Roertson et l. Microiome (218) 6:95 https://doi.org/1.1186/s4168-18-476-6 RESEARCH Open Access Mternl omeg-3 ftty cids regulte offspring oesity through persistent modultion of gut microiot Ruiri C. Roertson 1,2,3, Knkrju Klinnn 1, Conll R. Strin 2,3, R. Pul Ross 3, Ctherine Stnton 2,3 nd Jing X. Kng 1* Astrct Bckground: The erly-life gut microiot plys criticl role in host metolism in lter life. However, little is known out how the ftty cid profile of the mternl diet during gesttion nd lcttion influences the development of the offspring gut microiot nd susequent metolic helth outcomes. Results: Here, using unique trnsgenic model, we report tht mternl endogenous n-3 polyunsturted ftty cid (PUFA) production during gesttion or lcttion significntly reduces weight gin nd mrkers of metolic disruption in mle murine offspring fed high-ft diet. However, mternl ftty cid sttus ppered to hve no significnt effect on weight gin in femle offspring. The metolic phenotypes in mle offspring ppered to e medited y comprehensive restructuring of gut microiot composition. Reduced mternl n-3 PUFA exposure led to significntly depleted Epsilonproteocteri, Bcteroides, ndakkermnsi nd higher reltive undnce of Clostridi. Interestingly, offspring metolism nd microiot composition were more profoundly influenced y the mternl ftty cid profile during lcttion thn in utero. Furthermore, the mternl ftty cid profile ppered to hve long-lsting effect on offspring microiot composition nd function tht persisted into dulthood fter life-long high-ft diet feeding. Conclusions: Our dt provide novel evidence tht weight gin nd metolic dysfunction in dulthood is medited y mternl ftty cid sttus through long-lsting restructuring of the gut microiot. These results hve importnt implictions for understnding the interction etween modern Western diets, metolic helth, nd the intestinl microiome. Keywords: Microiome, Microiot, n-3 PUFA, Oesity, Mternl diet Bckground The gut microiot is complex microil ecosystem lining the intestinl trct tht criticlly regultes host metolism, immune responses nd numer of other key physiologicl pthwys [1 3]. The composition nd function of the gut microiot is profoundly influenced y environmentl fctors such s diet, mode of delivery t irth nd ntiiotic usge [4, 5]. Due to the essentility of the commensl microiot in host physiologicl homeostsis, * Correspondence: jxkng@mgh.hrvrd.edu 1 Lortory for Lipid Medicine nd Technology, Deprtment of Medicine, Msschusetts Generl Hospitl nd Hrvrd Medicl School, Boston, MA, USA Full list of uthor informtion is ville t the end of the rticle environmentl disturnce of gut microiot composition nd function cn ply cusl role in weight gin nd metolic dysfunction [3, 4, 6]. The microiot is primrily estlished t irth through verticl trnsmission from the mother, while some reports lso indicte microil exposures in utero [7 9]. Indeed, the mternl vginl microiot closely resemles tht of the infnt microiot soon fter irth [7, 1]. Hence, disruption of the initil structuring of the microiot in erly-life my interfere with host metolism nd increse risk of lter-life metolic disese. Disruption of the norml gut microiot composition ( dysiosis ) is often chrcterized y elevted reltive undnce of pthogenic cteri, such s lipopolyscchride The Author(s). 218 Open Access This rticle is distriuted under the terms of the Cretive Commons Attriution 4. Interntionl License (http://cretivecommons.org/licenses/y/4./), which permits unrestricted use, distriution, nd reproduction in ny medium, provided you give pproprite credit to the originl uthor(s) nd the source, provide link to the Cretive Commons license, nd indicte if chnges were mde. The Cretive Commons Pulic Domin Dediction wiver (http://cretivecommons.org/pulicdomin/zero/1./) pplies to the dt mde ville in this rticle, unless otherwise stted.
Roertson et l. Microiome (218) 6:95 Pge 2 of 14 (LPS)-producing Enteroctericee, or depletion of commensl species tht mintin gut homeostsis, such s Akkermnsi muciniphili [11 13]. Recent reserch hs exmined how mode of delivery t irth, restfeeding durtion, nd ntiiotic use influence this structuring of the offspring microiot [4, 14]. However, little is known out how the mternl prentl or erly-postntl diet ffects the phylogenetic rchitecture of the offspring microiot nd the susequent effects this my hve for metolic disese risk in lter life. Indeed, mternl nutritionl indequcies dversely ffect fetl metolic progrmming in the offspring leding to negtive helth consequences in lter life [15]; however, the role of the microiot in this trnsgenertionl process remins underexplored. Previous reserch hs shown tht prentl Western diet is ssocited with elevted plsm LPS nd heightened colonic immune responses in stndrd chow-fed offspring [16]. These immune-modulting responses re dependent upon the comprehensive chnges induced to the offspring microiot, s result of the prentl diet. Indeed, mternl high-ft diets lso induce compositionl chnges to the offspring microiot in non-humn primtes [17]. Consequently, s the mternl microiot correltes with tht of the infnt [7], optimiztion of mternl diet nd microiot composition my therefore enhnce infnt microiot development. Although it hs een shown tht high-ft diets contriute to perturtion in gut microil lnce nd inflmmtion-induced oesity, recent evidence suggests tht sturted nd polyunsturted ftty cids (PUFA) differ in their interction with the gut microiot nd susequent metolic outcomes [18, 19]. Omeg-3 (n-3) nd omeg-6 (n-6) PUFA ply opposing roles in the inflmmtory response [2]. n-6 PUFA typiclly upregulte inflmmtion y cting s precursors to pro-inflmmtory eicosnoids, while n-3 PUFA resolve inflmmtion y competing within the sme enzymtic pthwy. The evolutionry rtio of n-6/n-3 PUFA hs een estimted t 1:1; however, this hs incresed to 1 5:1 in the modern western diet which mny hve suggested hs contriuted to the epidemic of chronic inflmmtory diseses such s oesity [21]. Indeed, there is convincing evidence tht lowering the n-6/n-3 rtio cn restore disrupted metolism in the context of chronic disese [22, 23]. From n erly-life perspective, lowering the n-6/n-3 rtio in oese mothers cn reduce offspring weight gin nd ssocited inflmmtory outcomes in mice [24]. Similr results hve een reported for improving insulin sensitivity [25]. Furthermore, oservtionl nd intervention studies in humns hve found negtive correltions for n-3 PUFA sttus nd positive correltions for n-6 sttus nd offspring diposity [26, 27]. Indeed, n-6 nd n-3 PUFA pper to hve opposing effects on intestinl homeostsis. We hve recently reported tht, reltive to dietry n-6 PUFA, n-3 PUFA reduce Enteroctericee reltive undnce, elevte reltive Bifidocterium undnce, nd dmpen intestinl inflmmtion prtilly through the upregultion of certin nti-microil peptides in mice [28, 29]. Similrly, n-3 PUFA deficiency induces stte of gut dysiosis through ltertion of gut microiot composition nd impirment of short-chin ftty cid production [3]. Chnges to gut microiot composition induced y omeg-3 deficiency in mice re lso ssocited with ehviorl impirments even efore dolescence, suggesting n essentil role for n-3 PUFA within the developing gut-rin xis [31]. Dt on the role of mternl n-3 PUFA on offspring microiot re limited. Some evidence suggests tht high doses of mternl fish oil supplementtion impir the immune response in offspring nd led to the overgrowth of pthogenic cteri [32]. However, these studies utilized crude fish oil in phrmcologiclly excessive doses (18% energy), nd hence, further studies re wrrnted to exmine the reltionship etween nutritionlly relevnt mternl n-3 PUFA intke nd offspring microiot. In ddition, the role of mternl n-3 PUFA intke in the context of offspring oesity nd its ssocition with gut microiot hs not een previously exmined. Furthermore, there is lck of evidence s to the differences etween pre- nd postntl n-3 PUFA sttus for offspring metolic helth. Bsed on previous evidence reporting the criticl opposing roles tht n-3 nd n-6 PUFA ply in microiot-ssocited metolic endotoxemi [28], we im to investigte this in the context of mternl n-6/n-3 PUFA sttus nd susequent effects on oese offspring. To exmine this, we utilize the ft-1 trnsgenic mouse model, which is cple of endogenous conversion of n-6 PUFA to n-3 PUFA [33] nd therey elimintes confounding fctors of diet. Therefore, using single diet for oth tretment groups, this model llows the genertion of two phenotypes: () ft-1 mice with lnced tissue n-6/n-3 rtio (~ 1:1) nd () wild-type (WT) mice with high n-6/ n-3 rtio similr to the Western diet (> 1:1). By exmining the WT offspring of these two mternl genotypes nd cross-fostering the offspring to mothers of different genotype t irth, we re le to identify the effect of mternl n-3 PUFA sttus during gesttion nd/or lcttion on offspring helth outcomes. Our results show tht lower mternl n-6/n-3 rtio during gesttion nd/or lcttion significntly reduces weight gin nd metolic disruption in offspring fed high-ft diet (HFD), which ppers to e medited through persistent restructuring of the gut microiot. These studies provide novel evidence for the criticl role of the mternl diet in erly-life gut microiot development nd the susequent impct on lter-life metolic disese sttes.
Roertson et l. Microiome (218) 6:95 Pge 3 of 14 Results Mternl ftty cid sttus during gesttion nd lcttion differentilly modulte offspring ftty cid profile To exmine the effects of vrying mternl ftty cid sttus during gesttion nd lcttion on the tissue ftty cid profile of offspring, we creted cross-fostering protocol nd mesured offspring ftty cid composition t different time points (Fig. 1). Tissue ftty cid profiles of WT nd trnsgenic ft-1 mothers were ssessed in ddition to their WT offspring t wening (4 weeks old) nd following 3 months of HFD feeding (Fig. 1). As expected, the n-6/n-3 PUFA rtio in til tissue ws significntly greter in WT mothers thn in ft-1 mothers (p <.1, Fig. 1). Mternl ftty cid profiles ppered to e trnsferred to offspring wherey, following 4 weeks of lcttion, mle offspring whom hd WT mother during oth gesttion nd lcttion () displyed significntly greter n-6/n-3 rtio compred with ll other groups (p <.1, Fig. 1c). Conversely, offspring whom hd trnsgenic ft-1 mother during oth perintl periods () displyed the lowest til n-6/n-3 rtio. Interestingly, in the crossover groups, offspring (those who hd WT mother during gesttion nd ft-1 mother during lcttion) hd significntly lower n-6/n-3 rtio thn offspring, suggesting tht the mternl ftty cid profile during lcttion hd more pronounced effect on the offspring ftty cid profile thn the mternl ftty cid profile during gesttion. Interestingly, differences in dolescent til n-6/n-3 rtio s result of mternl genotype were eliminted following 3 months of HFD feeding (Fig. 1d), nd similrly, there were lso no differences in n-6/n-3 PUFA rtio in offspring liver (Additionl file 1: Tle S3). 5 4 3 2 1 Mothers 4 Pre-HFD c d n6/n3 rtio **** n-6/n-3 rtio 3 2 1 c n-6/n-3 rtio 4 3 2 1 Post-HFD WT ft-1 Fig. 1 Experimentl design nd til ftty cid profiles. ft-1 (n = 15) nd WT mothers (n = 9) were mted while on high-n-6 PUFA diet (1% corn oil). At irth, WT offspring were cross-fostered to different mothers for the period of lcttion (4 weeks) to produce four experimentl groups sed on the mothers genotype (iologicl mother s genotype/foster mother s genotype): (n = 8 mles, n = 7 femles), (n =9 mles, n = 1 femles), (n =9mles,n = 5 femles), nd (n = 1 mles, n = 1 femles). During lcttion, mothers were continued on high-n-6 PUFA diet. After 4 weeks of lcttion, offspring were wened onto high-ft diet (HFD) (6% kcl from ft) for 3 months during which ody weights were ssessed long with numer of other prmeters. WT mothers displyed significntly greter til n-6/n-3 rtio compred with ft-1 mothers. c Following lcttion for 4 weeks nd prior to HFD, mle offspring hd significntly greter til n-6/n-3 rtio compred with ll other groups. d However, fter 3 months on HFD, differences in til n-6/n-3 ftty cid rtio were eliminted nd there were no significnt differences etween groups. Dt shown s men ± SEM. n =5 15 per group, n =1 4 per cge. Brs with different letters re significntly different
Roertson et l. Microiome (218) 6:95 Pge 4 of 14 Perintl n-3 PUFA exposure reduces weight gin under high-ft diet To exmine whether mternl ftty cid sttus ffects offspring weight gin, offspring of WT nd ft-1 were fed HFD (6% kcl from ft) for 3 months. There were no significnt differences in offspring weights t week 4 prior to introduction of the HFD. In femle offspring, there were no significnt differences in weight gin etween groups following 3 months HFD feeding (Fig. 2, ). However, t 1 weeks of ge nd following 6 weeks of HFD feeding, mles hd gined significntly more weight thn mles (p <.5) suggesting tht the influence of mternl n-3 PUFA sttus on offspring weight gin is gender-dependent. This effect of mternl ftty cid sttus on mle offspring weight gin ws strengthened with continution of the HFD wherey continued to gin significntly more weight thn ll other groups throughout the experimentl period (Fig. 2c). At scrifice, following 3 months of HFD, mice hd gined 232% ody weight, which ws significntly greter (p =.12) thn ll other groups ( 189%; 195%; 197%; Fig. 2d). Importntly, food intke did not differ significntly etween groups. There were lso no significnt differences in ody composition etween groups (Additionl file 1: Figure S1). Mternl n-3 PUFA dmpen mrkers of metolic disruption in offspring fed high-ft diet We then exmined mrkers of metolic disruption following HFD feeding to ssess whether differences in offspring weight gin induced y mternl n-3 PUFA vilility medited metolic helth. Lipopolyscchride-inding protein (LBP), mrker of metolic endotoxemi, ws lowest in offspring (p <.5, Fig. 2e). Intestinl permeility (IP) is mesure of gut epithelil integrity relting to the pssge of microil endotoxins such s LPS into circultion. Agin, the group exhiited the lowest IP (p =.5, Fig. 2f). However, t this time point, there were no oserved differences etween groups in concentrtions of circulting serum cytokines IL-1, IL-1β, IL-6, MCP-1, or TNFα (Additionl file 1: Figure S2A-E). Anlysis of sucutneous dipose tissue inflmmtion reveled no differences in mrna expression of TNF, F4/8 or CCL2; however, TLR4 expression ws lower (p <.5) in ft- 1/WT compred with oth nd (Additionl file 1: FigureS2F-I). There were sutle differences in glucose tolernce pre- HFD wherey offspring hd significntly greter circulting glucose (p <.5) t two time points during glucose tolernce test (GTT). These differences were eliminted following HFD feeding (Additionl file 1: Femles Weight gin (%) 25 2 15 1 5 Mles c e Weight gin (%) 25 2 15 1 5 * # * * & & ## ## ## ## ** ** ** ** # ** LBP (ng/ml) 2 15 1 5 6 8 1 12 14 16 Age (Weeks) 6 8 1 12 14 16 Age (Weeks) d f Totl weight gin (%) 25 2 15 1 5 Totl weight gin (%) 25 2 15 1 Fig. 2 Weight gin nd metolic mrkers on HFD., There were no differences in weight gin etween femle groups. c, d mle offspring gined significntly more weight on HFD thn ll other groups (repeted-mesures two-wy ANOVA (time nd group) with Tukey s post-hoc test). e Ft-1/ft-1 displyed the lowest circulting LBP nd intestinl permeility (f). Dt shown s men ± SEM. : n = 8 mles, n = 7 femles; : n = 9 mles, n = 1 femles; : n = 9 mles, n = 5 femles; : n = 1 mles, n = 1 femles. n =1 4 nimls per cge. Brs with different letters re significntly different. *p <.5 nd **p <.1, vs. ; #p <.5 nd ##p <.1 vs ; & p <.5, vs. Intestinl permeility dextrn-fitc (ug/ml) 4 3 2 1
Roertson et l. Microiome (218) 6:95 Pge 5 of 14 Figure S3A-C). Insulin sensitivity lso ppered to e differentilly regulted post-hfd with the group displying the lowest insulin re under the curve (AUC) following glucose dministrtion (p =.2, Additionl file 1: FigureS3D-H). Mternl n-3 PUFA sttus hs profound influence on offspring microiot composition in erly life, which persists into dulthood To exmine whether differences in weight nd metolic helth etween offspring groups were medited y chnges in the gut microiot, 16S rrna gene sequencing ws performed on fecl DNA of mothers nd offspring oth prend post-hfd feeding. 16S rrna gene sequencing of the mothers, offspring pre-hfd, nd post-hfd fecl microiot generted totl of 23 million sequenced reds. Principl coordinte nlysis (PCoA) reveled distinct clustering of WT nd ft-1 mothers sed on microiot composition (Additionl file 1: Figure S5A). Whole microiome significnce testing using PERMANOVA with Bry- Curtis dissimilrity index showed significnt differences etween ft-1 nd WT mothers (p =.44, F = 4.874; Additionl file 1: Figure S5A).In the offspring,clustering ws evident mong tretment groups, s ssessed using PERMANOVA with Bry-Curtis dissimilrity index oth pre-hfd (p =.6; F = 3.497) nd post-hfd (p =. 1, F = 9.821; Fig. 3). Both prior to nd fter HFD feeding, offspring clustered significntly ccording to the foster mother s genotype such tht the nd groups clustered together, nd the ft-1/ ft-1 nd groups clustered together, ccording to Bry-Curtis dissimilrity index testing. Assessment of α-diversity using the Shnnon Index reveled no significnt differences etween WT nd ft-1 mothers (Additionl file 1: Figure S5B) or etween post-hfd groups (Fig. 3). However, microil diversity s mesured y the Shnnon Index ws significntly reduced in the group compred with prior to HFD feeding (Fig. 3). Reltive undnces of phylum level tx reveled distinct differences etween groups (Fig. 3c, d). In mothers, Proteocteri ws the only phylum significntly different etween groups following FDR correction (p =.5) nd ws present in greter reltive undnce in ft-1 femles (Additionl file 1: Figure S5C). Txonomic distriution ppered to e most significntly influenced y lctting/foster mother s genotype s opposed to iologicl mother s genotype. Hence, Proteocteri ws significntly greter in oth nd offspring oth pre-hfd (p =.18) nd post-hfd (p <.1). Verrucomicroi were lso highest in offspring post-hfd (p =.23). Conversely, Firmicutes were significntly greter in nd offspring post-hfd (p <.1). c d Pre-HFD Coordinte 2 (27.1%) PERMANOVA (p=.6) (F=3.497) Coordinte 1 (36.9%) Shnnon diversity index 4. 3.5 3. 2.5 2. Firm:Bc rtio 4 3 2 1 Reltive undnce (%) 1 8 6 4 2 1 Post-HFD Coordinte 2 (2.2%) PERMANOVA (p=.1) (F=9.821) Coordinte 1 (35.8%) Shnnon diversity index 4. 3.5 3. 2.5 Firm:Bc rtio 4 3 2 1 Reltive undnce (%) 8 6 4 2 Firmicutes Bcteroidetes Proteocteri Verrucomicroi Deferricteres Cndidtus_Scchricteri Actinocteri Lentisphere Other Fig. 3 Phylum compositionl nd diversity differences in fecl microiot etween offspring of ft-1 nd WT mothers. Anlysis of et diversity identified significnt clustering of offspring groups y foster mternl genotype. α-diversity s mesured y the Shnnon index did not differ etween post-hfd groups, however, ppered to e reduced in compred with efore HFD feeding. c The Firmicutes:Bcteroidetes rtio ws lowest in offspring. d Fecl microiot composition differed significntly etween offspring groups t phylum level nd ppered dependent on foster mother genotype. Significnt differences were determined y non-prmetric nlysis using the Kruskll-Wllis test followed y Mnn-Whitney test. n =5 1 per group, n =1 4 per cge. Groups with different letters re significntly different
Roertson et l. Microiome (218) 6:95 Pge 6 of 14 Due to the significnt PCoA clustering etween groups sed on foster mother s genotype, offspring were grouped ccording to the foster mother s genotype nd liner discriminnt nlysis effect size (LEfSe) ws performed on these two groups (1. + ; 2. + ) s iomrker discovery tool to identify tx tht my e contriuting to differences etween groups (Fig. 4 nd Additionl file 1: Figure S4). LEfSe nlysis elucidted the oserved phylum level differences such tht offspring whom hd WT foster mother were more undnt in species within the Firmicutes phylum (Clostridi) nd offspring whom hd ft-1 foster mother were more undnt in species of the Bcteroidetes (primrily Bcteroides) nd Proteocteri phyl (primrily Epsilonproteocteri) (Additionl file 1: FigureS5). Further nlysis t lower txonomic levels reveled numer of tx (13 gener pre-hfd nd 26 gener post- HFD) tht ppered to significntly differ etween groups (Fig. 4). Clostridi nd Prcteroides were significntly greter in nd offspring post-hfd (Fig. 4c). Conversely, Akkermnsi ppered to e significntly elevted in offspring post-hfd. Bcteroides were lso significntly lower in offspring with WT foster mothers. Interestingly, the reduced Proteocteri oserved in WT mothers nd offspring fostered to WT mothers ws driven primrily y depleted Epsilonproteocteri (Fig. 4c) nd Deltproteocteri, wheres reltive undnce of Gmmproteocteri ws unchnged (Additionl file 1: Figure S5). Helicocter, commensl memer of the Epsilonproteocteri order, Fig. 4 Genus level distriution of fecl microiot etween offspring of ft-1 nd WT mothers. Lest discriminnt nlysis of effect size (LEfSe) on clustered groups sed on mternl genotype reveled tht Firmicutes were more undnt in offspring of WT foster mothers wheres Bcteroidetes nd Proteocteri were more undnt in offspring of ft-1 foster mothers. Normlized dt of fecl microiot reltive undnce reveled distinct differences etween offspring groups. Microiot composition ppered similr in groups of the sme foster mother genotype. Dt represents OTUs with significntly different relevnt undnces etween tretment groups (s determined y Kruskll-Wllis testing nd Benjmni-Hocherg multiple correction testing). Ech row represents n OTU leled y lowest txonomic description nd OTU ID, normlized to the row mximum. Dt normlized per txonomic red. c Clostridi displyed significntly higher reltive undnce in offspring of WT foster mothers. Bcteroides displyed higher reltive undnce in offspring of ft-1 foster mothers. Akkermnsi displyed the highest reltive undnce in offspring. Epsilonproteocteri were lmost entirely depleted (<.1%) in offspring fostered to WT mothers. n =5 1 per group, n =1 4 per cge. Significnt differences were determined y non-prmetric nlysis using the Kruskll-Wllis test followed y Mnn-Whitney test nd FDR correction y Benjmni-Hochurg testing. Groups with different letters re significntly different
Roertson et l. Microiome (218) 6:95 Pge 7 of 14 which constituted 9 14% of the nd post-hfd microiot, ws lmost entirely depleted (<.1%) in offspring of WT foster mothers. Network interctions revel host-microiome interctions driven y ftty cid sttus To ssess the overll mesure of correltion etween the n-6/n-3 rtio-induced metolic chnges nd microiot, n RV coefficient ws clculted. An RV coefficient of.456 (p =.1) ws found etween the reltive undnce of microil gener (pre- nd post-hfd) nd host phenotypes (mother nd offspring pre-hfd n-6/n-3 rtio, LBP, IP, nd totl ody weight gin). Network-sed nlyticl pproches hve the potentil to help disentngle complex host-microe interctions [34]. Pirwise correltions etween offspring n-6/ n-3-induced chnges in microiot nd host prmeters with significnt Spermn s non-prmetric rnk correltion coefficient were employed to generte correltion networks for oth pre-hfd nd post-hfd (Fig. 5, ). Correltion nlysis with pre-hfd microiot dt resulted in correltion network of 92 microil prmeters (tx nd α-diversity indices) nd 4 host prmeters (pre-hfd n-6/n-3 rtio, finl ody weight gin, LBP, nd IP) for which t lest one correltion could e found. The network consists of 118 edges (7 nd 48 positive (green) nd negtive correltions (red) respectively) nd 98 nodes (microil nd host prmeters). Three modules were oserved in the pre-hfd network (1. n-6/n-3 rtio; 2. weight + LBP; 3. IP). Accordingly, the lrgest module (n-6/n3 rtio) showed high modulrity (lue nodes) comprised of oth positive (P; red lines) nd negtive correltions (N; green lines) with numer of tx, nmely Firmicutes (P), Bcteroidetes (N), Proteocteri (N), Clostridi (P), Epsilonproteocteri (N), Ruminococccee (P), Porphyromondcee (N), Bcteroides (N), nd Akkermnsi (P) (Fig. 5). A numer of similr outcomes persisted in the post-hfd correltion network, wherey ody weight ws negtively ssocited with Proteocteri nd LBP negtively correlted with Akkermnsi (Fig. 5). Networks were lso constructed etween mother s n-6/n-3 rtio, pre- nd post-hfd microiot dt, nd host prmeters (Additionl file 1: Figure S7). To further investigte the identified groups of correltions etween chnges in microiot composition nd host prmeters, we uilt prtil lest squres (PLS) models for selected correltions of interest creted (Additionl file 2; Goodness of fit Q 2 cum >.8). Comintion of offspring pre-hfd n-6/n-3 rtios nd ll microiot dt explined 68, 72, nd 61% vriility in the ody weight gin, LBP, nd IP, respectively (Additionl file 2: dt models 4 6). A model creted etween ll the host prmeters (weight gin, LBP, nd IP) nd ll microiot dt nd mothers nd offspring n-6/n-3 rtio (model 12) explined 45, 32, nd 35% vriility in the host prmeters respectively (Fig. 5c). Vrile importnce in the projection (VIP) scores for the vriles in ech PLS model cn e found in the Supplementry Dt File 6 (M1 14). Next, prmeters contriuting to the multivrite PLS models were compred with the corresponding identified modules in the correltion networks. Notly, 42 out of 5 pre-hfd microil prmeters, which exhiited direct correltions with offspring n-6/n-3 rtio in the network nlysis (lrgest module), were presented with VIP vlues 1 or > 1 in the PLS model (model 8). The selected microil prmeters hve een shown with R 2 nd VIP scores in the Supplementry dt File 6 (M8). Twenty-eight out of 44 pre-hfd nd 6 out of 9 post-hfd microil prmeters hving correltions with mother s n-6/n-3 rtio nd LBP in the network nlysis (lrgest module in the pre- nd post-hfd networks) were presented with VIP vlues 1 or > 1 in the PLS model (model 3 nd M3). Finlly, this ws visulized using multiple fctor nlysis (MFA) tht llowed microil prmeters to e superimposed onto the host prmeters (Fig. 5d). Superimposed microiome nd host dt were seprted depending upon the n-6/n-3 rtio of oth mother nd foster mother. Pirwise comprison (Monte Crlo simultions with P vlue =.8) etween groups with superimposed host nd microiot dt showed the importnce of lctting mothers n-6/n-3 rtio when offspring were orn from WT mothers. Discussion In this study, we oserved tht the perintl mternl tissue ftty cid profile profoundly nd persistently restructures the offspring gut microiot, which my hve long-term implictions for metolic helth. Mternl n- 3 PUFA significntly reduced offspring weight gin into dulthood during high-ft feeding s hs een demonstrted previously in humns nd nimls [26, 35, 36]. A numer of mechnisms my contriute to this. Firstly, mternl tissue nd milk n-3 PUFA correlte with umilicl cord nd infnt n-3 PUFA [37, 38]; therefore, the eneficil effects on weight my e ttriuted to direct nutrient trnsport from mother to infnt nd the susequent nti-dipogenic effects of n-3 PUFA [39]. The intriguing spect of these findings is tht the group hd higher n-6/n-3 rtio thn the group, suggesting tht the offspring n-3 sttus is more influenced y the dietry n-3 PUFA during lcttion, vi milk, rther thn mternl n-3 PUFA sttus during gesttion. It hs previously een reported tht n-3/n-6 content in milk from ft-1 femles is greter thn WT femles [33]. Additionlly, there ppered to e strong gender difference etween mle nd femle offspring regrding weight gin s result of mternl n-3 sttus.
Roertson et l. Microiome (218) 6:95 Pge 8 of 14 c d Fig. 5 Network nlyses of microiome nd host metolic phenotype interctions. Host-microiot interction network uilt from Spermn s non-prmetric rnk correltion coefficient (P <.5) etween host prmeters (mother nd offspring pre-hfd n-6/n-3 rtio, ody weight, IP, nd LBP) nd microil prmeters (pre- nd post-hfd OTUs with FDR-corrected p vlues <.5, FIR/BAC rtio, nd Shnnon ADI) for pre-hfd nd post-hfd. Ech node ws colored ccording to the modulrity score nd nodes were grouped s three () or four () modules. Lines represent sttisticlly significnt correltions nd re colored red for positive nd green for negtive correltions. c Prtil lest squre (PLS)-regression loding score plot illustrting the ssocition etween host prmeters (dependent vriles Y) nd microil prmeters (explntory vriles X; red dots). Explntory vriles of interest with vrile importnce in the projection (VIP) scores 1 or > 1 were leled with circles on the red dots. Smples from four different groups (,,, ) were oserved (green dots) nd grouped using circles sed on where they clustered on the plot. Leve-one-out cross-vlidtion (LOO-CV) ws pplied. d Multiple fctor nlysis (MFA) using Spermn type principl component nlysis of host nd microiot dt. One end of the ech connecting line for n oservtion indictes the host prmeters (differently colored to indicte the groups) nd nother end (red) indictes the microiot Mternl n-3 PUFA reduced weight gin in offspring mles ut not femles. Such gender-dependent differences hve een reported previously in humns, n effect which my e medited y the interction of femle sex hormones with dipogenesis nd ftty cid metolism [4 42]. We lso oserved tht mternl ftty cid sttus influenced immune regultion in the offspring, which my ffect weight. It ppered tht the lower n-6/n-3 rtio in ft-1 mothers my hve dmpened mternl nd plcentl inflmmtion, which induced n nti-inflmmtory nd nti-oesigenic environment in the offspring. This effect of mternl tissue ftty cid sttus on offspring inflmmtion hs een demonstrted previously. Using the sme ft-1 model, it hs een shown tht mternl oesity induces mternl nd plcentl inflmmtion, which is trnsmitted to the offspring resulting in numer of metolic disruptionsthtrenotevidentinoffspringofft-1 mothers [24].Indeed,inthisstudy,othIPndLBPweresignificntly reduced y mternl n-3 PUFA. This interesting finding suggests tht the chronic low-grde inflmmtion tht medites the oesigenic phenotype [43] my e trnsferred from the mother to the infnt during the perintl period. This inflmmtory phenotype ppers to originte in the intestines through degrdtion of the intestinl rrier nd hence trnsloction of cteril endotoxins.
Roertson et l. Microiome (218) 6:95 Pge 9 of 14 As hs een reported previously, this inflmmtory phenotype is induced y distured microiot [19, 44]. The composition of n infnt s microiot is strongly influenced y tht of the mother [7]. Therefore, the ntiinflmmtory ft-1 microiot tht hs een descried previously [28] my hve een trnsmitted verticlly to the offspring therey reducing the microiot-ssocited weight gin. The gut microiot hs well-estlished role in energy metolism nd oesity y regulting energy hrvest from mcronutrients [6]. Interestingly, the oserved differences in weight gin in this study were independent of dietry intke, which did not differ, suggesting tht the microiot displyed n incresed cpcity for energy hrvest. Previous hypotheses surrounding the oesigenic microiot hve concentrted on the energy-hrvesting cpcity of Firmicutes nd the production of short-chin ftty cids (SCFA). Indeed, we oserved this effect here wherey Firmicutes nd their SCFA-producing fmilies (Lchnospircee nd Ruminococccee) were significntly greter in WT mothers nd their foster offspring. We hve previously demonstrted tht differing rtios of dietry ftty cids significntly lter SCFA production in mice [3]. The role of the microiot in regulting gut epithelil integrity nd the susequent inflmmtory response hs lso een hypothesized to ply role in oesity [45, 46]. Here, we oserved tht IP nd LBP were lowest in offspring. These chnges to IP were independent of chnges to tight junctionproteinexpression(additionlfile1: FigureS6). Conversely, Akkermnsi ws greter in offspring of ft-1 mothers nd hence my medite the protective effect of mternl n-3 PUFA on weight gin in the offspring s hs een shown previously in fish oil-fed mice [19]. There is growing evidence tht Akkermnsi plys criticl role in metolic helth nd cn reduce weight gin nd metolic endotoxemiinmicendhumns[12, 13, 47]. Epsilonproteocteri, nditsgenushelicocter, werelsosignificntly depleted in offspring of WT foster mothers in response to lck of perintl n-3 PUFA. Our dt extend on work of M et l. who reported tht mternl HFD in primtes induced loss of non-pthogenic Helicocter nd Cmpylocter, nother memer of Epsilonproteocteri, in offspring [17]. Interestingly, offspring of non-humn primtes fed high-ft diet exhiit reduced plsm n-3 PUFA [48]. This unique model nd study design lso llowed us to distinguish the role of the prentl versus postntl mternl tissue ftty cid profile on offspring outcomes. Interestingly, s with the n-6/n-3 rtio, the gut microiot of the offspring ppered to strongly mtch tht of the lctting/foster mother rther thn the iologicl mother. Indeed, it hs previously een shown in crossfostering models tht the nursing mother cuses permnent shift in the offspring microiome [49]. It hs een ssumed tht the fetus is sterile; however, recent studies my suggest otherwise [9]. The iologicl mother imprints unique microiot on the infnt t irth [7]. The results reported here suggest tht the lile irth microiot is quickly nd comprehensively ltered y the foster mother, presumly through differences in milk/dietry ftty cids nd the foster mother's microiot following irth. These results would suggest tht n-3 PUFA sttus in milk during lcttion hs stronger impct on the infnt microiot thn mternl ftty cid sttus during gesttion. Hence, postntl n-3 PUFA exposure my rescue nd recover dysiotic offspring microiot induced y the mternl prentl n-3 PUFA insufficiency. Furthermore, despite the differences in n-3 nd n-6 PUFA disppering in dulthood fter HFD, the oserved differences in the microiot persisted, suggesting tht mternl ftty cid sttus nd erly neontl feeding regime my hve persistent effect on the offspring microiot throughout life. Conclusions Much evidence exists indicting tht oesity nd its ssocited disorders hve their origins in the fetl nd neontl periods. As the gut microiot plys criticl role in the pthogenesis of these disorders nd the chronic low-grde inflmmtion tht defines them, nutrition reserch must now focus on mternl nd erly-life interventions tht trget the gut microiot. This study hs demonstrted tht mternl ftty cid sttus persistently restructures the offspring microiot nd the ssocited metolic homeostsis relted to oesity. In ddition, the unique trnsgenic model used here chllenges the concept of direct diet-microiot interction in oesity nd insted uncovers the importnce of the underlying tissue ftty cid profile in microiot-metolic interctions. These results hve importnt implictions for the current chronic disese epidemic. Excessive n-6/n-3 rtios in the Western diet hve contriuted to trnsgenertionl epidemic of chronic metolic disese, which my e prtilly ttriuted to persistent gut microiot dysiosis. Consequently, mternl n-3 PUFA intke, especilly during lcttion, poses potentil s n effective therpeutic mesure to restore gut microiot homeostsis nd metolic disturnces ssocited with modern chronic disese. Methods Animls nd diets Genertion of trnsgenic ft-1 mice ws performed s previously descried [33] followed y ckcrossing onto C57BL/6 ckground. Ft-1 phenotype ws confirmed y gs chromtogrphy flme-ioniztion detection (GC-FID) following identifiction of incresed tissue n-3 PUFA compred with WT. Ft-1 genotype ws confirmed y RT-
Roertson et l. Microiome (218) 6:95 Pge 1 of 14 PCR. Mice were housed in the Msschusetts Generl Hospitl (MGH) niml fcility in iosfety room (level 2) in hrd top cges with filtered ir. Mice were mintined in temperture-controlled room (22 24 C) with 12-h light/drk diurnl cycle. Food nd wter were provided d liitum. A suset of 3-month-old femle C57BL/ 6 WT mice ws purchsed from Chrles River Lortories nd llowed to cclimtize to the fcility conditions for 1 week prior to mting. Ft-1 nd WT mting pirs were fed diet high in n-6 PUFA (AIN-76A with 1% corn oil) from LDiet in order to mintin ft-1 nd WT phenotypes. At postntl dy (PND) 28, mle nd femle offspring were wened onto high-ft diet (HFD) with 6% kcl from ft (D12492, Reserch Diets Inc.). Detiled ftty cid profiles of oth diets re outlined in Additionl file 1: Tle S1. Body weight nd food intke were mesured weekly using n electricl lnce. Body composition (ft mss, len mss, wter mss) ws ssessed on the dy of scrifice using Minispec mq ench-top NMR spectrometer (Bruker Instruments). Animls were scrificed using CO 2. Dissected tissues were flsh frozen in liquid nitrogen. All niml procedures in this study were performed in ccordnce with the guidelines pproved y the MGH Sucommittee on Reserch Animl Cre. Breeding nd cross-fostering Three-month-old femle ft-1 (n = 15) nd WT (n =9) mice were mted with ge-mtched WT mles. Mting pirs were housed in individul cges, nd mles were seprted from the femles following confirmtion of pregnncy. Within 48-h of prturition, neworn litters were fostered to new mothers until wening, t PND 28. Briefly, the neworn litter ws removed from the iologicl mother s cge then mixed with the edding of the foster mother in the hnd of the investigtor. The litter ws then plced in the empty nest of the new foster mother. The foster mother ws held ove the new litter until she urinted on the litter in order to disguise their scent. The foster mother pirs were chosen such tht offspring were fostered to mothers whom hd given irth within 48 h to litter of similr size. The crossfostering procedure ws crried out in order to generte offspring of four distinct experimentl groups s follows: n6-n3 group WT iologicl mother, cross-fostered to ft-1 mother; n3-n6 group ft-1 iologicl mother, cross-fostered to WT mother; n3-n3 group ft-1 iologicl mother, cross-fostered to new ft-1 mother; n6- n6 group WT iologicl mother, cross-fostered to new WT mother. Cross-fostering ws crried out in the n3- n3 nd n6-n6 groups s control to the cross-fostering procedure in the other two groups. The study design is outlined in Fig. 1. At PND 1, the tils of the offspring were clipped with scissors nd genotyping ws performed on the til tissue y RT-PCR. Following confirmtion of genotype, ft-1 mice were removed from the litter such tht only the WT offspring remined. At PND28, WT offspring were seprted from their mothers, grouped in seprte cges (rndomized y experimentl group, n = 1 4 nimls per cge, n =3 4 cges per experimentl group) nd wened onto the HFD. Ftty cid nlysis Ftty cid nlysis of til nd liver tissues ws performed s previously descried [5]. Briefly, frozen tissue smples were ground to powder under liquid nitrogen using mortr nd pestle. Lipid extrction nd ftty cid methyltion ws performed y the ddition of 14% (w/v) oron trifluoride (BF3)-methnol regent (Sigm-Aldrich) followed y heding t 1 C for 1 h. Ftty cid methyl esters (FAME) were nlyzed using fully utomted HP589 gs chromtogrphy system equipped with flme-ioniztion detector (Agilent Technologies, Plo Alto, CA). The ftty cid peks were identified y compring their reltive retention times with the commercil mixed stndrds (NuChek Prep, Elysin, MN), nd re percentge for ll resolved peks ws nlyzed y using PerkinElmer M1 integrtor. Intestinl permeility Intestinl permeility ws performed s descried previously [28]. Briefly, mice were fsted for 6 h nd then FITC-dextrn (7kDA, Sigm-Aldrich, in PBS solution) ws dministered to mice y orl gvge t dose of 6 mg/kg ody weight. Following gvge, lood smples were collected from the fcil vein fter 9 min. Serum ws diluted with n equl volume of PBS, nd fluorescence intensity ws mesured using fluorospectrophotometer (PerkinElmer) with n excittion wvelength of 48 nm nd n emission wvelength of 52 nm. Serum FITC-dextrn concentrtion ws clculted from stndrd curve generted y seril dilution of FITC-dextrn in PBS. Serum LBP Concentrtions of lipopolyscchride-inding protein (LBP) in serum were ssyed using commercil ELISA kit (NeoBioL, Cmridge, MA) ccording to the mnufcturer s instructions. Stool DNA extrction nd 16S rrna gene sequencing Bcteril genomic DNA ws extrcted from mice fecl pellets using the QIAmp DNA Stool Mini Kit (Qigen, UK) ccording to the mnufcturer s instructions. DNA ws quntified nd purifiction ws susequently ssessed y mesuring sornce nd determining the A26/ A28 rtio. DNA ws stored t 2 C until nlysis.
Roertson et l. Microiome (218) 6:95 Pge 11 of 14 16S rrna gene sequencing lirry preprtion ws performed on DNA smples ccording to the Illumin 16S rrna gene sequencing lirry protocol in order to generte V3-V4 mplicons. DNA smples were sujected to n initil PCR rection utilizing primers specific for mplifiction of the V3-V4 region of the 16S rrna gene (Forwrd primer 5 TCGTCGGCAGCGTCAGATGTG TATAAGAGACAGCCTACGGGNGGCWGCAG; reverse primer 5 GTCTCGTGGGCTCGGAGATGTGTATAA GAGACAGGACTACHVGGGTATCTAATCC). Clen-up nd purifiction of the PCR product ws performed using the Agencourt AMPure XP system (Lpln, Dulin, Irelnd). Following clen-up nd purifiction, second PCR rection ws performed in order to incorporte unique indexing primer pir to ech smple (Illumin Nexter XT indexing primers, Illumin, Sweden). The PCR products were purified second time using the Agencourt AMPure XP system. Quntifiction of smples ws performed using the Quit rod rnge DNA quntifiction ssy kit (Bio-Sciences, Dulin, Irelnd). Following quntifiction, smples were pooled in equimolr mounts (1 nm) nd sequenced t Clinicl-Microiomics (Copenhgen, Denmrk) using Illumin MiSeq 2 3 p pired-end sequencing. 16S rrna gene sequencing ioinformtics nlysis The 64-it version of USEARCH 8.1.1825 [51] nd mother v.1.36.1 [52] were used for ioinformtic nlysis of the sequence dt. These were used in comintion with customized in-house progrms, the process of which is detiled precisely elow. Following tg identifiction nd trimming (performed ntively in the sequencing mchine), ll sequences from ll smples were pooled using perl script (sh commnd "ct", e.g. "ct *.R1.fstq > pool.r1.fstq"). Using USEARCH s fstq_mergepirs commnd, pired-end reds were merged, truncting reds t qulity score of 4, requiring t lest 1 p overlp nd merged red length etween 3 nd 6 p in length nd less thn one expected errors in the merged red ("userch8.1. 1825_i86osx64 -fstq_mergepirs pool.r1.fstq -reverse pool.r2.fstq -fstq_minmergelen 3 -fstq_mxmergelen 6 -fstq_minovlen 1 -fstq_merge_mxee 1 -fstout merged.fst -relel -"). Using mothur's trim.seqs commnd, sequences with miguous ses, without perfect mtch to the primers, or homopolymer length greter thn 8 were discrded nd primer sequences trimmed. Using USEARCH s derep_fulllength commnd, sequences were then strictly dereplicted discrding clusters smller thn 5. Using USEARCH s cluster_otus commnd, sequences were clustered t 97% sequence similrity, using the most undnt strictly dereplicted reds s centroids (" userch8.1.1825_i86osx64 -cluster_otus ru16s.sorted.fst -otus ru16s.clustered.fst uprseout ru16s.clustered. tle"). Using USEARCH s uchime_ref [53] commnd,suspected chimers were discrded sed on comprison with the Riosoml Dtse Project clssifier trining set v9 [54]. Using mothur s clssify.seqs commnd, txonomic ssignment of OTUs ws performed using the method y Wng et l. [55] withmothur s PDSversionoftheRDP trining dtse v14 ("clssify.seqs(fst=recovery.fst, templte=otu.fst, txonomy=trinset14_3215.pds.fst, method=wng)"). The Wng prmeters used were ksize = 8, iters = 1, nd cutoff =. A ootstrp threshold of 8 ws susequently pplied using R nd vlues < 8 were ssigned s unclssified. Alph diversity ws clculted using mothur s Shnnon commnd. To reduce is from vrition in smple red numers, smples were rrefied to the smple with the lowest red count, 1,597 reds. Rrefction cn introduce is into dt nd therey ffect outcomes [56, 57]. To ensure our dt processing did not ffect dt outcomes, we performed PERMANOVA with Bry- Curtis dissimilrity testing to exmine whether rrifying cused differences to the sequencing dt. No significnt differences were oserved etween rrified nd non-rrified dt. Similrly no significnt differences were oserved in lph diversity (Shnnon index) or etween ny single OTU (Kruskll-Wllis followed y Mnn-Whitney testing) etween rrified nd non-rrified dt. Sttisticl nlysis Sttisticl nlysis ws performed using SPSS (v19, NY, USA), GrphPd Prism (v6, CA, USA) nd R (v3.2.4). One-wy nlysis of vrince (ANOVA) ws performed to ssess differences etween groups followed y Tukey s or LSD post-hoc test. Repeted-mesures two-wy ANOVA (time nd group) with Tukey s post-hoc test ws used for weight gin dt. For 16S rrna gene sequencing dt, principl coordinte nlysis ws conducted using PAST Softwre (v3.18) with Bry-Curtis dissimilrity testing. To ssess whether significnt differences existed etween specific tx, non-prmetric nlysis ws performed using the Kruskll-Wllis test followed y Mnn-Whitney test. Flse discovery rte (FDR) nlysis ws susequently performed using the Benjmni-Hochurg method nd significnce ws clculted t q <.5. Liner discriminnt nlysis (LDA) effect size (LEfSe) is iomrker discovery tool for high-dimensionl dt tht provides effect size estimtion [58]. Microiotsed iomrker nlysis ws performed with LEfSe using the online glxy server (https://huttenhower.sph. hrvrd.edu/glxy/). LDA scores (> 3.) derived from LEfSe nlysis were used to show the reltionship etween txon using cldogrm (circulr hierrchicl tree). Levels of the cldogrm represent, from the inner to outer rings, phylum, clss, order, fmily, nd genus.
Roertson et l. Microiome (218) 6:95 Pge 12 of 14 For host-microiome interction nlysis, n RV coefficient ( multivrite generliztion of the Person correltion coefficient) ws clculted etween the microil (pre- nd post-hfd) nd host prmeters (mother nd offspring pre-hfd n-6/n-3 rtio, LBP, IP, nd ody weight). The dt ws rescled from to 1 efore nlysis. Pirwise correltions etween ech txon nd host prmeter were clculted using the Spermn s nonprmetric rnk correltion coefficient. Network-sed nlyticl pproches hve een used previously to disentngle host-microe interctions [59, 6]. Dt ws rescled from to 1 efore nlysis. Bsed on these correltion coefficients, correltion network (lel djust nd no overlp lyout) ws uilt where nodes represent either txon or host prmeter. For ech txon nd host prmeter, n undirected edge ws dded etween the corresponding nodes in the correltion network. Edges (red links indicte positive nd drk green links indicte negtive ssocitions) represent sttisticlly significnt correltions (P <.5). Correltions were clculted in PAST softwre version 2.17 nd the network ws visulized in Gephi Grph Visuliztion nd Mnipultion softwre version.9.2. Nodes were colored sed on modulrity community detection lgorithm. A module in the network is set of nodes connected to ech other y mny links, while connected y few links to nodes of other groups, so modules re elementry units of ny iologicl network (ech ssigned unique color). Degree centrlity of nodes ws employed s n index of node centrlity. Prtil lest squres regression (PLS-R) ws used to ssocite the microil composition with host prmeters including jckknife-sed vrile selection s reported previously [6]. For ll models, the dt were rescled from to 1 efore PLS-R nd centered s well s reduced during PLS-R. Leve-one-out cross-vlidtion (LOO-CV) ws pplied. The Q 2 cumulted index (Q 2 cum) ws used s mesure of the glol goodness of fit, the predictive qulity of the models nd to test the vlidity of the model ginst over-fitting. A Q 2 cum threshold of >.8 ws pplied. The resulting plot displys the dependent vriles on the c vectors nd the explntory vriles on the w* vectors which llows visulizing the glol reltionship etween the vriles. The w* re relted to the weights of the vriles in the models. The results re lso presented in PLS sctter plots for suject clustering nd vriles. The R 2 (coefficient of determintion) indictes the percentge of vriility of the dependent vrile (Y) which is explined y the explntory vriles (X). The reltive importnce of ech x-vrile is expressed y vrile importnce in the projection (VIP) vlues. VIP-vlue 1 or > 1. is considered influentil nd > 1.5 s highly influentil. Results of ll of the ove mentioned sttistics were given in the Supplementry file 6. All nlyses were performed using precise lgorithm in the XLSTAT softwre version 217.6. Associtions etween the host nd microiome dt sets were lso ssessed y multiple fctor nlysis (MFA). The methodology of the MFA reks up into two phses: (i) principl component nlysis (PCA) (Spermn type) successively crried out for ech dtset (Dtset 1, host prmeters; Dtset 2, pre- nd post- HFD microiot dt) which stores the vlue of the first eigenvlue of ech nlysis to then weigh the vrious dtsets in the second prt of the nlysis. (ii) A weighted PCA on the columns of ll the dtsets leds to ech indictor vrile hving weight tht is function of the frequency of the corresponding ctegory. After these two phses, the coordintes of the projected points in the spce resulting from the MFA re displyed. The projected points correspond to projections of the oservtions in the spces reduced to the dimensions of ech dtset. Bsed on the eigenvlues of the weighted PCA, the first two fctors (F1/F2) lmost covered 6% of the vriility in this nlysis. To test whether the four groups with superimposed host nd microiome dt were seprted from ech other, Kruskl-Wllis testing ws performed on the coordintes of the projected points nd p vlues were otined using 1, Monte Crlo permuttions. One end of ech line for n oservtion indictes the host prmeters (differently colored to indicte the groups) nd nother end (red) indictes the microiot. Additionl file Additionl file 1: Figure S1. Mle nd femle ody composition. Figure S2. Serum cytokines nd sucutneous dipose tissue inflmmtory gene expression did. Figure S3. Glucose tolernce nd insulin tolernce testing. Figure S4. LDA scores following LEfSe nlysis of pre-hfd nd post-hfd microiot grouped y foster mother genotype. Figure S5. Mothers microiot nd offspring proteocteri undnce. Figure S6. Ilel tight junction protein expression. Figure S7. Correltion network of mternl ftty cid sttus nd offspring microiot. Tle S1. Ftty cid profile of diet. Tle S2. Til ftty cid profiles of mothers nd offspring efore nd fter high-ft diet feeding. Tle S3. Liver ftty cid profiles of offspring fter high-ft diet feeding. Tle S4. Primer sequences for qpcr. (DOCX 523 k) Additionl file 2: Supplementry Dt. (XLSX 75.6 k) Acknowledgements The uthors wish to cknowledge Dr. Chih-Yu Chen nd Ms. Jie Bin for their technicl ssistnce during the study. The uthors would lso like to thnk Nikolj Sørensen nd Clinicl-Microiomics (Kogle Allé 5, DK-297 Hørsholm, Denmrk) for their sequencing services nd technicl ssistnce. Funding The uthors re supported in prt y Fortune Eduction Foundtion nd Snsun Life Sciences; Science Foundtion Irelnd in the form of Centre grnt (APC Microiome Institute grnt numer SFI/12/RC/2273); the Helth Reserch Bord of Irelnd (Grnt Numers HRA_POR/211/23 nd HRA_POR/ 212/32); TODDLERFOOD: Food Solutions for Replenishing Disrupted Microiot in Toddlers (214-218); the Se Chnge Strtegy, NutrMr progrmme (Grnt-Aid Agreement No. MFFRI/7/1); the SMART FOOD