Depletion of the TRF1 telomere-binding protein leads to leaner mice with altered metabolic profiles

Mice lacking TRF1 in the whole organism show decreased body weight and adiposity

We previously showed that deleting the Trf1 gene in both wild-type and cancer-prone (due to p53 and Ink4Arf deficiencies) mouse models had no significant impact on mouse survival, although in some genetic backgrounds led to reduced body weight and mild hair and skin phenotypes [24]. To gain a deeper understanding on the weight reduction associated to TRF1 abrogation in mice, here we induced persistent Trf1 deletion from ten weeks of age until humane endpoint in Trf1^lox/lox, hUBC-CreERT2 mice by intraperitoneal injection of tamoxifen (Figure 1A). Consistent with previous observations, no significant differences in survival were found between genotypes (Supplementary Figure 1A). Interestingly, we observed a significantly decreased body weight starting at five months of age in Trf1^Δ/Δ females and at six months of age in Trf1^Δ/Δ males compared to age-matched wild-type controls (Figure 1B, 1C). The decreased body weight was maintained throughout their remaining lifespan (Figure 1D).

Trf1 is a target of the pluripotency factor OCT4 and has been previously shown by us to play a role in maintenance of pluripotency, as well as stemness [14]. Given the high proliferative capacity of the intestine, we investigated whether the reduced body weight observed in Trf1-deleted mice could be the result of intestinal atrophy owing to TRF1 abrogation. To this end, we measured the length of their intestinal villi. We did not observe any significant differences in villi length between wild-type and Trf1-deleted mice in both genders (Supplementary Figure 1B). Next, a comprehensive histological analysis of mice of both genotypes and sexes at the time of death was performed by a pathologist to assess the potential effects of Trf1 deletion in the organism which could explain the observed differences in body weight. However, we did not detect any differences between genotypes in pancreas, spleen, intestin, colon, kidney, heart, lung, brain and muscle. We next further explored metabolic organs, such as liver, as well as brown and white adipose tissues. We did not observe any differences between genotypes in liver or white adipose tissue. Of interest, the brown adipose tissue of Trf1 depleted mice exhibited fewer lipid droplets compared to wild-type controls. (Figure 1E).

Previous studies have linked telomere shortening with metabolic syndrome [17–19, 23]. Thus, we set to address whether TRF1 deficiency in mice leads to telomere shortening. To this end, we performed quantitative telomere fluorescence hybridization (Telomere Q-FISH). However, we did not detect any differences in telomere length between genotypes in the liver as well as the white adipose and brown adipose tissue (Supplementary Figure 1C). Consistent with prior studies reported, no change in telomere length upon TRF1 depletion (Piñeiro-Hermida et al., 2022; Bejarano et al., 2017).

To further investigate the origin of the decreased body weight of Trf1^Δ/Δ mice, we next determined total fat content in 40 weeks old males and females by using Dual-energy X-ray absorptiometry (DEXA). We found a striking reduction in fat mass and adiposity in both Trf1^Δ/Δ males and females compared to wild-type controls, which was maintained at 80 weeks of age.

Bone mineral density and bone mineral content, however, remained similar between genotypes (Figure 2A). Thus, the differences in body weight can be attributed to reduced fat content as the consequence of TRF1 abrogation.

Next, we set to measure several metabolic parameters in blood. Of note, LDL cholesterol levels were significantly higher in 40 weeks old Trf ^Δ/Δ male and female mice compared to wild-type controls, while plasma HDL cholesterol levels remained similar (Figure 2B).

This trend persisted at 80 weeks of age; however, differences in both LDL and HDL levels did not reach statistical significance (Figure 2B). Plasma triglycerides levels were similar in 50 weeks old Trf1^Δ/Δ and wild-type mice but were significantly decreased at 80 weeks of age in Trf1^Δ/Δ mice of both genders (Figure 2C). None of the other metabolic and blood chemistry parameters tested showed significant differences between genotypes (Supplementary Figure 1D). Altogether, these findings indicate that Trf1 deletion induced a persistent weight reduction due to decrease adiposity.

Mice lacking TRF1 in the whole organism show improved glucose and insulin metabolism with aging

Given the unprecedented effect of Trf1 deletion on reducing fat content and body weight, we next set to address whether Trf1 abrogation affected other major metabolic parameters, such as glucose and insulin metabolism. At thirty-six weeks of age, Trf1^Δ/Δ mice showed improved glucose tolerance compared to wild-type controls, reaching statistical significance in the case of females (Figure 3A). In turn, at eighty weeks of age, Trf1^Δ/Δ males showed better glucose tolerance compared to wild-type controls. Accordingly, both Trf1^Δ/Δ male and female mice showed an improved tolerance to insulin at eighty weeks of age compared to wild-type mice (Figure 3B). Of note, upon fasting, however, we found similar glucose and insulin levels in both genders at eighty weeks of age. The derived insulin resistance and insulin sensitivity indices, homeostatic model assessment (HOMA-IR), and quantitative insulin sensitivity check index (QUICKI), respectively, revealed improved insulin resistance and increased insulin sensitivity in Trf1 deleted mice of both genders at eighty weeks of age (Figure 3C). In summary, Trf1^Δ/Δ mice exhibit an enhanced glucose metabolism with aging.

Mice lacking TRF1 in the whole organism show normal fitness

We next determined whether decreased body weight and improved lipid, glucose and insulin metabolism in Trf1^Δ/Δ mice could be attributed to a lower food intake or to an increase physical activity as the consequence of TRF1 protein abrogation. To this end, we subjected forty-three weeks old females to metabolic cages. Due to the limited space, we could not study male mice. We found that Trf1-deficient females produce more volume of dioxide (VCO2) and tend to use more energy compared to wild-type females during light cycles. The production of more volume of CO2 and the use of more energy indicates an increase metabolic rate of females without Trf1 compared to wild-type controls (Figure 4A). Of interest, in spite of their lower weight, Trf1^Δ/Δ mice showed a higher food intake, indicating that the lower body weight is not the consequence of a lower food intake (Figure 4A). We did not find any differences between genotypes in the respiratory quotient (RQ), activity, and rearing (Figure 4A). At eighty-two weeks of age, we found significantly lower levels of VCO2, volume of oxygen (VO2), and energy consumed in Trf1^Δ/Δ mice of both genders compared to wild-types controls (Figure 4B). At this age, no differences between genotypes were observed in activity, rearing and respiratory quotient (Figure 4B). In summary, the leanness observed in Trf1^Δ/Δ mice cannot be attributed to reduced food intake or increased physical activity levels. Interestingly, in forty-three weeks old females, the absence of TRF1 leads to heightened metabolic activity as evidenced by increased production of CO2 volume, higher energy consumption, and elevated food intake compared to wild-type controls.

Gene expression changes in mice lacking TRF1

To gain a deeper understanding on the molecular mechanisms by which TRF1 can influence metabolism, we performed an RNA sequencing (RNA-seq) on the livers of fifteen weeks old Trf1^+/+ and Trf1^Δ/Δ female mice. This early time-point is previous to the differences in weight between genotypes (Figure 5A). Surprisingly, even at this early age, we observed increased liver weight, a phenomenon not previously noted in older mice, along with reduced mass of white adipose tissue in Trf1^Δ/Δ mice compared to wild-type controls (Figure 5B). The higher liver weight in mice lacking Trf1 was unexpected, as all observed phenotypes thus far have been associated with improved metabolic aging and lower body weight. The decreased mass of white adipose tissue is consistent with the lower adiposity observed in the older cohort. Similarly, in line with these observations, fifteen weeks Trf1-deficient mice exhibited fewer fat droplets in their brown adipose tissue compared to wild-type controls. (Figure 5D).

Gene set enrichment analysis (GSEA) of Trf1^Δ/Δ versus control Trf1^+/+ RNA sequencing data showed that cholesterol biosynthesis and inflammatory pathways: inflammatory response, IL6-JAK-STAT3 signaling and TNFA signaling via NFKβ (Figure 5C) (Supplementary Figure 2A) were upregulated in Trf1^Δ/Δ mice compared to wild-type mice. The upregulation of cholesterol biosynthesis was surprising; however, it explains the higher LDL serum plasma levels observed in Trf1-deficient mice. On the other hand, Trf1 depletion has been described to cause inflammation [14, 25], in agreement with the upregulation of inflammatory pathways found here. In line with these findings, we observed increased macrophage infiltration as determined by F4/80 immunohistochemistry staining in liver Trf1^Δ/Δ mice compared to the wild-type controls (Figure 5E). Furthermore, adiponectin, a hormone involved in glucose and fatty acid metabolism was found upregulated in the RNAseq data. On the other hand, we observed a significant downregulation of pathways related to myogenesis, the citric acid TCA cycle and respiratory electron transport, adipogenesis, oxidative phosphorylation and PPAR signaling in Trf1^Δ/Δ mice compared to wild-type mice (Figure 5C). Hepatic stellate cells (HSCs) are quiescent liver-resident cells responsible for vitamin A storage and extracellular matrix (ECM) homeostasis. Upon liver injury or stress, HSCs become activated and transdifferentiated into a myofibroblast-like cells characterized by increased proliferation, contractility, and ECM production. The observed downregulation of myogenesis-related genes in the liver of mice lacking Trf1, suggests a shift from normal regenerative processes towards fibrotic remodeling, possibly driven by HSC activation in response to TRF1 abrogation. To test this hypothesis, we performed Sirius Red staining to assess liver fibrosis in the young female mice used for RNA-seq; no differences were observed at this early age. However, livers from fifty-week-old Trf1^Δ/Δ female mice displayed a significantly increased Sirius Red-positive area compared to wild-type controls (Figure 5F).

To investigate the origin of the increased liver fibrosis, we assessed the burden of DNA damage in the livers of fifteen-weeks-old females mice using H2AX staining. We found that Trf1 deficiency resulted in higher levels of DNA damage compared to wild-types controls (Figure 5G). A sustained DNA damage response following TRF1 depletion has been previously reported (Martínez et al., 2009; Bejarano et al., 2019).

Finally, other downregulated pathways were key metabolic pathways related with cellular energy production and metabolism regulation: oxidative phosphorylation, citric acid TCA cycle and respiratory electron transport, and PPAR pathway. These pathways provide insights into the observed phenotype of reduced weight and adiposity. Decrease expression of genes related to these pathways’ leads to a reduction in fat storage or deposition. Regarding energy storage, Periodic acid-Schiff (PAS) immunohistochemistry staining without and with diastase on the liver of fifteen-week-old females showed a tendency towards reduced glycogen content in Trf1^Δ/Δ compared to wild-types (Supplementary Figure 2B).

In summary, young Trf1-deficient females displayed a higher liver weight, which was concomitant with an upregulation in pathways related to cholesterol biosynthesis and inflammation. We also found a downregulation on myogenesis pathways concomitant with increased DNA damage liver damage and fibrosis as determined by H2AX and Sirius red staining’s respectively. Major metabolic pathways related with energy production and regulation of metabolism homeostasis were also found downregulated in Trf1-deficient mice. In addition, white adipose tissue of Trf1^Δ/Δ mice weight less compared to wild-types.

Mice lacking TRF1 in the whole organism subjected to a high-fat diet show decreased body weight, lower adiposity and cholesterol levels

Next, we addressed the impact of a high-fat diet in the metabolic phenotypes found in TRF1-abrogated mice. Under a high-fat diet, Trf1^Δ/Δ males showed a significant lower body weight, lower total fat content, and lower adiposity compared to wild-type male controls, while this was not observed in female counterparts (Figure 6A). Again, the decreased body weight in males subjected to a high fat died was not associated to intestinal atrophy as the consequence of TRF1 abrogation as indicated by similar length of the intestinal villi in Trf1^Δ/Δ and wild-type males (Figure 6B).

At 18 weeks of age both Trf1^Δ/Δ females and males subjected to a HFD showed significantly lower levels of HDL compared to wild-type controls (Figure 6C). Similarly, LDL cholesterol levels were lower in Trf1^Δ/Δ mice of both genders, but this difference only reached statistical significance in Trf1^Δ/Δ females (Figure 6C).

Furthermore, Trf1^Δ/Δ males showed statistically significant lower levels of blood urea nitrogen (BUN) in plasma compared to wild-type controls (Figure 6D), indicating lower kidney damage.

The weight of the brown adipose tissue was also significantly decreased in Trf1^Δ/Δ males compared to wild-type controls (Figure 6E). In addition, the subcutaneous fat layer thickness was strikingly thinner in Trf1 deleted males compared to wild-type counterparts (Figure 6F). Together, these findings suggest that Trf1 plays a role in modulating body weight by reducing fat content.

Mice lacking TRF1 in the whole organism subjected to a high-fat died show improved glucose metabolism

To explore whether glucose metabolism was also improved in Trf1^Δ/Δ mice subjected to a high-fat diet, we performed a glucose tolerance test. Males without Trf1 showed a better tolerance to glucose compared to wild-type controls (Figure 7A), however this difference was not observed in similarly treated females (data not shown).

To further study the male phenotype, we studied glucose and insulin metabolism. Plasma glucose P levels were strikingly lower in Trf1 deleted mice compared to wild-type males (Figure 7B). When mice were subjected to fasting, glucose and insulin levels were similar between genotypes. In accordance with mice fed with high-fat diet, both genotypes showed resistance to insulin (Figure 7C). Next, we measured the Respiratory Quotient (RQ) in both genotypes by using metabolic cages. An RQ between 0.7 and 1 indicates that the sources of energy are protein and fat. In high-fat fed mice, the expect RQ should be under 0.7, as fat is the main source of energy. As expected, wild-type males showed an RQ below 0.7 (Figure 7D). In contrast, the RQ in Trf1-deficient mice was close to 0.7 (Figure 7D), suggesting that they use protein and fat as energy sources. No differences were observed in VO2, VCO2, energy expenditure, rearing and activity (Figure 7D). Food intake could not be determined as the mice were gnawing the food.

Next, we performed full histological analysis of both male and female of both genotypes mice subjected to high-fat diet mice. Fewer and smaller lipid droplets were observed in liver, white adipose tissue and brown adipose (Supplementary Figure 2C). Again, the Trf1^Δ/Δ mice exhibited a significantly larger H2AX-positive area relative to the total area compared to the wild-type controls. (Supplementary Figure 2D).

Finally, we observed that both Trf1 deficient male and female mice subjected to both normal and high-fat diets had lower fat accumulation in visceral and gonadal areas compared to wild-type controls at time of death (Supplementary Figure 2E).

These results suggest that Trf1 abrogation may impede the utilization of fat as main energy source, probably as the consequence to the reduced content of white adipose tissue.

Mouse models

TRF1^lox/lox mice [29] were crossed with the mouse strain carrying ubiquitously expressed, tamoxifen activated recombinase, hUBC-CreERT2 [36] to generate TRF1^+/+ or TRF1^lox/lox; hUBC-CreERT2 mice. All mice have been maintained at the Spanish National Cancer Centre under specific pathogen-free conditions, with a 12h light-dark cycle in a temperature-controlled room following the recommendations of the Federation of European Laboratory Animal Science Associations (FELASA). All the experiments were approved by the CNIO-ISCIII Ethics Committee and, by Consejería de Medio Ambiente, Administración Local y Ordenación del Territorio (Comunidad de Madrid). All experiments have been performed in accordance with the guidelines for ethical conduct in the care and use of animals as stated in the international guiding principles for biomedical research involving animals, developed by the Council for International Organizations of Medical Sciences (CIOMS). Littermates of the same sex were randomly assigned to either experimental or control groups. Developmental stage of mice is included appropriately in the text and figure legends. Food (Harlan Laboratories and Research Diets) and water were provided ad libitum, unless otherwise noted. The timepoints were selected according to the health status of the mice and the required recovery periods between experiments. This approach helped minimize stress on the animals and ensured stable conditions for the tests.

Mouse diets

Mice were randomized into test groups and given ad libitum access to either a chow (18% fat, 58% carbohydrates and 24% proteins) (Harlan Laboratories, 2018S) or high-fat diet (45% fat, 35% carbohydrates and 20% proteins) (Research Diets, D12451) at the age of 10 weeks and, 8 weeks in the case of high-fat diet fed mice. Cre-mediated recombination was activated by tamoxifen intraperitoneal injection at 10 weeks and, 12 weeks for high-fat diet mice.

Intraperitoneal injection of tamoxifen was administrated (1mg/injection) 3 times per week for 1-1,5 month and then 1 boost every 15 days. Body weight were measured every week.

Blood parameters

Blood was collected from the facial vein to an EDTA tube for hemogram analysis using Laser Cell. Metabolic parameters such as ALT, ALP, total bilirubin, creatinine, total cholesterol, albumin and blood urea nitrogen were quantified by Vetscan mammalian liver profile. Plasma was obtained after 4000g, 20 minutes centrifugation. TGC (triglycerides), glucose P, HDL and LDL levels were determined by ABX Pentra (Horiba Medical). Insulin was measured using ultrasensitive mouse insulin ELISA kit (Christal chem ELISA). HOMA-IR was calculated following the formula: fasting insulin [F06DU/mL] x fasting glucose [mg/dL]/22.5. Quicki formula followed was: 1/[log (fasting glucose [mg/dl]) + log (fasting insulin [μU/mL]). The conversion factor used for calculate HOMA and Quicki given by Christal chem ELISA was: 1ng/mL = 24 μIU/mL.

Body composition analysis

Animals were anaesthetised with 2% isoflurane (Isovet, Braun Vetcare) and total fat, lean and fat masses were measured by Dual energy X-ray Absorptiometry (DEXA) (PIXImus, Lunar Corporation). The analysis was performed using a region of interest (ROI) in the whole body using Lunar PIXImus 2.10 software.

Gene expression analysis

Total RNA from tissue was extracted with the RNeasy kit (QIAGEN). For RNA-seq experiments, total RNA samples were used. RNA quality scores were 7,51 on average assayed by a PerkinElmer LabChip analyzer. For library generation, reverse transcription using oligo-dT priming was performed and subsequently, a second-strand synthesis was performed using random primers and a DNA polymerase. Both sets of primers included Illumina-compatible sequences. The resulting cDNA libraries were purified and applied to an Illumina flow cell to generate clusters. Sequencing was carried out on an Illumina NextSeq 550 using v2.5 reagent kits, following the manufacturer’s protocols. The reads underwent analysis using a Snakemake-based pipeline developed at the CNIO Bioinformatics Unit (https://github.com/cnio-bu/cluster_rnaseq). The quality of sequencing was assessed using FastQC v0.11.9, and absence of sample contamination from other species was verified with FastQ Screen v0.15.1. To remove sequencing adaptors and poly(A) tails, the ‘bbduk.sh’ command was applied as per Lexogen recommendations. Alignment of the processed reads to the mouse reference genome (GRCm39) was performed using STAR v2.7.10a, and read counts were obtained using featureCounts [37]. For gene annotation, GENCODE GRCm39 vM27 was used.

Differential expression was done with DESeq2 [38], and those genes with an FDR < 0.05 were considered as significant.

Functional analysis was performed with Gene Set Enrichment Analysis (GSEA) employing 1,000 gene-set permutations only gene sets exhibiting significant enrichment levels (FDR q-value < 0.25) were considered.

Glucose tolerance test and insulin tolerance test

For GTT, mice were fast for 12 hours and injected intraperitoneally 2g of glucose/kg of body weight in PBS. For ITT, 0,75U insulin /kg of body weight in PBS (Humalog Insulin) was intraperitoneally administered. Glucose levels were measured from the tail vein before, and at 15, 30, 60, 90 and 120 minutes after the injection using a glucometer.

Immunohistochemistry analyses in tissue sections

Tissues were fixed overnight in 10% formalin solution (sigma) and embedded in paraffin. For immunohistochemistry staining, tissue sections were deparaffinized with 10mM sodium citrate for 2 minutes. Slides were washed in water, then in buffer TBS Tween20 0,5%, blocked with peroxidase, washed again with TBS Tween 20 0,5% and blocked with fetal bovine serum followed by another wash.

Primary antibodies

Hematoxylin and eosin (H&E) staining was performed for the quantification of intestinal villi and subcutaneous skin fat layer. H&E staining was also employed to evaluate tissue morphology and integrity of liver, white and brown adipose tissue, pancreas, spleen, intestine, colon, kidney, heart, lung, brain and muscle. Sirius red staining was used for determining liver fibrosis. H2AX staining was utilized for the evaluation of DNA damage in liver. Periodic acid-Schiff staining without and with diastase was performed to detect the presence of glycogen storage. Followed by incubation with secondary antibodies conjugated with peroxidase from DAKO.

Olympus AX70 microscope was used for taking pictures of immunohistochemistry tissue samples. The percentage of positive cells was identified using ZEISS ZEN 3.7 software.

Metabolic activity

Metabolic activity was determined using Oxylet system metabolic chambers (Panlab Harvard Apparatus). Mice were individually caged for 1 week for acclimation. During the 72-hour measurement period, the mice were provided unrestricted access to food and water. Throughout the entire study duration, mouse activity, food intake, and drink intake were observed and recorded at 2-minute intervals. The respiratory exchange ratio (RQ) was calculated by dividing the volume of consumed O2 (VO2) by the volume of eliminated CO2 (VCO2), which were measured every 24 minutes. Energy expenditure (EE) was determined using the formula EE = (3.815 + (1.232 × RQ)) × VO2 × 1.44. The data analysis was conducted using Metabolism 3.0 software.

Telomere length analysis in tissue sections

To analyze telomere length, a quantitative telomere fluorescence in situ hybridization (qFISH) was performed. Tissues were embedded in paraffin and were deparaffinized. Followed first by a digestion with pre-warmed pepsin/HCl and then 4% formaldehyde for fixation. Slides were dehydrated by increasing concentration of ethanol (70%, 90%, and 100%). Next, the samples were incubated with the telomeric probe for 3.5 minutes at 85°C followed by another incubation for two hours at room temperature in a wet chamber. The slides were washed extensively with 10mM Tris in 70% formamide and with 0,08% TBS-Tween. DAPI was added and the slides were mounted with Vectashield. Immunofluorescence images were obtained using a CS SP8 STED 3X (Leica-Microsystems) under the Navigator software integrated in the LAS X v3.5.7 (Leica-Microsystems).

A robust methodology was developed to identify and characterize telomere signals using Cellpose [39] algorithm, which leverages deep learning techniques. Our approach consisted of three key components: telomere/nucleus identification, telomere-nucleus pairing, feature extraction and characterization to assess telomere-nucleus interactions. The model for telomere segmentation was fine-tuned on our specific dataset to enhance performance, then telomeres were accurately delineated, providing precise localization and segmentation results. In parallel, nuclei segmentation was crucial for associating telomeres with their respective cellular context. In this regard, a custom pipeline in Groovy, utilizing ImageJ libraries, linked each telomere to its corresponding nucleus, establishing telomere-nucleus pairs. This pairing process was meticulously validated for accuracy and consistency. Subsequently, for each telomere-nucleus pair, telomere feature extraction and characterization were computed to quantify telomere fluorescence signal strength, size and spatial distribution with their corresponding descriptive statistics to assess telomere quantification. To determine telomere localization, fluorescence intensity distribution was analyzed in telomeres across all cells in our dataset. Then, positive thresholds were computed based on mean intensity and deviation from the average. Telomeres with fluorescence marker intensities significantly deviated from the norm were classified as positive.

Statistical analysis

Survival data was analyzed by Kaplan Meier survival curves, and comparisons were conducted via the Log Rank test. The statistical analysis was assessed through the use of GraphPad Prism 10.1.1 software. Quantification of immunohistochemistry was carried out using ZEISS ZEN 3.7 software, while telomere length was determined through the application of the Cellpose algorithm as previously detailed in earlier sections. Unpaired Student’s t-test was used to determine statistical significance, with P-values below 0.05 considered as significant. The statistical analysis was executed using Microsoft Excel 16.66.1.

Data availability

The RNA-seq data has been archived in the Gene Expression Omnibus of the National Center for Biotechnology Information under the accession number GSE273596.