Background
Diabetes of Mendelian inheritance constitute a heterogeneous group of conditions that cause hyperglycemia, mainly due to molecular defects in genes that are critical for beta cell or adipocyte development and functions. There is a growing interest in identifying the molecular and cellular mechanisms responsible for these inherited diabetes to improve genetic counseling and personalize treatment [
1]. Indeed, knowledge of their genetic cause will assist patient treatment, prognosis, and genetic counseling [
2,
3]. Among Mendelian diabetes, the most frequent and well-known subgroup is maturity-onset diabetes of the young (MODY) [
4]. Lipoatrophic diabetes (LD) constitute another class of inherited diabetes [
5,
6]. These lipodystrophic syndromes are characterized by clinical lipoatrophy due to a defect in adipose tissue storage of triglycerides. This results in ectopic lipid infiltration of non-adipose tissues leading to insulin resistance, increased liver glucose production, hypertriglyceridemia, and liver steatosis [
3,
5]. About 30 genes have been implicated in LD, but a majority of cases remains genetically unexplained [
5,
6].
In the present study, we identified by whole-exome sequencing (WES) two affected siblings with LD and carrying a homozygous splice-site pathogenic variant in
TYMP, the gene encoding thymidine phosphorylase (TP). A number of
TYMP loss-of-function mutations have previously been implicated in Mitochondrial NeuroGastroIntestinal Encephalomyopathy (MNGIE; MIM #603041), a rare metabolic autosomal recessive disease [
7‐
9]. MNGIE is mostly characterized by gastrointestinal and neurological manifestations, including severe cachexia, gastrointestinal dysmotility, peripheral neuropathy, leukoencephalopathy, ophthalmoplegia, and ptosis [
10,
11]. The disease is progressively degenerative and usually leads to death in early adulthood [
12]. Currently, there are no specific therapies for patients with MNGIE and the disease management aims to treat the symptoms evidenced in each individual [
13]. The genetic result obtained in the patients investigated herein was unexpected since classical MNGIE manifestations were absent at referral, though the disease evolved towards a more complex phenotype. A third patient with a similar disease onset and carrying a
TYMP homozygous missense variant was subsequently investigated.
It is well established that the TP enzyme catalyzes the reversible phosphorylation of the deoxyribonucleosides, thymidine (also known as deoxythymidine) and 2′-deoxyuridine, into thymine and uracil, respectively, accompanied by the release of 2-deoxyribose 1-phosphate [
10]. The loss of TP activity due to pathogenic variants results in increased cellular levels of thymidine and 2′-deoxyuridine [
14,
15]. Although
TYMP is a nuclear gene and TP a cytosolic protein, mutations affect mitochondrial DNA and function. High deoxyribonucleoside levels are indeed toxic for mitochondrial DNA (mtDNA) and lead to mitochondrial failure due to a progressive acquisition of secondary mtDNA mutations and mtDNA depletion [
16]. This ultimately leads to dysfunction of the respiratory chain, and thus, inadequate cellular energy production. The involvement of TP in LD would imply a key role of this enzyme in adipose tissue physiology. In this regard, the expression of TP in adipocytes remains uncertain. It was reported in one study more than 20 years ago that TP was not expressed in adipose tissue, although the corresponding data obtained by an enzyme immunoassay were not presented [
17]. Since then, and despite the presence in expression databases of information arguing for the contrary, the idea of an absence of TP in adipocytes has remained persistent in the literature [
7,
10,
18]. This could have led to the disregard of a possible role of TP in adipocytes. Most in vitro models of MNGIE described in the literature have used HEK293 cells, HeLa, or fibroblast cells and have facilitated the understanding of the cellular effect of deoxyribonucleoside pool imbalances [
19‐
21]. However, studies in cellular systems with relevance to the organs affected in MNGIE are still lacking.
The aim of this study was to describe the first reported patients with atypical forms of monogenic LD and carrying homozygous pathogenic variants in TYMP. This led us to investigate the role of TP in preadipocytes and adipocytes in vitro. The impact of the loss of TP activity on mitochondrial functions, adipogenesis, and insulin sensitivity, as well as cellular senescence, was evaluated using a CRISPR-Cas9-mediated genome-editing approach.
Methods
Genetic analyses
Genomic DNA was extracted from peripheral blood leukocytes using standard procedures.
Gene panel
A panel containing the following genes involved in LD was analyzed:
ADRA2A,
AGPAT2,
AIRE,
AKT2,
BANF1,
BLM,
BSCL2,
CAV1,
CAVIN1,
CIDEC,
DYRK1B,
EPHX1,
ERCC3,
ERCC6,
ERCC8,
FBN1,
INSR,
LEMD2,
LIPE,
LMF1,
LMNA,
LMNB2,
MDM2,
MFN2,
MTX2,
NSMCE2,
PCNT,
PCYT1A,
PIK3R1,
PLIN1,
POC1A,
POLD1,
POLR3A,
PPARG,
PTPN11,
POMP,
PSMA3,
PSMB4,
PSMB8,
PSMB9,
PSMG2,
OTULIN,
SLC29A3,
SPRTN,
WRN, and
ZMPSTE24. Exons and flanking intronic sequences were captured from fragmented DNA with the SeqCapEZ enrichment protocol (Roche NimbleGen, WI, USA) followed by paired-end massively parallel sequencing on a MiSeq platform (Illumina, CA, USA) [
22]. Bioinformatic analysis was performed using the Sophia DDM pipeline® (Sophia Genetics, Switzerland).
Whole exome sequencing (WES)
Library preparation, exome capture, sequencing, and variant annotation were performed by IntegraGen SA (Evry, France). Genomic DNA was captured using the Twist Human Core Exome Enrichment System (Twist Bioscience, OR, USA) and IntegraGen Custom, followed by paired-end 75 bases massively parallel sequencing on Illumina HiSeq4000. Analysis of exome data was performed using Sirius software (IntegraGen SA). TYMP variants were described based on the longest isoform (NM_001953.4) using Alamut 2.11 (Sophia Genetics, Lausanne, Switzerland) and Human Genome Variation Society guidelines.
Sanger sequencing
The Big Dye Terminator v3.1 sequencing kit (Thermo Fisher Scientific, MS, USA) was employed after PCR amplification and data were analyzed on a 3500xL Dx device with the SeqScape v2.7 software (Thermo Fisher Scientific).
Computational analysis of TYMP variants
We evaluated the pathogenicity of
TYMP variants using bioinformatic tools available online. These bioinformatic tools are not the same for splice site and missense variants, as detailed in supplementary materials for each of the two variants identified (Additional file
1: Table S1).
Adipose stem cell (ASC) isolation, culture, and adipocyte differentiation
ASC were isolated from surgical samples of subcutaneous abdominal adipose tissue from a 25-year-old healthy woman with a normal body mass index (BMI). Adipose tissue was enzymatically digested with collagenase B (0.2%). After centrifugation, stromal vascular fraction was filtered, rinsed, plated, and cultured in α-MEM with 10% fetal calf serum (FCS), 1% GlutaMAX (#35050061, Thermo Fisher Scientific), 1% Penicillin/streptomycin (PS - 10,000 UI/mL), 1% HEPES, and fibroblast growth factor-2 (FGF-2 -145 nmol/L). After 24 h, only ASC adhered to plastic surfaces, while other cells were removed after culture medium replacement. ASC were maintained in an undifferentiated state in α-MEM supplemented with 10% newborn calf serum (#CA-1151500; Biosera, MI, USA), 1% GlutaMAX, HEPES and P/S, and FGF-2 (145 nmol/L). Adipocyte differentiation was induced by treating 2-day post-confluent cultures with high-glucose (25 mmol/L) DMEM supplemented with 10% FCS, 1% PS, 1 μmol/L dexamethasone (#D4902; Sigma-Aldrich, MI, USA), 1 μM rosiglitazone (#D4902; Sigma-Aldrich), 250 μM 3-isobutyl-1-methyl xanthine (IBMX) (#I7018; Sigma-Aldrich), and 0.17 μmol/L insulin (#I0516; Sigma-Aldrich) for 10 days. The medium was then replaced with high-glucose DMEM supplemented with 10% FCS, 1% PS, 1 μmol/L rosiglitazone, and 0.17 μM insulin and changed to fresh medium every 2 days until the 20th day.
ASC osteoblastic differentiation
Osteoblast differentiation was induced by treating 2-day post-confluent cultures with α-MEM supplemented with 10% FCS, 1% PS, 1 nmol/L vitamin D3 (#C9756; Sigma-Aldrich), 170 μmol/L ascorbic acid (#A8960; Sigma-Aldrich), and 10 mmol/L β-glycerophosphate (#G9422; Sigma-Aldrich) for 14 days.
The lentiviral plasmid plentiCRISPRv2 was a gift from Zhang lab (Addgene, MA, USA; plasmid #52961) and contains hSpCas9, a guide RNA (gRNA), and a puromycin resistance sequence. The gRNA targeting exon 5 of
TYMP was designed with a well-recognized tool (
http://cistrome.org/SSC) to ensure specificity and high cleavage efficiency. Its sequence was the following: sense 5′ CAGAGATGTGACAGCCACCG 3′; antisense 5′ CGGTGGCTGTCACATCTCTG 3′. The web-based tool, CRISPOR (
http://crispor.tefor.net) [
23] was used to avoid off-target sequences (Additional file
2: Table S2). Lentiviruses dedicated to TP knockdown were produced by the VVTG platform (Federative Research Institute, Necker, France). ASC were infected with viral particles at a minimal titer of 10
8 units per mL. 48 h post infection, transduced cells were selected with 0.5 μg/mL puromycin dihydrochloride (#P9620; Sigma-Aldrich). Surviving cells were propagated, and the heterogeneous cell pool was used for experiments. The percentage of on-target recombination including insertions and deletions (indels) in the genomic DNA from this cell population was evaluated by Sanger sequencing of
TYMP exon 5 followed by analysis using the Synthego web-based tool (
https://ice.synthego.com).
Measurement of TP activity
TP activity was measured in white blood cells using a spectrophotometric method. Whole blood was collected at the time of the medical consultation, not necessary at fasting state, for the patients of family A and their parents. The normal reference range for TP activity was determined in the same laboratory using samples from healthy volunteers. Pellets of white blood cells were first homogenized in lysis buffer (50 mmol/L Tris–HCl, pH 7.2, containing 1% Triton X-100, 2 mmol/L phenylmethylsulfonyl fluoride, and 0.02% mercaptoethanol) and sonicated for 10 s, before centrifugation at 20,000g for 30 min at 4°C. Protein concentration in supernatants was determined according to the bicinchoninic acid method on a multiparametric analyser (Indiko™ Clinical Chemistry Analyze; Thermo Fischer Scientific). The reaction mixture containing 100 μg of protein, 15 mmol/L thymidine in 0.1 mol/L Tris–arsenate, pH 6.5, was incubated at 37°C for 1 h. The reaction was stopped by the addition of 1 mL of 0.3 N NaOH. The amount of thymine formed was measured at 300 nm wavelength and determined based on the 3.4 × 103 L/mol/cm difference in the molar extinction coefficient between thymidine and thymine at alkaline pH. Enzyme activity was expressed as μmol of thymine formed per hour per mg of protein.
Western blot
Cells were suspended in NP-40 lysis buffer. Thirty micrograms of protein extracts were separated by sodium dodecyl sulfate - polyacrylamide gel electrophoresis (SDS-PAGE), transferred to polyvinylidene difluoride membrane and analyzed by immunoblotting using appropriate antibodies (see below for a detailed list). Western blot quantification was performed in triplicate using Fiji software (Open source), and results were normalized to the tubulin protein levels. Uncropped and unedited Western blots seen in the different figures are available in Additional file
3.
Oil Red-O staining, image processing, and quantification
Intracellular lipids were stained by Oil Red-O (#O0625; Sigma-Aldrich). Cells were washed with phosphate-buffered saline (PBS) and fixed with 4% paraformaldehyde (PFA) in PBS, for 10 min. Fixed cells were incubated with Oil Red-O solution for 1 h at room temperature and then with DAPI (Thermo Fischer Scientific) for 5 min. Fluorescence images were generated with IX83 Olympus microscope, acquired with Cell-Sens V1.6, and analyzed with FIJI software. Images of 8–10 different areas per condition were visualized by fluorescence microscopy using mCherry filter, followed by computer image analysis using FIJI software. Analysis was performed by threshold converting the 8-bit Red-Green-Blue image into a binary image, which consists only of pixels representing lipid droplets (i.e., red). Importantly, after separation, the binary image was manually compared with the original image for consistency and correct binary conversion. The area occupied by lipid droplets in the image was displayed by FIJI software as surface area in μm2 and normalized to cell number by semi-automated counting of DAPI-stained nuclei.
Seahorse analysis
Measurement of mitochondrial respiration (oxygen consumption rates - OCR) was done using a Seahorse XFe96 BioAnalyser (Agilent Technologies, CA, USA). WT, control, and KO ASC were seeded at an optimized density of 10 000 cells/100 μL/well in a 96-well XFe96 cell culture microplate, incubated 24 h, and equilibrated for 1 h in unbuffered XF assay medium (Agilent Technologies) supplemented with 2 mM glutamine, 10 mM glucose, and 1 mM sodium pyruvate. Successive OCR measurements were performed in each well: 3 basal measurements, 3 measurements following the automated injection of 1 μM oligomycin (ATP synthase inhibitor to measure respiration associated with cellular ATP production), 3 following the injection of 1 μM carbonyl cyanide p-trifluoromethoxyphenyl hydrazone (FCCP) (uncoupling agent to measure the maximal respiration capacity), and 3 following the injection of 1 μM antimycin A (electron transport chain inhibitor to measure non-mitochondrial respiration). The data were normalized to the protein content measured in each well using the bicinchoninic acid assay (BCA; Thermo Fisher Scientific) according to the manufacturer’s instructions.
Quantification of intracellular triglyceride content
Intracellular lipids were extracted from differentiated ASC using hexane/isopropyl alcohol (3:2). Cells were washed and incubated with hexane/isopropyl alcohol (3:2, vol/vol) using 500 μL per well in 6-well culture plates, in a shaker (80 rpm/minute) at room temperature for 60 min. The content of each well was then transferred into a glass tube for nitrogen evaporation of the organic solvent. After evaporation, lipids were resuspended in isopropyl alcohol and transferred into duplicate 96-well plates for analysis after drying. Triglycerides were measured using the Infinity™ Triglyceride kit (Thermo Fischer Scientific) according to manufacturer’s instructions. The absorbance of each well was measured using a Tecan microplate reader (TECAN, Männedorf, Switzerland) and converted to concentration based on a standard curve. Results were normalized to the cell protein content.
Oxidative stress and cellular senescence
The oxidation of the fluorogenic probe 2,7-dichlorodihydrofluorescein diacetate (CM-H2DCFHDA) (2 μg/mL, #C6827; Sigma-Aldrich) was used to evaluate intracellular levels of reactive oxygen species (ROS) on a 200-plate fluorescence reader (TECAN) at 520–595 nm. The blue staining of β-galactosidase (β-gal) at pH 6 was used as a biomarker of cellular senescence. Cells were fixed with 4% PFA in PBS for 5 min at room temperature. Cells were washed twice with PBS and incubated overnight in fresh SA-β-gal staining solution containing 1 mg/mL of X-gal (5-bromo-4-chloro-3-indolyl-β-Dgalactopyranoside) (#3117073001; Sigma-Aldrich), 5 mmol/L potassium ferrocyanide, 5 mmol/L potassium ferricyanide, 150 mmol/L NaCl, 2 mmol/L MgCl2, and 0.4 mmol/L phosphate buffer, at pH 6.0, in darkness at 37°C without CO2. For positive staining controls, fixed cells were treated with the same solution, but at pH 4.0. After imaging with an IX83 Olympus microscope, stained cells were resuspended with 2% SDS, scratched, and sonicated. Absorbance (630 nm) was read with a Tecan Infinite 200-plate reader, and the pH 6.0/pH 4.0 staining ratio was calculated.
Statistics
Data are presented as means ± SD (standard deviation). GraphPad Prism software (CA, USA) was used to evaluate statistical significance. Gaussian distribution was tested with the Kolmogorov–Smirnov test. Multiple comparisons were conducted by one-way analysis of variance (ANOVA) with Bonferroni-test or Kruskal-Wallis test for post hoc analysis. p < 0.05 was considered statistically significant.
Key resources table
Reagent type or resource
|
Designation
|
Source and reference
|
Identifiers
|
Additional information
|
Adipose stem cells | ASC | Pr. Fève lab at CRSA, Paris | N/A | Female, from subcutaneous abdominal adipose tissue |
Antibody | Anti-adiponectin | Thermo Fisher Scientific | Cat# MA1-054 | WB (1:1000) |
Antibody | Anti-AKT | Cell Signaling Technology | Cat# #9272 | WB (1:1000) |
Antibody | Anti-C/EPBα | Protein Tech | Cat# 18311-1-1P | WB (1:1000) |
Antibody | Anti-ERK | Cell Signaling Technology | Cat# 9102 | WB (1:1000) |
Antibody | Anti-FAS | Cell Signaling Technology | Cat# 3180 | WB (1:1000) |
Antibody | Anti-IRΒ | Cell Signaling Technology | Cat# 3025 | WB (1:1000) |
Antibody | Anti-IRS1 | Protein Tech | Cat# 17509-1-AP | WB (1:1000) |
Antibody | Anti-leptin | Thermo Fisher Scientific | Cat# PA1-051 | WB (1:1000) |
Antibody | Anti-osteocalcin | Santa Cruz Biotechnology | Cat# sc-74495 | WB (1:1000) |
Antibody | Anti-P16 | Protein Tech | Cat# 10883-1-AP | WB (1:1000) |
Antibody | Anti-P21 | Protein Tech | Cat# 10355-1-AP | WB (1:1000) |
Antibody | Anti-P53 | Abcam | Cat# ab1101 | WB (1:1000) |
Antibody | Anti-P-AKT | Cell Signaling Technology | Cat# #9271 | WB (1:1000) |
Antibody | Anti-perilipin | Abcam | Cat# ab3526 | WB (1:1000) |
Antibody | Anti-P-ERK | Cell Signaling Technology | Cat# 9101 | WB (1:1000) |
Antibody | Anti-P-P53 | Abcam | Cat# ab38497 | WB (1:1000) |
Antibody | Anti-PPARg | Protein Tech | Cat# 16643-1-AP | WB (1:1000) |
Antibody | Anti-Runx2 | Protein Tech | Cat# 20700-1-AP | WB (1:1000) |
Antibody | Anti-SREBP-1 | Santa Cruz Biotechnology | Cat# sc-366 | WB (1:1000) |
Antibody | Anti-Tubulin | Sigma-Aldrich | Cat# T5168 | WB (1:10,000) |
Antibody | Anti-TP | GeneTex | Cat# GTX23151 | WB (1:1000) |
Antibody | Anti-P-Tyr | Santa Cruz Biotechnology | Cat# sc-7020 | WB (1:500) |
Antibody | Anti-rabbit-HRP | Cell Signaling Technology | Cat# 7074 | WB (1:3000) |
Antibody | Anti-mouse-HRP | Cell Signaling Technology | Cat# 7076 | WB (1:3000) |
Recombinant DNA reagent (plasmid) | lentiCRISPR v2 | Addgene | Cat# 52961 | A gift from Zhang lab |
Software algorithm | FIJI software | NIH | N/A | |
Software algorithm | Prism | Graphpad Software | N/A | |
Discussion
This translational study identifies biallelic TYMP pathogenic variants as a new genetic cause of monogenic lipoatrophic diabetes due to mitochondrial dysfunction. Our data also demonstrate the key role of TP in adipocytes.
MNGIE is characterized by a complex clinical picture involving multiple organs to differing extents in different individuals. The onset of MNGIE disease is usually between the first and second and decade of life [
8], as seen in the patients reported herein. Based on a review of the literature, Pacitti et al. proposed a classification of the major and minor MNGIE clinical features [
10]. Major signs included severe gastrointestinal dysmotility, cachexia, peripheral neuropathy, ocular symptoms, and diffuse leukoencephalopathy. Minor clinical criteria comprised neurological, muscular, cardiac, and endocrine features. The originality of the current study is to report an inaugural presentation of the disease in the form of isolated lipoatrophic diabetes, thereby showing the heterogeneous clinical onset of MNGIE. In patients from families A and B, generalized lipoatrophy indeed appeared several years before the first gastrointestinal or neurological manifestations. Consequently, the extreme thinness observed in so many patients with MNGIE could correspond to both lipoatrophy due to a primitive adipose tissue dysfunction like in the patients investigated herein, to cachexia secondary to gastrointestinal features as described in other cases, or to a combination of the two. It might seem curious that the generalized lipoatrophic phenotype present in the three patients investigated herein appeared in adolescence. Indeed, the historical classification of lypodystrophies distinguishes the generalized forms with a generalized fat loss apparent at birth, and the partial forms of lipodystrophic syndromes beginning later in life, frequently in adolescence. However, this dichotomy does not apply to all situations and some patients with a genetically defined lipodystrophic syndrome have already been reported with normal fat distribution at birth and appearance of generalized fat loss later in life [
37]. In addition, in MNGIE syndrome, the disease phenotype is known to appear progressively, when a threshold level of mutated mtDNA is reached, which is generally when more than 80–90% of total mitochondria are affected [
38,
39]. This threshold effect very likely contributes to the protracted interval before the condition manifests and to the disease phenotypic heterogeneity.
We undertook a systematic search in the literature to identify patients with a genetic diagnosis of MNGIE and manifestations of LD. To the best of our knowledge, lipoatrophy has never been reported, though several reports described patients with metabolic alterations evocative of LD. Hypertriglyceridemia was mentioned in a few studies [
8,
40‐
44]. Liver steatosis or cirrhosis associated with hepatomegaly and increased liver enzymes were described in several reports [
8,
40‐
43,
45‐
50]. Early diabetes was mentioned in a few patients [
8,
40,
51‐
53], although its etiology has never been investigated. In one patient, the disease was reported to start by metabolic manifestations including diabetes, hypertriglyceridemia, and liver steatosis [
54], which is reminiscent of current observations. All these data show that the metabolic part of the MNGIE clinical spectrum might be underestimated.
The patients investigated herein carry in the homozygous state two novel
TYMP molecular defects, including a splice site and a missense variant, whose pathogenicity was confirmed by biochemical assays. This study shows that
TYMP analysis should enter genetic routine diagnosis of monogenic lipoatrophic diabetes.
TYMP pathogenic variants are found in ethnically diverse populations [
10,
55]. It is currently not possible to state the prevalence of MNGIE as the disorder is probably underdiagnosed due to its multisystem presentation [
54]. The condition is not familiar to a majority of clinicians, and patients frequently undergo referral to several different specialties over a protracted period of time before a diagnosis is achieved. This work should increase clinical awareness of the clinical heterogeneity and atypical presentations of MNGIE, thereby reducing diagnostic delay and improving patient care since management of MNGIE requires the coordinated effort of different clinical specialties. Genetic counseling is fundamental in this autosomal recessive disease and prenatal diagnosis should be proposed, with a 25% risk for offspring of carrier parents to be affected.
A few mice models of MNGIE syndrome have been generated over the last past 20 years [
56‐
58]. They were used to study if the pathogenesis of MNGIE involving mitochondrial DNA defects could be attributable to aberrant thymidine metabolism [
56], to characterize the biochemical, genetic and histological features of MNGIE [
57], and to study the role of deoxynucleoside accumulation in the pathogenesis of MNGIE [
58]. Several limitations were encountered when using these animal models. In the mouse, thymidine is not only phosphorylated by thymidine phosphorylase, but also by uridine phosphorylase 1 and uridine phosphorylase 2. In contrast, in the human, thymidine is solely metabolized by thymidine phosphorylase. To partially circumvent this problem, authors established murine models based on double KO of
Tymp−/− and
Upp1−/− genes, but to the best of our knowledge, the triple KO has not been generated to date. Consequently, KO animals have a 10-fold increase in plasma thymidine and deoxyuridine, compared to a more than 100-fold increase in the human. Mice models only display minor cerebral signs, but no gastrointestinal or skeletal muscle involvement. This might be explained by the lower increase in deoxyribonucleoside levels, by the fact that mice may not live long enough to accumulate sufficient mtDNA damage, or by a potentially stronger impact of deoxyribonucleoside imbalance in humans. In order to recreate the phenotype of MNGIE in this model the exogenous administration of thymidine and deoxyuridine by dietary supplementation is required [
58].
TP was originally identified as platelet-derived endothelial cell growth factor (PDECGF), an angiogenic factor [
59]. TP, which is found in a wide range of normal tissues, plays an important role in angiogenesis [
60], as well as inflammation [
61,
62]. The protein is highly expressed in many types of cancers, including the lung and breast [
29,
63,
64]. Nevertheless, more than 30 years after the identification of the TP protein, its multifaceted role is far from being elucidated. In addition, tissue-specific models of MNGIE relevant to organs affected in this syndrome are scarce [
65], and besides the nervous and enteric system, the cellular consequences of the lack of TP deficiency are not well addressed. In particular, there is no data on the role of
TYMP in the adipose tissue and we are the first to demonstrate that TP is present in adipocytes. CRISPR-Cas9 KO of TP in ASC led to a major defect in adipocyte differentiation and function, with a major decrease in intra-cellular lipid levels, triglyceride content, and decreased expression of adipogenesis and mature adipocyte markers. Insulin signaling was also altered, even in pre-adipocytes. These data are consistent with the lipoatrophic and insulin-resistant phenotype of the patients investigated herein. Such an adipocyte differentiation defect has been reported in other lipoatrophic diabetes of various genetic origins [
66‐
68]. Lipoatrophic diabetes are indeed characterized by an incapacity of adipose tissue to store triglycerides, leading to ectopic fat depots and severe insulin resistance. The profound serum leptin and adiponectin deficiency observed in patients further confirms an endocrine defect of adipose tissue, since these hormones are secreted by mature adipocytes.
What is the cellular link between the loss of TP activity and adipogenesis defect? The current study demonstrates the key role of TP for mitochondrial homeostasis and associated adipocyte differentiation and functions. TP KO in ASC induce oxidative stress with a high rate of ROS production, associated with altered mitochondrial respiration. Recent reports emphasize the importance of mitochondria in white adipose tissue biology. In addition to their crucial role in energy homeostasis, mitochondria are the main site of ROS generation. When moderately produced, ROS function as physiological signaling molecules and promote adipocyte differentiation. Primary human mesenchymal stem cells undergoing differentiation into adipocytes indeed display an early increase in mitochondrial metabolism, biogenesis, and ROS generation [
69]. However, under different stress conditions, mitochondrial ROS overproduction induces the expression of adipogenic repressors and inhibits adipocyte differentiation [
32]. Balanced mROS production is thus at the core of proper metabolic maintenance, and unbalanced mROS production appears as an important trigger of metabolic disorders [
70]. Our study provides an additional illustration to these basic research data by the description of a monogenic disease known to induce an accumulation of mitochondrial DNA mutations and leading to major adipocyte dysfunction. In addition to its deleterious effect on adipocyte differentiation, emerging evidence points to ROS overproduction as a direct cause of insulin resistance. As an example, a previous study has shown that adipocytes subjected to oxidative stress induced by chemical agents exhibited an insulin-resistant state [
71]. In the same line and more generally, enhanced ROS production is observed in insulin resistance states and clinical diabetes [
72]. Hyperglycemia indeed increases the mitochondrial electron transfer chain’s activity until saturation, finally resulting in disruption and uncoupling of several key oxidative reactions and excessive ROS production [
72]. This is consistent with our data showing that TP KO undifferentiated ASC do not respond properly to insulin stimulation.
The causal link between ROS, cellular senescence, aging, and senescence-associated pathologies is intensely studied [
33]. It is widely assumed that ROS produced by mitochondria are involved in senescence [
73‐
76]. Therefore, we might speculate that the cellular senescence observed in TP KO cells is partly due to mitochondrial dysfunction and ROS overproduction, though direct evidence is lacking. In addition, cellular senescence has already been reported in a few cellular models of lipodystrophic syndromes or in cells from certain patients with LD [
34,
35,
77,
78]. More generally, increased cellular senescence has been functionally linked to fat-related metabolic dysfunction, which is underlined by the data of the current study [
36].
The role of the mitochondria in monogenic forms of diabetes is beginning to be better understood. In recent years, a few genes involved in monogenic lipoatrophic diabetes and encoding proteins playing a key role for mitochondrial function have been identified. As an example,
MFN2 encoding mitofusin 2 has been involved in the lipodystrophic Launois-Bensaude syndrome [
67,
79,
80], and
SLC25A24 encoding a calcium-binding mitochondrial carrier protein has been involved in a complex progeroid syndrome with lipoatrophy [
81]. These monogenic lipoatrophic diabetes have to be distinguished from the so-called mitochondrial diabetes due to molecular defects in mitochondrial DNA. Mitochondrial diabetes usually leads to dysfunction of pancreatic islet beta-cells due to their poor ability to resist oxidative stress induced by mitochondrial chain dysfunction and to a consequent defect in insulin secretion [
82,
83]. This situation is different from that of the patients investigated herein, who displayed severe insulin resistance and a major demise in adipose tissue, thereby underlining the key role of mitochondrial homeostasis for proper adipose tissue function. The spectrum of genes playing a key role for mitochondrial function and involved in adipose tissue dysfunction should therefore continue to expand in the near future.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.