page 1 of 46 diabetes · page 1 of 46 diabetes,* diabetes publish ahead of print, published online...

1

ADIPOCYTE ATP-BINDING CASSETTE G1 PROMOTES TRIGLYCERIDE STORAGE, FAT MASS

GROWTH AND HUMAN OBESITY

Eric FRISDAL1,2,3

, Soazig LE LAY4, Henri HOOTON

2,6, Lucie POUPEL

3, Maryline OLIVIER

1,2, Rohia

ALILI2,3,6

, Wanee PLENGPANICH1,8

, Elise F. VILLARD1,2,3

, Sophie GILIBERT1,2,3

, Marie LHOMME3,

Alexandre SUPERVILLE1,2,3

, Lobna MIFTAH-ALKHAIR1, M. John CHAPMAN

1,2, Geesje M. DALLINGA-

THIE5, Nicolas VENTECLEF

2,3,6, Christine POITOU

2,3,6,7, Joan TORDJMAN

2,3,6, Philippe LESNIK

1,2,3, Anatol

KONTUSH1,2,3

, Thierry HUBY1,2,3

, Isabelle DUGAIL2,3,6

, Karine CLEMENT 2,3,6,7

, Maryse GUERIN1,2,3

and

Wilfried LE GOFF1,2,3

1- INSERM, UMR_S1166, Team 4, F-75013 Paris, France;

2- Université Pierre et Marie Curie-Paris6, F-75005 Paris, France;

3- Institute of Cardiometabolism and Nutrition (ICAN), Pitié-Salpêtrière hospital, F-75013 Paris, France;

4- INSERM, U1063, F-49933 Angers, France;

5- AMC Amsterdam, Laboratory of Vascular Medicine, Amsterdam, The Netherlands;

6- INSERM, U872, Nutriomique team 7, Cordeliers Research Center, F-75006, Paris, France;

7- Assistance-Publique Hôpitaux de Paris, Heart and Metabolism, Pitié-Salpêtrière hospital, F-75013, Paris,

France;

8- King Chulalongkorn Memorial Hospital, Thai Red Cross Society, Patumwan, Bangkok 10330, Thailand.

Word count: 6610

Key Words: ABCG1, LPL, PPARγ, RNAi, adipocyte, obesity

*Address correspondence to: Wilfried LE GOFF, PhD. INSERM UMR_S1166 Team 4 : Integrative Biology of Atherosclerosis Hôpital de la Pitié Pavillon Benjamin Delessert 83, boulevard de l’Hôpital 75651 Paris Cedex 13 France Tel: +33 1 42 17 79 77 Fax: +33 1 45 82 81 98 email: [email protected]

of 46 Diabetes

,*

Diabetes Publish Ahead of Print, published online September 23, 2014

2

Abstract.

The role of ATP-binding Cassette G1 (ABCG1) transporter in human pathophysiology is still

largely unknown. Indeed, beyond its role in mediating free cholesterol efflux to HDL, ABCG1

transporter equally promotes lipid accumulation in a triglyceride (TG)-rich environment through

regulation of the bioavailability of Lipoprotein Lipase (LPL).

As both ABCG1 and LPL are expressed in adipose tissue, we hypothesize that ABCG1 is

implicated in adipocyte TG storage and could be then a major actor in adipose tissue fat accumulation.

Silencing of Abcg1 expression by RNAi in 3T3-L1 preadipocytes compromised LPL-dependent

TG accumulation during initial phase of differentiation. Generation of stable Abcg1 Knockdown 3T3-L1

adipocytes revealed that Abcg1 deficiency reduces TG storage and diminishes lipid droplet size

through inhibition of Pparγ expression. Strikingly, local inhibition of adipocyte Abcg1 in adipose tissue

from mice fed a high fat diet led to a rapid decrease of adiposity and weight gain. Analysis of two

frequent ABCG1 SNPs (rs1893590 (A/C) and rs1378577 (T/G)) in morbidly obese individuals indicated

that elevated ABCG1 expression in adipose tissue was associated with an increased PPARγ

expression and adiposity concomitant to an increased fat mass and BMI (haplotype AT>GC). The

critical role of ABCG1 regarding obesity was further confirmed in independent populations of severe

obese and diabetic obese individuals.

For the first time, this study identifies a major role of adipocyte ABCG1 in adiposity and fat

mass growth and suggests that adipose ABCG1 might represent a potential therapeutic target in

obesity.

of 46Diabetes

3

Introduction.

The ATP-binding cassette G1 (ABCG1) transporter has been proposed to promote cellular

cholesterol efflux to HDL (1) and targeted disruption of Abcg1 was shown to induce massive tissue

neutral lipid accumulation in mice fed a high-fat/high-cholesterol diet (2). However the precise role of

ABCG1 is still matter of debate, especially in human pathophysiology (3).

We recently reported that two frequent ABCG1 SNPs (rs1893590 and rs1378577) were

significantly associated to plasma lipoprotein lipase (LPL) activity in the Regression Growth Evaluation

Statin Study (REGRESS) population (4). Analysis of the relationship between ABCG1 genotype and

LPL led us to propose a mechanism by which ABCG1 controls macrophage LPL activity through

modulation of membrane lipid rafts to promote intracellular lipid accumulation and foam cell formation

in a triglyceride (TG)-rich context (4). Thus, beyond a role in sterol export to HDL, ABCG1 may equally

contribute to intracellular fatty acid accumulation and lipid storage in metabolic situations associated

with elevated levels of circulating TG-rich lipoproteins. Consistent with a role of Abcg1 in lipid storage,

random insertion of modified transposable elements of the P-family in Drosophila melanogaster

identified the CG17646 locus, the Drosophila ortholog of Abcg1, as a candidate gene for TG storage

(5). Moreover, total ablation of Abcg1 in mice fed a high fat diet devoid of cholesterol (5) reduced TG

accumulation in the adipose and liver tissues. However the cellular mechanisms underlying this

phenotype, and more specifically the tissue-specific contribution of Abcg1, were not elucidated.

Considered together, these data prompted us to evaluate the function of ABCG1in adipocytes which

are professional cells for TG storage.

ATP-Binding Cassette G1 is expressed in adipocytes and in adipose tissue of mice, which

develop diet-induced obesity (5; 6). Moreover, adipose tissue is a major source of LPL (7) which

critically controls TG accumulation by generating free fatty acids from circulating lipoproteins (8).

Our data demonstrate that silencing of Abcg1 expression in adipocyte reduced LPL activity

and alters lipid homeostasis. Moreover, Abcg1 deficiency resulted in inhibition of Pparγ expression

and alteration of adipocyte maturation. In vivo, local lentiviral-mediated adipose tissue targeting of

Abcg1 rapidly reduced adiposity and high-fat diet-induced weight gain in mice. More strikingly, we

observed that ABCG1 genotype in humans was associated to fat mass formation and obesity in

independent populations of obese individuals, thereby highlighting the critical role of ABCG1 in the

of 46 Diabetes

4

context of human obesity. Taken together, the present study suggests that adipose tissue ABCG1

might represent a future therapeutic target in metabolic disorders associated to obesity.

of 46Diabetes

5

Research Design and Methods.

Morbidly obese population.

Middle-aged (42.0 ± 0.04 years) morbidly obese patients (n=1320; BMI = 45.5 ± 0.07 Kg/m2) of

Caucasian origin (Sex ratio M/F = 0.33) were recruited at the Department of Nutrition at the Pitié-

Salpêtrière hospital, Paris, France (9). Patients were phenotyped for a series of bioclinical variables.

Body composition in this population was evaluated by biphotonic absorptiometry (DXA) as described

(10). All subjects gave their informed written consent to participate in the genetic study (Clinical

Research Contracts), which was approved by the local ethic committee.

The severe obese and diabetic obese populations are described in Supplementary Materials.

In a subset of patients’ candidate for bariatric surgery (ABCG1 AT haplotype), subcutaneous adipose

tissue pieces were sampled, after an overnight fast, in the s.c. peri-umbilical by needle biopsy under

local anesthesia (1% xylocaïne). Biopsies were washed and stored in RNA Later preservative solution

(Qiagen) at −80°C until analysis. Total RNA was extracted from adipose tissue biopsies using the

RNeasy total RNA minikit (Qiagen). Total RNA concentration and quality was confirmed using the

Agilent 2100 bioanalyzer (Agilent Technologies).

Genotyping.

Genotyping of ABCG1 SNPs (rs1893590 and rs1378577) was performed using TaqMan® SNP

genotyping assay (Applied Biosystems). Hardy-Weinberg equilibrium was respected for both ABCG1

SNPs in the obese populations studied.

Culture and differentiation of adipocytes.

The 3T3-L1 preadipocytes (Dr J. Pairault, Paris) were maintained in Dulbecco’s modified Eagle

medium (DMEM) supplemented with 10% calf serum and 2 mM glutamine. Differentiation of confluent

preadipocytes was initiated with 0.25 µM insulin, 1.25 µM dexamethasone and 250 µM 3-isobutyl-

methyl-1-xanthine) in DMEM (4.5 g/L glucose) supplemented with 10% FBS. After 3 days, the culture

medium was switched to DMEM (4.5 g/L glucose) supplemented with 10% FBS and 100 nM Insulin for

2 days. Then, 3T3-adipocytes were allowed to differentiate in DMEM (4.5 g/L glucose) containing 10%

FBS, replaced every other day for 15 days.

of 46 Diabetes

6

Silencing of Abcg1 expression was performed by application of siRNA oligonucleotides (Dharmacon)

targeted to the cDNA sequence of mouse Abcg1 gene (NM_009593). Transfection of 3T3-L1

preadipocytes and differentiated adipocytes with siRNA was achieved using the Nucleofector

technology (Lonza) according to the manufacturer’s protocol. For each experiment, 2 x 106 cells and

100 pmol siRNA were diluted in 100 µl of V solution and processed with A-033 program.

Human preadipocytes (Promocell, Heidelberg, Germany) were cultured and differentiated as

recommended by the manufacturer. Differentiation efficiency was validated by quantifying the

induction of adipocyte marker mRNA levels (ADIPOQ, LEP and PPARγ).

Generation of stable Abcg1 Knockdown 3T3-L1 adipocytes.

Control shRNAs and validated oligonucleotides encoding shRNAs targeting the cDNA sequence of

mouse Abcg1 gene (NM_009593) (R1 sense: 5’- GAT CCC CGG AAA GGT CTC CAA TCT CGT TCA

AGA GAC GAG ATT GGA GAC CTT TCC TTT TTG GAA A-3’ and R1 antisense: 5’- AGC TTT TCC

AAA AAG GAA AGG TCT CCA ATC TCG TCT CTT GAA CGA GAT TGG AGA CCT TTC CGG G-3’;

R2 sense: 5’-GAT CCC CGA GAA GAC CTG CAC TGC GAT TCA AGA GAT CGC AGT GCA GGT

CTT CTC TTT TTG GAA A-3’ and R2 antisense: 5’-AGC TTT TCC AAA AAG AGA AGA CCT GCA

CTG CGA TCT CTT GAA TCG CAG TGC AGG TCT TCT CGG G-3’) were annealed and cloned into

pSUPER as previously described (11). The shRNA expression cassette was then transferred into the

XhoI/EcoRI site of the pRVH1-puro retroviral vector and recombinant knockdown viruses were

generated using the human Phoenix gag-pol packaging cell line (obtained from the National Gene

Vector Biorepository, Indianapolis) as previously described (12). Preadipocytes 3T3-L1 were plated in

6-well plates (1 x 105 cells per plate) in DMEM supplemented with 10% calf serum. After 48 hours,

media was aspirated and cells were infected with 1mL of either supernatant from Phoenix cells

containing control (R-Ctrl) or Abcg1 KD (R1 and R2) retroviral particles supplemented with 4 µg/mL of

hexadimethrine bromide (Polybrene, Sigma) or lentiviral particles expressing control shRNA (L-Ctrl) or

shRNAs targeting the cDNA sequence of mouse Abcg1 gene (NM_009593) (L1, L2, L3) (Sigma).

Selection of virus-transduced 3T3-L1 preadipocytes was achieved by incubation with 4 µg/mL

puromycin (Invitrogen) for 6 days. Stable Ctrl and Abcg1 KD 3T3-L1 clones were then trypsinized and

reseeded into DMEM (4.5 g/L glucose) supplemented with 10% FBS and 4 µg/mL puromycin

(Invitrogen) and differentiated into adipocytes as described above.

of 46Diabetes

7

RNA extraction, reverse-transcription and quantitative-PCR.

Total RNA from cell culture or tissues were extracted using the NucleoSpin RNA II kit (Macherey-

Nagel) or TRIzol reagents (Euromedex), respectively, according to the manufacturer's instructions.

Reverse transcription and real-time qPCR using a LightCycler LC480 (Roche) were performed as

previously described (13). Expression of mRNA levels was normalized to human non-POU domain

containing, octamer-binding housekeeping gene (NONO), human α-tubulin (TUBA) and human heat

shock protein 90kDa alpha (cytosolic), class B member 1 (HSP90AB1) or mouse hypoxanthine

phosphoribosyltransferase 1 (Hprt1), mouse non-POU domain containing, octamer-binding

housekeeping gene (Nono), mouse heat shock protein 90kDa alpha (cytosolic), class B member 1

(Hsp90ab1), mouse cyclophilin A (CycA), mouse beta-glucuronidase (Gusb) and mouse 18S

ribosomal RNA (18S rRNA). Data were expressed as a fold change in mRNA expression relative to

control values.

Adipocyte diameter measurements.

Adipose tissue pieces were minced and immediately digested by 200 µg/mL collagenase (Sigma) for

30 min at 37°C. For cell size measurements, adipocyte suspensions were then visualized under a light

microscope attached to a camera (TriCCD, Sony, France) and computer interface. Adipocyte

diameters were measured by using PERFECT IMAGE software (Numeris, Nanterre, France). Mean

adipocyte diameter and volume were defined as the median value for the distribution of adipocyte

diameters of at least 250 cells by the same investigator.

Quantification of apoE secretion. Secreted apoE in the culture media of 3T3-L1 adipocytes were

quantified by ELISA (Cloud-Clone Corp., Houston, USA) according to the manufacturer’s instructions

and normalized to cell protein levels.

Lipoprotein lipase activity measurement and cellular lipid quantification.

LPL activity was determined with a 50-µl aliquot of culture medium using a LPL activity assay kit

(Roar, New York, USA) according to the manufacturer’s instructions. Intracellular total lipase activity

was measured using a lipase activity assay kit III (Sigma, Saint-Quentin Fallavier, France). Results

of 46 Diabetes

8

were normalized to cell protein levels. When indicated, 3T3-L1 adipocytes were incubated for 16 hours

at 37°C with either 250 µM sphingomyelin (from chicken egg yolk, Sigma) or with 3 U/mL

sphingomyelinase (from Staphylococcus aureus, Sigma) in order to enrich or deplete membranes in

sphingomyelin, respectively, or with 1 mM methyl-β-cyclodextrin (in PBS, Sigma) to remove free

cholesterol at the plasma membrane. Quantification of cellular triglyceride, total and free cholesterol

mass was performed as previously described (14).

Quantification of sphingomyelin mass. Control and Abcg1 SKD adipocytes (D10) were incubated

for 16 hours at 37°C with 0.2% bovine serum albumin (endotoxin-, free fatty acid-free) as an efficient

acceptor for Abcg1-mediated sphingomyelin (SM) efflux (15). Media were collected and cells were

washed extensively with cold PBS. Extraction was adapted from Ivanova et al. (16). In brief, cells (30

µg of enzymatically-quantified phospholipids) were supplemented with 3.2 ml of methanol acidified

with 0.1N HCl (1:1 v/v) containing 80 ng of phosphatidylcholine (PC) 16:0/16:0 d9 and 3 ng of

lysophosphatidylcholine (LPC) 15:0. Cell media (600µl) were supplemented with 1 ml of

methanol/0.1N HCl (1:1 v/v), 66 µL of 1N HCl and 660 µL methanol containing 80 ng of PC 16:0/16:0

d9 and 3 ng of LPC 15:0. The mixtures were vortexed for 1 min. Blank and control samples were

extracted in parallel with each batch to ensure quality control; each sample was corrected for blank

readings. Chloroform was added to cells (800 µl) and media (1.16 ml), and the mixtures were vortexed

for 1 min and centrifuged at 3,600 g for 10 min at 4°C. Lower organic phases were dried under

nitrogen, resuspended and transferred into LC/MS amber vials with inserts. LC/MS analysis and SM

quantification were performed as previously described (17). The percentage of SM efflux was

calculated as 100 x (medium SM) / (medium SM + cell SM).

Free cholesterol efflux assays.

Differentiated 3T3-L1 adipocytes were incubated for 24 h with [3H]-cholesterol-labeled (1 µCi/mL) SVF

(10%) in DMEM (4.5 g/L glucose) media. Then, the labeling medium was removed and cells were

equilibrated in serum-free media containing 0.2% BSA for an additional 16 h period. Cellular free [3H]-

cholesterol efflux to 20µg/mL free apoAI (Sigma) or 30 µg/mL of HDL-PL was assayed in serum-free

media containing 0.2% BSA for a 4-hour chase period. Finally culture media were harvested and

cleared of cellular debris by a brief centrifugation. Cell radioactivity was determined by extraction in

of 46Diabetes

9

hexane-isopropanol (3:2), evaporation of the solvent and liquid scintillation counting (Wallac Trilux

1450 Microbeta). The percentage of cholesterol efflux was calculated as 100 x (medium cpm) /

(medium cpm + cell cpm).

Quantification of lipid rafts.

Lipid rafts in differentiated adipocytes was detected following incubation with 1µg/mL Alexa Fluor 594–

conjugated cholera toxin subunit-β (CT-B) (Molecular Probes) for 15 minutes at 4°C as previously

described (4; 6). Indeed, binding of CT-B to the pentasaccharide chain of plasma membrane

ganglioside GM1 which selectively partitions into lipid rafts (18), allows a reliable detection of lipid rafts

in live cells. After washing twice with cold PBS, cells were detached from plates with trypsin and

subjected to flow cytometry analysis on LSR II FORTESSA SORP (BD Biosciences). When indicated,

images were captured using a Zeiss Axio Imager.M2 microscope with a 63x objective.

Injection of siRNA targeting ABCG1 expression in adipose tissue in vivo.

Four-week aged male C57BL/6 mice (Janvier, Le Genest Saint Isle, France) were fed on a high fat

diet (45% fat, Brogaarden Diet#TD12451) for 4 weeks before the day of injection. At the day of

injection, mice were weighted and anesthetized with isoflurane and maintained under anesthesia

during the surgical procedure. A sub-abdominal incision was operated and epididymal fat pads was

injected with 100µl of lentiviral particles (1.4x105 lentiviral transducing particles per milliliter) encoding

either a short-hairpin RNA (shRNA) designed to knock down mouse Abcg1 expression (Santa Cruz) or

control shRNA lentiviral particles encoding a shRNA that will not lead to the degradation of any known

cellular mRNA (Santa Cruz) using a 30 gauge needle. Cell targeting of lentiviral particules into adipose

tissue was visualized by injection of copGFP control lentiviral particles (Santa Cruz). Dispersion of the

injected volume in the whole organ by this procedure was validated using a colored dye in preliminary

experiments (Supplementary Figure 2A). Then injected epididymal fat pads were replaced in the sub-

abdominal cavity and the incision was sutured. Mice were fed for an additional 4-week period on a

high fat diet (60% fat, Brogaarden Diet#TD12492) until the day of sacrifice. Food intake was monitored

for a 3-day period and locomotion and activity was monitored over a 24 hours period using the

Activmeter actimetry system (Bioseb, Chaville, France). At the day of sacrifice, blood samples were

collected in Microvette tubes (Sarstedt) by retro-orbital bleeding under isoflurane anesthesia. Mice

of 46 Diabetes

10

were weighted, euthanized and epididymal fat pads were isolated for RNA extraction,

immunohistochemistry analysis and adipocyte diameter measurements. Liver and intestine were

collected, weighed, and flash frozen. Plasma samples were analyzed with an Autoanalyzer (Konelab

20) using reagent kits from Roche (total cholesterol), Diasys (free cholesterol, free fatty acids),

ThermoElectron (HDL-cholesterol, triglycerides, glucose).

Adipose tissue cell sorting.

Epididymal adipose tissue was excised from mice, minced and digested in Hank’s balanced salt

solution (HBSS, Gibco, Invitrogen, Cergy Pontoise, France) with 2.5 mg/mL collagenase D (Roche,

Boulogne Billancourt, France) for 30 min at 37°C and dissociated through a 200 µm pored cell strainer

(Franklin lakes, NJ, USA). After decanting, adipocytes (supernatant) were washed with a 10% sucrose

solution and used for subsequent analyses. Stromal vascular fraction cells (SVF, bottom) were

suspended in cold HBSS containing 3% foetal bovine serum and centrifuged at 1,500 rpm for 5

minutes. Recovered cells (200 µl) were stained with 1 µg/mL purified anti-CD16/32 (Becton Dickinson,

Franklin lakes, NJ, USA) for 10 minutes at 4°C and for an additional 30 minutes with the appropriate

dilution of specific antibodies. The panel of antibodies used was: anti-CD45 (clone 30-F11), anti-

CD11b (clone M1/70), anti-F4/80 (clone BM8), anti CD64 (clone 290322), anti-CMH II (clone M5/114)

and anti-CD31 (clone 390). Propidium iodide (PI) was used as a viability marker. Adipose tissue

macrophages (ATM) were defined as PI-CD45+CD11b+F4/80+CD64+CD31- and endothelial cells as

PI-CD31+. Cells were sorted using the MoFlo Astrios (Beckman Coulter, Villepinte, France) using

Summit acquisition software and stored in RLT buffer (QIAgen, Courtaboeuf, France) at -80°C until

used.

Immunofluorescence.

A portion of epididymal adipose tissue was fixed in 10% formalin overnight at 4°C before being

embedded in paraffin. Five-micrometre-thick paraffin tissue sections were dewaxed with xylene and

graded ethanol, and antigen unmasking was performed by heating the sections in 10 mmol/l citrate

buffer, pH 6.0, at 750W for 15 min in a domestic microwave. Then, sections were washed twice in

phosphate-buffered saline and saturated with 3% bovine serum before staining with primary antibodies

against: ABCG1 (Novus, Littleton, USA), Perilipin (Progen, Heidelberg, Germany). Alexa Fluor 568-

of 46Diabetes

11

conjugated anti-guinea-pig or Alexa Fluor 568 and 488-conjugated anti-rabbit were used as secondary

antibodies (Life Technologies, Saint Aubin, France). Immunostained sections were examined on a

Zeiss Axio Imager.M2 microscope. Microscopy images were captured using AxioCam digital

microscope cameras and AxioVision image processing (Carl Zeiss Vision, Germany). The specificity of

antibodies was tested with their isotype controls.

Western blot analysis.

Cell proteins were extracted using 200µL M-PER reagent (Pierce) containing protease inhibitors and

were subsequently separated on a 4-12% Bis-Tris gel (Invitrogen). Proteins (25 µg per lane) were

transferred to nitrocellulose and the membrane was blocked with Casein blocker solution for 1h.

Membranes were then incubated overnight at 4°C with a rabbit anti-Abcg1 (NB400-132; Novus), or

anti-Pparγ (C26H12) (2435, Cell Signaling), or anti-Fabp4 (2120, Cell Signaling) or with guinea pig

anti-perilipin (GP29, Progen) antibody diluted at 1:500 and revealed with either IRDye 800CW-

conjugated goat anti-rabbit or donkey anti-guinea pig (Li-Cor) at 1: 10000 for 1 hour. Detection was

performed using an Odyssey infrared imaging system (Li-Cor).

Statistical analysis.

Linkage disequilibrium between both SNPs was calculated with Haploview 4.1. Associations between

phenotypes and genotypes were tested with multivariate linear regression models. All phenotypes

were transformed to log10 before testing for associations. Genotype-phenotype association tests were

performed with R 2.8.2. Association between haplotypes and phenotypes were performed with

Hapstat 3.0. All models were adjusted for age and sex; models testing associations with CRP, IL6 and

Adiponectin were also adjusted for BMI. Finally, models testing associations with triglycerides, total

cholesterol, HDL, Lp(a), ApoA1 and ApoB were also adjusted for BMI and medical treatment for

dyslipidaemia. HOMA index was calculated using the Homa2 method

(http://www.dtu.ox.ac.uk/Homacalculator/index.php) which led to the calculation of three different

HOMA (HOMA2S, HOMA2B and HOMAIR).

Data are shown as mean ± SEM. Experiments were performed in triplicate and values correspond to

the mean from at least three independent experiments. Comparisons of 2 groups were

performed by a

2-tailed Student’s t test and comparisons of 3 or more groups were performed by ANOVA with

of 46 Diabetes

12

Newman–Keuls post-test. All statistical analyses were performed using Prism software from GraphPad

(San Diego, CA, USA)

of 46Diabetes

13

Results.

RNAi-mediated Abcg1 targeting decreases LPL-dependent cell triglyceride storage in

preadipocytes.

Analysis of Abcg1 expression during the course of differentiation of 3T3-L1 preadipocytes into

mature adipocytes indicated that Abcg1 mRNA levels were increased by ~40-fold during adipocyte

differentiation (Figure 1A). In agreement with a role for Abcg1 in TG storage, the accumulation of TG

was significantly correlated with Abcg1 mRNA levels (Figure 1B; r2=0.9, p<0.0001).

We previously reported that macrophage ABCG1 promotes TG storage by modulating LPL

activity (4). In order to test whether the same mechanism is operative in adipocytes, Abcg1 expression

was silenced in 3T3-L1 preadipocytes using specific siRNAs which did not alter adipocyte

differentiation and viability. Inhibition of Abcg1 expression at both mRNA and protein levels in 3T3-L1

preadipocytes (Figure 1C-D) was without effect on Lpl mRNA expression (Fig. 1E). We observed a

marked reduction of LPL activity (-81%, p<0.05) however in media from Abcg1 Knockdown (KD)

adipocytes (Fig. 1F) as compared to control cells. Decreased LPL activity was moreover associated

with changes in cell surface membrane properties, as the binding of Cholera toxin Beta subunit, which

associates with lipid rafts preferentially, increased in cells transfected with Abcg1 RNAi as visualized

by fluorescence microscopy and quantified by flow cytometry (Figure 1G; +24%, p<0.05).

Although the silencing of Abcg1 in preadipocytes (Abcg1 KD) was initiated prior to the addition

of the adipocyte differentiation cocktail (Day 0, D0), a marked reduction in intracellular TG

accumulation occurred during subsequent adipocyte conversion (Figure 1H; -22%, p<0.05 after 4 days

of differentiation). Remarkably, addition of tetrahydrolipstatin (THL), an inhibitor of LPL activity over a

period of 24h, compromised TG accumulation in control cells, indicating LPL-dependency of TG

secretion in preadipocytes (Fig. 1I, Day 6). Conversely, addition of increasing amounts of exogenous

recombinant bovine LPL (bLPL) led to dose-dependent elevations in intracellular TG mass in control

3T3-L1 preadipocytes that paralleled those of LPL activity in culture media (Fig. 1I, Day 6).

Together, these data indicate that TG accumulation during the initial phase of adipocyte

differentiation is LPL-dependent and is compromised by Abcg1 invalidation.

of 46 Diabetes

14

Impaired maturation in stable Abcg1 knockdown 3T3-L1 adipocytes.

To explore the impact of prolonged inhibition of Abcg1 expression on adipocyte maturation,

stable Abcg1 KD 3T3-L1 fully differentiated adipocytes were generated using either lentiviral or

retroviral particles expressing shRNAs targeted to distinct regions of the Abcg1 mRNA (Figure 2).

Stable knockdown of Abcg1 (Abcg1 SKD), either by lentiviruses (L1 to L3 vs L-Ctrl) or by

retroviruses (R1-R2 versus R-Ctrl), led to a marked reduction in Pparγ, Fabp4, C/ebpα, Perilipin, Cd36

and Hsl mRNA expression (Fig. 2B-G). The expression of some others genes, including Fas or

C/ebpβ, wich are also known to participate in adipocyte differentiation, remained unaffected. To note

that expression of Abca1 in adipocyte was very recently reported to influence adipocyte lipid

homeostasis (19). This observation may be important since Abcg1 deficiency in mouse macrophages

was proposed to be compensated by an increase of Abca1 expression (20). However such elevation

of Abca1 expression was not observed in our conditions when Abcg1 was knocked down in 3T3-L1

adipocytes (Fig. 2J). The reduction of Pparγ, Perilipin and Fabp4 expression was confirmed at the

protein level (Figure 3A-E). Strikingly, lipid accumulation was reduced markedly in Abcg1 SKD

adipocytes as compared to stable 3T3-L1 control cells (Fig. 3F) and was confirmed by decrease in

lipid droplet diameter (-46%, p<0.05; Fig. 3G) and lower intracellular TG storage (-45%, p<0.005; Fig.

3H). Moreover, 24h-treatment with exogenous bLPL partially rescued TG storage in Abcg1 SKD

adipocytes, while total restoration was observed when bLPL was added throughout the course of

adipocyte maturation (from D0 to D10), (Fig. 3I).

Consistent with the well-established role of Abcg1 in cholesterol transport, Abcg1-deficient

adipocytes exhibited reduced capacity to promote free cholesterol efflux to HDL (-31%, p<0.005, Fig.

3J), even if cell free cholesterol mass was lowered (-60%, p<0.005; Fig 3K). Such a decrease in

intracellular cholesterol levels appeared to be accompanied by increased mRNA amounts of genes

involved in cholesterol synthesis in Abcg1 SKD adipocytes (Supplementary Figure 1A-D).

Accumulation of sphingomyelin in stable Abcg1 knockdown adipocytes reduces LPL-

dependent TG storage.

ABCG1 was reported to promote export not only of free cholesterol but also phospholipids,

such as sphingomyelin (SM) (15) which was described to inhibit LPL activity (21; 22). Therefore, we

next address the hypothesis that an accumulation of SM, found in large amounts in lipid raft domains,

of 46Diabetes

15

was responsible of the reduced LPL activity and the subsequent impaired TG storage in Abcg1 SKD

adipocytes. Quantification of SM mass by LC-MS revealed that stable Abcg1 knockdown in 3T3-L1

adipocytes was accompanied by a reduced SM efflux (-14%, p<0.0005; Fig. 4A) which led to a marked

increase of intracellular SM content (+48%, p<0.0005; Fig. 4B) in those cells in comparison to control

adipocytes. More strikingly, depletion of SM by sphingomyelinase in Abcg1 SKD adipocytes restored

LPL activity in a comparable level to that observed in control adipocytes (Fig. 4C) and promoted TG

storage (Fig. 4D). The contribution of SM in this mechanism was further strengthened by the

observation that enrichment of control adipocytes with SM led to an impaired TG storage similar to

Abcg1 SKD adipocytes (Fig. 4D). A treatment with 1 mM methyl-β-cyclodextrin for 16 hours which

removes cholesterol but not SM from plasma membrane (23; 24) was without effect on intracellular TG

levels in Abcg1 SKD adipocytes (Supplementary Figure 1E), suggesting that free cholesterol in lipid

rafts was not responsible for the reduced TG storage in those cells. Finally, because Abcg1 deficiency

may be associated to an increased apoE secretion (25), which may affect LPL activity (26), apoE

secretion from control and Abcg1 SKD adipocytes was examined. As shown in Supplementary Figure

1F, no difference in apoE secretion was detected in Abcg1 SKD adipocytes in comparison to control

cells.

Taken together, those results support the mechanism through which Abcg1 deficiency led to a

reduced sphingomyelin efflux and a concomitant increased SM content at the plasma membrane,

likely associated to lipid rafts, which decreases LPL activity and subsequent TG storage.

In vivo silencing of Abcg1 expression locally in adipose tissue attenuates fat storage upon

high-fat diet.

In order to further investigate the in vivo role of adipocyte Abcg1, adipose tissue Abcg1 was

silenced by local delivery of lentiviral particles encoding either control shRNAs (L-Ctrl) or Abcg1

shRNAs (L-Abcg1) in epididymal adipose tissue. Injected C57BL/6 mice were maintained on a high-fat

diet according to the experimental design presented on Figure 5A. Expression of Abcg1 in adipocytes

in adipose tissue from mice fed a high fat diet (W4) was visualized by immunofluorescence (Figure

5B). Injection of lentiviral particles expressing GFP alone into epididymal adipose tissue confirmed that

this strategy was efficient in targeting adipocytes from adipose tissue (W4, Figure 5C-D. In an

independent control experiment, a marked decrease of Abcg1 staining was observed in an epididymal

of 46 Diabetes

16

fat pad injected with L-Abcg1 four weeks following injection (W4) as compared to an epididymal fat

pad injected with L-Ctrl (Figure 5E), thus validating the knockdown of Abcg1 in adipose tissue. Such

reduced Abcg1 expression in L-Abcg1 fat pads was highly reproducible when tested in multiple tissue

samples treated simultaneously under identical experimental conditions. Quantification of adipose

tissue Abcg1 mRNA indicated that a significant reduction of Abcg1 expression (-45%, p<0.05; Fig. 5F)

was observed in injected fat pads from L-Abcg1 mice, which mostly reflected the specific silencing of

Abcg1 expression in adipocytes (-47%, p<0.0005) and in a lesser degree in adipose tissue

macrophages (ATM); the expression of Abcg1 in those cells being ~3-fold less abundant than in

adipocytes (Fig. 5G). No effect was observed in endothelial cells isolated from adipose tissue (Fig. 5G)

or in others organs such as intestine (Fig. 5H) or liver (Fig. 5I).

All mice were maintained on a high-fat diet for four weeks following injection and weight gain

was evaluated in the two groups (W4). L-Abcg1-injected mice gained less weight than L-Ctrl injected

mice (-24%, p<0.05, respectively, Fig. 6A) and had a decreased epididymal fat mass (-26%, p<0.05,

Fig. 6B), suggesting lower adipose tissue fat storage. Indeed, mean adipocyte diameter was

significantly smaller than that in epididymal adipose tissue from L-Ctrl mice (Fig. 6C). In agreement,

mRNA levels of leptin were decreased in adipose tissue from L-Abcg1 mice as compared to L-Ctrl

mice (Fig. 6D). Analysis of food intake (Fig. 6E) and locomotor activity (Fig. 6F-G) indicated that

reduced weight gain in locally injected mice did not result from overt alteration of energy balance.

Additional analysis of metabolic parameters in L-Abcg1 and L-Ctrl mice is presented in Supplementary

Table 1. The genes whose expression was downregulated in stable Abcg1 SKD mouse adipocytes

generated in culture were also strikingly affected in Abcg1-silenced fat pads, namely Pparγ, Perilipin,

Fabp4, Cd36, Hsl and C/ebpα (Fig. 6H-M) whereas that of C/ebpβ or Fas was not altered (Fig. 6N-O).

Expression of inflammatory and insulin-resistance genes in epididymal adipose tissue from L-Ctrl and

L-Abcg1 mice is shown in Supplementary Figure 2.

Taken together, these results support a critical role of Abcg1 in adipocyte lipid storage and

indicate that local inhibition of Abcg1 in adipose tissue of mice impaired fat storage under a high-fat

diet.

A higher expression of ABCG1 in human adipose tissue is associated with increased fat mass

and corpulence.

of 46Diabetes

17

We presently demonstrate that adipocyte Abcg1 contributes to TG storage and adiposity and

hypothesize that an elevated expression of ABCG1 might be associated with an increased fat mass in

obese subjects. To evaluate this question, we took the opportunity to examine functional ABCG1

SNPs and their association with adipose tissue gene expression and body composition in obese

individuals. Two frequent ABCG1 SNPs (rs1378577 and rs1893590) located in the human ABCG1

gene promoter were genotyped in a population of 1320 middle-aged morbid obese patients (BMI>40

Kg/m2; mean BMI = 45.5 ± 0.04 Kg/m

2). The relative allele frequencies for both ABCG1 SNPs

(rs1893590, -204A/C : -0.73/0.27 and rs1378577, -134T/G : 0.78/0.22) were similar to those observed

in the REGRESS cohort (4). As previously described (4), in vitro analysis of ABCG1 promoter activity

according to ABCG1 haplotypes confirmed that the frequent AT haplotype was associated with a

higher transcriptional activity than the rare CG haplotype (Figure 7A). Analysis of ABCG1 expression

in biopsies of adipose tissues isolated from a subset of obese patients displaying either the AT or CG

haplotypes revealed that mRNA levels of ABCG1 were 27% (p<0.05) more elevated in adipose tissues

from patients carrying the AT haplotype relative to those carrying the CG haplotype (Fig. 7B). It is to

note that ABCG1 expression in primary human preadipocytes was significantly induced upon

differentiation into adipocytes (Supplementary Figure 3A).

In agreement with our data in Abcg1-deficient adipocytes, levels of mRNAs coding for genes

involved in adipocyte differentiation (PPARγ , CD36, PLIN1) were increased in adipose tissue from

obese patients carrying the AT haplotype as compared to those carrying the CG haplotype (Fig. 7C-

E); by comparison those of LPL were unchanged (Fig. 7F). Adipose tissue mRNA levels of

macrophage markers were not different between patients carrying the AT haplotype and those

carrying the GC haplotype (Supplementary Figure 3B-D), suggesting that the macrophage cell number

present in adipose tissue was similar.

ABCG1 expression in adipose tissue from obese patients was positively correlated to the

adipocyte diameter (r=0.51, p=0.023, Fig. 7G). Importantly, obese individuals carrying the AT

functional haplotype displayed a significantly higher DXA-evaluated fat mass together with a lower fat-

free mass (Fig. 7H and 7I) (p<0.05) than those carrying the CG haplotype.

Moreover, the two ABCG1 SNPs were significantly associated with BMI; individuals carrying

the -204AA or the-134TT genotype displaying the highest BMI (p=0.0034 and p=0.011, respectively)

(Figure 7J). Haplotype analysis confirmed that the AT haplotype (-204A / -134T) was significantly

of 46 Diabetes

18

associated with BMI (p=0.006); moreover, BMI increased in parallel with increase in the amount of the

AT haplotype (Figure 7K). Association of both ABCG1 SNPs with BMI was still observed after

adjustment with diabetes or HOMA index (Supplementary Table 2). To note that ABCG1 SNPs

rs1378577 and rs1893590 were not associated with HOMA index after adjustment or not with BMI

(data not shown). In addition, obese individuals carrying the -134TT genotype (rs1378577, most

frequent genotype) also displayed the highest fat mass index (p=0.0242; Fig. 7L); this effect was

equally observed in subjects carrying the -204AA genotype (rs1893590, most frequent genotype)

(p=0.0331). Of note, Adiponectin and CRP levels were significantly associated with both ABCG1

SNPs in this population (Supplementary Figure 3). However no significant effect of ABCG1 SNPs on

plasma lipid and apolipoprotein levels was observed (Supplementary Table 3).

Thus elevated adipose expression of ABCG1 in obese individuals carrying the AT haplotype is

linked with increased fat cell size, fat mass and obesity and links ABCG1 genotype to obesity in

humans for the first time.

Association of ABCG1 genotype with obesity was replicated in two independent populations

with distinct grades of obesity composed of severe obese ((35<BMI<40 Kg/m2; mean BMI = 39.2 ± 8.5

Kg/m2) and diabetic obese subjects ((30<BMI<35 Kg/m

2; mean BMI = 30.9 ± 4.9 Kg/m

2), respectively.

Genotyping of both ABCG1 SNPs rs1378577 and rs1893590 in 216 type 2 diabetic obese subjects

(30<BMI<35 Kg/m2; mean BMI = 30.9 ± 4.9 Kg/m

2) from the Diabetes Atorvastatin Lipid Intervention

(DALI) Study (27) revealed that both ABCG1 SNPs were significantly associated with Waist Hip Ratio

(WHR) (Supplementary Fig. 4A-B), with individuals carrying the AT haplotype displaying a significant

higher WHR (p<0.005) than those carrying the CG haplotype (Supplementary Fig. 4C), confirming the

deleterious role of the AT haplotype regarding obesity. Finally, genotyping of the ABCG1 SNPs

rs1378577 in an independent population of 595 severe obese subjects (35<BMI<40 Kg/m2; BMI = 39.2

± 8.5 Kg/m2) (28) confirmed the association of ABCG1 genotype with both BMI and fat mass

(Supplementary Fig. 4D-E).

of 46Diabetes

19

Discussion.

In the present study, we unraveled an unexpected role for Abcg1 in the control of lipid

homeostasis in adipocytes. Indeed, through a mechanism involving modulation of both LPL activity

and regulation of Pparγ, Abcg1 appears to be a critical component in TG storage and adiposity. The

central role of Abcg1 in adipocyte is further strengthened following the targeted silencing of its

expression locally in adipose tissue in mice fed a high fat diet which led to a rapid reduction in

adiposity and weight gain. Consistent with these findings, ABCG1 is associated with adiposity, fat

mass and obesity in obese individuals.

These results may initially appear conflicting with respect to the widely described role of Abcg1

in cellular free cholesterol efflux, where its inhibition leads to increase rather than decrease in

intracellular lipid accumulation (2). Indeed, in this way, Abcg1 can protect the cell from the

accumulation of free cholesterol which is toxic for cell survival. In agreement with such a role for

Abcg1, we presently demonstrated that silencing of Abcg1 expression in differentiated 3T3-L1

adipocyte was accompanied by a significant reduction in free cholesterol efflux to HDL, thereby

indicating that this mechanism is also operative in fat cells. However, in contrast to plasma membrane

free cholesterol, the large content of free cholesterol associated with lipid droplets, and which is

closely proportional to TG storage in adipocytes and fat cell size, is not mobilized for efflux (29).

Mobilization of free cholesterol in lipid droplets therefore provides an alternative pathway for protection

of adipocytes from toxicity. The present study supports the notion that although the role of Abcg1 in

cellular free cholesterol efflux mechanisms is crucial in maintaining tissue lipid homeostasis in a

cholesterol-rich environment (2), Abcg1 contributes to cellular lipid, mostly triglycerides, accumulation

and storage in high fat-rich metabolic states (4; 5). Thus, Abcg1 deficiency (30-32) as well as

expression of human ABCG1 in mice (2; 33) fed an atherogenic diet enriched in cholesterol highlight

the role of Abcg1 in protecting tissues, especially the lung, from lipid accumulation without any

apparent changes in adipose tissue mass or adiposity. By contrast and coherent with the mechanism

described in the present study, Abcg1 KO mice fed a high fat diet devoid of cholesterol did not

accumulate lipids in tissues but rather exhibited a reduced adipose tissue formation and were

protected against diet-induced obesity (5).

Our data supports a mechanism by which Abcg1 promotes TG storage through the

requirement of bioactive LPL (4), as addition of exogenous LPL totally rescued the impaired TG

of 46 Diabetes

20

storage observed in Abcg1-deficient adipocytes. Our data led us to propose a mechanism by which

Abcg1-mediated export of SM contributes to sustain LPL activity and TG storage in adipocytes.

Indeed, LPL hydrolyzes triacylglycerol-rich lipoproteins allowing the subsequent release and uptake of

fatty acids for intracellular TG synthesis. LPL is expressed at very early stages of adipocyte

differentiation when cell-cell contact occurs (34) and adipocyte-derived LPL was reported to be a key

determinant for efficient TG storage and adipocyte hypertrophy (8). In agreement with this mechanism,

silencing of Abcg1 expression in 3T3-L1 adipocytes was accompanied by marked reduction in TG

storages; an effect observed at early stages of adipocyte differentiation. Beyond its role in TG

hydrolysis, LPL was also reported to facilitate lipoprotein uptake and thereby to contribute to cellular

cholesterol accumulation through this mechanism (14; 35). Consistent with such a role for LPL,

intracellular concentrations of free cholesterol were markedly reduced in Abcg1 KD adipocytes; a

finding in agreement with the reduced lipid droplet size observed in Abcg1 SKD adipocytes. The role

of ABCG1 in adiposity and obesity through its action on adipocyte LPL activity is supported by studies

in human subjects indicating that adipose tissue LPL activity is elevated in obesity (36). Furthermore,

several variants in the LPL gene have been associated with obesity (37; 38) and ob/ob mice with a

specific Lpl deficiency in adipose tissue displayed reduced weight and fat mass as compared to

control mice (39).

Alteration of Pparγ and Pparγ-target gene expression likely results from the lower abundance

of intracellular cholesterol and fatty acid derivatives delivered by LPL hydrolysis upon silencing of

Abcg1 expression in adipocytes. Indeed, fatty acid derivatives are ligands for Pparγ activation and

Pparγ expression in adipocytes was reported to be induced by its own activators and /or ligands, such

as fatty acids (40-42), and cholesterol derivatives through LXR activation (43). In agreement with this

mechanism, some fatty acid derivatives, whose delivery into cells is mediated by LPL, are activators of

Pparγ (41). Interestingly, overexpression of a dominant negative mutant of Pparγ that lacks the 16

COOH-terminal amino acids in 3T3-L1 adipocytes led to a reduction in the rate of free fatty acid

uptake, TG storage and adipocyte size, and more interestingly to a decrease in the expression of the

perilipin, Fabp4, Cd36, Hsl and C/ebpα genes (44); a similar phenotype to that observed in the

present study when Abcg1 expression is silenced in those cells. However, the modest reduction of

adipocyte diameter observed in Abcg1 knockdown epididymal adipose tissue in comparison to the

of 46Diabetes

21

more pronounced decrease of the tissue weight leads us to suggest that the decreased Pparγ

expression could also alter adipocyte differentiation and adipocyte cell number in vivo.

The participation of Pparγ in the overall mechanism by which Abcg1 contributes to adiposity

and obesity is supported by the observation that the adipose-specific Pparγ knockout mice displayed a

diminished weight gain and were protected against high fat diet-induced obesity (45). Taken together,

our findings indicate that modulation of LPL activity by Abcg1 in adipocyte might act as an intracellular

signaling pathway that controls adipocyte growth through activation of Pparγ and contributes by this

mechanism to fat mass formation and development of obesity in humans. Moreover, we presently

reported that a higher expression of ABCG1 in adipose tissue from obese individuals carrying the AT

haplotype was associated with increased PPARγ, adiposity, fat mass and BMI. However, although

targeted deletion of Pparγ in adipose tissue in mice led to an impaired adipose tissue growth, it must

be kept in mind that those mice also exhibited deleterious metabolic consequences such as lipid

accumulation in liver and in muscle and potentially insulin resistance (45; 46).

In a previous study, Buchman et al. first linked Abcg1 to obesity in a mouse model in which

Abcg1 was knocked out in the whole body (5). Thus, ablation of Abcg1 in Abcg1-/-

mice reduces

adipose cell size and hepatic steatosis and protects against diet-induced obesity. Although the

mechanisms underlying those effects were not elucidated in this earlier study, the authors proposed

that resistance of Abcg1-/-

mice to diet-induced obesity likely resulted to an increased energy

expenditure as compared to Abcg1+/+

animals. In addition, a slight reduction of food intake in Abcg1-/-

mice was also reported in this latter study and could thus contribute to the protective effect of Abcg1

deficiency. Our findings indicated that the contribution of Abcg1 in adiposity and weight gain results

from the critical role of adipocyte Abcg1 in adipose tissue, as testified by the targeted silencing of

Abcg1 expression by RNAi locally in adipose tissue in mice fed a high fat diet. Moreover, energy

expenditure as well as food intake were not altered upon inhibition of Abcg1 expression in adipose

tissue which further reinforces the specific contribution of adipose tissue Abcg1 in those effects.

Despite the fact that silencing of Abcg1 expression was restricted to adipose tissue, we

observed that the injection of lentiviral particles locally in adipose tissue not only targets adipocytes but

equally macrophages present in this tissue. Although Abcg1 appears to be more expressed in

adipocytes than in adipose tissue macrophages under our experimental conditions, this point may be

critical as Abcg1 promotes LPL-mediated lipid accumulation in macrophage in a TG-rich environment

of 46 Diabetes

22

(4). However, a recent study reported that macrophage Lpl does not contribute to adiposity and weight

gain (47). Indeed epididymal fat mass, lipid droplet size and gene expression levels in adipose tissue

as well as body weight were not altered in macrophage Lpl knockout mice as compared to control

mice.

Although further investigations are required in order to explore the whole spectrum of effects

consecutive to adipocyte Abcg1 silencing, our findings nonetheless support the contention that

ABCG1 might represent an interesting pharmacological target in obesity, notably by reducing fat mass

growth and weight gain in obese patients or in individuals prone to develop morbid obesity.

of 46Diabetes

23

Author Contributions

E.F., S.L.L., H.H., L.P., M.O., R.A., W.P., E.F.V., S.G., M.L., A.S., L.M-A, M.J.C., G.M.D-T., N.V., C.P.,

J.T., I.D., P.L., A.K., T.H., M.G. and W.L.G. contributed to the experimental work and/or data analysis.

E.F., S.L.L., I.D., T.H., K.C., M.G. and W.L.G. contributed to the development of the study. KC and CP

contributed to patients’ recruitment and phenotyping and biobank constitution. All authors contributed

to the development of the manuscript. W.L.G. wrote the manuscript, conceived, designed and

supervised the study.

Acknowledgements

INSERM and UPMC (Parinov program) provided generous support of these studies. M.O., E.F.V. and

A.S. were recipient of a Research Fellowship from the French Ministry of Research and Technology.

W.L.G. was the recipient of a PNRC award from INSERM. W.P. was the recipient of a junior research

fellowship from the French Embassy in Thailand. The INSERM U872 team thanks Assistance

publique/hôpitaux de Paris, Programme Hospitalier de recherche clinique (PHRC 1996 and 2002) for

supporting the genetic DNA bank on obesity. The ethic committee (Comité Protection des personnes

N° 1 Hôtel-Dieu) provided the ethic agreements. This work was supported by the Fondation de France

(W.L.G., P.L. and T.H.) and by the French National Agency through the national program

“Investissements d’avenir” with the reference ANR-10-IAHU-05. The authors are indebted to the

patients for their cooperation. Dr. Wilfried Le Goff is the guarantor of this work and, as such, had full

access to all the data in the study and takes responsibility for the integrity of the data and the accuracy

of the data analysis.

Disclosures

None

of 46 Diabetes

24

References.

1. Wang N, Lan D, Chen W, Matsuura F, Tall AR: ATP-binding cassette transporters G1 and G4 mediate cellular cholesterol efflux to high-density lipoproteins. Proc Natl Acad Sci U S A 2004;101:9774-9779 2. Kennedy MA, Barrera GC, Nakamura K, Baldan A, Tarr P, Fishbein MC, Frank J, Francone OL, Edwards PA: ABCG1 has a critical role in mediating cholesterol efflux to HDL and preventing cellular lipid accumulation. Cell Metab 2005;1:121-131 3. Le Goff W, Dallinga-Thie GM: ABCG1: Not as good as expected? Atherosclerosis 2011;219:393-394 4. Olivier M, Tanck MW, Out R, Villard EF, Lammers B, Bouchareychas L, Frisdal E, Superville A, Van Berkel T, Kastelein JJ, Eck MV, Jukema JW, Chapman MJ, Dallinga-Thie GM, Guerin M, Le Goff W: Human ATP-binding cassette G1 controls macrophage lipoprotein lipase bioavailability and promotes foam cell formation. Arterioscler Thromb Vasc Biol 2012;32:2223-2231 5. Buchmann J, Meyer C, Neschen S, Augustin R, Schmolz K, Kluge R, Al-Hasani H, Jurgens H, Eulenberg K, Wehr R, Dohrmann C, Joost HG, Schurmann A: Ablation of the cholesterol transporter adenosine triphosphate-binding cassette transporter G1 reduces adipose cell size and protects against diet-induced obesity. Endocrinology 2007;148:1561-1573 6. Umemoto T, Han CY, Mitra P, Averill MM, Tang C, Goodspeed L, Omer M, Subramanian S, Wang S, Den Hartigh LJ, Wei H, Kim EJ, Kim J, O'Brien KD, Chait A: Apolipoprotein AI and high-density lipoprotein have anti-inflammatory effects on adipocytes via cholesterol transporters: ATP-binding cassette A-1, ATP-binding cassette G-1, and scavenger receptor B-1. Circ Res 2013;112:1345-1354 7. Mead JR, Irvine SA, Ramji DP: Lipoprotein lipase: structure, function, regulation, and role in disease. J Mol Med 2002;80:753-769 8. Gonzales AM, Orlando RA: Role of adipocyte-derived lipoprotein lipase in adipocyte hypertrophy. Nutr Metab (Lond) 2007;4:22 9. Spielmann N, Mutch DM, Rousseau F, Tores F, Hager J, Bertrais S, Basdevant A, Tounian P, Dubern B, Galan P, Clement K: Cathepsin S genotypes are associated with Apo-A1 and HDL-cholesterol in lean and obese French populations. Clin Genet 2008;74:155-163 10. Ciangura C, Bouillot JL, Lloret-Linares C, Poitou C, Veyrie N, Basdevant A, Oppert JM: Dynamics of change in total and regional body composition after gastric bypass in obese patients. Obesity (Silver Spring) 2010;18:760-765 11. Brummelkamp TR, Bernards R, Agami R: A system for stable expression of short interfering RNAs in mammalian cells. Science 2002;296:550-553 12. Schuck S, Manninen A, Honsho M, Fullekrug J, Simons K: Generation of single and double knockdowns in polarized epithelial cells by retrovirus-mediated RNA interference. Proc Natl Acad Sci U S A 2004;101:4912-4917 13. Larrede S, Quinn CM, Jessup W, Frisdal E, Olivier M, Hsieh V, Kim MJ, Van Eck M, Couvert P, Carrie A, Giral P, Chapman MJ, Guerin M, Le Goff W: Stimulation of cholesterol efflux by LXR agonists in cholesterol-loaded human macrophages is ABCA1-dependent but ABCG1-independent. Arterioscler Thromb Vasc Biol 2009;29:1930-1936 14. Milosavljevic D, Kontush A, Griglio S, Le Naour G, Thillet J, Chapman MJ: VLDL-induced triglyceride accumulation in human macrophages is mediated by modulation of LPL lipolytic activity in the absence of change in LPL mass. Biochim Biophys Acta 2003;1631:51-60

of 46Diabetes

25

15. Kobayashi A, Takanezawa Y, Hirata T, Shimizu Y, Misasa K, Kioka N, Arai H, Ueda K, Matsuo M: Efflux of sphingomyelin, cholesterol, and phosphatidylcholine by ABCG1. J Lipid Res 2006;47:1791-1802 16. Ivanova PT, Milne SB, Byrne MO, Xiang Y, Brown HA: Glycerophospholipid identification and quantification by electrospray ionization mass spectrometry (Chapter 2). Methods in Enzymology 2007;432:21-57 17. Camont L, Lhomme M, Rached F, Le Goff W, Negre-Salvayre A, Salvayre R, Calzada C, Lagarde M, Chapman MJ, Kontush A: Small, dense high-density lipoprotein-3 particles are enriched in negatively charged phospholipids: relevance to cellular cholesterol efflux, antioxidative, antithrombotic, anti-inflammatory, and antiapoptotic functionalities. Arterioscler Thromb Vasc Biol 2013;33:2715-2723 18. Janes PW, Ley SC, Magee AI: Aggregation of lipid rafts accompanies signaling via the T cell antigen receptor. The Journal of cell biology 1999;147:447-461 19. de Haan W, Bhattacharjee A, Ruddle P, Kang MH, Hayden MR: ABCA1 in adipocytes regulates adipose tissue lipid content, glucose tolerance, and insulin sensitivity. J Lipid Res 2014;55:516-523 20. Yvan-Charvet L, Ranalletta M, Wang N, Han S, Terasaka N, Li R, Welch C, Tall AR: Combined deficiency of ABCA1 and ABCG1 promotes foam cell accumulation and accelerates atherosclerosis in mice. J Clin Invest 2007;117:3900-3908 21. Arimoto I, Saito H, Kawashima Y, Miyajima K, Handa T: Effects of sphingomyelin and cholesterol on lipoprotein lipase-mediated lipolysis in lipid emulsions. J Lipid Res 1998;39:143-151 22. Saito H, Arimoto I, Tanaka M, Sasaki T, Tanimoto T, Okada S, Handa T: Inhibition of lipoprotein lipase activity by sphingomyelin: role of membrane surface structure. Biochim Biophys Acta 2000;1486:312-320 23. Atger VM, de la Llera Moya M, Stoudt GW, Rodrigueza WV, Phillips MC, Rothblat GH: Cyclodextrins as catalysts for the removal of cholesterol from macrophage foam cells. J Clin Invest 1997;99:773-780 24. Jablin MS, Flasinski M, Dubey M, Ratnaweera DR, Broniatowski M, Dynarowicz-Latka P, Majewski J: Effects of beta-cyclodextrin on the structure of sphingomyelin/cholesterol model membranes. Biophysical journal 2010;99:1475-1481 25. Ranalletta M, Wang N, Han S, Yvan-Charvet L, Welch C, Tall AR: Decreased atherosclerosis in low-density lipoprotein receptor knockout mice transplanted with Abcg1-/- bone marrow. Arterioscler Thromb Vasc Biol 2006;26:2308-2315 26. Rensen PC, van Berkel TJ: Apolipoprotein E effectively inhibits lipoprotein lipase-mediated lipolysis of chylomicron-like triglyceride-rich lipid emulsions in vitro and in vivo. The Journal of biological chemistry 1996;271:14791-14799 27. The effect of aggressive versus standard lipid lowering by atorvastatin on diabetic dyslipidemia: the DALI study: a double-blind, randomized, placebo-controlled trial in patients with type 2 diabetes and diabetic dyslipidemia. Diabetes Care 2001;24:1335-1341 28. Villard EF, P EIK, Frisdal E, Bruckert E, Clement K, Bonnefont-Rousselot D, Bittar R, Le Goff W, Guerin M: Genetic determination of plasma cholesterol efflux capacity is gender-specific and independent of HDL-cholesterol levels. Arterioscler Thromb Vasc Biol 2013;33:822-828 29. Le Lay S, Ferre P, Dugail I: Adipocyte cholesterol balance in obesity. Biochemical Society transactions 2004;32:103-106 30. Kennedy MA, Venkateswaran A, Tarr PT, Xenarios I, Kudoh J, Shimizu N, Edwards PA: Characterization of the human ABCG1 gene: liver X receptor activates an internal promoter that

of 46 Diabetes

26

produces a novel transcript encoding an alternative form of the protein. The Journal of biological chemistry 2001;276:39438-39447 31. Out R, Hoekstra M, Meurs I, de Vos P, Kuiper J, Van Eck M, Van Berkel TJ: Total body ABCG1 expression protects against early atherosclerotic lesion development in mice. Arterioscler Thromb Vasc Biol 2007;27:594-599 32. Baldan A, Tarr P, Vales CS, Frank J, Shimotake TK, Hawgood S, Edwards PA: Deletion of the transmembrane transporter ABCG1 results in progressive pulmonary lipidosis. The Journal of biological chemistry 2006;281:29401-29410 33. Burgess B, Naus K, Chan J, Hirsch-Reinshagen V, Tansley G, Matzke L, Chan B, Wilkinson A, Fan J, Donkin J, Balik D, Tanaka T, Ou G, Dyer R, Innis S, McManus B, Lutjohann D, Wellington C: Overexpression of human ABCG1 does not affect atherosclerosis in fat-fed ApoE-deficient mice. Arterioscler Thromb Vasc Biol 2008;28:1731-1737 34. Ntambi JM, Young-Cheul K: Adipocyte differentiation and gene expression. The Journal of nutrition 2000;130:3122S-3126S 35. Babaev VR, Fazio S, Gleaves LA, Carter KJ, Semenkovich CF, Linton MF: Macrophage lipoprotein lipase promotes foam cell formation and atherosclerosis in vivo. J Clin Invest 1999;103:1697-1705 36. Wang H, Eckel RH: Lipoprotein lipase: from gene to obesity. Am J Physiol Endocrinol Metab 2009;297:E271-288 37. Radha V, Vimaleswaran KS, Ayyappa KA, Mohan V: Association of lipoprotein lipase gene polymorphisms with obesity and type 2 diabetes in an Asian Indian population. Int J Obes (Lond) 2007;31:913-918 38. Jemaa R, Tuzet S, Portos C, Betoulle D, Apfelbaum M, Fumeron F: Lipoprotein lipase gene polymorphisms: associations with hypertriglyceridemia and body mass index in obese people. Int J Obes Relat Metab Disord 1995;19:270-274 39. Weinstock PH, Levak-Frank S, Hudgins LC, Radner H, Friedman JM, Zechner R, Breslow JL: Lipoprotein lipase controls fatty acid entry into adipose tissue, but fat mass is preserved by endogenous synthesis in mice deficient in adipose tissue lipoprotein lipase. Proc Natl Acad Sci U S A 1997;94:10261-10266 40. Lehmann JM, Moore LB, Smith-Oliver TA, Wilkison WO, Willson TM, Kliewer SA: An antidiabetic thiazolidinedione is a high affinity ligand for peroxisome proliferator-activated receptor gamma (PPAR gamma). The Journal of biological chemistry 1995;270:12953-12956 41. Schoonjans K, Peinado-Onsurbe J, Lefebvre AM, Heyman RA, Briggs M, Deeb S, Staels B, Auwerx J: PPARalpha and PPARgamma activators direct a distinct tissue-specific transcriptional response via a PPRE in the lipoprotein lipase gene. The EMBO journal 1996;15:5336-5348 42. Bouaboula M, Hilairet S, Marchand J, Fajas L, Le Fur G, Casellas P: Anandamide induced PPARgamma transcriptional activation and 3T3-L1 preadipocyte differentiation. European journal of pharmacology 2005;517:174-181 43. Seo JB, Moon HM, Kim WS, Lee YS, Jeong HW, Yoo EJ, Ham J, Kang H, Park MG, Steffensen KR, Stulnig TM, Gustafsson JA, Park SD, Kim JB: Activated liver X receptors stimulate adipocyte differentiation through induction of peroxisome proliferator-activated receptor gamma expression. Molecular and cellular biology 2004;24:3430-3444 44. Tamori Y, Masugi J, Nishino N, Kasuga M: Role of peroxisome proliferator-activated receptor-gamma in maintenance of the characteristics of mature 3T3-L1 adipocytes. Diabetes 2002;51:2045-2055

of 46Diabetes

27

45. Jones JR, Barrick C, Kim KA, Lindner J, Blondeau B, Fujimoto Y, Shiota M, Kesterson RA, Kahn BB, Magnuson MA: Deletion of PPARgamma in adipose tissues of mice protects against high fat diet-induced obesity and insulin resistance. Proc Natl Acad Sci U S A 2005;102:6207-6212 46. He W, Barak Y, Hevener A, Olson P, Liao D, Le J, Nelson M, Ong E, Olefsky JM, Evans RM: Adipose-specific peroxisome proliferator-activated receptor gamma knockout causes insulin resistance in fat and liver but not in muscle. Proc Natl Acad Sci U S A 2003;100:15712-15717 47. Takahashi M, Yagyu H, Tazoe F, Nagashima S, Ohshiro T, Okada K, Osuga J, Goldberg IJ, Ishibashi S: Macrophage lipoprotein lipase modulates the development of atherosclerosis but not adiposity. J Lipid Res 2013;54:1124-1134

of 46 Diabetes

28

Figure Legends.

Figure 1. Abcg1-silencing in preadipocytes affects TG storage in a LPL-dependent manner. A.

Levels of Abcg1 mRNA and (B) correlation between Abcg1 mRNA and cellular TG storage during

adipocyte differentiation (from D0 to D13). The efficiency of the Abcg1 knockdown in 3T3-L1

adipocytes was assessed by quantification of mRNA (C) and protein (D) levels. E. Evaluation of Lpl

mRNA levels (F), secreted LPL activity and (G) membrane lipid raft formation in Ctrl and Abcg1

Knockdown (KD) 3T3-L1 adipocyte. Representative photographs of lipid rafts visualized by

fluorescence microscopy (x63). Cellular triglyceride content was quantified (H) during maturation (from

D0 to D4) of Ctrl and Abcg1 KD 3T3-L1 preadipocytes into adipocytes. I. Impact of a 24h-treatment

with either tetrahydrolipstatin (THL) or increasing doses of bovine LPL (bLPL) on cellular triglycerides

mass and secreted LPL activity during control adipocyte differentiation (D6). Data are shown as mean

± SEM. Experiments were performed in triplicate. *p<0.05 and **p<0.005 versus control cells.

Figure 2. Gene expression profile in stable Abcg1 Kd 3T3-L1 mature adipocyte. Quantification of

protein (A) and mRNA levels (B-J) levels in different stable 3T3-L1 adipocytes generated following the

infection with lentiviral (L) or retroviral (R) particles expressing control shRNA (L-Ctrl and R-Ctrl,

respectively) or shRNAs targeting the cDNA sequence of mouse Abcg1 gene (L1, L2, L3 and R1, R2,

respectively) following 10 days of differentiation. A similar gene expression pattern was observed in all

the Abcg1 KD adipocytes generated. Decreased expression of Pparγ and Pparγ-target genes in stable

Abcg1 KD adipocytes clones compared to stable control adipocytes. Data are shown as mean ± SEM.

Experiments were performed in triplicate. *p<0.05, **p<0.005 and ***p<0.005 versus respective control

cells.

Figure 3. Stable Abcg1 Knockdown compromises 3T3-L1 adipocyte lipid storage. A. Total

protein levels were assessed by Western blots analysis. Quantification of Abcg1, Pparγ, Perilipin and

Fabp4 protein levels in stable control (Ctrl) and Abcg1 SKD 3T3-L1 adipocytes following 10 days of

differentiation. F. Phase-contrast photographs representative of lipid droplets in control (Ctrl) and

Abcg1 SKD 3T3-L1 adipocytes visualized by microscopy (x20). G. Measurement of lipid droplets size

and quantification of cellular (H) triglycerides and (K) free cholesterol masses in stable control (Ctrl)

and Abcg1 SKD 3T3-L1 adipocytes following 10 days of differentiation. I. Rescue of impaired TG

of 46Diabetes

29

storage in Abcg1 SKD adipocytes upon incubation with exogenous bovine LPL (bLPL) for the last 24h

(Days: 1) or all along the 10 days of the maturation period (Days: 10). J. Cellular [3H]Cholesterol efflux

to apoAI or HDL from mature control (Ctrl) and Abcg1 SKD adipocytes (D10). Data are shown as

mean ± SEM. Experiments were performed in triplicate. *p<0.05 and **p<0.005 versus control cells.

Figure 4. Increased sphyingomyelin content in Abcg1 knockdown adipocytes is associated

with an altered LPL-dependent TG storage. Sphingomyelin (SM) efflux to BSA (A) and intracellular

SM mass (B) in mature control (Ctrl) and Abcg1 SKD adipocytes (D10). Secreted LPL activity (C) and

intracellular TG mass (D) in adipocytes (D10) enriched or depleted with either SM or

sphingomyelinase (SMase) for 16 hours, respectively.

Figure 5. Knockdown of adipose tissue Abcg1 following shRNA lentiviral local delivery. A.

Scheme of the experimental procedure. B. Abcg1 expression in epididymal adipose tissue from a

C57BL/6 mouse was visualized by fluorescence microscopy (x63). Arrows indicate Abcg1 expression

(green) in adipocytes (red, perilipin). Nuclei were counterstained with DAPI (blue). Recovery of GFP

fluorescence in adipose tissue from an individual C57BL/6 mouse fed a high fat diet (60% fat) 4 weeks

following the local injection in the epididymal adipose tissue of lentiviral particles encoding either a

shRNA control (C, left fat pad) or the full length copGFP gene (D, right fat pad). Fluorescence was

visualized by microscopy (x400). E. Visualization of Abcg1 (red) by fluorescence microscopy (x63) in

epididymal fat pads following the local injection in the epididymal adipose tissue of lentiviral particles

encoding either a shRNA control or a shRNA inhibiting mouse Abcg1 expression. Nuclei were

counterstained with DAPI (blue). Quantification of Abcg1 mRNA levels in (F) adipose tissue (mean Ct :

25,18 in L-Ctrl), (H) intestine (mean Ct : 31.78 in L-Ctrl) and (I) liver (mean Ct : 29.95 in L-Ctrl) from

C57BL/6 mice fed a high fat diet (60% fat) after 4 weeks following the local injection in the epididymal

adipose tissue of lentiviral particles encoding either a shRNA inhibiting mouse Abcg1 expression (L-

Abcg1) or a shRNA control (L-Ctrl). G. Quantification of Abcg1 mRNA levels in adipocytes (mean Ct :

27.95 in L-Ctrl), adipose tissue macrophages (ATM, mean Ct : 31.01 in L-Ctrl) and endothelial cells

(mean Ct : 30.15 in L-Ctrl) isolated in adipose tissue from L-Abcg1 and L-Ctrl mice. n=11 mice per

group. Data are shown as mean ± SEM. *p<0.05 and ***p<0.0005 versus L-Ctrl.

of 46 Diabetes

30

Figure 6. Abcg1 deficiency in adipose tissues affects high-fat feeding response of mice.

C57BL/6 mice fed a high fat diet (40% fat) were injected with lentiviral particles encoding either a

shRNA inhibiting mouse Abcg1 expression (L-Abcg1) or a shRNA control (L-Ctrl) locally in the

epididymal adipose tissue. (A) Weight gain, (B) epididymal fat mass, (C) adipocyte diameter and (D,

H-O) mRNA levels in epididymal adipose tissue was measured after 4 weeks following the day of the

injection (n=10 mice per group). (E) Food intake were measured in C57BL/6 mice fed a high fat diet

(60% fat) after 4 weeks following the local injection in the epididymal adipose tissue of lentiviral

particles encoding either a shRNA inhibiting mouse Abcg1 expression (L-Abcg1, n=6) or a shRNA

control (L-Ctrl, n=6). (F-G) Locomotor activity in L-Ctrl (n=8) and L-Abcg1 (n=8) mice was monitored

throughout the 4 weeks following the injection. Data are shown as mean ± SEM. *p<0.05 and

**p<0.005 versus L-Ctrl.

Figure 7. Elevated adipose tissue ABCG1 expression and increased fat mass and obesity in

obese individuals carrying the AT haplotype. A. Human ABCG1 promoter activity according to the

CG and AT haplotypes. HepG2 cells were transiently transfected with a construct containing the

proximal 1056 bp of the human promoter with either the -204A / -134T (AT) haplotype or the -204C / -

134G (CG) haplotype. Luciferase activity is expressed in RLU after normalization for β-galactosidase

activity. Values are means±SEM of 5 independent experiments performed in triplicate. *p<0.0005. B-F.

Quantification of mRNA levels isolated in adipose tissue biopsies from 10 morbid obese women

carrying either the AT or the CG haplotype. (G) Correlation between adipocyte diameter and ABCG1

mRNA levels in adipose tissue from morbid obese women. n=20. Fat mass (H) and fat-free mass (I) in

obese individuals carrying either the AT (n=102) or the CG haplotype (n=22). Data are shown as mean

± SEM. *p<0.05 versus CG haplotype, adjusted for age and sex. Association of the rs1378577 (-

134T/G) and rs1893590 (-204A/C) ABCG1 SNP with (J) BMI (L) and fat mass index (FMI) in a

population of 1320 middle-aged severely morbid obese patients (BMI = 45.47 ± 0.002 Kg/m2). (K)

Amount of -204A / -134T (AT) haplotypes relative to BMI in obese individuals. AT/AT = 2. The effect of

each SNP on BMI was analyzed by linear regression in an additive, dominant and recessive manner.

The best model fitting the data is shown (dominant). All models were adjusted for age and sex. Data

are shown as mean ± SEM. *p<0.05.

of 46Diabetes

of 46 Diabetes

of 46Diabetes

of 46 Diabetes

of 46Diabetes

of 46 Diabetes

of 46Diabetes

of 46 Diabetes

Supplementary Table 1. Metabolic consequences of epididymal adipose tissue Abcg1 knockdown in C57BL/6 mice fed a high fat diet.

Metabolic parameters L-Ctrl L-Abcg1 p value (Fasting state) Plasma lipid levels Triglycerides (mg/dL) 88.61 ± 4.92 103.7 ± 6.77 0.121 Total cholesterol (mg/dL) 119.6 ± 6.53 141.2 ± 4.07 0.006* Free cholesterol (mg/dL) 23.78 ± 3.36 24.86 ± 2.41 0.790 HDL-cholesterol (mg/dL) 111.6 ± 5.96 133.5 ± 3.7 0.002** Free fatty acids (mmol/L) 1.98 ± 0.24 1.98 ± 0.27 0.987 Insulin resistance Glucose (mg/dL) 179.7 ± 6.03 210.5 ± 9.66 0.017* Insulin (ng/mL) 0.13 ± 0.02 0.16 ± 0.02 0.445 HOMA-IR 1.25 ± 0.33 1.39 ± 0.21 0.712

Data represent Mean ± S.E.M. of L-Ctrl (n=16) and L-Abcg1 (n=21) mice at 4 weeks following the injection of lentiviral particle. p value versus L-Ctrl.

of 46Diabetes

Supplementary Table 2. Association of rs1378577 and rs1893590 ABCG1 SNPs with BMI in a population of 1320 middle-aged morbid obese patients.

ABCG1 SNP BMI (kg/m²) BMI (kg/m²) /

diabetes BMI (kg/m²) /

HOMA2B BMI (kg/m²) /

HOMA2S BMI.kg.m.² / HOMA2IR

rs1378577 G/G vs G/T P value 0,40474 0,284993 0,32257 0,35605 0,63625 G/G vs G/T β -0,92 -1,02 -0,91 -0,83 -0,33 G/G vs T/T P value 0,55378 0,43940 0,40062 0,35993 0,22551 G/G vs T/T β 0,65 1,31 1,38 1,46 1,80 G/T vs T/T P value 0,00232** 0,00017*** 0,00019*** 0,00018*** 0,00029*** G/T vs T/T β -1,57 -2,33 -2,30 -2,29 -2,13

rs1893590 A/A vs A/C P value 0,04237* 0,02776* 0,03007* 0,03363* 0,04308* A/A vs A/C β -1,08 -1,43 -1,41 -1,37 -1,25 A/A vs C/C P value 0,01958* 0,01022* 0,00883* 0,00989* 0,00758* A/A vs C/C β -2,15 -2,72 -2,75 -2,70 -2,80 A/C vs C/C P value 0,27832 0,28020 0,25157 0,25539 0,17610 A/C vs C/C β 1,074 1,28 1,34 1,32 1,56

rs1378577

G/G vs G/T & T/T P value 0,95916 0,54119 0,916 0,8569 0,8 G/G vs G/T & T/T β 0,06 1,05 0,48 0,57 0,65 G/G & G/T vs T/T P value 0,00339** 0,00026*** 0,00031*** 0,00031*** 0,00027*** G/G & G/T vs T/T β 1,45 2,09 2,17 2,16 2,17

rs1893590

A/A & A/C vs C/C P value 0,06083 0,02497* 0,0379* 0,03258* 0,0349* A/A & A/C vs C/C β -1,69 -2,32 -2,16 -2,21 -2,17 A/A vs A/C & C/C P value 0,01102* 0,00693* 0,00471** 0,00477** 0,00561* A/A vs A/C & C/C β -1,29 -1,59 -1,74 -1,73 -1,69

All analysis were adjusted for age and sex. β, regression coefficient. *p<0.05, **p<0.005 and ***p<0.0005. Association of ABCG1 SNPs with BMI was adjusted for diabetes or HOMA index (HOMA2B, HOM2S and HOMAIR).

of 46 Diabetes

Supplementary Table 3. Association of rs1378577 and rs1893590 ABCG1 SNPs with plasma lipid and apolipoprotein levels in a population of 1320 middle-aged morbid obese patients.

ABCG1 SNP

Total cholesterol

(g/L) HDL-C (g/L)

Triglycerides (g/L)

ApoA-I (g/L)

ApoB (g/L)

Lp(a) (g/L)

rs1378577 G/G 3.94 ± 0.25 0.94 ± 0.08 1.50 ± 0.09 1.41 ± 0.05 1.09 ± 0.05 0.25 ± 0.05 G/T 4.38 ± 0.10 1.02 ± 0.03 1.65 ± 0.07 1.43 ± 0.02 1.13 ± 0.04 0.26 ± 0.02 T/T 4.37 ± 0.07 0.97 ± 0.02 1.61 ± 0.04 1.40 ± 0.01 1.10 ± 0.01 0.26 ± 0.01

p G/G vs G/T 0.12 0.14 0.57 0.70 0.81 0.64 p G/G vs T/T 0.09 0.35 0.79 0.55 0.73 0.98 p G/T vs T/T 0.79 0.23 0.51 0.67 0.84 0.34

p G/G+G/T vs TT 0.09 0.24 0.70 0.59 0.75 0.85 p G/G vs G/T+T/T 0.45 0.43 0.61 0.57 0.78 0.39

rs1893590 A/A 4.29 ± 0.08 0.96 ± 0.02 1.61 ± 0.04 1.40 ± 0.01 1.10 ± 0.02 0.27 ± 0.02 A/C 4.28 ± 0.11 0.99 ± 0.03 1.65 ± 0.08 1.42 ± 0.02 1.14 ± 0.04 0.27 ± 0.02 C/C 4.34 ± 0.2 1.02 ± 0.06 1.52 ± 0.08 1.45 ± 0.04 1.13 ± 0.03 0.23 ± 0.04

p A/A vs A/C 0.90 0.37 0.80 0.33 0.59 0.37 p A/A vs C/C 0.78 0.44 0.78 0.17 0.21 0.19 p A/C vs C/C 0.73 0.85 0.67 0.47 0.39 0.48

p A/A+A/C vs C/C 0.10 0.30 0.91 0.18 0.36 0.21 p A/A vs A/C+C/C 0.75 0.58 0.72 0.24 0.25 0.26

All models were adjusted for age, sex, BMI and treatment for dyslipidemia.

of 46Diabetes

Supplementary Figure Legends.

Supplementary Figure 1. Characterization of Abcg1 stable knockdown 3T3-L1 adipocytes. A-D.

Quantification of cholesterol synthesis and uptake gene mRNA levels in stable control (Ctrl) and

Abcg1 SKD 3T3-L1 adipocytes following 10 days of differentiation (D10). E. Removal of plasma

membrane cholesterol with 1 mM methyl-β-cyclodextrin for 16 hours on intracellular TG mass in

control (Ctrl) and Abcg1 SKD adipocytes (D10). Measurement of apoE secretion (F) and intracellular

total lipase activity (G) in stable control (Ctrl) and Abcg1 SKD 3T3-L1 adipocytes (D10). *p<0.05

versus control cells.

Supplementary Figure 2. Analysis of inflammatory and insulin-resistance gene expression in

adipose tissue Abcg1 knockdown mice. A. Dispersion of the injected volume in the whole

epididymal fat pad using a colored dye (trypan blue). B-F. C57BL/6 mice fed a high fat diet (40% fat)

were injected locally in the epididymal adipose tissue with lentiviral particles encoding either a shRNA

inhibiting mouse Abcg1 expression (L-Abcg1) or a shRNA control (L-Ctrl). Levels of mRNA l in

epididymal adipose tissue was measured after 4 weeks following the day of the injection (n=10 mice

per group). Data are shown as mean ± SEM. *p<0.05 and **p<0.05 versus L-Ctrl.

Supplementary Figure 3. Effect of rs1378577 and rs1893590 ABCG1 SNPs in a large cohort of

severely obese patients.

A. Analysis of the relative expression of ABCG1 mRNA levels in primary human preadipocytes before

and after differentiation into adipocytes (10 Days) and in primary human monocyte-derived

macrophage. *p<0.05 versus preadipocytes. B-D. Quantification of macrophage markers mRNA levels

isolated in adipose tissue biopsies from 10 morbid obese women carrying either the AT or the CG

haplotype. Association of the rs1378577 (-134T/G) and rs1893590 (-204A/C) ABCG1 SNP with (E)

adiponectin and (F) CRP levels in a population of 1320 middle-aged severely morbid obese patients

(BMI = 45.5 ± 0.04 Kg/m2). The best model fitting the data is shown (dominant in F and recessive in

E). All models were adjusted for age and sex. *p<0.05.

of 46 Diabetes

Supplementary Figure 4. Analysis of the -134T/G and -204A/C ABCG1 SNPs in independent

populations of type 2 obese diabetic (DALI) and severe obese patients. Association of the

rs1378577 (-134T/G) (A) and rs1893590 (-204A/C) (B) ABCG1 SNP with Waist Hip Ratio (WHR) in

217 type 2 diabetic patients from the DALI Study (1). *p<0.05 versus TG and #p<0.05 versus GG

(rs1378577), *p<0.05 versus AC and #p<0.05 versus CC (rs1378577). (C) Waist Hip Ratio in

individuals carrying either the AT (n=101) or the CG haplotype (n=11). **p=0.0057. Association of the

rs1378577 (-134T/G) ABCG1 SNP with BMI (D) and (E) fat mass in 595 morbid obese individuals (2).

*p<0.05. The effect of each SNP was analyzed by linear regression in an additive, dominant and

recessive manner. The best model fitting the data is shown (additive in A-B, dominant in D and

recessive in E). All models were adjusted for age and sex.

of 46Diabetes

of 46 Diabetes

of 46Diabetes

of 46 Diabetes

of 46Diabetes

page 1 of 46 diabetes · page 1 of 46 diabetes,* diabetes publish ahead of print, published online...

Documents