Dicer1 deficient mice exhibit premature aging and metabolic perturbations in adipocytes.

Age-related diseases are major concern in developed countries. To avoid disabilities that accompany increased lifespan, pharmaceutical approaches are considered. Therefore, appropriate animal models are required for a better understanding of aging processes and potential in vivo assays to evaluate the impact of molecules that may delay the occurrence of age-related diseases. Few mouse models exhibiting pathological aging exist, but currently, none of them reproducibly mimics human diseases like osteoporosis, cognitive dysfunctions or sarcopenia that can be seen in some, but not all, elders. Here, we describe the premature aging phenotypes of Dicer-deficient mature animals, which exhibit an overall deterioration of many organs and tissues (skin, heart, and adipose tissue) ultimately leading to a significant reduction of their lifespan. Molecular characterization of transcriptional responses focused on the adipose tissue suggested that both canonical and non-canonical functions of DICER are involved in this process and highlight potential actionable pathways to revert it.

Introduction

Aging can be defined at the molecular, cellular or organismal levels but its most understandable and tangible definition is the inevitable, time-dependent “physiological deterioration”. Indeed, several changes that accumulate with time are quite obvious: hair loss, cognitive decline, reduced reproductive capacities, impaired muscle strength, to cite a few. Yet, these manifestations are highly variable and while “super-agers” (people in their 70s and 80s who have the mental or physical capability of their decades-younger counterparts) with preserved memory (or physical) performance abilities were described (Sun et al., 2016), others exhibit accelerated deteriorations, as that seen, in the most extreme cases, in progeria patients (Ahmed et al., 2018). This illustrates that our current knowledge of aging is still limited and providing a universal description of this process and its markers remains a challenging task (de Magalhaes and Passos, 2018; Johnson and Stolzing, 2019; Neves and Sousa-Victor, 2019). This points to the necessity to better describe the molecular and cellular processes that are at play during aging in vivo.

A crucial observation which sustains aging and the limitation of lifespan is the finite capacity of mammalian cells to divide (Hayflick and Moorhead, 1961), which led to the concept of “cellular senescence” characterized by permanent withdrawal from the cell cycle and acquisition of a pro-inflammatory, proteolytic secretome (Childs et al., 2017). This chronic low-grade inflammatory phenotype, or “inflammaging” (a term defining a biomarker, not interchangeable with “immunosenescence” (Pawelec et al., 2020)), is postulated to be an important contributor to tissue dysfunctions that accompanies age-related decline (van Deursen, 2014). Of note, a causal relationship between inflammation and aging has not yet been formally established. Several stressors, such as reactive oxygen species (ROS) or DNA damage are known inducers of senescence, but it has also recently been suggested that the accumulation of self-molecules (which are normally processed and eliminated by mechanisms such as autophagy) might be an important source of inflammation (Franceschi et al., 2017). In this regard, the role of DICER1 appears of particular importance. DICER1 is essentially known for its canonical role in the RNAi pathway through processing of pre-miRNA transcripts into mature miRNAs (Ha and Kim, 2014). However, additional (non-canonical) functions have been attributed to this RNAse III, among which the processing of long non-coding RNAs such as those derived from the transcription of Alu sequences (Song and Rossi, 2017). Although the role of miRNAs in aging has been amply investigated (Kinser and Pincus, 2020), the impact of Alu RNAs accumulation caused by DICER1 deficiency on inflammation and an age-related disease has been demonstrated in human retinal cells (Gelfand et al., 2015; Kaneko et al., 2011; Kim et al., 2014; Tarallo et al., 2012). Furthermore, recent data in mice (Aryal et al., 2019) (Reis et al., 2016) and in humans (Borras et al., 2017) point to an important contribution of Dicer and other component of the miRNA machinery (Proshkina et al., 2020) in aging. Importantly, mutant mice (Dicer1d/d) characterized by reduced Dicer1 expression are more prone to develop joint inflammation during experimental arthritis (Alsaleh et al., 2016) and have recently been shown to also exhibit age-related macular degeneration (Wright et al., 2020).

In this manuscript, we describe reduced lifespan and increased age-related diseases in Dicer1d/d mice. Next, we focused our investigation on the adipose tissue and explored in more details the mechanistic relationships between reduced Dicer1 expression, inflammation and premature aging. Altogether, our data point to a major role of Dicer1 in aging, probably through its canonical and non-canonical functions.

Results

Dicer1-deficient mice (Dicer1d/d) die prematurely and exhibit age-related diseases

Increased susceptibility to viral infections (Ostermann et al., 2015; Ostermann et al., 2012; Otsuka et al., 2007) and induced colon cancer (Yoshikawa et al., 2013) of Dicer1d/d mice have previously been described. However, in the absence of any obvious infectious or tumorigenic trigger (as assessed on careful examination of necropsied animals), we observed a reduced lifespan of Dicer1d/d mice (Figure 1A). Animals were kept in the same room of our animal facility and both genotypes (Dicer1+/+ and Dicer1d/d) were co-housed in the same cages to avoid microbiota-specific effects. In these conditions, we could calculate a median survival of 82 weeks for Dicer1 mutant mice, whereas that of their littermate controls exceeds 2 years. For practical reasons, we could not keep the control animals until their natural death, but the median survival of C57Bl/6J males mice is considered to be of 125 weeks (Kunstyr and Leuenberger, 1975). This means that the average life time of Dicer1d/d mice is 35% shorter than controls. Importantly, mutant mice develop more severe and/or more prematurely a morbid state and a set of age-related defects/pathologies. Indeed, aging of mutant mice is accompanied with the development of severe dermatitis (Figure 1B) and kyphosis (Figure 1C), both conditions frequently encountered in aged humans (Ailon et al., 2015; Hahnel et al., 2017) and associated with adverse health effects. Of note, these manifestations were not observed in mice until they were aged at least 40 weeks (Table S1). Additional signs of premature aging were also occasionally seen in Dicerd/d mutants following necropsy (Table S1): conjunctivitis (illustrated Figure 1D) which appears at 90 weeks in normative aging conditions (Coursey et al., 2017), splenomegaly (seen at 80 weeks in aging C57Bl/6 mice (Loukov et al., 2016)), cardiomegaly (starting in 100-week-old C57Bl/6 mice (Eisenberg et al., 2016)) or enlarged seminal vesicles (a feature of 70 week-old mice (Fritz et al., 2005)), although the occurrence of such events was too rare to reach statistical significance. Notably, other defects unrelated to aging, such as alopecia (caused by social behavior (Kalueff et al., 2006)) or rectal prolapses (resulting from intestinal dysbiosis (Buonocore et al., 2010)) were not modified in Dicer1d/d mice.

Figure 1

Dicer1 mutant mice exhibit visible phenotypes evocative of premature aging

(A) The survival of a cohort of 62 wild-type (Dicer1+/+) and 60 mutant (Dicer1d/d) mice was followed for 80 weeks.

(B) The occurrence of ulcerative dermatitis was quantified and expressed as a percentage (%) of mutant (60) and control (65) animals showing the phenotype.

(C) Kyphosis in Dicer1d/d mice (bottom) is shown by a red arrow.

(D) Conjunctivitis in Dicer1d/d mice (bottom). Control mice (Dicer1+/+) are shown in the top pictures.

(E) Weight (expressed in grams, g) was measured for 30 controls (Dicer1+/+) and 30 mutants (Dicer1d/d) for 100 weeks.

(F) Violin plot showing circulating (Plasma) Leptin (expressed in ng/ml) quantified by ELISA in 40 (N = 16) and 80 weeks-old (N = 13) control and Dicer1d/d mice.

(G) Violin plot showing the amount of adipose tissue (expressed as a percentage of the total body weight) quantified in 20-, 40- and 80-week-old control (Dicer1+/+) and mutant (Dicer1d/d) mice following necropsy. In panels A and B, data were analyzed with the Log-rank test. In panels E, F and G, Data were analyzed by comparing two datasets with a Mann-Whitney U test. ∗p<0.05; ∗∗p<0.01; ∗∗∗p<0.001; ∗∗∗∗p<0.0001.Data are represented as mean ±SEM. See also Figure S1 and Table S1.

We observed lower body weight of aged mutant mice, a phenotype that can be seen in 60-weeksold animals and beyond (Figure 1E). We also noted increased leptin production in the serum of aged wild-type mice, which is in agreement with their elevated body weight (Oswal and Yeo, 2010). Surprisingly however, leptin levels were even higher in 80-week-old Dicer 1 mutants (Figure 1F), despite reduced adipose tissue in these same mice Figure 1G). Indeed, although control mice gained weight and accumulated fat (which is the main source of leptin) with age, this was not the case in Dicer1d/d animals. Importantly, age-dependent weight loss is now considered as a key biological marker of aging and reflects the fragility and vulnerability to pathogens seen in elderly people as well as aged animals (Dutta and Sengupta, 2016). Of note, controls and Dicer1d/d mice reach a similar maximum weight (38 g), yet at different times, i.e., week 30 for mutants and week 70 for wild-type animals. The difference (40 weeks) is similar to the loss of life expectancy seen in Dicer1 mutants. This observation is in line with previous reports showing that the age at which the maximum body weight is reached, rather than the body weight itself, correlates with lifespan (Wagener et al., 2013).

Besides these visible signs, we also quantified several parameters, such as blood pressure and glucose tolerance. As seen in Figure S1, 20 weeks-old Dicer1d/d mutants exhibited a slight increase of blood pressure, which appeared significantly augmented in older mice (40- and 80-week-old). This suggests that Dicer1 deficiency might promote hypertension, a classical feature of aging in humans (Buford, 2016).

Altered glucose metabolism in Dicer1 mutants

To gain more insights into potential metabolic dysfunctions in Dicer1 mutants, we first performed glucose tolerance tests in cohorts of 20-, 40- and 80-weeksold mice. Figure 2 shows that, following intraperitoneal injection (performed in mice that were fasted for 5 h), glucose clearance is more efficient in Dicer1-deficient animals, regardless of their age, as determined by the measure of the area under the curve (AUC) (Panels A to C, i and ii). However, we also noted that glycemia is significantly reduced 15 min after glucose injection only in 80-week-old Dicer1 mutants compared to wild-types (Panel Ci), suggesting increased glucose tolerance in these animals. In addition, fasting glycemia is systematically reduced in Dicer1d/d mice (iii). Noticeably, ELISA quantification of plasma insulin showed increased levels in 80-week-old Dicer1 mutants (Figure 3A), in line with improved glucose clearance. Next, we realized an Insulin Tolerance Test (ITT) and observed a steady glycemia 30 min following insulin injection in fasting 40 weeks-old Dicer1-deficient mice (panels B and C), suggesting premature unresponsiveness. Indeed, 80-week-old animals of both genotypes also appeared unresponsive, likely reflecting an age-dependent effect, as seen by others (Ehrhardt et al., 2019). Inconsistencies between our observations and those of others reporting normal glucose metabolism in mice carrying another Dicer1 hypomorphic mutation (Morita et al., 2009) or hyperglycemia in pancreas-specific Dicer1 KO mice (Martinez-Sanchez et al., 2015), likely reflect differences in animal models and Dicer expression levels. Nevertheless our data illustrate an interesting paradoxical situation whereas insulin resistance should drive hyperglycemia and obesity, our Dicer1d/d mice exhibit the opposite phenotype.

Figure 2

Dicer1 mutant mice exhibit improved clearance in a glucose tolerance test

(A) Glucose tolerance assay performed in control (Dicer1+/+, N = 13) and mutant (Dicer1d/d, N = 13) 20 weeks-old mice. (1) After Glucose intraperitoneal (ip) injection, glycemia (expressed in mg/dL) was quantified during a 120 min (min) period. Wild-type and mutant mice were compared at each time point using an unpaired t-test. (2)The area under the curve (AUC, expressed in arbitrary units-AU) was calculated and differences between wild-types and mutants was analyzed with and unpaired t-test. (3) Quantification of the fasting glycemia (expressed in mg/dL) in the same animals and differences between wild-types and mutants was analyzed with and unpaired t-test.

(B) Similar experiments were performed in 40-week-old mice.

(C) Similar experiments were performed in 80-week-old mice. ∗p<0.05; ∗∗p<0.01; ∗∗∗p<0.001. Data are represented as mean ±SEM.

Figure 3

Insulin-resistance in Dicer1 mutant mice

(A) Violin plot showing the quantification of blood insulin (expressed in ng/ml) by ELISA in 40 and 80 weeks-old wild-type (Dicer1+/+, N = 7) and mutants (Dicer1d/d, N = 6 or N = 7). Statistical significance is evaluated with a Mann-Whitney U test.

(B) Insulin-tolerance test (ITT). Following insulin ip injection in fasting 20 -, 40 - and 80-week-old wild-type (Dicer1+/+, N = 4) and mutants (Dicer1d/d, N = 4), glucose (expressed in mg/dL) was quantified in the blood after 15, 30, 45, 60, 90 and 120 min (min).

(C) Violin plots showing glycemia values at 0 and 30 min after insulin injection in 20-, 40- and 80-week-old Dicer1+/+ (top) and Dicer1d/d (bottom) animals. Data were analyzed with a Mann-Whitney U test. ∗p<0.05.Data are represented as mean ±SEM.

Adipose tissue dysfunctions in Dicer1d/d mice

Defects in glucose homeostasis seen in aged Dicer mutant animals, in conjunction with reduced body weight (Figure 1), pointed to changes in the energetic balance toward consumption, rather than synthesis or storage of nutrients. SIRT1 is a key regulator of energy production through its involvement in insulin secretion as well as maturation and remodeling of the adipose tissue (Hui et al., 2017). Furthermore, Sirt1 reduced expression in muscle, liver, brain, and adipose tissue has been observed during aging (Gong et al., 2014). Of interest, we observed diminished Sirt1 expression in the epididymal white adipose tissue (eWAT) of Dicer1d/d mice (Figure 4B). Energy metabolism occurs in mitochondria in which glucose and lipid degradation enables ATP production, a process that can be influenced by Un-Coupling Proteins (UCP). In humans, increased UCP1 expression in visceral adipose tissue is associated with more efficient glucose uptake, reduced blood glycemia and triglyceride levels, combined to a diminished ratio of visceral/subcutaneous adipose tissue (Lim et al., 2020; Tews et al., 2019). Of note, 20 weeks-old Dicer1mutant mice exhibit increased Ucp1 expression in eWAT (Figure 4C). Such changes in SIRT1 and UCP1 gene expression in the visceral adipose tissue have been observed in obese patients and might induce metabolic dysfunctions in order to adapt to excessive nutrients availability (Lim et al., 2020). However, a chronic hyper activation of the energy metabolism also induces pathogenic situations, such as cachexia on increased expression of Ucp1 (Petruzzelli et al., 2014). In addition, activation of the NLRP3 inflammasome drives reduced Sirt1 expression (Chalkiadaki and Guarente, 2012) and Ucp1 increased expression is the consequence of ROS production (Han et al., 2016). Of interest, accumulation of SINEs transcripts (B1 and B2 RNAs in mice) resulting from reduced DICER expression has been shown to promote NLRP3-dependent inflammation in retinal cells, thereby inducing age-related macular degeneration (Kaneko et al., 2011; Kim et al., 2014; Tarallo et al., 2012). As seen in Figures 4D and 4E, B1 and B2 RNAs accumulate in eWAT, in young as well as in aged Dicer1 mutants compared to controls. This important observation indicates that this phenomenon, not restricted to the retinal pigment epithelial cells, could potentially represent a more conserved mechanism linking reduced Dicer1 expression, increased inflammation and aging.

Figure 4

Dicer1-specific changes in adipose tissue gene expression

(A–E) Dicer1, B. Sirt1, C. Ucp1, D. B1 RNA and E. B2 RNA transcripts were quantified by RT-qPCR in the epididymal white adipose tissue of wild-type (Dicer1+/+) and mutants (Dicer1d/d). Data, represented in violin plots, were analyzed using a Mann-Whitney U test. ∗p<0.05; ∗∗p<0.01; ∗∗∗p<0.001. Data are represented as mean ±SEM.

Because Dicer1mutants likely exhibit altered metabolic responses in the white adipose tissue, we decided to analyze their behavior in response to high fat diet (HFD) challenge. In these experiments, 10-week-old wild-type and Dicer1-deficient mice were subjected to an obesogenic diet for 9 weeks. In these conditions, Dicer1-deficient mice exhibited a surprising resistance to weight gain and adipose tissue expansion (Figures 5A–5C). Unexpectedly, Dicer1-deficient animals ingested more food than littermate controls and had and increased caloric intake (Figure 5D). In addition, histological examination of the epididymal white adipose tissue shows that adipocytes from Dicer1-deficient animals are significantly smaller compared to controls (Figures 5E and 5F).

Figure 5

Dicer1 mutants are resistant to high fat diet

(A and B) Absolute weight (expressed in grams, g) or B. expressed as a percentage of the initial weight was measured in 10-week-old wild-type (Dicer1+/+) and mutants (Dicer1d/d) mice fed with normal (chow) or high fat (HFD) diet for 9 weeks.

(C)The amount of white adipose tissue (expressed as a percentage of the body mass) was measured in selected wild-type (Dicer1+/+) and mutants (Dicer1d/d) mice at the end of the experiment.

(D)The caloric intake (expressed in kcal/week/mouse) was measured in the course of the experiment for wild-type (Dicer1+/+) and mutants (Dicer1d/d) animals.

(E) Histological examination of epididymal white adipose tissue harvested from a wild-type (Dicer1+/+) and mutant (Dicer1d/d) mouse 9 weeks after high fat diet (HFD) and following hematoxilin/eosin staining.

(F) Quantification of the size (expressed in μm) of the adipocytes from control (Dicer1+/+) and mutant (Dicer1d/d) adipocytes on HFD. Data, represented as violin plots, were analyzed using a Mann-Whitney U test. Curves in panels A and B were compared following determination of the area under the curve. ∗p<0.05; ∗∗p<0.01; ∗∗∗p<0.001. Data are represented as mean ±SEM.

Such chronic exposure to an excess of nutrients induced an increased fasting glycemia and blood insulin concentration in wild-type mice (Figures 6A and 6B), whereas Dicer1d/d animals appeared resistant, maintaining a reduced basal glycemia and exhibiting a limited rise in plasma insulin levels. Glucose tolerance tests also showed that Dicer1d/d mice exhibited a markedly improved glucose clearance even 8 weeks post HFD (Figures 6C and6D).

Figure 6

Improved glucose tolerance in Dicer1 mutants after HFD

(A) Fasting glycemia in the blood (expressed in mg/dL) of wild-type (Dicer1+/+) and mutants (Dicer1d/d) animals following normal (chow) or high fat diet (HFD).

(B) Blood insulin (expressed in ng/ml) was quantified by ELISA in the same animals as in A.

(C) Glucose tolerance test before high fat diet regimen. Blood glucose (expressed in mg/dL) following glucose ip injection was quantified for 120 min (min) in wild-type (Dicer1+/+, in gray) and mutant (Dicer1d/d, in red) 10 weeks-old animals that will be fed with normal (chow, dotted line) or high fat diet (HFD, plain line). The right panel shows the area under the curve (AUC, expressed in arbitrary units-AU).

(D) Same experiment as in C was performed in mice following 9 weeks of normal (chow, dotted line) or high fat diet (HFD, plain line). Data, represented as violin plots, were analyzed with a Mann Whitney U test. ∗p<0.05; ∗∗p<0.01.Data are represented as mean ±SEM.

Because HFD is known to induce inflammatory stress in adipocytes, we quantified by RT-qPCR the expression levels of prominent inflammatory markers. As seen in Figure 7A, Il-1β, Il-6, Il-18 and Tnf-α gene expression is markedly increased in the adipose tissue harvested from Dicer1d/d mice subjected to HFD. Importantly, the expression of B1 and B2 RNAs also appeared increased in these samples (Figure 7B). Yet, targeted gene expression analysis did not provide answers regarding the reduced amount of adipose tissue in Dicer mutants. Indeed, expression of β-oxydation genes (cdvl, Acox1, Hadha, Hadhb), of mitochondrial respiratory chain (Idh3a, Ogdh, Sdhd, Uqcrq), of intestinal barrier integrity (Zo1, Ocl, Muc2) remained comparable in mutants and wt tissues. Furthermore, no liver steatosis could be observed in necropsied animals.

Figure 7

Increased inflammatory signature in Dicer1 mutant adipocytes on high fat diet

(A) Relative gene expression of inflammatory genes (Il-1β, Il-6, Il-18 and Tnfα) quantified by RT-qPCR in the epididymal adipose tissue harvested from wild-type (Dicer1+/+) and mutants (Dicer1d/d) animals following normal (chow) or high fat diet (HFD).

(B) Relative quantification of Dicer1, B1 and B2 transcripts in the same conditions as in A.

(C) Relative quantification of Fgf21 and Gdf15 gene expression in the quadriceps muscle of 20 weeks-old wild-type (Dicer1+/+) and mutants (Dicer1d/d) animals, on normal (chow) or high fat diet (HFD). Data, represented in violin plots, were analyzed with a Mann-Whitney U test. ∗p<0.05; ∗∗p<0.01.Data are represented as mean ±SEM See also Figures S2–S7.

The myokines FGF21 and GDF15 were recently reported as endocrine, stress induced factors capable of establishing a functional crosstalk with the adipose tissue, respectively inducing energy expenditure and reducing food intake and body weight (Keipert and Ost, 2021). Interestingly, we observed increased Fgf21 and Gdf15 expression in muscle tissue harvested from 20-week-old Dicer1 mutant mice compared to controls under normal (chow) diet (Figure 7C). Of note, high fat diet also promoted increased expression of Gdf15 in Dicer1d/d mice. These data, although preliminary, provide additional insights in favor of the involvement of these factors in the metabolic disturbances seen in Dicer1-deficient animals.

Finally, we used an RNAseq approach to obtain a more comprehensive view of the changes in gene expression that occur in adipocytes during aging and evaluate the concomitant impact of reduced Dicer1 expression. A first hint into global transcriptomic changes is illustrated by the Volcano plots in Figure S2. Panel A shows that aging (in that case between 20 and 80 weeks) of adipocytes in control (wild-type) mice is accompanied by a statistically significant (P adj<0.01) up regulation (2-fold) of 31 genes and down regulation (2-fold) of 52 genes. In contrast, 792 genes are upregulated and 151 down regulated in 80-week-old, compared to 20-week-old Dicer1d/d mice (Panel B). Of interest, only 9 genes are differentially expressed between 20-week-old Dicer1 mutants compared to controls, clearly showing that the mutation has almost no impact in this tissue in young animals (Panel C). This is in sharp contrast to the comparison between 80 weeks-old Dicer1d/d and controls (Panel D), which reveals that the expression of 747 genes is increased, whilst that of 99 is reduced. To stratify some of the relevant genes whichexhibit differential gene expression, we used heatmap representations and performed Ingenuity Pathways Analysis (IPA, Qiagen) and Gene Set Enrichment Analysis (GSEA, http://www.webgestalt.org/). Using these tools, we first managed to highlight several genes involved in aging of wild-type adipocytes (Figure S3), among which HNF4A (Hepatocyte Nuclear Factor 4 Alpha) and Pyruvate Kinase L/R (PKLR) that participate in the Maturity Onset Diabetes of Young (MODY) Signaling and are down-regulated in aged mice. Of note, reduced Hnf4a expression has been noted in pancreatic islet of aged rats (Sandovici et al., 2011). Because of the limited amount of differentially expressed genes, GSEA did not allow to identify specific enrichment in gene sets, which likely indicates that 80-week-old animals does probably not really correspond to an advanced age in these wild-type mice. The second dataset that we analyzed compared aged to young Dicer1d/d mice (Figure S4). IPA revealed, amongst the top affected pathways, the apelin signaling pathway (p = 1.44 × 10−4). Given the role of the adipokin apelin and its receptor in age-related diseases (Luo et al., 2020; Vinel et al., 2018), reduced expression of both genes in aged Dicer1-deficient adipocytes appears particularly relevant and interesting. In addition, GSEA pointed to significant (FDR<0.05) negative Normalized Enrichment Scores (NES) for gene sets involved in biological processes such as ossification, tissue migration or response to wounding which reflects an ongoing aging process. For instance, the gene set involved in ossification (NES= − 2.31; FDR= 0.0019949) includes leptin (Lep), Secreted acidic cysteine rich glycoprotein (Sparc), Noggin (Nog), Collagen, type I, alpha 1 (Col1a1), Refilin B (Rflnb), Transformation related protein 53 inducible nuclear protein 2 (Trp53inp2), Asporin (Asp) and cAMP responsive element binding protein 3-like 1 (Creb3l1). We also noted the down regulation of several genes encoding extracellular matrix components (Collagens, Tenascin-C), which is also a hallmark of aging (Birch, 2018). Altogether, these data point to premature aging phenotype in these mutant mice, as a result of reduced expression of many genes involved in tissue renewal and homeostasis. A first comparison applied to adipocytes harvested from young animals exhibited only minor changes, mostly involved in interferon signaling, as seen in Figure S5. On the contrary, when we compared the transcriptome of aged mutant adipocytes to that of controls (Figure S6), IPA identified overexpressed genes such as aldolase or glyceraldehyde-3P dehydrogenase that highlighted deregulation of metabolic pathways (panel B) such as gluconeogenesis (p = 3.7 × 10−5) or glycolysis (p = 4.2 × 10−4) in mutant adipocytes. Furthermore, diseases (dermatological, injury) which were identified by IPA (panel C) in this dataset are evocative of ageing-associated processes. GSEA also identified negative NES for genes involved in immune functions (phagocytosis, cytokine production, panel D) or tissue homeostasis, correlating with negative NES indicating abnormal skeleton physiology or immune responses (panel E). Neither in these datasets, nor by RTqPCR could we detect Dicer1-specific transcriptional changes in beta-oxydation (Acdvl, Acox1, Hadha, Hadhb, not shown) genes expression. Along with the absence of steatotic liver in 80 weeks-old Dicer d/d mice, the metabolic reason for the “lean” phenotype seen in mutants remains unanswered and will require additional investigations.

In line with these observations, analysis of our RNAseq data with ImmQuant (Frishberg et al., 2016) revealed that in contrast to wild-type aged mice in which the adipose tissue exhibits increased NKT cells and eosinophils and a reduced number of immune cells progenitors and pDCs, aged Dicer d/d mutants are characterized by more Pro and Early B-cells as well as CD8+ effector memory cells infiltrating this tissue (Figure S7), likely contributing to the senescence-associated secretory phenotype (Frasca et al., 2021).

Discussion

Dicer1d/d mice, a relevant model for aging studies

Several animal models have been developed to investigate aging at cellular, molecular and system levels, among which mice exhibit many advantages due to physiological and genetic similarities with humans. However, genetically engineered mice with modified (reduced or increased) lifespan either mimic extreme phenotypes (such as progeroïd syndromes which are extremely rare in humans) or affect targeted functions (such as inflammation, metabolism or DNA repair) or specific organs (adipose tissue, liver) (Koks et al., 2016). Dicer1-deficient mice appear as an attractive alternative for several reasons: (1) Mutant mice exhibit many defects affecting many organs (retina (Wright et al., 2020); joints (Alsaleh et al., 2016); ovaries (Otsuka et al., 2008); skin and adipose tissue (this study) and, as demonstrated here, are also characterized by reduced lifespan (Figure 1). Of course, more investigations will be necessary to explore in more details phenotypes (like kyphosis) observed in aged Dicer1d/d mice. These future studies will enable a better understanding of age-related pathologies (osteoarthritis) of major importance in humans. (2) Dicer1d/d mice are hypomorphic, which means that the animals exhibit only a reduced Dicer1 expression. Indeed, gene deletions are uncommon in humans, in which diversity is more caused by genetic variations and subsequent changes in gene expression and/or activity. In addition, the mutation, which affects pre-mRNA splicing (Otsuka et al., 2007), induces inter-individual heterogeneity, as demonstrated by variable Dicer transcript levels between mutant mice, and with one animal, between organs; whereas such variations make the analysis of Dicer1d/d mice complex (requiring large cohorts), they mimic the wide spectrum of age-related phenotypes in humans in which aging is not an evolution with stereotypic features. (3) The dynamic of appearance of the phenotypes seen in Dicer1d/d mice is evocative of an accelerated aging. Indeed, we could calculate that 50% of Dicer1-deficient mice can reach an age (80 weeks) which is equivalent to 62 years for humans (Dutta and Sengupta, 2016), while 50% of the controls can reach 120 weeks (92 years for humans). This likely represents the physiological evolution and diversity seen in human populations. Our findings are in line with observations made using another knock-in mouse model in which DICER1 activity was reduced by constitutive phosphorylation (Aryal et al., 2019). Of note, accelerated aging, kyphosis and reduced abdominal fat were also observed in this model. Using embryonic fibroblasts, the authors reported increased metabolic rates, possibly explaining fat reduction. Finally, another hypomorphic Dicer1-deficient mice has been generated (Morita et al., 2009) and exhibited histological abnormalities in the pancreas, leaving all the other organs intact. These differences likely illustrate the phenotypic variability inherent to hypomorphic mutations, a feature that has been already discussed in the case of Dicer d/d mice (Ostermann et al., 2012; Otsuka et al., 2007). On the other hand, to circumvent the embryonic lethality caused by Dicer deficiency, conditional, tissue-specific Dicer KO mice have been generated using the CRE/lox technology. Initial observations were made in isolated murine pre-adipocytes in which Dicer ablation was obtained on transduction of a CRE-expressing adenovirus (Mori et al., 2012). In these cells, increased sensitivity to cells stressors and gene expression analysis suggested that Dicer ablation promoted senescence. Conditional Dicer KO mice were also generated and exhibited reduced survival rate following oxidative stress induction by paraquat injection. Of interest, identical conditional Dicer KO mice (using the aP2-Cre driver) were analyzed by two different groups (Mori et al., 2014; Mudhasani et al., 2011), who observed severe post-natal lethality, a feature that was attributed to off-target effects and Cre expression in non-adipose tissues. In contrast, using the adiponectin promoter to drive Cre expression (Mori et al., 2014), mutant mice exhibited lipodystrophy, loss of white fat and insulin resistance, phenotypes that are also seen in our Dicer d/d mice. Finally, we would like to stress that our observations in mice are in line with human data showing that centenarians, exhibiting delayed aging, are characterized by miRNA and Dicer overexpression compared to octogenarians with ordinary aging features (Borras et al., 2017). Altogether, these data point to an important role of Dicer in the adipose tissue and for a healthy organism.

Nevertheless, the mechanism by whichDicer1-reduced expression triggers premature aging remains to be fully elucidated. A first clue is the overexpression of B1 and B2 RNAs detected in adipocytes of Dicer1d/d mice fed with a lipid-rich diet (Figure 4). This observation, in line with those performed in retinal epithelial pigment cells (Kaneko et al., 2011; Kim et al., 2014; Tarallo et al., 2012), suggests that non-canonical functions of DICER1 (Alu sequences processing) could participate in the inflammatory setting of adipocytes and mediate the inflammageing phenotype. Alternatively, reduced Dicer1 expression also entails an impairment of nuclear DICER1 functions, such as the maintenance of telomeres or the response to DNA damages (Burger and Gullerova, 2015), whose dysfunction might also drive cellular senescence. Finally, the canonical role of DICER1 (i.e, the maturation of miRNAs) is certainly an important player in the accelerated aging seen in Dicer1d/d mice. This is visualized by the large number of deregulated genes in adipocytes harvested from aged mutant mice compared to controls. Indeed, many of these genes might be under transcriptional control by miRNAs, as demonstrated by others (Mori et al., 2014).

Transcriptomic analyses of adipocytes highlight multiple pathways involved in aging

Adipocytes have long been under-estimated and explored for their energy storage functions only. Yet, the adipose tissue represents an important proportion (20%, depending on age, sex) of the total body mass in humans (Jackson et al., 2012) (30% in mice, depending on strains (Reed et al., 2007)) and is able to secrete many signaling molecules (adipokines) involved in the regulation of multiple processes like wounding (Lee et al., 2018), bone density (Reid et al., 2018) or immune responses (Francisco et al., 2018). Furthermore, the relationships between lipid metabolism and age-related disease appear more and more functionally related (Arai et al., 2019). To better characterize the changes that occur in these cells during aging and the impact of reduced Dicer1 expression in this process at the molecular level, we performed a transcriptomic analyses by RNAseq. First, we focused on “normal” aging by comparing gene expression levels between 80 weeks-old and 20 weeks-old wild-type mice (Figure S3). Setting the cutoff for significant changes to a p adj<0.01 and a minimal fold change of 2, we found few (83) deregulated genes, which likely indicates that 80 weeks does not correspond to old animals, but rather to “mature adults” which are not so different, at least with regards to the adipose tissue, from 20-week-old, “young adults”. Nevertheless, we observed that the top pathways which were predicted to be deregulated by IPA play a role in glucose metabolism. Importantly, the MODY pathway is among them, in line with premature diabetes and insulin-resistance, two importance features of aging. We also observed increased expression of several immunoglobulin-encoding genes, possibly reflecting the accumulation of B cells in the adipose tissue during aging, as previously reported (Camell et al., 2019). Of interest, GSEA highlighted the reduced expression of many genes encoding components of the extracellular matrix (Tenascin-C, various Collagens, Asporin) in Dicer1d/d mice. Such alterations drive probably the most visible effect of aging in humans, the loss of skin elasticity and the appearance of wrinkles. Surprisingly, the comparison of 20 weeks-old Dicer1 mutants and age-matched controls (Figure S4) revealed almost no difference (8 differentially expressed genes), except an overexpression of Irf7 and Ifit1 genes, evocative of an interferon signature that we described earlier (Ostermann et al., 2012). Transcriptomic differences were much more marked when we compared 80 and 20 weeks-old Dicer1d/d adipocytes (Figure S5) and highlighted several deregulated genes (Nrep, Fabp3, Col1a1, H1f2, Lyz1, Col3a1, B2m, S100a4) common to human, rat and mouse which are listed in the Human Aging Genomic Resources database (https://genomics.senescence.info/) (de Magalhaes et al., 2009). Several deregulated metabolic pathways were identified by IPA (i.e., neoglucogenesis), as well as the Apelin signaling cascade (shown panel B). Indeed, reduced expression of Apelin and its receptor have been recently described as important events promoting senescence (Zhou et al., 2018). GSEA identified significant (FDR<0.05) negative enrichment scores for genes participating in processes such as ossification or response to wounding (panel C), providing additional evidence of accelerated aging in Dicer1 mutants. Finally, we analyzed the impact of Dicer1-reduced expression in aged animals and compared 80-week-old mutants to age-matched controls (Figure S6). First, we noted that most genes were up-regulated (Figure S2D), reflecting the reduced expression of miRs (which exhibit essentially negative regulatory roles) as a result of low DICER1 levels. Next, IPA confirmed the deregulation of metabolic pathways (gluconeogenesis, glycolysis) that occurs in Dicer1-deficient aged animals and GSEA confirmed the negative enrichment of genes participating in extracellular matrix and phenotypes (immune, bone-related), reflecting an overall degradation of major physiological functions and premature aging. Interestingly, we also showed that Dicer deficiency, by reducing Fgf21 and Gdf15 expression in muscles (Figure 7C), impacts endocrine signaling emanating from this tissue, thereby participating in altered lipid metabolism and feeding behavior seen in mutant animals. Of course, more work will be necessary to uncover the changes in muscle/adipocytes connections that results from Dicer1-reduced expression.

Delaying aging and the occurrence of age-related diseases

The proportion of elderly is constantly increasing in developed countries. This aging population is at risk to develop chronic diseases, such as cardiovascular, rheumatic and neurodegenerative conditions, thereby creating a massive economic burden to the social security systems. Hence, the development of drugs enabling enhanced lifespan, in the absence of age-related diseases is a major challenge and the focus of intense research. To this end, rejuvenation strategies are considered and can be divided into two main categories: (1) Those that are based on dietary restriction and (2) the ablation of senescent cells by chemotherapy (senolytics) (Mahmoudi et al., 2019). In this work, we showed that high fat diet (HFD) promoted enhanced inflammatory genes expression in Dicer1mutant mice (Figure 7). Concomitantly, we observed that Dicer1d/d mice are also resistant to adipogenesis and weight gain in these conditions (Figure 5) and are still able to control their glycemia better than wild-type animals (Figure 6). These contrasting observations will require additional investigations, like a quantification of lipid absorption in the intestinal lumen. However, they can be connected to recent observations in humans suggesting that a reduced BMI is associated with a greater risk of mortality in elderly (Winter et al., 2014). In this work, neither did we evaluate the impact of HDF on mortality in Dicer1 mutants, nor the consequences of diet restrictions. These experiments will have to be performed in the future.

An important message from our transcriptomic analyses is the potential impact of adipokines on the accelerated aging seen in Dicer1d/d mice. Indeed, in addition to apelin and its receptor, the genes encoding leptin, resistin, chemerin are also down regulated in aged mutant animals. These observations suggest that, in addition to senolytics (Xu et al., 2018) or anti-inflammatory molecules (Neves and Sousa-Victor, 2020), the modulation of physiological processes such as food intake or energy expenditure by adipokines (which also possess some neuroprotective activities (Arnoldussen et al., 2014) might become an attractive pharmacological tool to enhance lifespan in healthy conditions.

In conclusion, our work enabled us to provide a comprehensive description of the functions of Dicer1 in adipocytes and during aging. Although still imperfectly characterized, Dicer1d/d mice might represent a physiologically-relevant model to thoroughly investigate aging in vivo and evaluate the efficacy of molecules which will be used in humans to postpone the occurrence of age-related diseases.

Limitations of the study

This study is essentially observational and premature aging seen in Dicer1-deficient mutants needs to be sustained by biochemical, histological (e.g., SA-β-Gal staining) and molecular (telomere length) investigations to consider senescence as a mechanism accounting for the observed phenotypes. Furthermore, metabolic perturbations suggested by our transcriptomic analysis of adipocytes should be documented with measures of energy expenditure to better apprehend the subsequent mechanisms driven by reduced Dicer1 expression. Finally, quantification of microRNAs expression, with our transcriptomic analysis, should enable the identification of miRNA-regulated mRNAs in adipocytes.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Philippe Georgel (pgeorgel@unistra.fr).

Materials availability

This study did not generate new unique reagents.

Experimental model and subject details

Dicer1-deficient mouse line

Dicer1-deficient strain was established by other (Otsuka et al., 2007). Briefly, the model was generated by using a gene-trap method with a β-galactosidase-neomycin (β-geo) insertion in intron 24 of the Dicer1 locus, abolishing Dicer1’s activity. However, a low-frequency alternative splicing event results in a wild-type-processed transcript in Dicer1d/d mice, leading to a hypomorphic phenotype (low DICER1 protein expression). The Dicer1-deficient strain, backcrossed to C56Bl/6Jcrl (Charles River Laboratories) was maintained in a heterozygous state, and homozygous wild-type and mutant littermates were utilized for the experiments.

Housing and husbandry

All animals were kept under specific pathogen-free conditions in the animal facility of the Center deRecherche d’Hématologie et d’Immunologie. Only males were used in the experiments described in this paper. Unless stated otherwise, mice were housed in groups (2–7 animals) containing both genotypes to minimize potential microbiota-specific effect and were kept under stabilized temperature (20–22°C) with a 12h dark-light cycle and fed water/food ad libitum. Cages were enriched with cotton tubes for nests and red polycarbonate tunnels. Handling and experiments were done by a single experimenter, aware of mice genotype at all time. Our experiments were approved by the local ethic committee (CREMEAS, protocol reference APAFIS#21782-2019061317371610 v4).

Method details

High fat diet (HFD)

Ten-week-old males were housed in individually cages and provided with standard food pellets (SAFE, A03) or a high-fat diet (SAFE, 292HF, U8954 60% calories from lipids) and tap water ad libitum throughout the study (10 weeks). Animals and food were weight weekly to follow body weight and food consumption. Blood collection on the submandibular vein was performed on conscious animals at week 1 and before sacrifice for Insulin quantification (as described below). An intraperitoneal glucose tolerance test (as described below) was performed at week 1 and week 9.

Glucose/insulin tolerance test

The glucose tolerance of animals was assessed by an intraperitoneal glucose tolerance test (ipGTT) at designated times. Mice were transferred into new clean cage and fasted for 5 h, typically from 8 a.m. to 1 p.m. The first blood samples were collected from tail tip and blood glucose was measured using a glucometer (CareSens), which was considered as time 0 min. Subsequently, a 20% glucose solution was administrated intraperitoneally into mice (2 g/kg of body weight) and blood glucose was monitored at 15, 30, 60, 90 and 120 min after glucose administration.

The insulin tolerance was assessed by an intraperitoneal insulin tolerance test (ipITT). After 5 h of fasting, blood samples were collected from the mouse tail tip and blood glucose was measured. Subsequently, mice were intraperitoneally injected with 0.75 U/kg body weight of recombinant human insulin (Sigma-Aldrich). Blood glucose was monitored at 15, 30, 45, 60, 90 and 120 min after insulin administration.

Insulin and leptin quantification

At designed time/age, blood was collected on submandibular vein on conscious animals. Samples were centrifuged at 5000× g for 10 min to separate plasma and stored at −80°C for analysis. Leptin and Insulin concentration were estimated by ELISA following the manufactures’ guidelines (Mouse/Rat Leptin Quantikine ELISA Kit, R&D Systems; Rat/Mouse Insulin ELISA kit, Merck Millipore).

Adipose tissue quantification

At described time or age, mice were euthanized by cervical dislocation. Fat depots were immediately collected and weighted to compare visceral (perirenal, epididymal, retroperitoneal) and subcutaneous (dorsolumbar, inguinal) depots. Epididymal white adipose tissue was then processed for freezing in dry ice (protein and mRNA analysis) or for embedding in paraffin (Leica, 39,601,006) after 4% formalin fixation.

Histological analysis

Embedded tissue was then sliced to 8-μm thickness sections on glass slides, and subjected to Hematoxilin and Eosin staining following manufacturer’s instructions (H&E staining kit, Abam ab245880). Sections were examined using a Zeiss Axio Imager A1; pictures were taken with an Axiocam IC3 and further analyzed using ImageJ software.

RT-qPCR

Total RNA was prepared as described (Nehmar et al., 2017). Briefly, adipose tissue was minced and then grinded into TRIzol reagent for RNA extraction following manufacturer’s instructions (Invitrogen N015596026). Total RNA was then reverse transcribed using the cDNA synthesis kit (Biorad). Real-time quantitative RT-qPCR was performed in a total volume of 10μL using the Sso-advanced universal SYBR-Green supermix (Biorad) and gene-specific primers (list available upon request). After a denaturing step at 95°C for 30 s, 40 cycles were performed (95°C for 5s and 60°C for 20s) using a Rotor-Gene 6000 real-time PCR machine (Corbett Life Science). Results were obtained using the SDS Software (Perkin Elmer). Melting-curve analysis was performed to assess the specificity of PCR products. Relative expression was calculated using the comparative threshold cycle (Ct) method. The Ct of the gene of interest was adjusted to the average Ct of three housekeeping genes (18S, Actin and Gapdh) to obtain a Δct.

RNA-sequencing

Total RNA integrity was determined with the Agilent total RNA Pico Kit on a 2100 Bioanalyzer instrument (Agilent Technologies, Paolo Alto, USA). Library construction was performed with the SMARTer Stranded Total RNA-Seq Kit v2 - Pico Input Mammalian" (TaKaRa Bio USA, Inc., Mountain View, CA, USA) with a final multiplexing of 11 libraries according to the manufacturer’s instructions. The library pool was sequenced on a NextSeq 500 (Illumina Inc., San Diego, CA, USA) following the manufacturers protocol.

Analysis of RNA-seq data

The transcriptome dataset, composed of sequencing reads, was generated by an Illumina NextSeq instrument. For every sample, quality control was carried out and assessed with the NGS Core Tools FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Sequence reads were mapped using STAR (Dobin et al., 2013) and unmapped reads were remapped with Bowtie2 (Langmead and Salzberg, 2012) using the very sensitive local option. The total mapped reads were finally available in a BAM (Binary Alignment Map) format for raw read counts extraction. Read counts were generated by the htseq-count tool of the Python package HTSeq (Anders et al., 2015) with default parameters to generate an abundance matrix. At last, differential analyses were performed by the DESEQ2 (Love et al., 2014) package of the Bioconductor framework. Up- and down-regulated genes were selected based on the adjusted p value and the fold-change information. Gene expression was analyzed using dedicated R scripts to build volcano plots and. Heatmaps were built using the online application heatmapper (http://www.heatmapper.ca/). IPA (Ingenuity pathway analysis, Qiagen) was used for pathway analysis. Predictions of immune cell population from RNAseq data were made with the ImmQuant deconvolution software (Frishberg et al., 2016).

Quantification and statistical analysis

Following normality tests (Kolmogorov-Smirnov and Shapiro-Wilk), data were analyzed with a Mann-Whitney U-test or an unpaired t test (two-tailed unpaired) to compare two independent groups and non-parametric (Spearman) test for correlation analysis. Statistics were calculated with GraphPad 9.3.1 software. A probability (p) value of <0.05 was considered significant. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Critical commercial assays
Custom diet for Rat & Mice, 61% of the energy from fat (lard)	SAFE custom Diets	SAFE® U8954 version 205
Insulin solution human	Sigma-Aldrich	CAS 11061-68-0
Mouse/Rat Leptin Quantikine ELISA Kit	R&D systems	CAT#MOB00
Rat/Mouse Insulin ELISA	Merck	EZRMI-13K
Hematoxylin and Eosin Staining Kit	Abcam	Ab245880
TRIzol® Reagent	Invitrogen	CAT#15596026
IScript™ cDNA Qynthesis Kit	BIO-RAD	1708891
SsoAdvanced Universal SYBR® Green Supermix	BIO-RAD	1725274
Agilent RNA 6000 Pico kit	Agilent	5067-1513
SMARTer Stranded total RNA-seq ki v2 – Pico input mammalian	Takara	634417
Experimental models: Organisms/strains
Dicer1gt(β-geo)han mouse strain	Otsuka et al., 2007	N/A
Software and algorithms
FastQC	Babraham Institute	RRID:SCR-014583http://www.bioinformatics.babraham.ac.uk/projects/fastqc/
Image J		RRID:SCR_003070https://imagej.net/
STAR	Dobin et al., 2013	RRID:SCR_004463http://code.google.com/p/rna-star/
Bowtie2	Langmead and Salzberg, 2012	RRID:SCR_016368http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
HTseq	Anders et al., 2015	RRID:SCR_005514http://htseq.readthedocs.io/en/release_0.9.1/
DESeq2	Love et al., 2014	RRID:SCR_015687https://bioconductor.org/packages/release/bioc/html/DESeq2.html
Ingenuity Pathway Analysis IPA	Ingenuity	RRID:SCR_008653
ImmQuant	Frishberg et al., 2016	http://csgi.tau.ac.il/ImmQuant/
Other
Databases to study the genetics of ageing	Human Ageing Genomic Resources	https://genomics.senescence.info/