR-19b-3p], Id2
   [mmu-miR-19a-3p], Myo5b [mmu-miR-6240], Atxn1 [mmu-miR-101a-3p,
   mmu-miR-125a-5p, and mmu-miR-141-3p]) ([191]Figure 5E; [192]Data S6B).
   Collectively, these data show that DEsncRNAs carried by sperm are
   correlated with the transgenerational transmission of metabolic and
   reproductive phenotype in male offspring.

Comparison between DEsncRNAs blood of PCOS-sons and mouse sperm

   Next, we aligned the DEsncRNAs identified in sons of women with and
   without PCOS and each mouse lineage ([193]Figure 6A) and selected those
   with >90% sequence homology to investigate if the human DEsncRNAs
   overlap with the identified transgenerational DEsncRNAs in mice.
   Overall, the obese lineage transgenerational DEsncRNAs has the greatest
   number of overlaps with DEsncRNAs from sons of women with PCOS
   ([194]Figure 6B; [195]Data S7). The transgenerational DEtsRNA in the
   androgenized and the obese lineages, respectively, overlap with the
   human DEtsRNA specifically derived from the 5′ end of tRNA-Val-CAC
   ([196]Figure 6C; [197]Data S7). Finally, we also identified several
   shared biological processes of DEmiRNA target genes in serum of
   PCOS-sons and sperm of male F1 offspring in the different lineages such
   as male gonad development, primary male sexual characteristic
   development, embryo organ development in the androgenized lineage, etc.
   ([198]Figure 6D; [199]Data S8). Common biological processes of DEmiRNA
   target genes in serum of PCOS-sons and sperm of male F1 offspring in
   the different lineages are vasculature development, male gonad
   development, primary male sexual characteristics development, and
   embryo organ development in the androgenized lineage; in utero
   embryonic development, neuron development, cell differentiation, and
   cell catabolic process in the obese linage; and protein
   phosphorylation, nervous system development, cell activity, and
   vasculature development pathways in the combined linage.

Figure 6.

   [200]Figure 6
   [201]Open in a new tab

   sncRNA analysis in whole blood of sons from mothers with PCOS and
   comparison with sperm of mice

   (A) Schematic illustration of workflow showing how we identify human
   and mouse homologous DEsncRNA.

   (B) Number and biotype of homologous pairs of human and mouse
   transgenerational DEsncRNAs in each lineage.

   (C) Cloverleaf structure of tRNA-Val-CAC showing the region (blue)
   where the human and mouse common DEtsRNAs (as in B) are derived, with
   arrows pointing to the cutting sites.

   (D) GO enrichment for target genes of overlapped DEsncRNAs between
   mouse sperm (F[1]) and human blood.

   Representative pathways with log[10] (p value) >2 are presented. Size
   of the bubble represents the number of genes enriched in the pathway.
   Orange color represents up-regulated expression of DEsncRNA in human
   and mouse, while light yellow color represents down-regulation. CD+Veh
   (F[1]: n = 4, litters = 4; F[2]: n = 4, litters = 4; F[3]: n = 4,
   litters = 4); CD+DHT (F[1]: n = 3, litters = 3; F[2]: n = 4, litters =
   4; F[3]: n = 3, litters = 3); HFHS+Veh (F[1]: n = 4, litters = 4; F[2]:
   n = 4, litters = 4; F[3]: n = 4, litters = 4); and HFHS+DHT (F[1]: n =
   4, litters = 4; F[2]: n = 4, litters = 4; F[3]: n = 4, litters = 4).

   Taken together, these common DEsncRNAs found in PCOS-sons and offspring
   of androgenized and obese lineages suggest their roles in regulating
   PCOS-like phenotypic traits.

Discussion

   As to other complex diseases such as type 2 diabetes,[202]^56^,[203]^57
   PCOS is a highly heritable disorder.[204]^58 In addition to genetic
   factors, growing evidence from clinical and preclinical studies
   suggests that epigenetic regulation triggered by an adverse
   maternal-fetal environment could result in phenotypic transmission
   similar to conventional genetic effects as demonstrated by us and
   others.[205]^6^,[206]^18 More intriguingly, a recent study showed that
   men who carry high polygenic risk scores for PCOS develop an increased
   risk of obesity, type 2 diabetes, and cardiovascular disease, as well
   as male-pattern baldness, indicating that PCOS-sons could also be
   adversely affected.[207]^17

   To identify common features in PCOS-sons, we here explored a large
   Swedish register-based cohort study and a Chilean case-control study
   and found that PCOS-sons are more likely to be diagnosed with obesity
   and display an altered lipid profile with high circulating total
   cholesterol and LDL cholesterol. In the Chilean cohort, we did observe
   significantly increased body weight and lipid profile alterations
   likely reflecting insulin resistance that is translated into obesity
   during adult life as we have previously reported.[208]^8 These findings
   demonstrate that sons of obese women are more metabolically affected
   and highlight the importance of weight counseling and preferably weight
   reduction prior to pregnancy, especially in women with PCOS. Testicular
   volumes were comparable due to the initial study design for recruiting
   the pubertal group; nevertheless, previous work by our group showed
   that testicular volumes are lower at adulthood in PCOS-sons.[209]^9
   Moreover, our serum sncRNA analyses of sons from women with and without
   PCOS in the longitudinal Chilean clinical cohort study allowed us, for
   the first time, to identify potential miRNAs that underlie phenotypic
   transmission in humans by performing comparisons with published miRNA
   profiles from women with or without
   PCOS.[210]^38^,[211]^59^,[212]^60^,[213]^61^,[214]^62^,[215]^63^,[216]^
   64^,[217]^65^,[218]^66 Several overlapped miRNAs are enriched in
   pathways that could contribute to the metabolic and reproductive
   phenotypic features observed in PCOS-sons including insulin resistance
   (hsa-miR-122-5p, hsa-miR-4678), sex differentiation (hsa-miR-199a-5p,
   hsa-miR-532-3p, hsa-miR-548ar-3p), and response to hormones
   (hsa-miR-122-5p, hsa-miR-199a-5p, hsa-miR-199b-3p, hsa-miR-4685-3p). In
   addition, miRNAs observed in serum of PCOS-sons appear to regulate six
   PCOS-risk genes identified in GWASs. The AOPEP that encodes
   zinc-dependent metallopeptidase catalyzing the removal of an amino acid
   from a protein[219]^30^,[220]^33 is also linked to type 2
   diabetes,[221]^67 testosterone levels,[222]^3 and endometrial
   cancer[223]^68; TOX3 encodes a nuclear protein of the high-mobility
   group (HMG) box family and is associated with insulin resistance and
   metabolic syndrome in women with PCOS[224]^32^,[225]^45; ERBB4 is a
   member of epidermal growth factor receptors (EGFRs) and is a risk locus
   for PCOS, BMI and visceral adipose tissue, type 2 diabetes, and
   epithelial ovarian cancer, as well as sperm
   motility[226]^31^,[227]^32^,[228]^34^,[229]^35^,[230]^45; the SNP
   rs1159315 is located near γ-aminobutyric acid A receptor β1 (GABRB1)
   and linked to PCOS, obesity-related traits, and depression;[231]^46
   ADGRB3 is linked to BMI, triglyceride, adiponectin, and
   follicle-stimulating hormone levels in PCOS;[232]^33 MYRIP is related
   to insulin levels and PCOS[233]^36; and ADGRB3 is linked to BMI,
   triglyceride, adiponectin, and follicle-stimulating hormone levels in
   PCOS.[234]^33

   These findings implicate that PCOS-sons may carry circulating factors
   that underlie PCOS susceptibility for the development of phenotype.
   Together with our recent findings that daughters of women with PCOS are
   five times more likely to be diagnosed with PCOS[235]^6 and that
   prenatal androgen-exposed F[1] male offspring develop an aberrant
   reproductive and metabolic phenotype,[236]^19^,[237]^20 our current
   findings strengthen the hypothesis that maternal PCOS could induce
   fetal programming, predisposing not only daughters but also their sons
   to adult disease due to adverse maternal-fetal environment.

   Recently, we and others showed that PCOS-like traits induced by
   maternal androgen or AMH exposure in mice can be passed down to the
   third generation in female offspring,[238]^6^,[239]^18 but it remains
   unexplored whether such traits in F[1] male offspring could be
   transmitted across generations in their male progeny. In agreement with
   the clinical findings, our mouse models confirmed that maternal obesity
   or prenatal androgen exposure affects their F[1] male offspring,
   resulting in reproductive dysfunctions with increased anogenital
   distance, decreased testis weight, and aberrant mitochondrial
   morphology. Moreover, metabolic dysfunctions in F[1] male offspring are
   presented by increased fat mass and epididymal adipocytes and altered
   energy metabolism in both obesity and androgenized lineages, except for
   impaired glucose homeostasis, which is only in the obese lineage. Of
   note, the indirect calorimetry results showed lower metabolic activity
   (RER and EE) both in light and dark phases in the F[1] and F[3] males
   in the androgenized, obese, and obese-androgenized lineages,
   respectively. Importantly, the decrease in RER and EE is independent of
   food intake and total activity, suggesting that it is caused by
   increased body weight and fat mass. Thus, the lower metabolic
   activities (RER and EE) are likely consequences of in utero
   programming, either by androgen- and/or diet-induced obesity, resulting
   in inheritable metabolic pattern alterations such as different
   preferences of energy substrate.