Molecular alterations induced by a high-fat high-fiber diet in porcine adipose tissues

17 Pages·2017·1.25 MB·English

Molecular alterations induced by a high-fat high-fiber
diet in porcine adipose tissues: variations according to
the anatomical fat location
Florence Gondret, Annie Vincent, Magalie Houée-Bigot, Anne Siegel,
Sandrine Lagarrigue, David Causeur, Isabelle Louveau
To cite this version:
Florence Gondret, Annie Vincent, Magalie Houée-Bigot, Anne Siegel, Sandrine Lagarrigue, et al..
Molecular alterations induced by a high-fat high-fiber diet in porcine adipose tissues: variations ac-
cording to the anatomical fat location. BMC Genomics, 2016, 17 (1), pp.120. ￿10.1186/s12864-016-
2438-3￿. ￿hal-01286555￿
HAL Id: hal-01286555
https://hal.science/hal-01286555
Submitted on 26 Sep 2017
HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est
archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Gondretetal.BMCGenomics (2016) 17:120
DOI10.1186/s12864-016-2438-3
RESEARCH ARTICLE Open Access
Molecular alterations induced by a high-fat
high-fiber diet in porcine adipose tissues:
variations according to the anatomical fat
location
Florence Gondret1,2*, Annie Vincent1,2, Magalie Houée-Bigot3, Anne Siegel4, Sandrine Lagarrigue1,2,
Isabelle Louveau1,2 and David Causeur3
Abstract
Background: Changing the energy and nutrient source for growinganimals may be an effective way of limiting
adipose tissue expansion, a response which may depend onthegenetic background oftheanimals. Thisstudy
aims to describe thetranscriptional modulations present in theadiposetissues of two pig lines divergently selected
for residual feedintake which were either fed a high-fat high-fiber (HF) diet or an isocaloric low-fat high-starch diet
(LF).
Results: Transcriptomic analysis using a porcine microarraywas performed on 48 pigs (n=12 perdiet and per line)
inboth perirenal (PRAT) and subcutaneous (SCAT) adipose tissues. There was no interaction between diet and line
oneither adiposity or transcriptional profiles,so that thediet effectwas inferred independently oftheline.
Irrespective ofline, therelative weights of the twofatdepots were lower in HF pigs than inLF pigs after 58 days
ondietary treatment. Inthe two adipose tissues, themost apparent effect of the HF diet was thedown-regulation
of several genes associated with the ubiquitin-proteasome system, which therefore may be associated with
dietary-induced modulations in genes acting in apoptotic and cell cycle regulatory pathways. Genes involved
in glucose metabolic processes were also down-regulated by the HF diet, with no significant variation or
decreased expression of important lipid-related genes such as the low-density lipoprotein receptor and leptin
in the two fat pads. The master regulators of glucose and fatty acid homeostasis SREBF1 and MLXIPL, and
peroxisome proliferator-activated receptor (PPAR)δ and its heterodimeric partner RXRA were down-regulated by
the HF diet. PPARγ which has pleiotropic functions including lipid metabolism and adipocyte differentiation,
was however up-regulated by this diet in PRAT and SCAT. Dietary-related modulations in the expression of
genes associated with immunity and inflammation were mainly revealed in PRAT.
Conclusion: A high-fat high-fiber diet depressed glucose and lipid anabolic molecular pathways, thus
counteracting adipose tissue expansion. Interaction effects between dietary intake of fiber and lipids on gene
expression may modulate innate immunity and inflammation, a response which is of interest with regard to
chronic inflammation and its adverse effects on health and performance.
Keywords: Adipose tissue, Dietary fiber, High-fat diet, Gene module, Multiple factor analysis, Transcriptome,
Upstream regulators
*Correspondence:[email protected]
1INRA,UMR1348Pegase,F-35590Saint-Gilles,France
2Agrocampus-Ouest,UMR1348Pegase,F-35000Rennes,France
Fulllistofauthorinformationisavailableattheendofthearticle
©2016Gondretetal.OpenAccessThisarticleisdistributedunderthetermsoftheCreativeCommonsAttribution4.0
InternationalLicense(http://creativecommons.org/licenses/by/4.0/),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedyougiveappropriatecredittotheoriginalauthor(s)andthesource,providealinkto
theCreativeCommonslicense,andindicateifchangesweremade.TheCreativeCommonsPublicDomainDedicationwaiver
(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthisarticle,unlessotherwisestated.
Gondretetal.BMCGenomics (2016) 17:120 Page2of16
Background total body fat,perirenal (visceral) fatgrows fasterinpigs;
White adipose tissue (AT) has a central role in various thus, the distinction between these two anatomical fat
physiological processes related to lipid metabolism and sites is also economically important. Hence, similarities
energy homeostasis in mammals [1, 2] and is considered and particularities in the molecular responses to diet of
to be a significant contributor to systemic low-grade in- subcutaneous and perivisceral ATs are worthy of investi-
flammation [3]. Adipose tissue development is triggered gation. It has recently been shown that AT expansion
bythe imbalance between energyintake andexpenditure was reduced at the subcutaneous and perirenal locations
and most related studies have exploited high-fat hyper- when growing pigs of two divergent RFI lines were fed a
caloric diets to generate adipose tissue expansion and high-fat high-fiber diet as compared with pigs fed an
document molecular alterations [4–6]. The nutrient isocaloric low-fat high-starch diet [21]. Molecular mech-
source is also important in determining adiposity. anisms and processes associated with both diet and line
Noticeably, dietary fibers, which are poorly digested by in the dorsal subcutaneous and perirenal ATs are ex-
monogastric animals, are associated with energy dilution plored in the present study. Pathways common to both
of the diet, hence reducing the adiposity that accompan- ATs and also specificities within each AT in the re-
ies the chronic intake of high levels of fat [7, 8]. This re- sponses to diet, are highlighted using dedicated inte-
duction results from a lower food intake, enhanced grated statistical methods.
insulin sensitivity, increased energy expenditure, and
changes in expression levels of some genes involved in Results
lipid anabolism and fatty acid β-oxidation, in different Summaryofphenotypicdata
organs [7, 9]. However, the diversity in molecular path- Forty-eightgrowingpigsoftwolinesdivergentlyselected
ways in AT which can be altered by a high-fat high-fiber on RFI, a measure of feed efficiency, were fed diets for-
diet remains still to be clarified, together with the ways mulated at isocaloric and isoproteic bases but differing
by which these changes relate to adiposity variations. in energy source and nutrients (lipids and fibers vs.
The pig is recognized as an experimental model to in- starch). Full description of performance of experimental
vestigate genetic and mechanistic aspects of human dis- pigs(n=12 perline andper diet) after 58days ofdietary
ease such as obesity [10–12], and is a leading source of treatment can be found in an associated paper [21] and
animal protein in human diets. The introduction of are briefly summarized here. Importantly, there was no
more fiber to cereal-based diets formulated for pigs has interaction between diet and line on performance and
regained interest due to both economic considerations body composition. Irrespective of RFI line, pigs fed the
and the potential benefits on health and welfare [13]. high-fat high-fiber (HF) diet ate 12 % less (p<0.001)
While, the addition of fat to a fiber-rich diet is also re- than pigs fed the low-fat low-fiber starch-based (LF) diet
quired to retain a dietary energy value compatible with during the trial and they developed less body fat, as
good animal performance [14]. Therefore, the primary illustrated by 1.4-fold lower proportions of ATs at the
objective of this study was to decipher the transcrip- perirenal (%PRAT) and dorsal subcutaneous (%SCAT)
tomic changes that take place in AT when pigs received locations. The differences between diets were significant
either a high-fat high-fiber diet or an isocaloric low-fat when the adiposity data was corrected for individual
high-starch diet, thus changing energy source and nutri- daily feed intake, indicating that AT development was
ents rather than energy content in diet. The availability triggered not only by differences in feed intake between
of pig lines divergently selected for residual feed intake pigs but also by the nutrient composition. Irrespective of
(RFI), a measure of feed efficiency, offers an additional diet, divergent selection for RFI did not affect AT
opportunity to consider line-associated differences in weightsinthis experiment (Table1).
basal metabolism and energy expenditure [15, 16] when
reasoning body composition and dietary responses. Both Dietmarkedlyaffectedadiposetissuemolecularprofiles
independent and possible interaction effects between Microarray analyses were separately performed in PRAT
dietandlineontheATtranscriptionalprofileshavethen and in SCAT from the 48 experimental pigs. After
tobe considered. normalization,thesignalintensitywasfoundtobeabove
The anatomical location of AT is also important when background noise for 38,142 spots in PRATand 38,987
analyzing the molecular responses to diet. In many spots in SCAT, respectively. The linear model to analyze
species [17, 18] including pigs [19, 20], differences in the log2-transformed signals revealed no significant
adipogenic factors, responses to hormones, metabolic interaction (p>0.01) between diet and line on the ex-
properties or secretion of inflammatory cytokines have pression levels of the molecular probes in the two ATs,
been observed between subcutaneous AT and visceral so that diet and line effects could be inferred independ-
ATs which are located within the abdominal cavity. ently. The global effects of diet and of line on the num-
Whereas subcutaneous fat grows at the same rate as ber of differentially-expressed probes (DEP) were more
Gondretetal.BMCGenomics (2016) 17:120 Page3of16
Table1Growthandadiposityasaffectedbydietandline
Diet HF LF Statistics
Line Low High Low High p-diet p-line
Pigsa,n 12 12 12 12
Initialage(d) 73 73 75 75 0.77 0.04
Initialweight(kg) 25.7 26.6 26.5 27.1 0.62 0.54
Finalage(d) 132 131 134 133 0.49 0.12
Finalbodyweight(kg) 72.8 69.0 82.7 77.6 <0.001 0.008
Feedintake(kg/d) 2.13 2.12 2.45 2.40 <0.001 0.45
Gaintofoodratio 0.38 0.34 0.39 0.36 0.13 <0.001
Adiposetissuemass(%bodyweight)
PerirenalAT 0.59 0.54 0.83 0.83 <0.001 0.46
SubcutaneousAT 3.90 4.10 5.90 5.70 <0.001 0.41
aPigswerefedahigh-fathigh-fiberdiet(HF)oralow-fathigh-starchdiet(LF).Durationofthefeedingtrialwas58.5±0.5daysforallpigs.Twodivergentpiglines
thathavebeenselectedovereightgenerationsforresidualfeedintake(RFI),ameasureoffeedefficiency,wereused.PigsselectedforlowRFIwerethemost
efficientintheconversionoffoodtoweightgain.Therewasnosignificantinteraction(p>0.10)betweendietandlineonperformanceandadipositytraits.
Attheendofthetrial,weightsofperirenalanddorsalsubcutaneousadiposetissues(AT)wereusedassurrogatesofbodyadiposity
pronounced in PRAT than in SCAT (Fig. 1). In the two and of 2,107 DEP in SCAT were found as over-
ATs,diethadalsoamorepronouncedeffectonthetran- expressed by HF diet. These corresponded to 1,251
scriptome than did the line. Thus, only data related to unique differentially-genes (DEG) up-regulated in PRAT
dietary effects are further considered in this manuscript. and to 825 unique DEG up-regulated in SCAT. Con-
By using cut-offs of fold-change between diets>|1.1| versely, 4,820 DEP corresponding to 2,440 unique DEG
and p-value<0.01, while ensuring Benjamini-Hochberg in PRAT, and 2,689 DEP corresponding to 1,279 unique
(BH) multiplicity correction of the p-values for a control DEG in SCAT were under-expressed by HF diet.
of the FDR at level<0.08, a total of 3,313 DEP in PRAT Altogether, 1,330 DEG were commonly affected by diet
across the two ATs (Fig. 1), which corresponded to 36 %
of the DEG in PRAT and 63 % of the DEG in SCAT,
respectively.
Similaritiesinthemolecularresponsestodietacrosstwo
adiposetissues
Multiple Factor Analysis (MFA) was performed to derive
a common framework of molecular responses to diet
and identify the communalities across AT. The basic
interest of MFA as a data integration procedure [22] was
to ensure that each tissue influence was equally weighed
in a manageable number of comprised factors, which
can then be related to external phenotypic characteris-
tics. Figure 2 shows the diagnostic MFA plot. The first
dimension(Dim1)oftheMFAaccountedfor31%ofthe
Fig1Numberofdifferentially-expressedprobesandcorresponding
variation, and correlated significantly to variations in AT
unique genes in adipose tissues as affected by diet and line.
weights (r=0.78 with %PRAT, and r=0.76 with %SCAT,
Microarray data obtained in subcutaneous (SCAT) or in perirenal
(PRAT) adipose tissues were separately analyzed for the main p<0.001), so that pigs fed the HF diet and pigs fed the
effects of diet (HF: high-fat high-fiber vs. LF: low-fat high-starch), LF diet were oppositely represented along Dim1. The
line (low RFI: residual feed intake below the average; high RFI: second dimension of the MFA (Dim2) explained 8.2 %
residualfeedintakeabovetheaverage),andtheinteractionbetween
of the variation and did not clearly separate the data ac-
dietandline.Molecularprobesweredeclaredasdifferentially-expressed
cording todiet oradiposity phenotypic traits.
betweendietsorbetweenlinesaccordingtocutoffsforfold-change
betweenconditions>|1.1|andp<0.01(Benjamin-Hochbergadjusted As illustrated (Fig. 2), the two first comprised variables
p-value<0.08).Venndiagramsillustratethenumberofdifferentially- synthesizing the molecular information of the originally-
expressedprobesineachexperimentalgroupforSCATandforPRAT. distinct data tables (dim_1_PRATand dim_1_SCAT, re-
Thecorrespondingnumberofdifferentially-expresseduniquegenesis
spectively) were projected together along Dim1 in the
indicatedintobrackets
MFA plot. This means that large similarities between
Gondretetal.BMCGenomics (2016) 17:120 Page4of16
Fig2(Seelegendonnextpage.)
Gondretetal.BMCGenomics (2016) 17:120 Page5of16
(Seefigureonpreviouspage.)
Fig2Multi-waydatasetsanalysis:consensusinmicroarraydatarelativetodietaryeffectacrosstwoadiposetissues.Thefirsttwosynthetic
variablesobtainedfortheperirenal(Dim_1_PRAT)andsubcutaneousadiposetissue(Dim_1_SCAT)transcriptomeswereprojectedinthe
correlationcircleofthemultiplefactoranalysis(MFA),anintegratedstatisticalmethodusedtorevealcommunalitiesacrossseparatedatasets.
Largesimilaritiesacrossadiposetissuescanbededucedfrommolecularvariablescontributingtothefirstdimension(Dim1)ofMFA(Fig.1a).
Relativeweightsofperirenalfat(%PRAT)andsubcutaneousfat(%SCAT)weresuperimposedontheplot,showingstrongcorrelationbetweenthe
molecularprobescontributingtoDim1andadiposityvariations.PigswererepresentedonthescatterMFAplot(Fig.1b)andcoloredfollowing
thediettheyreceived(HF:high-fathigh-fiber;LF:low-fathigh-starch).Thisshowsawell-definedpartitionofpigsbetweendietsalongDim1
ATs could be inferred by focusing on the DEP that con- or linking glycolysis and tricarboxylic acid cycle
tributed mainly to Dim1. A total of 1,128 DEP showing (PDK1, PDHA1, PDHA2, DLAT). Different factors in
the highest correlation with Dim1 (r>|0.70|; p<0.001) coenzyme metabolic processes such as those related
were thus commonly regulated by diet across the two to the synthesis of cytosolic acetyl-CoA (ACLY) and
ATs. The identity of these DEP together with their lipid anabolic process (ME1, HMGCR) were also all
fold-changes between HF and LF diets are listed in down-regulated by the HF diet, with the noticeable
Additional file 1: Table S1 for PRAT and for SCAT, exception of glutathione S-transferase kappa 1 (GSTK1)
respectively; as indicated in this additional file, all these which functions in cellular detoxification. Conversely,
DEP were altered by diet (BH adjusted p-value<0.05 in cellular responses to stress included some genes up-
PRATandBHadjustedp-value<0.08inSCAT).Theycor- regulated in the HF diet (e.g.; IFI16 and JMY, two genes
responded to 436 unique differentially-expressed genes known to interact with p53 in regulating cell cycle pro-
(DEG). This statistical approach was thus useful in redu- gression)aswellasgenesthatweredown-regulatedbythe
cing the number of DEG commonly regulated by diet same diet (e.g.; CDKN1A, another regulator of cell cycle
across the two ATs to those having the strongest correl- progression). Other gene groups with a lower enrichment
ationwithvariationsinATweights. score (E<1.3) but p<0.05 could be also regarded as po-
A functional analysis was performed to understand the tentially interesting in deciphering the molecular mecha-
biological meaning behind this subset of 436 DEG, by nismsmodulatedbythediet,i.e.theresponsetohormone
using automatic functional annotation tools to identify stimulus, regulation of cell death and cell mitotic process
biological gene ontology (GO) terms and clustering re- (Table 2). Indeed, although dietary-modulated transcripts
dundant annotation terms in enriched biological path- in ATs were not primarily enriched in lipid metabolism,
ways. Enrichment (E) score≥1.3 and p-value<0.5 individual genes clustered in the response to hormone
were used to select the top-enriched clusters among stimulus had known relationships with lipid metabolic
these DEG. As indicated in Table 2, the modification- process: the cytosolic enzyme malic enzyme (ME1) that
dependent protein catabolic process, intracellular pro- generates NADPH for fatty acid biosynthesis and which
tein transport, coenzyme metabolic process, cellular was down-regulated>2 fold (BH-p value<0.01) by the
response to stress, phosphorus metabolic process and HF diet, the leptin (LEP) coding for a protein secreted in
hexose metabolic process were the top-enriched bio- ATs to regulate food intake and energy expenditure and
logical pathways in the common molecular responses whichexhibited>1.5-foldreductioninitsexpressionlevel
of ATs to diet. The detailed list of genes included in in HF pigs (BH-p value<0.001), the low-density lipopro-
these clusters can be found in Additional file 2: tein receptor (LDLR) reduced by>1.4-fold (BH-p<0.001)
Table S2. To have details on fold-change of these by HF diet, and the adiponectin receptor (ADIPOR2) that
genes between diets, one may also consult Additional mediateslipidoxidationandwasalsosignificantlylowered
file 1: Table S1. (BH-corrected p<0.001) for pigs fed the HF diet.
Among protein catabolic process, the ubiquitin- Conversely, IGF1R coding for the insulin-growth
dependent proteolysis was the most represented factor 1 receptor binding IGF-I, a well-known regu-
pathway with a total of 20 genes encoding ubiquitin- lator of cell development, and which was first listed
conjugating enzymes, different ubiquitin ligases, vari- in the significantly-enriched phosphorus metabolic
ous ubiquitin-like proteins and different proteasome process, was also included in the less enriched clus-
complex subunits responsible for the degradation of ters corresponding to the response to hormone
polyubiquitinated proteins, which were down- stimulus and the regulation of cell death; this gene
regulated by the HF diet (Table 2). Similarly, the was one of the top genes being up-regulated by HF
glucose metabolic process assembled 13 genes, all diet. Altogether, different genes were thus listed in
down-regulated by the HF diet, and coding for en- more than one of these pathways, suggesting that
zymes acting in the first step of glucose use (HK2), these various biological processes were at least in
catalyzing glycogen synthesis (GSK3B, GYS2, UGP2) part, inter-connected.
Gondretetal.BMCGenomics (2016) 17:120 Page6of16
Table2Enrichedbiologicalprocessescommonlyregulatedbydietacrossadiposetissues
Functionalannotationofresponsestodiet Escorea NbDEG p-value HFvs.LFdiet
GO:0019941~modification-dependentproteincatabolicprocess 4.31 34 <0.001 ↓:31↑:3
(GO:0006511~ubiquitin-dependentproteincatabolicprocess) (21)
GO:0046907~intracellulartransport 3.45 37 <0.001 ↓:29↑:8
(GO:0006886~intracellularproteintransport) (26)
GO:0006511~ubiquitin-dependentproteincatabolicprocess 2.76 21 <0.001 ↓:20↑:1
(GO:0010498~proteasomalproteincatabolicprocess) (11)
GO:0051186~coenzymemetabolicprocess 1.72 11 0.003 ↓:10↑:1
GO:0033554~cellularresponsetostress 1.58 28 <0.001 ↓:18↑:10
GO:0016310~phosphorylation 1.42 31 0.017 ↓:21↑:10
GO:0019318~hexosemetabolicprocess 1.32 13 0.003 ↓:13
(GO:0006006~glucosemetabolicprocess) (11)
(hsa00620:Pyruvatemetabolism) (7)
(hsa00010:Glycolysis/Gluconeogenesis) (5)
GO:0017038~proteinimport 1.25 9 0.017 ↓:8↑:1
GO:0009725~responsetohormonestimulus 0.99 16 0.042 ↓:10↑:6
GO:0010941~regulationofcelldeath 0.94 33 0.007 ↓:26↑:7
GO:0000278~mitoticcellcycle 0.66 18 0.010 ↓:15↑:3
aThemostimportantgenesinthecommondietary-relatedresponsesacrossperirenalandsubcutaneousadiposetissuesandcorrelatingwithadiposityvariations
wereconsideredtolistenrichedfunctionalclusters.Foreachclusterbasedongeneontology(GO)termsforbiologicalprocessesandGenomes(KEGG)biological
pathways,theenrichment(E)score,thenumber(Nb)ofdifferentially-expressedgenes(DEG)andassociatedp-valuewereprovided.Thedirectionofchangesin
geneexpressionlevelsbytheHFdietwassymbolized(↑:up-regulated;↓:down-regulated)togetherwiththenumberofuniquegenesconcernedbyeachchange
Particularitiesofperirenaladiposetissueresponsetodiet included a higher number of DEP from PRAT than from
To provide another representation of the transcriptional SCAT, and the second one (blue) had almost an equal
changesinATsresponsetodiet,theWeighedGeneCor- numberofDEPfromthetwoadiposetissues.Eigengenes
relation Network Analysis (WGCNA) was used to cap- in the turquoise and blue modules were highly corre-
ture strong relationships between transcripts in modules lated to the first comprised dimension Dim1 in MFA
of interconnected genes [23]. In each module, the eigen- (data not shown), suggesting that these modules in-
gene (the weighed mean of the transcripts providing the cluded not much more supplementary biological mean-
best univariate summary of the within-module variabil- ing than that deduced from Dim1. Two smaller network
ity) was calculated to relate transcriptional changes to modules represented 11 % and 7 % of the DEP data set,
external phenotypic traits and to deduce the biological respectively. The brown module corresponded mainly to
meaning of the module. Four distinct network modules a co-expression network in DEP from SCAT; however,
were thusobtained (Table3). most of these DEP were also found in PRAT including
The first two big modules represented 48 % and 20 % gene transcripts related to protein catabolic process,
of the DEP, respectively. The first module (turquoise) protein transport, pyruvate metabolism and cell respir-
ation. The pattern of expression in the yellow module
was mainly assigned to PRAT; only 4 % of the DEP in
Table3Co-expressedgenenetworksinadiposetissuesin
this module were also listed as differentially-expressedin
responsetodiet
SCAT in response to diet. In addition, the eigengene of
Inter-dependent NumberofDEPineachmodule
modules this yellow module was not highly correlated with
Total OfPRAT OfSCAT %PRAT (r=−0.25; p=0.09), suggesting that molecular
Turquoisea 6,720 5,302 1,418
mechanisms unrelated to adipose expansion were pre-
Blue 2,738 1,416 1,322 dominant in this module. Thus, biological particularities
Brown 1,238 79 1,159 at the PRAT location in the dietary-associated responses
Yellow 973 931 42 can be revealed by further examining the interconnected
aFiveco-expressedgenenetworkmoduleswerededucedfromthedatasetof genesintheyellow module.
differentially-expressedprobes(DEP)inresponsetodiet(high-fathigh-fibervs. A total of 973 DEP corresponding to 411 unique DEG
low-fathighstarch)andobtainedinperirenal(PRAT)andsubcutaneous(SCAT)
were listed in the yellow module. Most of the DEP (969
adiposetissues.ThetotalnumberofDEPincludedwithineachmoduleand
thenumberofDEPfromeachtissueinthemodulewereindicated out of 973) were up-regulated by the HF diet (see
Gondretetal.BMCGenomics (2016) 17:120 Page7of16
Additional file 3: Table S3). Altogether, 207 DEP corre- Table5Mostimportantgenesparticipatingtoparticularitiesin
sponding to 91 unique DEG, all up-regulated by the HF responsetodietofperirenaladiposetissue
diet, were highly correlated to the eigengene of the Genesymbol Fullname Connectivitya
yellow module (r>|0.70|, p<0.001), so that they could IL10 Interleukin10 205.8
be consideredasvery relevantforthe biologicalmeaning C1QB Complementcomponent1,q 204.9
of this module. Functional clustering of these 91 DEG subcomponent,Bchain
revealed defense response including inflammatory and CYBB Cytochromeb-245,betapolypeptide 201.9
innate immune responses, membrane organization,
C1QC Complementcomponent1,q 201.9
natural killer cell mediated cytotoxicity, and protein subcomponent,Cchain
amino-acid phosphorylation as the top-enriched bio- C1QA Complementcomponent1,q 199.2
logical processes (Table 4); the detailed list of genes subcomponent,Achain
within each cluster can be found in Additional file 4: ARRB2 Arrestin,beta2 195.2
Table S4.
CYTH4 Cytohesin4 194.8
Intra-modular connectivity can measure how con-
CYBA Cytochromeb-245,alphapolypeptide 194.7
nected a gene is to the other genes of its module. Intra-
modular connectivity values pointed the following genes NCKAP1L NCK-associatedprotein1-like 193.9
as among the 20 first most important nodes in the FCER1G FcfragmentofIgE,highaffinityI, 193.3
receptorforgammapolypeptide
yellow module (Table 5): the cytokine interleukin-10
(IL10) and its receptor (IL10RA) which are involved in VAV1 Vav1guaninenucleotideexchange 192.9
factor
anti-inflammatory response, three q sub-components of
the complement component 1 system (C1QA, C1QB, FOLR1 Folatereceptor1 191.6
C1QC) and the cytochrome b-245 (CYBB, CYBA) which CD53 CD53molecule 190.9
are involved in reactions participating to innate immun- VSIG4 V-setandimmunoglobulindomain 190.6
ity, arrestin beta-2 (ARRB2) causing specific dampening containing4
of cellular responses to stimuli, and a member of the IL10RA Interleukin10receptor,alpha 189.2
VAV gene family (VAV1) as a natural killer playing role CFP Complementfactorproperdin 188.8
inT-cellandB-cell development. VSIG4 V-setandimmunoglobulindomain 188.2
containing4
Upstreamregulatorycandidates CADM1 Celladhesionmolecule1 188.1
The two above-described analyses have clearly shown
NCF4 Neutrophilcytosolicfactor4,40kDa 187.8
that different biological pathways were involved in re-
F13A1 CoagulationfactorXIII,A1polypeptide 187.7
sponses to diet; therefore, it is important to identify
upstream regulators which coordinate these molecular aThefirst20geneswererankedaccordingtotheirconnectivityvaluewithin
theyellowmodule,whichassembledatotalof411uniquegenesinperirenal
changes in porcine ATs. For this purpose, the transcripts adiposetissue,allup-regulatedbyahigh-fathigh-fiberdiet.Thehigheristhe
of genes identified as differentially-expressed by diet connectivityvalue,thestrongeristheevidencethatthegeneispartof
themodule
Table4Biologicalmeaningofco-expressedgenesinperirenaladiposetissueofpigs
BiologicalProcess Nbofgenes Escore p-value HFvs.LFdiet
GO:0006952~defenseresponse 21 6.96 <0.001
(GO:0006954~inflammatoryresponse) (15)
(GO:0045087~innateimmuneresponse) (10)
GO:0016044~membraneorganization 11 3.92 <0.001
(GO:0006897~endocytosis) (10)
GO:0045087~innateimmuneresponse 10 3.27 <0.001 ↑
(GO:0006956~complementactivation) (5)
hsa04650:Naturalkillercellmediatedcytotoxicity 9 2.54 <0.001
(hsa04664:FcepsilonRIsignalingpathway) (6)
GO:0006468~proteinaminoacidphosphorylation 11 2.34 0.004
(GO:0007243~proteinkinasecascade)
aUniqueannotatedgenesbeinghighlycorrelated(correlations>|0.70|)totheeigengeneoftheyellowsmallmodule(Table3)wereinferredfortheirgene
ontology(GO)termsforbiologicalprocesses.Functionalclustershavingenrichment(E)score>2andmodifiedFisherexactp-valuelessthan0.01wereindicated.
Allgeneswereup-regulated(↑)inperirenaladiposetissuewhenpigswerefedahigh-fathigh-fiber(HF)dietcomparedwithalow-fathigh-starch(LF)diet
Gondretetal.BMCGenomics (2016) 17:120 Page8of16
within each AT were submitted to a dedicated algorithm This included the under-expression by the HF diet of
(http://keyregulatorfinder.genouest.org/) which matches the MLX interactingprotein-like MLXIPL (alias ChREBP)
thelistsofDEPtoacausalitygraphofreactionsandreg- andthesterol-regulatorybindingprotein-1(SREBF1),two
ulations built from Transpath® database information, transcriptionalfactorsknowntoregulateglucoseandlipid
and ranks upstream candidates in accordance to their metabolism. The peroxisome proliferator activated recep-
ability to regulate significant numbers of the DEP. The tordelta(PPAR)δandtheretinoicXreceptoralphaRXRA
obtained candidates were listed in Additional file 5: forming heterodimers to bind promoters of target genes
Table S5 for PRATand in Additional file 6: Table S6 for in energy metabolism and apoptotic process were also
SCAT. A biological post-prioritization in these candi- down-regulated by the HF diet in the two ATs;
dates was manually made by adding GOBP terms. Only however, PPARα isotype was confirmed as under-
a summary of the most relevant candidates was given in expressed (p<0.05) only in SCAT. PPARγ, the last
Table 6, but more background information on the bio- isotype of the PPAR family having pleiotropic func-
logical processes in which the candidates may play arole tions in lipid metabolism and adipocyte differentiation
is provided in Additional file 5: TableS5 and Additional was conversely up-regulated by the HF diet in the
file6:TableS6. two ATs. It is noteworthy that the direction of change
Dietary-related changes in expression levels of 16 in PPARγ mRNA levels between HF and LF diet by
genes were studied by qPCR in the 48 samples of qPCR was inconsistent with the sign of variation pre-
PRAT and the 48 samples of SCAT to provide a func- dicted by the algorithm (Additional file 5: Table S5),
tional validation among the graph driven-hypothesis so that PPARγ was regulated by diet in a way oppos-
upstream candidates. In the two ATs, the majority of ite to the two other PPAR isotypes. The O-linked N-
the examined candidates were significantly affected by acetylglucosamine transferase OGT regulating protein
diet (Fig. 3). ubiquitination and cell death and the protein phos-
phatase 1 regulatory subunit 1A (PPP1R1A) known to
Table6Candidatesforupstreamregulatoryrolesinfunctional regulate glycogen metabolism were also confirmed as
pathwaysalteredbydiets being up-regulated by the HF diet. Surprisingly, the
retinoic acid receptor alpha (RARA) involved in apop-
Genesproposedasregulatory Associatedbiologicalprocesses
candidatesa totic process was regulated in an opposite manner be-
SREBF1,MLXIPL,PPARδ,PPARα,PPARγ, Glucosemetabolicprocess tween the two ATs. Specificities for PRAT were
USF1,OGT,GALNT1,PPP1R1A,PPP1R3C, further enlightened for three upstream candidates that
B4GALT5,FGFR4
were over-expressed in HF pigs, without any signifi-
SREBF1,MLXIPL,PPARδ,PPARα,PPARγ, Co-factormetabolicprocesses cant variations in their expression levels compared
RXRA,RXRG,SREBF2,GATA4 relatedtolipidsandcholesterol with LF pigs in SCAT: RXRG involved in mediating
RXRA,RXRG,RARA,OGT Responsetoorganicsubstances the anti-proliferative effects of retinoic acid, the nu-
MLXIPL,GLOD4,NRF1,PPARGC1A, Phosphorusmetabolicprocess clear respiratory factor NRF1 involved in cell response
ACOX2
to stress, and the myocyte enhancer factor MEF2A in-
PPARδ,OGT,NFE2L2,RARA,MEF2A, Regulationofapoptosisandcell volved in innate immune response. Finally, the rela-
MEBEAFG29C,,DMLECF12,DM,AASSPN1S,ZNF268, death tive expression levels of USF1, SREBF2, NEF2L2 and
GALNT1 did not differ between diets in the two por-
PPARδ,MLXIPL,MAFG,KLF13,CCDC85B, Cellcycleprocess
ZNF268,IKZF1,COPS2,NFYA cine ATs. Except for GALNT1 in PRAT and NFE2L2
in SCAT, this was in agreement with microarray data
OGT,GALNT1,NFE2L2,UBL4A,ASNS, Proteinmetabolicprocess
NR1H3,B4GALT5,RFN10,ZNF268, (Additional file 5: Table S5 and Additional file 6:
FEM1A,TOMM7 Table S6) which indicated no significant variations in
SREBF1,UBL4A,TOMM7 Proteintransport these gene transcripts when pigs were fed HF or LF
PPARδ,NRF1,NFE2L2,PPARGC1A Responsetostress diets. We observed that sirtuin 1 (SIRT1) was not
proposed as an upstream regulatory candidate of the
PPARα,PPARγ,NFE2L1,IL25 Defenseresponse,inflammatory
response AT transcriptomes, although this gene currently at-
PPARγ,MEF2A,CNIH,MASP1,FGFR4 Innateimmuneresponse tract much attention in the literature; a detailed
examination of the list of DEG revealed its up-
aThelistsofdifferentially-expressedgenesinporcineperirenal(PRAT)or
subcutaneous(SCAT)adiposetissuesinresponsetodietwereautomatically regulation by the HF diet notably in PRAT (x1.25 fold
confrontedtotheacademicinformationonregulationsandreactionsin as compared with LF diet, BH-adjusted p<0.03).
mammaliancellsignaling.Genesthatwereabletoregulateasignificant
numberofthedifferentially-expressedgeneswereproposedasupstream
regulatorycandidates.Thebiologicalprocessesinwhichthesegenes Discussion
couldbeinvolvedwereindicatedasreferencedinthegeneontology(GO)
Irrespective of species, there are still a limited number
Entrezdatabase.ThecompletelistsofcandidatesaregiveninAdditional
file5:TableS5forPRATandAdditionalfile6:TableS6forSCAT of studies which have deciphered the effects of changing
Gondretetal.BMCGenomics (2016) 17:120 Page9of16
Fig.3ExpressionlevelsoftranscriptionfactorsexaminedbyqPCRasaffectedbydietinadiposetissues.The%variationratioinexpressionlevel
ofeachtargetgenewasshownforHFdiet(high-fathigh-fiber)relativetoLFdiet(low-fathigh-starch)forperirenal(PRAT)andsubcutaneous
(SCAT)adiposetissues.Whenagenewasdown-regulatedbytheHFdiet,thevaluewasprecededbyaminussign.***p<0.001;**p<0.01,
*p<0.05,t0.05<p<0.10
dietary energy source rather than dietary energy level on Depressedcatabolicproteinprocessandlowerglucose
gene expression levels in AT [7, 24, 25]. A recent study useforanabolicpathwayswereprimarilyassociatedwith
on Iberian pigs revealed only minor influences of dietary reducedbodyadipositybyahigh-fathigh-fiberdiet
energy source (sunflower oil and fiber opposed to carbo- Among the 436 unique genes which exhibited close-
hydrates) on dorsal subcutaneous fat transcriptome [26]. similar dietary-induced variations in their expression
In the present study, feeding pigs a diet rich in lipids levels across the two ATs, protein catabolic process was
(rapeseed and soybean oils) and fibers (insoluble wheat- suggested as the top- down-regulated pathway acting in
straw) during 58 days led to 1.4-fold reduction in adiposity variation by diet. Down-regulation of the
adiposity as opposed to pigs fed a ration characterized ubiquitin-mediated proteolysis and proteasome pathway
by starch as the main energy source. Large numbers of has been previously reported in the liver for a specific
genes in perirenal AT and, to a lesser extent in sub- strain of mice fed a high-fatdiet compared with acarbo-
cutaneous AT, exhibited dietary-related changes in ex- hydrate chow diet [27]. In the present study, it is
pression levels. Complementary integrative statistical possible that molecular changes related to protein catab-
methods and functional bioinformatics tools were thus olism have also resulted from the introduction of insol-
used on the transcript data sets to identify the most im- uble fiber in the pig diet, which may have compromised
portantmetabolicpathwaysaccountingfordietary-related dietary protein digestibility and nitrogen retention
variationsadiposityacrossthetwoATs. [28]. The finding of lower plasma level of lysine, an