Article
Celiac Disease Genomic, Environmental, Microbiome, and Metabolomic (CDGEMM) Study Design: Approach to the Future of Personalized Prevention of Celiac Disease
Maureen M. Leonard 1,2,*, Stephanie Camhi 1,2, Tania B. Huedo-Medina 3 and Alessio Fasano 1,2
Received: 16 September 2015 ; Accepted: 4 November 2015 ; Published: 11 November 2015
1 Center for Celiac Research, Massachusetts General Hospital for Children, Boston, MA 02114, USA; sscamhi@partners.org (S.C.); afasano@mgh.harvard.edu (A.F.)
2 Mucosal Immunology and Biology Research Center, Massachusetts General Hospital and Division of Pediatric Gastroenterology and Nutrition, Massachusetts General Hospital for Children, Boston, MA 02114, USA
3 Allied Health Sciences Department, University of Connecticut, Storrs, CT 06269, USA; tania.huedo-medina@uconn.edu
* Correspondence: mleonard7@mgh.harvard.edu; Tel.: +1-617-724-4155; Fax: +1-617-724-3248
Abstract: In the past it was believed that genetic predisposition and exposure to gluten were necessary and sufficient to develop celiac disease (CD). Recent studies however suggest that loss of gluten tolerance can occur at any time in life as a consequence of other environmental stimuli. Many environmental factors known to influence the composition of the intestinal microbiota are also suggested to play a role in the development of CD. These include birthing delivery mode, infant feeding, and antibiotic use. To date no large-scale longitudinal studies have defined if and how gut microbiota composition and metabolomic profiles may influence the loss of gluten tolerance and subsequent onset of CD in genetically-susceptible individuals. Here we describe a prospective, multicenter, longitudinal study of infants at risk for CD which will employ a blend of basic and applied studies to yield fundamental insights into the role of the gut microbiome as an additional factor that may play a key role in early steps involved in the onset of autoimmune disease.
Keywords: celiac; gluten; microbiome; metabolomic; environmental; genetic; personalized medicine; prospective; cohort; design

1. Introduction
Celiac disease (CD) is an autoimmune enteropathy triggered by the ingestion of gluten containing grains (i.e., wheat, barley, and rye) in genetically susceptible individuals [1]. CD represents a unique model of autoimmune disease as, in contrast to other autoimmune diseases, the triggering environmental factor (gluten), a close genetic association with human leukocyte antigen (HLA) genes (DQ2 or DQ8), and a highly specific humoral autoimmune response (autoantibodies to tissue transglutaminase) are known [1]. However, the early steps following intestinal mucosal exposure to gluten leading to the loss of tolerance and the development of the autoimmune process are still largely unknown. Studies now suggest that this loss of gluten tolerance does not always occur at the time of gluten introduction in the diet of genetically susceptible individuals; rather it can occur at any time in life as a consequence of other unknown environmental stimuli [2].
A proof of concept study, published by our group, has shown that a unique interplay between a peculiar microbiota and host may lead to alterations in metabolic pathways that result in the production of specific metabolites prior to the onset of autoimmune disease [3]. The intestinal

Nutrients 2015, 7, 9325–9336; doi:10.3390/nu7115470

www.mdpi.com/journal/nutrients

Nutrients 2015, 7, 9325–9336
microbiome is essential to the development of the immune system and begins to assemble in utero. As early as the first year after birth, the microbiota has begun to develop into an adult-like pattern, suggesting that environmental influences in infancy may have important lasting effects [4]. Many environmental factors known to influence the composition of the intestinal microbiota are also thought to play a role in the development of CD. These include birthing delivery mode, infant feeding type, history of infection, and antibiotic use [5–11].
The composition of the microbiota in infants at risk of developing autoimmune disease is altered when compared to infants without the genetic risk [3,12]. A previous study found that compared to control infants with a non-selected genetic background, at-risk subjects had a decreased representation of Bacteriodetes and a higher abundance of Firmicutes [3]. Their microbiota showed a delay in maturation at two years of age [3] while the maturation was complete in not at-risk infants at one year [4]. Additionally, this same study showed that infants who developed autoimmunity had decreased lactate signals in their stools coincident with a diminished representation in Lactobacillus species in their microbiome and that preceded the first detection of positive antibodies [3]. Early microbiota alterations in genetically-predisposed infants were also suggested in a recent study that compared microbial communities in infants carrying the DQ2 haplotype with infants who did not carry a compatible haplotype. Distinct differences between the microbiota composition at one month of age were observed, with infants who carry DQ2 showing a higher abundance of Firmicutes and Proteobacteria compared to infants without a genetic predisposition [12].
Two recent landmark studies which prospectively screened infants, with a first degree family member with CD, from birth found that CD develops quite early in life in this risk group, further supporting the notion that early environmental factors may be very important in the development of CD. These studies found that 16% of infants who have a first degree relative with CD and who carry HLA DQ2 and/or DQ8 will develop CD by age five, most of whom will be diagnosed by age three [5,6]. These studies further demonstrated that 38% of infants who are first-degree relatives of CD patients and who carry two copies of DQ2 will develop CD by age five [5,6].
The major limitation in performing pre-clinical studies pertinent to human gastrointestinal inflammatory diseases is the recent appreciation that animal models do not completely recapitulate the complexity of host-microbiota interactions that influence activation of specific metabolic pathways dictating the tolerance to immune response balance in humans. To date, no large-scale, longitudinal studies have defined if and how gut microbiota composition and metabolomic profiles may influence the loss of gluten tolerance and subsequent onset of CD in genetically susceptible subjects. Therefore, we propose to investigate the role of the developing intestinal microbiome and the resulting metabolome as additional factors that may play a key role in the onset of and predisposition to CD autoimmunity.
1.1. Objective
The objective of Celiac Disease Genomic, Environmental, Microbiome, and Metabolomic (CDGEMM) is to understand the role of the gut microbiome as an additional factor that may play a key role in early steps involved in the development of autoimmune disease. This prospective, longitudinal, observational study will serve as a basis for investigating the natural history of CD and other autoimmune disorders. We hypothesize that the combination of introduction of gluten into the diet and particular microbiota composition of infants genetically at risk for CD activates specific metabolic pathways that can contribute to the loss of gluten tolerance and to the onset of autoimmunity, as reflected by specific metabolomic phenotypes. Our ultimate goal is to identify and validate specific microbiome and metabolomic profiles that can predict loss of tolerance in children genetically at risk of autoimmunity. This study will serve to provide the foundation of knowledge necessary in order to one day implement early preventive interventions to re-establish tolerance and prevent CD.
9326

Nutrients 2015, 7, 9325–9336
Nutrients 2015, volumeme, page–page 
1.2. Specific Aims 1.2. Specific Aims 
The three specific aims of CDGEMM are: The three specific aims of CDGEMM are:  1. 1T.oT sotusdtuyd my modoidfiicfiactiaotniosn osfo tfhteh einifnafnatnst’s m’ micircorobbioiomme einin rreelalatitoionn ttoo ssppeecciiffiicc eennvviirroonnmmeennttaall ffaaccttoorrss,,  prepsreensceen coer oarbsaebnsceen coef oHfLHAL DAQD2Q a2ndan/odr/ DorQD8 Qpr8epdriespdoisspinogsi nggengeesn, easn,da nind rienlarteiloanti oton ttooletoralenrcaen vcse.  imvms.uinme mreusnpeonressep loeandseinlgea tdoi tnhge taoutthoeimaumtouinmem inutnesetiinnatel sitninsuallti ntyspuilctatly opfi cCaDl o; f CD; 2. 2T.oT  sotustduyd  ythteh einifnafnatnst’s  ’mmeteatbaobloolmomici cpphheennootytyppee  vvaarriaiattioionn  iinn  rrelation  to ttoolleerraannccee vvss..  immune  resrpeospnosen slealdeaindgin tgo tohet haeuatouitmoimmumnuen inetienstteinstainl ainlsiunsltu tlyt ptyicpailc aolf oCfDC;D a;nadn  d
3. 3T.oT  ionvinevsteisgtaigtea ttehteh  eimimpapcatc  tofo fspspeceicfific cbbaactceteriraia‐d-deerirviveedd  mmeetatabbooliltiteess  oonn  gguutt  mucosal  molecular 
patphawthawysa ylesaldeaindgin tgo tohet heearelayr lsytesptes posf oCfDC pDapthaothgoengensiess. is.
22.. SSttuuddyy DDeessiiggnn 
TThhee CCDDGGEEMMMM ssttuuddyy isis a ammulutilctiecnetnetre srtustduyd cyomcopmripsreidse odf coof llcaoblloarbaotorarsto irns thine tUhneitUedn iStetadteSst aatneds   aItnadly.I tIatl yis. sIutpiesrvsuispeedr vbiys eMdabsys GMeansesraGl eHnoersaplitHalo fsopri tCahl iflodrreCnh ailtd HreanrvaatrdH MarvedaricdalM Secdhiocoall iSnc Bhooostloinn,  BMoasstosanc,hMusaeststasc  h(uCsleinttisca(Cl  lTinriicaalsl  Tirdiaelnstiidfieenr:tifiNerC:TN02C0T6012300661).3 0C6)D. GCEDMGMEM  Maimasim  tsot osstutuddyy  ggeennoommiicc,,  eennvviirroonnmmeennttaall,, mmiiccrroobbiioommee,, aanndd mmeettaabboolloommiicc ffaaccttoorrss tthhaatt mmaayy ccoonnttrriibbuuttee ttoo tthhee ddeevveellooppmmeenntt ooff CCDD  lloonnggiittuuddiinnaallllyy.. InIna ddaidtidointiotonr  etpoe atreedpeCaDtesde roClDog icsaelrsoclroegeincainl gsucnretielnaignegfi  vuen, tdile taaigleed  efnivvei,r ondmeteanilteadl  iennfvoirrmonamtioenntiasl ionbftoarimneadtiofnre iqsu oebnttalyin, eadn dfresqtuooenl tilsy,c aonllde cstteodole ivse croylltehcrteeed emvoenryth tshrfeoer mthoentfihrss tfotrh trheee  yfierastr sthorfeel iyfeeaarnsd ofe lvifeer yansdix evmeoryn tshisx mthoenretahfst ethr eurneatifltearg uenfitivl eag(eF ifgivuer e(F1i)g.urIen f1a)n. Itns’famntisc’r ombiicormobeioamnde  manedta  mboeltoamboelowmilel  bweilcl obme pcaormedpalroendg iltoundgiintualdlyinpalalyyi npgaypinargt ipcualratircuatltaern  atitotennttioond itfofe  rdeinffceersenbceefso rbeefaonrde  aafntder atfhteeri nthtreo dinutrcotidouncotifognl uotfe  gnl,ubteenfo, rbeeafonrde aafntedr athfteerd  tehvee ldoepvmeleonptmoef nCt Dofw  ChDen wahpepnl icaapbpllei,caabslew, ealsl  awsemll aansy  motahneyr eontvhierro nemnveinrotanlmfaecntotarls .fAacdtodristi. oAnadldlyi,tiwonitahlilny,t hweitlohning itthued inloanlgsittuuddyinwale  swtuildl yp ewrfeo rwmilal  npeesrtfeodrmca  ase  nceosntetrdo lcaansea lycosinst.roInl faannatslytshiast. gInofaonntsto  thdaetv egloo ponC Dtow  diethvebleopm  CatDch  ewditwh ibthe cmonattrcohleidn fwanitths  wcointhtroal ginenfaentitcs wpriethd ias pgoesnietitoicn ptroe,dbisuptowsihtioonh atov,e bnuot twdheov helaovpee ndo, tC dDe.veAlospeecdo,n CdDa.n Aal ysescisonwdi lalnmalaytscihs  iwnfilaln  mtsawtchho  ignofaonntst owdheov egloop  oCnD  tow idthevceolnotpro  Cl iDnf awnittshw  choontdrooln  ointfcaanrtrsy  wthheoH  dLoA  nporte dciasrproys itnhge gHenLeAs  tporeaddidsrpeosssienngv  igreonnems etnot aal dfadcrteosrss tehnavtimroanymceonnttarilb  fuatcetotorsa lttheraat timonasyi ncothnetrmibiucrteo btioo maeltearnadtiponresd  iisnp  othsee  tmoitchreobdieovmeleo apnmde pnrteodfisCpDo.se to the development of CD. 

FFiigguurree 11. .S  cShchememataitcico voevrevrivewiewof  dofa tdaaatnad  asnadm  psalemcpollele cctoiollnecptiroonce  dpurorceesdinuvreoslv  iendvionlvthede  Cinel itahceD  Ciseelaisaec  GDeisneoamsei cG, eEnnovmiroicn, mEnevnitraol,nMmiecnrotabli,o Mmiec,roanbdiomMee,t aabnodl oMmeitca(bCoDloGmEicM (MCD) GStEuMdyM. ) Study. 

 

retenCCtiDDonGG. EEIMMnnMMov wwataaivss edd  eedssaiitggann eecddo llwweciittthhio ntthh  eete  icinnhttneeinnqttu  tteoos  mmaliilnnoiiwmm iifzzoeer  ssrttueumddyyo tvvei isssiitttuss daaynn ddr emmcraauxxiitimmmiiezzneet  ppaaacrrrttoiiccsisipp  aathnnett   rSTrawtrfmseUeheeoeirtmlhtenhatcersraeeooec teeiat lnsetsteiuaoriiev eseltmrerrgt.dineic o oawahtTri thh nSnttlehrisah ttetdm.tceo hoatehyeIicatesiooononteerraehttnnuisets rehvmeeaw   oetlaetremysvoapan iri ew tmac ttadethncaehchtrho miotr epetorv siueisuiohaticteerrsnhnotulil ed spdetteg mag iorbtaratlbachihhe aontectecy a eetetesttombeoyhl. smcuyflrms peaoooFian aa ifalortorsyltrliolrederarloscot i fc m oiezaiereacbdtmmelnsaiciwelyron.pised m aa lnFephcaimawittocnelhotoyti oerdoa tolf´pa e mctinierlhctbfhar 8a , oowpetaonas0saerfmltetneieh eln˝oqvyt drctpoCodhmuee o n lfferepcel r[ryaoaclso1oterce loltie3mftlzha ocmauttf]eltqinoie.clpilrnmouodp,rinBlul  wneea’liaeenlssettosl,t dyoee ltt pfpoi−i  taoqoooovtd ea8nrhtunnnidilo0dnamlare,ln i   eitt° ausaqnoa samCtrttlsu iveiyrlehrdeo  oieet[qeedc nbuu1tsors iueaetnn3carrnietoihn]nisaoclan.sgntiiei ontnuglBruhcdidlnep eodollts ws’oea enalsydvdcloisa pnhertipear.draeeceaielteecT ts diraicmidv.h rohd oroeacei uvSaniiassontrsseiytat yltt aatorrla mmrpeeirbnotilscceacbloeiltpeiho tgpuau ndiclcwonhoneedtotiouie rtsn sylqalc ud.trll  cuoerfdetssSorirectlceeiebetolhttcrtssoeleuissehri.olcoevms odtionTtnaemleuit  riahrdoahrqcgmciyrynti eotuhhets a  ahlae meUfltta tcelreosioseelnoeomo crlt rmcnonuitu tiottitaewatdegohauettlildyoolhnyyesrsr.       
study and routine pediatric care overlap. All samples collected remotely are shipped to our research  9327

3

Nutrients 2015, 7, 9325–9336
a single blood draw at several time points during which sample collection for the study and routine pediatric care overlap. All samples collected remotely are shipped to our research facility overnight, to maintain sample viability, where they are ultimately stored at ´80 ˝C for processing.
2.1. Participants
CDGEMM aims to enroll 500 infants aged 0–6 months with a first-degree family member with CD. No more than half of the enrollees will be recruited in Italy. The first study samples must be collected prior to the introduction of solid foods, thus children who have been introduced to solid foods are excluded. The diagnosis of CD in the family member is confirmed by the recruiting institution by review of the pathology report obtained during the confirmatory or diagnostic biopsy. Patients seeking enrollment whose family member did not undergo a confirmatory endoscopic procedure with biopsy are still evaluated for inclusion. Those infants whose family member meets diagnostic criteria based on the published guidelines are accepted for inclusion into the study [14].
2.2. Data Collection
The timetable of data collection is shown in Table 1. Health status, anthropometrics, nutritional information, household, and environmental information is obtained regularly. Data is collected through questionnaires, which are distributed directly to the parent’s email address according to a pre-programmed schedule designed around the child’s date of birth. When data is collected in this format, all information gathered is entered directly by participants into a central database using the web-based data collection and management application REDCAP. Parents who are unfamiliar with, or do not have regular access to, a computer are able to receive all questionnaires in pen-and-paper format directly to their home address (through regular mail) according to the study schedule.
2.3. Parental and Child Questionnaires
Questionnaires pertain to parental and infant nutrition including infant feeding type, timing of gluten exposure and introduction to other foods, and types and frequency of foods ingested. Information on infant and maternal feeding habits is collected using the Infant Feeding Practices Study II (IFPS II) validated questionnaires created by the Food and Drug Administration (FDA) and Centers for Disease Control and Prevention, in collaboration with other federal agencies [15], modified in order to best assess quantity and frequency of gluten containing and other foods. For the Italian cohort, all instruments were translated and altered slightly for colloquial appropriateness by Italian study staff, with supervision of the principal investigator who is a native speaker of both English and Italian. At enrollment and every six months thereafter, questions regarding medical problems, medication use, including antibiotic use, probiotic use, household members, daytime activity, and other environmental factors are addressed. A symptom diary is included starting at six months of age and vaccinations are documented yearly. A monthly food diary and monthly antibiotic diary during the first year after birth and at 15 months of age collects information to ensure that introduction of foods and any use of antibiotics are recorded accurately.
2.4. Serological Markers
Serum is collected every six months for the first three years and then yearly for the remainder of the study (Table 1). An initial sample is obtained prior to the introduction of any solid food for baseline studies. It is analyzed for Immunoglobulin A (IgA) tissue transglutaminase (tTG), using QUANTA Lite Rh-tTG IgA ELISA (INOVA Diagnostics, San Diego, CA, USA). Serology for IgA and Immunoglobulin G (IgG) antideamidated gliadin antibody (dGP) using QUANTA Lite Celiac DGP Screen (INOVA Diagnostics, San Diego, CA, USA) will also be performed. If a patient tests positive for IgG dGP, an IgA level will be determined to investigate the possibility of IgA deficiency. Parents are informed of their child’s serological status throughout the study period—specifically, after each blood draw.
9328

Nutrients 2015, 7, 9325–9336

Table 1. Detailed schedule of data and samples collected throughout the CDGEMM Study.

Age in Months

0 1 2 3 4 5 6 7 8 9 10 11 12 15 18 21 24 27 30 33 36 42 48 54 60

Maternal Stool and Breast Milk Sample X

Cord Blood

X

Blood Sample

X X* X X X X X X

Stool Sample

XXXX

X XXXXXXXXXXXX

Food Diary

XXXXXXXXXXX X XX X X XXXXX

Antibiotic Diary

XXXXXXXXXXX X X

Maternal Diet

XXXXXXXXXX

X

X XXX

Anthropometrics

X X X X X XXXXX

Medical History

X

X

X X X X XXXXX

Parent and Child Demographics

X

X X XXX

Assessment of Sleep and Activity

X X X X X XXXXX

* HLA testing performed.

9329

Nutrients 2015, 7, 9325–9336
2.5. Whole Blood
For infants enrolled during gestation, whole blood is collected at birth via cord blood for storage in our biorepository and epigenetic studies. An additional tube of whole blood is collected at one year of age for HLA testing using the DQ-CD screening kit (Biodiagene, Palermo, Italy). Results of the child’s HLA testing are conveyed to parents when the child reaches 18 months of age. DNA is isolated from the remaining blood volume (QIAmp DNA Blood Maxi Kit, QIAGEN, Hilden, Germany) and stored at ´20 ˝C for future use. Excess clotted blood obtained every six months is also stored at ´80 ˝C for future use.
2.6. Stool
Stool is collected from the infant at enrollment, which may be as early as seven days after birth. (Figure 1). For children enrolled later in life, enrollment stool samples may be collected anywhere after 15 days and before three months of age. Stool is then collected every three months for the first three years of life and finally every six months thereafter until completion of the study when the child is aged five years. Each stool specimen is collected—from the diaper or from a provided collection hat, depending on the child’s age—into three cryo vials. Two cryo vials contain RNA later solution (Fischer Scientific, NY, USA) to preserve DNA and RNA. A third empty cryo vial will be used for future metabolomic studies. Stool is frozen for 24 h in the participant’s home to ensure sample integrity during shipment and then packaged with a frozen ice pack for overnight delivery to the coordinating research facility [13,16]. Upon arrival to the research facility, all three cryo vials are immediately stored at ´80 ˝C for microbiome and metabolomic studies.
2.7. Maternal Samples
Parents who enroll their child during gestation have the option to provide maternal stool and breast milk samples to coincide with the child’s enrollment samples at seven days of age. Like the child’s stool, maternal stool is voided into the stool collection hat and collected into three cryo vials, two of which contain RNA later solution. A small amount of breast milk is also collected concurrently for storage in our biorepository at ´80 ˝C.
2.8. Diagnosis of Celiac Disease
Patients testing positive for IgA tTG or IgA dGP or those with IgA deficiency who test positive for IgG dGP will undergo a repeat serological test three months after the initial positive test. If the second serological test returns positive, an anti-endomysial antibody (EMA) will be performed for confirmation. Following two positive serological tests, the patient will be referred to the participating institution, or a pediatric gastroenterologist nearby (if participating remotely), for confirmatory biopsy in order to make a diagnosis of CD.
3. Factors of Interest
3.1. Environmental
Historically, birthing delivery mode, method of infant feeding and introduction of gluten to the infant diet have been examined for their contributions to the development of CD. Though vaginal delivery and increased duration of maternal breast feeding were once thought to be protective against the development of CD, to date evidence is inconsistent. In terms of birthing delivery mode, the previous association suggested between cesarean delivery and increased incidence of CD [7] seems to no longer be supported [8]. Emilsson and colleagues extracted data from the Norwegian Mother and Child Cohort Study (MoBa), which included approximately 114,000 individuals, to investigate this association and concluded that birth by cesarean section was not associated with increased risk for CD [17]. A recent meta-analysis by Szajewska and colleagues considered results of 21 studies, among
9330

Nutrients 2015, 7, 9325–9336
which large-scale interventional and observational cohorts were included [9]. This meta-analysis revealed that breastfeeding, whether exclusive or in combination with formula feeding, did not reduce the risk of developing CD. As a collective, the same meta-analysis concluded, in agreement with the Lionetti et al., and Vriezinga et al., prospective cohort studies [5,6], that timing of introduction of gluten to the infant diet did not significantly influence the risk of developing CD by age three or five years [9]. This is in contrast to the Swedish epidemic of CD presented in 2000, in which a sharp increase in the number of CD cases was noted to be temporally associated with revision of the Swedish national guidelines recommending the introduction of gluten to the infant diet after six months of age [18]. At that time, incidence of CD (per 100,000 births) at two years of age increased from 1.7 to 3.7 cases.
Antibiotic exposure in early infancy has also emerged as a contributory factor to development of CD. Canova et al., found that infections in the first year of life were significantly associated with histologically confirmed CD [10]. A dose dependent effect of antibiotic use has also been reported previously, with increased use of antibiotics serving to increase the risk of developing CD [11]. Socioeconomic status has also been shown to influence development of CD, with CD more frequently observed in children from high SES as compared to low [10]. In fact, rates of autoimmune disease in general were found to be lower in children from lower SES, providing support for the longstanding but controversial hygiene hypothesis [10].
The CDGEMM Study has been carefully designed to address the aforementioned variables, and more, frequently through prospective parent report from the child’s time of birth. Birthing delivery mode is reported by the parent during completion of study enrollment forms. A diary of antibiotic usage is completed by the parents monthly for the first year of life, and food diaries completed at the same time points assess duration of breastfeeding (or other preferred feeding mode) and timing of gluten introduction. After the first year of life, detailed, though less frequent, records of antibiotic use are obtained with each stool sample for the remaining duration of the study. Though the food diary is discontinued at this time, pertinent dietary habits of both the mother and child continue to be carefully recorded approximately every six months.
3.2. Genetic
We hypothesize that studies examining environmental factors to date have been inconsistent largely due to the fact that environmental factors contribute to disease development differently based on underlying host factors. While many of the above mentioned environmental factors are likely to be important to the development of CD, it may be the number or the combination of environmental “hits” in the setting of a particular genetic compatibility which influences the microbiome and resulting metabolome to ultimately alter host physiology and lead to the loss of tolerance and development of autoimmune disease. Genetic compatibility with HLA DQ2 and or HLA DQ8 are necessary for the development of CD. Interestingly, the particular genetic makeup has recently been shown to be one of the strongest links to the disease to date. A large-scale study by Lionetti and colleagues recently found that CD developed more frequently in children homozygous for a variant of HLA DQ2 (DQA1*05-DQB1*02) compared to those carrying it only heterozygously [5], confirming previous studies that suggest homozygosity for this allele to be a “high risk” genotype for development of CD [11]. The findings by Lionetti and colleagues are particularly compelling as they suggest a diagnostic utility of early HLA typing of infants at risk for CD and suggest that early environmental factors may contribute to the development of disease differently based on genotype.
3.3. Microbiome/Metabolome
It is difficult to isolate the effects of specific environmental variables on the development of CD without taking into account the interplay with the microbiome. The previously reported protective effect of vaginal delivery from developing CD is likely attributable to the variability in microbiome composition that results from different birthing delivery modes [19]. Similarly, antibiotic use
9331

Nutrients 2015, 7, 9325–9336 Nutrients 2015, volumeme, page–page 
eeedtp[tpttohhh2axxear0erorpperto]lisasrdloyn.euoao isv Ticungwiuenenthaghnriesivleeh tvdtlihre iiz ifge[C rmeetm2saohdtDu0thinhroec]lcGooma.trdaiusroEtneueohgbTlMagnsu ehilhta otgssopMalemoerh lepep  Crfebserrcatmu,eeenDuo cdvelmvhtritGncoeibiiocnerrnrEatroostsgtittMahbv n i CagopoipemMnoenDwermdmnesor n tno ivcoecutnmtooird,tar doehg beliteamcoelahrst,fnrtte ee eatioemla  otcantda ptpniteoougecersrtvarvoeneastltneevvbioloqllyoieo oadui inmtlnnppnhoeetdtricsim eenCcerog,edgadiDcpcar irepm,asmsue vipteonaanieeidorvntcitlgetqsoaredureeoub pi nnnndbsogeiiekeon iletiottovyognnwiommce pdefmatloptieoildvircl oici ky,citne crdhrsvattopo auauenetmrbntgoanlae oidrl bcilodsdaptyoi.ymitne itmsginivvepceetge eeo onet tlsoCehewncnetcpavDoidiehtc tmitahnrwic[ on iab2opqanodn0imenunrm,iar2kvteeesli1eysisobd]wne tnitiuntotacoaioatavtnl ilhliicid tszdrne d.oaeeecvcandnxhvoettm paensucitlb oialeoiqwdisgunpouuasittbetrataihaeeecsll     44.. SStatatitsistitcicaall AApppprrooaacchh 
44..11.. PPoowweerr AAnnaallyyssiiss  OOuurr  ppoowweerr asassusmumptpiotinosn (sFig(Fuirgeu r2e)  a2re)  baarseedb  oanse douro  rneceonutrly rpeucebnlitslyhedp udbaltias h[5ed,6]. dAatsasum[5i,n6]g. 
Athsastu: m(ai)n 5g00 tihnafta:nt(sa )w5il0l 0bein efnanrotsllewdi ldl ubreinegn trhoell epdilodtu preinrigodth aendp iltohtatp tehreio ddroapn dratthe awt itlhl ebed nroop mroartee  wthialln b2e0%n ogimveonr ethteh adnur2a0t%iong oivf etnhet hsteudduy raantido nsimofiltahre tos ttuhdayt waen dhasdim wiliathr  otourt hpartevwioeuhs asdtuwdiietsh,[5o,u6r]  p40re0v iinofuasntsst uwdilile cso, m[5p,6l]et4e0 t0hein pfailnotts pwroitlloccoolm; (pbl)e ttheet hexeppeicltoetdp prrootobcaobli;li(tby) otfh CeDe xspereucmted auptroo banabtiibliotydieosf  CpoDsistievriutmy aat u1t8o–2a4n tmiboondtihess pino sfiirtisvt‐idtyegarte1e8 r–e2l4atmivoens tihs s1i0n%fi irrsrte-sdpeegcrteiever eolaf ttihvee sHiLs A10 c%omirpreastpibeicltitivye; (ocf)  t7h0e%H oLf Afircsot mdepgarteieb irleitlya;ti(vce)s7 a0r%e HofLfiAr sDt Qde2g arnede /roerl aDtiQve8s paorseitHivLe A(bDasQed2 oann do/uorr pDreQli8mpinoasrityi vdeat(ab)a; s(edd)  obnasoedu ropnr eploiminitnsa  ary  adnadt ab),;  t(hde) bexapseedcteodn  pprooinbtasbailiatyn dofb  C, tDh eseerxupmec taeudtop raonbtaibboildityy  poof sCitDivisteyr uamt  18a–u2to4  amnotinbtohds yinp oHsLitAiv‐ictyomatp1a8t–ib2l4em fiorsntt hdseginreHe LrAel-actoivmesp aisti b13le.3fi%rs; t(de)e gthreee mreinlaitmivaels diset1e3c.t3a%bl;e( ec)htahnegme iinni mCaDl  dauetteocimtabmleunchitayn  gperoinbaCbDiliatyu tiosi m10m%u,n  witye parnotbicaibpialittey  tisha1t0 %d,uwrinega ntthicei psatutedtyh aptedruiordin  gapthperosxtuimdyatpeelyri o5d0  ainpfparnotxs iwmiallt edleyv5e0loinpf CanDts.  will develop CD.
 
Figure 2. Study scheme outlining the recruitment and projected incidence of celiac disease (CD) and  related HLA genotypes for participants in the Celiac Disease Genomic, Environmental, Microbiome,  and Metabolomic Study.. 
CCDD iiss uunniiqquuee aammoonngg gguutt iinnffllaammmmaattoorryy ddiisseeaasseess iinn tthhaatt tthhee eennvviirroonnmmeennttaall ffaaccttoorr ttrriiggggeerriinngg tthhee  iinnflflaammmmaattoorryy eenntteerrooppaatthhyy iiss kknnoowwnn ((gglluutteenn)) aanndd,, tthheerreeffoorree ttrraacceeaabbllee iinn tteerrmmss ooff ttiimmiinngg ooff eexxppoossuurree..  GGiivveenn iittss ssttrroonngg aasssosocciaiatitoionn wwitihth HHLALA DQD2Q/D2/QD8Q, t8h,ist hsitsudsytu pdryovpirdoevsi dthees itdheeali dcoeanltrcool ngtrrooul pgsr,o buoptsh,  bHoLthA HDLQA2/DDQQ82 /nDegQa8tivnee gsautbivjeectssu fbojre cstusbftorarcstiuobnt raancatiloynsisa naas lwyseisll aass gweenlel taicsalgleyn seutisccaelplytibslues scuepbtjeibcltes  swuhboje chtsavweh  onohta  vloesnt ottolloesrtantoclee,r atnoc  eid, teontiidfyen  tmifeytambeotlaitbeosl itaesssoacsisaotceida tewditwhi tlhosloss  sofo fgglluutteenn  ttoolleerraannccee..  AAddddiittiioonnaallllyy,,  iinnffaannttss  tthhaatt  ggoo  oonn  ttoo  ddeevveelloopp  CCDD  wwiillll  uunnddeerrggoo  lloonnggiittuuddiinnaall  aannaallyyssiiss,,  aatt  mmuullttiippllee  ttiimmeeppooiinnttss,, bbeeffoorree aanndd aafftteerr tthhee ddeevveellooppmmeenntt ooff CCDD.. 
44..22.. DDeevveellooppiinngg aann IInntteeggrraattiivvee MMuullttiilleevveell MMooddeell ttoo PPrreeddiicctt CCeelliiaacc DDiisseeaassee  TThhee CCDDGGEEMMMM SSttuuddyy,, wwiitthh aallll ooff iittss iinncclluuddeedd mmeeaassuurreess,, wwiillll sseerrvvee ttoo  ccrreeaattee aa  llaarrggee ddaattaabbaassee 
iinncclluussiivvee ooff eennvviirroonnmmeennttaall aanndd mmuullttii-‐oommiicc iinnffoorrmmaattiioonn.. TThhuuss iitt wwiillll rreeqquuiirree eexxtteennssiivvee mmaannaaggeemmeenntt  aanndd  ssoopphhiissttiiccaatteedd ccoommppuuttaattiioonnaall ssttaattiissttiiccaall aannaallyyssiiss..  WWee  iinntteenndd  ttoo  aappppllyy  iinntteeggrraatteedd  ddaattaa  aannaallyyssiiss  ((IIDDAA)) wwhhiicchh wwiillll ccoommbbiinnee bbiioollooggiicc,, ggeenneettiicc,, aanndd eennvviirroonnmmeennttaall ddaattaa ffrroomm ppaarrttiicciippaannttss iinn CCDDGGEEMMMM..  IInntteeggrraattiioonn ooff rraaww mmuullttii-‐oommiicc ddaattaa ttoo ccrreeaattee pprreeddiiccttiivvee mmooddeellss ffoorr tthhee ddeevveellooppmmeenntt ooff CCDD wwiillll bbee aa  ccrruucciiaall iinnnnoovvaattiioonn iinn oorrddeerr ttoo aaddvvaannccee kknnoowwlleeddggee aabboouutt tthhiiss mmuullttii-‐ffaaccttoorriiaall ddiisseeaassee.. 
Multilevel  mediation‐moderation  models  allow  for  modeling  the  correlation  of  clustered  data  (e.g.,  time  points  within  individuals)  and  to  analyze  the  effect  of  a  factor  as  a  possible  bridge  (mediator) or attenuator/amplifier (moderator) of a relationship between two other variables. These 
9332 8 

Nutrients 2015, 7, 9325–9336
Multilevel mediation-moderation models allow for modeling the correlation of clustered data (e.g., time points within individuals) and to analyze the effect of a factor as a possible bridge (mediator) or attenuator/amplifier (moderator) of a relationship between two other variables. These mediation-moderation models will be used with data from infant and parent samples, integrating serologic data, genetic data, and environmental factors—including mode of infant delivery, dietary regimen, antibiotic intake, and many others—with microbial information and metabolic pathway phenotypes to elaborate causation models that can detect specific individual patterns leading to the loss of tolerance to gluten.
Specifically, we will examine (1) the genomics effects on proteomics mediated by the microbiome and metabolic activity on the final proteome; (2) the effect of the environment (region, rural versus urban) on the intestinal microbiome and the metabolites produced; and (3) the mediation effect of metatranscriptomics on the onset of CD. Further description and identification of the mechanism by which environmental and lifestyle factors (e.g., diet, antibiotic intake) contribute to the loss of tolerance to gluten and development of CD are desperately needed. Factors involved are likely to vary on individual aspects such as HLA genotype; timing, frequency, and type of antibiotic intake; dietary regimen; and birthing delivery mode, but studies to date have not examined these factors in combination. The proposed research aims to test potential mechanisms by examining the interaction among a wide range of factors. Our findings may help to establish personalized strategies to address the etiology of losing tolerance to gluten through the identification of predictive biomarkers thus allowing us to predict CD. The model will additionally be applicable to other autoimmune diseases such as type-1 diabetes mellitus.
The conceptual framework proposed (Figure 3) draws on biological and environmental literature [22–25] examining the intersection between biomarkers, environmental factors, genetic factors, and individual factors. The onset of loss of tolerance to gluten will be measured as a continuous dependent variable based on levels of tTG IgA. Ultimately the diagnosis of CD will be confirmed by an endoscopy with duodenal biopsy. Given this, we expect to observe the following:
(1) Path a: a causal path between the genetics and the final effect on the onset of CD, with the microbiome as the mediator (a1 ˆ a2), such that the intestinal microbiome will be altered in order for the genetics to ultimately influence the development of disease (microbiome-mediated epigenetic pressure);
(2) Path b: a causal path of the microbiome’s effect on CD mediated (b1 ˆ b2) by the metabolic activity of intestinal bacteria, which, through the production of metabolites, affects the intestinal transcriptome and proteome. Including a bidirectional path between microbiome and metabolomics to capture how metabolism could alter the microbial metabolism and ultimately the risk for CD;
(3) Path c: A bidirectional path between the transcriptome and the microbiome as their causal relationship will depend on each other;
(4) Path d: a direct relationship between lifestyle factors (here represented by dietary regimen and antibiotic intake) and the onset of the disease mediated (d1 ˆ d2) by alterations in the microbiome;
(5) Path e: The environment (including region characteristics, urban compared to rural, number of individuals living in the household, birth order, pets present) affecting the onset of the disease as a mediator and affecting the composition of the microbiome and resulting metabolome, in turn, influencing gene and protein expression;
(6) Full Model: Time points will be clustered within individual characteristics, analyzing all the paths listed above (Figure 3), individuals will be clustered within families, and families within regions.
9333

Nutrients 2015, 7, 9325–9336
Figure 3. Celiac Disease Genomic, Environmental, Microbiome, and Metabolomic (CDGEMM) Integrative Multilevel Model to Predict Celiac Disease. 5. Conclusions The implementation of primary prevention strategies for CD through manipulation of the microbiota would represent a complete shift of paradigm in autoimmune pathogenesis and treatment of life-long autoimmune disorders. The identification of specific CD metabolomic phenotypes can also help to define additional diagnostic tools and therapeutic interventions. Additionally, CDGEMM’s biorepository will allow for future epigenetic studies and validation of biomarkers. Our findings may have a far-reaching impact on other pediatric autoimmune diseases in which the diet-genome-microbiome interaction in the pathogenesis of the disease has been hypothesized. This, in turn, could help to set up strategies aimed at readdressing the process of oral tolerance to gluten and to other environmental antigens, opening the way to novel approaches of prevention/treatment of autoimmune diseases. Since 3 million people in the U.S. are affected by CD and approximately 17 million people suffer from other autoimmune diseases, and there currently are no effective strategies to prevent these conditions, the findings from CDGEMM can potentially have a tremendous impact on pediatric public health [26]. Acknowledgments: This work was conducted with support from The Ellison Foundation, Mead Johnson Nutrition, The Harvard Catalyst, The Harvard Clinical and Translational Science Center (NCRR and NCATS, NIH Award UL1 TR001102), and financial contributions from Harvard University and its affiliated academic healthcare centers. Author Contributions: All authors contributed to writing the manuscript and reviewing the manuscript. Conflicts of Interest: The authors declare no conflict of interest. References 1. Green, P.H.R.; Cellier, C. Celiac disease. N. Engl. J. Med. 2007, 357, 1731–1743. [CrossRef] [PubMed] 2. Catassi, C.; Kryszak, D.; Bhatti, B.; Sturgeon, C.; Helzlsouer, K.; Clipp, S.L.; Gelfond, D.; Puppa, E.; Sferruzza, A.; Fasano, A. Natural history of celiac disease autoimmunity in a USA cohort followed since 1974. Ann. Med. 2010, 42, 530–538. [CrossRef] [PubMed]
9334

Nutrients 2015, 7, 9325–9336
3. Sellitto, M.; Bai, G.; Serena, G.; Fricke, W.F.; Sturgeon, C.; Gajer, P.; White, J.R.; Koenig, S.S.; Sakamoto, J.; Boothe, D.; et al. Proof of concept of microbiome-metabolome analysis and delayed gluten exposure on celiac disease autoimmunity in genetically at-risk infants. PLoS ONE 2012, 7, e33387. [CrossRef] [PubMed]
4. Palmer, C.; Bik, E.M.; DiGiulio, D.B.; Relman, D.A.; Brown, P.O. Development of the human infant intestinal microbiota. PLoS Biol. 2007, 5, e177. [CrossRef] [PubMed]
5. Lionetti, E.; Castellaneta, S.; Francavilla, R.; Pulvirenti, A.; Tonutti, E.; Amarri, S.; Barbato, M.; Barbera, C.; Barera, G.; Bellantoni, A.; et al. Introduction of gluten, HLA status, and the risk of celiac disease in children. N. Engl. J. Med. 2014, 371, 1295–1303. [CrossRef] [PubMed]
6. Vriezinga, S.L.; Auricchio, R.; Bravi, E.; Castillejo, G.; Chmielewska, A.; Crespo Escobar, P.; Kolacek, S.; Koletzko, S.; Korponay-Szabo, I.R.; Mummert, E.; et al. Randomized feeding intervention in infants at high risk for celiac disease. N. Engl. J. Med. 2014, 371, 1304–1315. [CrossRef] [PubMed]
7. Decker, E.; Engelmann, G.; Findeisen, A.; Gerner, P.; Laass, M.; Ney, D.; Posovszky, C.; Hoy, L.; Hornef, M.W. Cesarean delivery is associated with celiac disease but not inflammatory bowel disease in children. Pediatrics 2010, 125, e1433–e1440. [CrossRef] [PubMed]
8. Emilsson, L.; Magnus, M.L.; Stordal, K. Perinatal risk factors for development of celiac disease in children, base on the propsective norwegian mother and child cohort study. Clin. Gastroenterol. Hepataol. 2015, 13, 921–927. [CrossRef] [PubMed]
9. Szajewska, H.; Shamir, R.; Chmielewska, A.; Piescik-Lech, M.; Auricchio, R.; Ivarsson, A.; Kolacek, S.; Koletzko, S.; Korponay-Szabo, I.; Mearin, M.L.; et al. Systematic review with meta-analysis: Early infant feeding and coeliac disease—Update 2015. Aliment. Pharmacol. Ther. 2015, 41, 1038–1054. [CrossRef] [PubMed]
10. Canova, C.; Zabeo, V.; Pitter, G.; Romor, P.; Baldovin, T.; Zanotti, R.; Simonato, L. Association of maternal education, early infections, and antibiotic use with celiac disease: A population-based birth cohort study in northeastern Italy. Am. J. Epidemiol. 2014, 180. [CrossRef] [PubMed]
11. Liu, E.; Lee, H.S.; Aronsson, C.A.; Hagopian, W.A.; Koletzko, S.; Rewers, M.J.; Eisenbarth, G.S.; Bingley, P.J.; Bonifacio, E.; Simell, V.; et al. Risk of pediatric celiac disease according to HLA haplotype and country. N. Engl. J. Med. 2014, 371, 42–49. [CrossRef] [PubMed]
12. Olivares, M.; Neef, A.; Castillejo, G.; Palma, G.D.; Varea, V.; Capilla, A.; Palau, F.; Nova, E.; Marcos, A.; Polanco, I.; et al. The HLA-DQ2 genotype selects for early intestinal microbiota composition in infants at high risk of developing coeliac disease. Gut 2015, 64, 406–417. [CrossRef] [PubMed]
13. Feigelson, H.S.; Bischoff, K.; Ardini, M.E.; Ravel, J.; Gail, M.H.; Flores, R.; Goedert, J.J. Feasibility of self-collection of fecal specimens by randomly sampled women for health-related studies of the gut microbiome. BMC Res. Notes 2014, 7, 204. [CrossRef] [PubMed]
14. Catassi, C.; Fasano, A. Celiac disease diagnosis: Simple rules are better than complicated algorithms. Am. J. Med. 2010, 123, 691–693. [CrossRef] [PubMed]
15. Division of Nutrition, Physical Activity, and Obesity, National Center for Chronic Disease Prevention and Health Promotion. Infant Feeding Practices II. Available online: http://www.cdc.gov/nccdphp/ dnpao/index.html (accessed on 14 July 2015).
16. Gorzelak, M.A.; Gill, S.K.; Tasnim, N.; Ahmadi-Vand, Z.; Jay, M.; Gibson, D.L. Methods for Improving Human Gut Microbiome Data by Reducing Variability through Sample Processing and Storage of Stool. PLoS ONE 2015, 10, e0134802. [CrossRef] [PubMed]
17. Magnus, P.; Igrens, L.M.; Haug, K.; Nystad, W.; Skjaerven, R.; Stoltenberg, C.; MoBa Study Group. Cohort profile: The norwegian mother and child cohort study (NoBa). Int. J. Epidemiol. 2006, 35, 1146–1150. [CrossRef] [PubMed]
18. Ivarsson, A.; Persson, L.A.; Nystrom, L.; Ascher, H.; Cavell, B.; Danielsson, L.; Dannaeus, A.; Lindberg, T.; Lindquist, B.; Stenhammar, L.; et al. Epidemic of coeliac disease in Swedish children. Acta Paediatr. 2000, 89, 165–171. [CrossRef] [PubMed]
19. Dominguez-Bello, M.G.; Costello, E.K.; Contreras, M.; Magris, M.; Hidalgo, G.; Fierer, N.; Knight, R. Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns. Proc. Natl. Acad. Sci. USA 2010, 107, 11971–11975. [CrossRef] [PubMed]
20. Marlid, K.; Weimin, Y.; Lebwohl, B.; Green, P.H.R.; Blaser, M.J.; Card, T.; Ludvigsson, J.F. Antibiotic exposure and the development of coeliac disease: A nationwide case-control study. BMC Gastroenterol. 2013, 13. [CrossRef]
9335

Nutrients 2015, 7, 9325–9336 21. Cox, L.M.; Yamanishi, S.; Sohn, J.; Alekseyenko, A.V.; Leung, J.M.; Cho, I.; Kim, S.G.; Li, H.; Gao, Z.;
Mahana, D.; et al. Altering the intestinal microbiota during a critical developmental window has lasting metabolic consequences. Cell 2014, 158, 705–721. [CrossRef] [PubMed] 22. Fairchild, A.J.; MacKinnon, D.P. A general model for testing mediation and moderation effects. Prev. Sci. 2009, 10, 87–99. [CrossRef] [PubMed] 23. Blood, E.A.; Cheng, D.M. The use of mixed models for the analysis of mediated data with time-dependent predictors. J. Environ. Public Health 2011, 2011, 435078:1–435078:12. [CrossRef] [PubMed] 24. Baron, R.M.; Kenny, D.A. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J. Pers. Soc. Psychol. 1986, 51, 1173–1182. [PubMed] 25. Kenny, D.A.; Korchmaros, J.D.; Bolger, N. Lower level mediation in multilevel models. Psychol. Methods 2003, 8, 115–128. [CrossRef] [PubMed] 26. US Department of Health and Human Services, National Institutes of Health National Institute of Allergy and Infectious Diseases. Progress in Autoimmune Diseases Research; 2005. Available online: http://www.niaid.nih.gov/topics/autoimmune/Documents/adccfinal.pdf (accessed on 16 June 2015).
© 2015 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
9336