Differential Gene Expression in Adipose Tissue From Obese Human Subjects During Weight Loss and Weight Maintenance
Differential Gene Expression in Adipose Tissue From Obese Human Subjects During Weight Loss and Weight Maintenance
Differential Gene Expression in Adipose Tissue From Obese Human Subjects During Weight Loss and Weight Maintenance
ABSTRACT substantial weight loss seen after bariatric surgery, are powerful in
Background: Differential gene expression in adipose tissue during protecting against at least some of these risk factors, particularly
diet-induced weight loss followed by a weight stability period is against diabetes (6–8).
poorly characterized. Markers of these processes may provide a deep- Obesity is a complex trait to which both genes and envi-
er understanding of underlying mechanisms. ronment provide important contributions. Thus far, very few
Objective: We aimed to identify differentially expressed genes in genetic variants have been convincingly shown to be associated
human adipose tissue during weight loss and weight maintenance with obesity (9). Among the more investigated genes, for which
196 Am J Clin Nutr 2012;96:196–207. Printed in USA. Ó 2012 American Society for Nutrition
TABLE 1
Clinical characteristics of study subjects at baseline (before weight loss), after 10% body-weight reduction by
consumption of an LCD, and after group therapy to achieve weight stability (65%)1
Baseline Weight loss Maintenance of reduced weight
2
Men/women (n) 3/6 3/6 3/6
Age (y) 40 6 3.73 40 6 4.2 40 6 3.7
Weight (kg) 129 6 7.3 105 6 7.8 106 6 6.7
Weight loss (%)4 — 218.8 6 1.9 217.7 6 1.8
BMI (kg/m2) 42.7 6 1.4 34.9 6 1.95 35.2 6 1.56
fP glucose (mmol/L) 5.0 6 0.2 4.6 6 0.2 4.7 6 0.1
fS insulin (pmol/L) 10.9 6 2.0 3.9 6 0.85 6.2 6 0.96,7
HOMA-IR 2.4 6 0.4 0.8 6 0.25 1.3 6 0.26,7
fP HDL (mmol/L) 1.1 6 0.1 1.0 6 0.1 1.3 6 0.16,7
fP triglycerides (mmol/L) 1.5 6 0.3 0.8 6 0.15 0.8 6 0.26
1
fP, fasting plasma; fS, fasting serum; LCD, low-calorie diet.
2
Biopsy from one woman after weight loss was not included in the microarray analysis.
3
Mean 6 SEM (all such values).
4
Comparison of weight change in percentage from baseline.
5
P , 0.05 between baseline and weight loss (Wilcoxon’s rank-sum test).
6
P , 0.05 between baseline and weight stability (Wilcoxon’s rank-sum test).
7
P , 0.05 between weight loss and weight stability (Wilcoxon’s rank-sum test).
198 JOHANSSON ET AL
second assessment) was excluded from the analysis because of (Affymetrix), and detection of hybridized complementary RNA
improper sampling (Figure 1). was performed according to manufacturer’s recommendation at
the SCIBLU Microarray Resource Centre, Lund University. The
Adipose tissue biopsies and RNA extraction HG-U133 Plus 2.0 GeneChip DNA microarray contains 54,675
probe sets for interrogation, including known genes and ex-
Biopsies were immediately put in RNAlater solution (Ambion),
pressed sequenced tags. For each array, the percentage present
and after overnight incubation at 4°C, biopsies were snap frozen
call was .50%, and the 3# to 5# ratio of housekeeping genes
at 280°C until additional processing. RNA was extracted by using
GAPDH and HSAC07 was ,3, which showed a good and uni-
Tri reagent (Sigma-Aldrich) and an RNeasy Midi kit (Qiagen).
form in vitro transcription. All gathered array data were uploaded
The RNA was further concentrated by using RNeasy MiniElute
to the Gene Expression Omnibus database at the National Center
kit for purifying and concentrating RNA (Qiagen) and a SpeedVac
for Biotechnology (http://www.ncbi.nlm.nih.gov/geo/). The ac-
centrifugal evaporator (DNA 120 SpeedVac; Thermo Savant).
cession number to the dataset is GSE35411.
RNA quality was assessed by using an Agilent 2100 Bioanalyzer
platform for sizing, quantification, and quality control for RNA
(Agilent). RNA was stored at 280°C until additional analysis. Microarray data extraction and analysis
We used the AffyProbeMiner web resource (14) to regroup
Microarray individual probes into consistent probe sets and remap probe
The synthesis of biotin-labeled complementary RNA, hy- sets to the correct sets of messenger RNA transcripts for the HG-
bridization to DNA microarray HG-U133 Plus 2.0 GeneChip U133 Plus 2.0 array, which resulted in 23,798 probe sets. The
kg kg
Top 15 upregulated (fold)
RARRES1 221872_at 76.45 6 134 222.51 6 52.57 2.91 0.018
RARRES1 206391_at, 206392_s_at 62.33 6 7.51 151.54 6 31.92 2.43 0.036
IGSF6 206420_at 25.73 6 1.92 58.31 6 11.38 2.27 0.045
ATP1B1 227556_at, 201242_s_at 418.5 6 41.75 837.56 6 153.5 2.00 0.047
C7 202992_at 240.19 6 48.21 477.1 6 111.33 1.99 0.024
EXPH5 214734_at 28.07 6 2.15 54.14 6 4.59 1.93 0.014
DTNA 205741_s_at 31.75 6 1.66 60.65 6 7.15 1.91 0.025
MT1M 217546_at 134.28 6 7.65 249.72 6 55.84 1.86 0.050
SRPX 204955_at 1851.62 6 174.61 3382.64 6 419.41 1.83 0.020
RGS17 220334_at 37.8 6 4.58 67.36 6 9.45 1.78 0.045
EGFLAM 226911_at 199.98 6 17.49 353.47 6 67.09 1.77 0.040
CIDEA 221295_at 830.75 6 99.15 1437.16 6 132.03 1.73 0.013
CHKB and CPT1B 210069_at 62.39 6 5.03 107.71 6 16.67 1.73 0.028
CA12 214164_x_at, 215867_x_at 127.55 6 10.12 215.25 6 34.04 1.69 0.026
C1orf51 227285_at 45.82 6 8.08 77.11 6 13.26 1.68 0.028
Top 15 downregulated (fold)
NQO1 201467_s_at 650.52 6 68.38 334.76 6 28.57 0.51 0.011
NQO1 201468_s_at, 243763_x_at 4175.27 6 308.98 2134.49 6 205.06 0.51 0.013
CLCA2 206164_at, 206165_s_at 42.87 6 5.57 20.56 6 2.91 0.48 0.017
THBS4 204776_at 802.79 6 168.59 354.06 6 82.08 0.44 0.014
LOC55908 220437_at 508.72 6 156.46 222.57 6 75.11 0.44 0.021
SCD 200832_s_at, 217200_x_at, 207986_x_at, 200831_s_at, 4544.31 6 475.67 1965.77 6 559.11 0.43 0.017
223839_s_at, 211162_x_at,
ALDOC 202022_at 1562.86 6 213.02 674.17 6 143.27 0.43 0.017
MXRA5 209596_at 942.3 6 108.5 393.21 6 67.38 0.42 0.011
FADS1 208963_x_at, 208964_s_at 313.81 6 64.66 125.82 6 28.78 0.40 0.031
FADS1 208962_s_at, 208963_x_at 292.92 6 68.01 105.45 6 25.28 0.36 0.023
FADS2 202218_s_at 257.67 6 42.25 91.05 6 19.22 0.35 0.025
B4GALT6 206233_at 161.22 6 32.05 52.71 6 9.09 0.33 0.021
LTF 202018_s_at 286.95 6 126.6 93.25 6 23.94 0.32 0.017
SCD 211708_s_at, 211162_x_at 1835.13 6 310.64 532.18 6 238.55 0.29 0.019
SFRP2 223121_s_at, 223122_s_at 959.02 6 210.7 251.85 6 54.29 0.26 0.014
1
Data were obtained by using the robust multiarray average method for normalization. Data from baseline measurements were based on 9 subjects. Data
from weight-loss measurements were based on 8 subjects. The gene list was sorted on the fold change for the comparison between baseline and weight loss.
2
Affymetrix.
3
FDR, false discovery rate defined by using the OCplus R package (Bioconductor; http://www.bioconductor.org) (18).
4
Mean 6 SEM (all such values).
WEIGHT LOSS AND ADIPOSE TISSUE GENE EXPRESSION 199
TABLE 3
Top 15 downregulated and upregulated genes during weight maintenance1
Acute weight Weight Fold
Gene Probe set2 loss maintenance change FDR3
kg kg
Top 15 upregulated (fold)
FADS1 208962_s_at, 208963_x_at 105.45 6 25.284 492.01 6 137.87 4.67 0.022
FADS1 208963_x_at, 208964_s_at 125.82 6 28.78 561.97 6 159.21 4.47 0.023
SCD 211708_s_at, 211162_x_at 532.18 6 238.55 2198.82 6 313 4.13 0.020
LTF 202018_s_at 93.25 6 23.94 352.9 6 172.45 3.78 0.021
ELOVL6 204256_at, 210868_s_at 88.67 6 57.11 318.21 6 92.56 3.59 0.023
FADS2 202218_s_at 91.05 6 19.22 318.14 6 35.67 3.49 0.018
THBS4 204776_at 354.06 6 82.08 1135.03 6 177.47 3.21 0.023
LOC55908 220437_at 222.57 6 75.11 700.24 6 206.13 3.15 0.025
B4GALT6 206233_at 52.71 6 9.09 160.59 6 32.5 3.05 0.020
FASN 212218_s_at 633.08 6 142.53 1848.95 6 108.19 2.92 0.019
SFRP2 223121_s_at, 223122_s_at 251.85 6 54.29 669.38 6 110 2.66 0.020
ALDOC 202022_at 674.17 6 143.27 1768.51 6 174.07 2.62 0.024
SCD 200832_s_at, 217200_x_at, 207986_x_at, 200831_s_at, 1965.77 6 559.11 5128.9 6 357.66 2.61 0.021
223839_s_at, 211162_x_at
KRT14 209351_at 27.61 6 4.48 70.44 6 15.01 2.55 0.028
SLC2A5 204430_s_at, 204429_s_at 77.82 6 18.83 196.24 6 40.76 2.52 0.037
normalization of the intensity of the microarray data and removal use of modified t statistics. We considered only those probe sets
of technical variation can have a major impact on results of as differentially expressed between groups that fulfilled each of
differential expression analyses, and many methods are available the following 2 requisites: had a global FDR ,0.05 by using the
(15). We applied the robust multiarray average method, which empirical operating characteristics function and a confidence
implements model-based background adjustment, quantile-based level 0.95 [which is equivalent to 5% of the FDR (19)] by using
normalization, and summarization on the basis of a multiarray the patterns from gene expression method. Both of these methods
model fit by using a median polish algorithm (16). We conducted provided correction for multiple comparisons. Gene lists were an-
filtering on the basis of the Microarray Suite 5.0 present/absent alyzed by using the WebGestalt system (http://bioinfo.vanderbilt.
calls, which classified each probe set as expressed above the edu/webgestalt/option.php) for the enrichment analysis of the Kyoto
background (present call) or not (absent or marginal call). We Encyclopedia of Genes and Genomes (KEGG) pathways by using
included probe sets, which have a detection call as present call in the hypergeometric test (20).
50% of arrays in one group (17), which left 13,463 of 23,798
probe sets for additional analysis. To identify differentially ex-
pressed genes for interclass paired comparisons, we used a permu- Quantitative real-time polymerase chain reaction analysis
tation that was based on the false discovery rate (FDR) approach of gene expression
calculated by using the empirical operating characteristics RNA from a total of 26 samples was transcribed by using the
function in the OCplus R package (Bioconductor; http://www. QuantiTect Whole Transcriptome kit (Qiagen) according to man-
bioconductor.org) (18) as well as the patterns from gene ex- ufacturer’s recommendations. Gene expression was analyzed by
pression method (19) for log (base 2)–transformed data with the using a TaqMan assay on demand on the 7900HT Fast Real-Time
200 JOHANSSON ET AL
TABLE 4
Top 15 downregulated and upregulated genes during the complete weight-loss program1
Weight Fold
Gene Probe set2 Baseline maintenance change FDR3
kg kg
Top 15 upregulated (fold)
CETP 206210_s_at 77.91 6 14.314 286.98 6 88.46 3.68 0.039
CA3 204865_at 26.64 6 6.62 80.95 6 19.19 3.04 0.041
PNPLA3 233030_at, 220675_s_at 86.48 6 13.63 250.29 6 61.68 2.89 0.050
SLC2A5 204430_s_at, 204429_s_at 91.98 6 15.04 202.68 6 36.52 2.20 0.041
AZGP1 209309_at, 217014_s_at 131.24 6 21.76 282.26 6 39.64 2.15 0.048
ADSSL1 226325_at 117.86 6 14.13 241.79 6 21.78 2.05 0.040
GLUL 200648_s_at, 217202_s_at 573.21 6 31.05 1052.17 6 114.84 1.84 0.037
STOX1 1553202_at 15.6 6 1.22 28.61 6 4.47 1.83 0.040
CDKN2C 211792_s_at, 204159_at 1403.41 6 149.11 2434.63 6 190.72 1.73 0.048
ENPEP 204845_s_at, 204844_at 72.72 6 8.41 126.12 6 15.68 1.73 0.048
AGTR1 205357_s_at, 208016_s_at 457.28 6 38.48 784.75 6 86.18 1.72 0.042
APOB 223579_s_at, 205108_s_at 46.12 6 5.9 77.6 6 11.4 1.68 0.048
NID2 204114_at 630.94 6 61.29 1061.2 6 95.44 1.68 0.045
ABCG1 211113_s_at, 204567_s_at 40.99 6 4.52 68.66 6 9.27 1.68 0.041
DBP 209782_s_at 120.78 6 4.33 193.84 6 22.02 1.60 0.044
Top 15 downregulated (fold)
PCR system (Applied Biosystems) according to the manufactu- were performed with Number Cruncher Statistical Systems 2004
rer’s recommendation. Five genes (CIDEA: Hs00154455_m1; software (NCSS). The correlation between CETP/ABCG1 gene
MMP9: Hs00234579_m1; PNPLA3: Hs00228747_m1; SCD: expression and blood HDL concentrations was tested by using
Hs01682761_m1; and TNMD: Hs00943212_m1) were selected for Spearman’s rank correlation.
confirmation of the microarray data by using real-time polymerase
chain reaction (PCR) on the basis of a significant fold change and
biological significance. Data were analyzed with the use of the DCt RESULTS
(comparative cycle time) method by using cyclophilin A as an
endogenous control (4326316E; Applied Biosystems). Cyclophilin Clinical characteristics
A was selected as an appropriate and stable control after testing
Clinical characteristics of study subjects at the 3 OGTT and
a subset of complementary DNA samples on a Human Endoge-
biopsy occasions (at baseline, after weight loss, and after group
nous Control Array (4367563; Applied Biosystems).
therapy to maintain the reduced weight) are shown in Table 1.
The LCD treatment resulted in an 18.8 6 5.4% (from 129 6 22
to 106 6 20 kg) loss of initial body weight (P = 0.0094) during
Statistical analysis of clinical data an average time of 101 6 26 d in 9 subjects from whom biopsies
The significance of differences in variables between different were analyzed by using a microarray. The achieved weight loss
time points was calculated by using the following robust tests: the was maintained throughout the group-therapy period (167 6 37
Kruskal-Wallis 1-factor ANOVA on ranks and Wilcoxon’s rank- d; P = 0.70; from 105 6 22 to 106 6 20 kg after group therapy;
sum test for difference in medians. All statistical calculations 1.1 6 2.1%; range: 29.3 to 10.6%). Insulin sensitivity improved
WEIGHT LOSS AND ADIPOSE TISSUE GENE EXPRESSION 201
KEGG analysis was performed on all significantly differen-
tially expressed genes shown during weight loss, during weight
maintenance, and for the whole process (ie, from baseline to
follow-up after weight maintenance). Results are shown sepa-
rately for pathways that were significantly upregulated or
downregulated (Table 5). During weight loss, genes involved in
metabolism, including fatty acid metabolism, were down-
regulated, whereas genes that are important for biogenesis and
cell proliferation were upregulated. On the contrary, during the
weight-maintenance phase, upregulated genes seemed to belong
to pathways involved in metabolism, whereas downregulated
genes belonged to pathways involved in immune regulatory
function. As for the individual transcripts, the results of the
pathway analysis allowed for groupings on the basis of regu-
lation (Table 6). After conservative Benjamini-Hochberger
adjustment, only changes for ribosomes, citric acid cycle, lyso-
some, and glycosylphosphatidylinositol-anchor biosynthesis were
considered significant, and all other pathways were considered to
be reciprocally regulated (ie, returned toward baseline at follow-
FIGURE 2. Venn diagram. Several regulated genes showed overlap up after weight maintenance regardless of whether they were
between comparisons.
FIGURE 3. Scatter plots. The prevailing HDL concentration correlated to the relative expression of CETP (r = 0.42, P = 0.034) (A) and the relative
expression of ABCG1 (r = 0.55, P = 0.0004) at 3 assessments (B).
TABLE 5 202
KEGG pathway analysis of genes significantly regulated in relation to weight loss and weight maintenance1
Reference Observed Expected Ratio of P (hypergeometric P (Benjamini-Hochberger
Regulation KEGG pathway KEGG ID genes genes genes enrichment test) adjusted)
Up Oxidative phosphorylation 190 114 46 7.26 6.34 1.15 · 10225 8.74 · 10224
Up Parkinson disease 5012 111 43 7.06 6.09 3.11 · 10223 1.58 · 10221
Up Alzheimer disease 5010 155 47 9.87 4.76 4.43 · 10220 1.68 · 10218
Up Huntington disease 5016 167 45 10.63 4.23 4.47 · 10217 1.36 · 10215
Up Citrate cycle (TCA cycle) 20 30 17 1.91 8.9 2.25 · 10213 5.70 · 10212
Up Pyruvate metabolism 620 40 19 2.55 7.46 5.92 · 10213 1.29 · 10211
Up Glycolysis/gluconeogenesis 10 60 22 3.82 5.76 5.48 · 10212 1.04 · 10210
Up Propanoate metabolism 640 32 15 2.04 7.36 2.12 · 10210 3.58 · 1029
Up Valine, leucine, and isoleucine 280 43 17 2.74 6.21 3.58 · 10210 5.44 · 1029
degradation
Down Ribosome 3010 85 50 7.61 6.57 8.20 · 10231 1.30 · 10228
Down Lysosome 4142 116 45 10.39 4.33 2.41 · 10218 1.90 · 10216
Down Viral myocarditis 5416 67 22 6 3.67 4.01 · 1028 2.11 · 1026
Down Toll-like receptor 4620 99 26 8.87 2.93 3.86 · 1027 1.52 · 1025
signaling pathway
Down Leukocyte transendothelial 4670 114 27 10.21 2.64 2.11 · 1026 4.17 · 1025
migration
Down MAPK signaling pathway 4010 265 48 23.74 2.02 1.87 · 1026 4.17 · 1025
(Continued)
Down Antigen processing and 4612 81 22 7.26 3.03 1.61 · 1026 4.17 · 1025
presentation
Down Fc c R-mediated phagocytosis 4666 93 24 8.33 2.88 1.51 · 1026 4.17 · 1025
Down Regulation of actin cytoskeleton 4810 208 39 18.63 2.09 7.21 · 1026 0.0001
Down Intestinal immune network for 4672 48 15 4.3 3.49 1.14 · 1025 0.0002
IgA production
Baseline to weight maintenance
Up Ribosome 3010 85 13 1.21 10.75 2.17 · 10210 1.50 · 1028
Up Citrate cycle (TCA cycle) 20 30 4 0.43 9.37 0.0008 0.0276
Up Tyrosine metabolism 350 44 4 0.63 6.39 0.0035 0.0604
Up Fatty acid metabolism 71 41 4 0.58 6.86 0.0027 0.0604
Up Glycerolipid metabolism 561 44 3 0.63 4.79 0.0246 0.0998
Up Glycerophospholipid metabolism 564 67 4 0.95 4.2 0.0152 0.0998
Up Lysine degradation 310 42 3 0.6 5.02 0.0217 0.0998
Up Oxidative phosphorylation 190 114 5 1.62 3.08 0.0235 0.0998
Up Pyruvate metabolism 620 40 3 0.57 5.27 0.0191 0.0998
Up Glycolysis/gluconeogenesis 10 60 4 0.85 4.69 0.0105 0.0998
Down Lysosome 4142 116 20 2.82 7.08 5.79 · 10212 5.56 · 10210
Down Proteasome 3050 42 7 1.02 6.85 6.24 · 1025 0.003
Down GPI-anchor biosynthesis 563 24 5 0.58 8.56 0.0002 0.0064
Down Other glycan degradation 511 15 4 0.37 10.96 0.0004 0.0096
Down Porphyrin and chlorophyll 860 33 5 0.8 6.22 0.0011 0.0211
metabolism
Down Nicotinate and nicotinamide 760 24 4 0.58 6.85 0.0025 0.04
metabolism
Down Glycosphingolipid 604 15 3 0.37 8.22 0.0052 0.0555
biosynthesis – ganglio series
WEIGHT LOSS AND ADIPOSE TISSUE GENE EXPRESSION
Weight loss
Up-regulation No pathways Ribosomes [2 No pathways
No pathways Y
Down-regulation Oxidative phosphorylation [ No pathways Glycosylphosphatidylinositol-
anchor biosynthesis pathway2
Metabolic pathways Y
No regulation Pathways for the citric acid cycle2 Pathways for lysosome*
Pyruvate metabolism Pathways for actin cytoskeleton
Glycolysis/gluconeogenesis
1
[ indicates that the end level was higher than the baseline level; Y indicates that the end level was lower than the baseline level.
2
Significant using the Hypergeometric test with the adjusted P value using Benjamini-Hochberger.
biologically distinct events. Most genes and pathways in the mediate the release and activation of growth factors as well as
adipose tissue identified by the complementary DNA microarray cleave cell-surface receptors (25). TNMD constitutes another
appeared to be reciprocally regulated during weight loss and interesting example. Kolehmainen et al (13) investigated the
antiatherogenic process termed reverse cholesterol transfer (52). vention Study (DPS): lifestyle intervention and 3-year results on diet
The relation between this process (ie, HDL functionality) and and physical activity. Diabetes Care 2003;26:3230–6.
8. Pories WJ, Swanson MS, MacDonald KG, Long SB, Morris PG,
decreased risk of coronary artery disease was recently confirmed Brown BM, Barakat HA, deRamon RA, Israel G, Dolezal JM, et al.
in human subjects (53). CETP encodes the cholesteryl ester Who would have thought it? An operation proves to be the most ef-
transfer protein that catalyzes the exchange of cholesterol for tri- fective therapy for adult-onset diabetes mellitus. Ann Surg 1995;222:
glycerides between HDL and very-low-density and intermediary- 339–50, discussion 50–2.
density lipoproteins to produce LDL particles that can be cleared 9. Bouchard L, Rabasa-Lhoret R, Faraj M, Lavoie M-E, Mill J, Pérusse L,
Vohl M-C. Differential epigenomic and transcriptomic responses in
by the liver. thus complementing the reverse cholesterol transfer subcutaneous adipose tissue between low and high responders to ca-
that HDL stands for by itself (54). The upregulation of both of loric restriction. Am J Clin Nutr 2010;91:309–20.
these transcripts may have simply reflected the greater concentra- 10. Capel F, Klimčáková E, Viguerie N, Roussel B, Vı́tková M, Kováčiková
tions of HDL cholesterol after the weight-maintenance phase, but is M, Polák J, Kováčová Z, Galitzky J, Maoret J-J, et al. Macrophages and
adipocytes in human obesity. Diabetes 2009;58:1558–67.
tempting to speculate about a role for them in the beneficial effects 11. Clément K, Viguerie N, Poitou C, Carette C, Pelloux V, Curat CA,
of losing weight and remaining weight stable. The correlation be- Sicard A, Rome S, Benis A, Zucker JD, et al. Weight loss regulates
tween CETP and ABCG1 expression amounts and HDL concen- inflammation-related genes in white adipose tissue of obese subjects.
trations and the increase in these amounts from baseline to weight FASEB J 2004;18:1657–69.
maintenance strengthened this hypothesis. The upregulation of 12. Dahlman I, Linder K, Arvidsson Nordstrom E, Andersson I, Liden J,
Verdich C, Sorensen TI, Arner P. Changes in adipose tissue gene ex-
ABCG1 but not of CETP was recently demonstrated in adipose pression with energy-restricted diets in obese women. Am J Clin Nutr
tissue after weight loss induced by bariatric surgery (55). Po- 2005;81:1275–85.
tential explanations for this difference include greater weight 13. Kolehmainen M, Salopuro T, Schwab US, Kekalainen J, Kallio P,
loss, shorter follow-up time, and different diets. Laaksonen DE, Pulkkinen L, Lindi VI, Sivenius K, Mager U, et al.
Weight reduction modulates expression of genes involved in extracel-