microbiota of the alimentary tract of children - … · microbiota of the alimentary tract of ......

Microbiota of the alimentary tract of children

-implications for allergy and inflammatory bowel disease

FeiSjöberg

孙红飞

DepartmentofInfectiousMedicine,ClinicalBacteriologyInstituteofBiomedicine,

SahlgrenskaAcademyatUniversityofGothenburg

Gothenburg2014

Coverillustration:ChinesecalligraphybymymotherYunWang

Microbiotaofthealimentarytractofchildren©FeiSjöberg2014Fei.sjoberg@microbio.gu.seISBN978-91-628-9090-2PrintedinGothenburg,Sweden2014Kompendiet.AidlaTradingAB

To my family

Microbiota of the alimentary tract of children

-implications for allergy and inflammatory bowel disease

FeiSjöberg

DepartmentofInfectiousMedicine,ClinicalBacteriology,InstituteofBiomedicine,

SahlgrenskaAcademyatUniversityofGothenburgGöteborg,Sweden

ABSTRACT

Allergy, which is the most common chronic disease in Swedishchildren and adolescents, is associated with a high standard oflivingandWestern lifestyle.According to thehygienehypothesis,allergyisduetoinadequatestimulationoftheimmunesystembymicrobes during early childhood, leading to failedmaturation ofthe immune system. The incidences of inflammatory boweldiseases, i.e., ulcerative colitis and Crohn’s disease, have alsoincreased dramatically in Western countries over the last fewdecades,andcurrently,thesediseasesareoftendiagnosedalreadyin childhood. Epidemiological evidence suggests links betweenalterations to the intestinalmicrobiota and these diseases. Thus,studies of the composition of the bacterial microbiota in infantsandyoungchildrenareof relevance for thepathogenesisofbothallergicdiseasesandinflammatoryboweldiseases.

In this thesis, quantitative culture and DNA‐based methods arecompared for their abilities to characterize the gutmicrobiota ininfants. Terminal‐RestrictionFragmentLengthPolymorphism (T‐RFLP) is based on differences in the 16S rRNA gene sequencesbetween bacteria, as revealed by differences in fragment sizesafterrestrictionenzymedigestion.Adatabasewasconstructedtoidentify bacteria based on their fragment sizes. Multi‐parallelsequencingof the16SrRNAgenebypyrosequencingandT‐RFLPwere compared for sensitivitywith quantitative culture of infantfecal samples. Bacterial genera thatwere present at >106 colony

forming units/g feces, as determined by culture, were generallyreadily detected by DNA‐based methods, with the exception ofbifidobacteria, which generated only one sequence read per 108viablebacteria.Clinicallyandimmunologicallyrelevantfacultativebacteria, e.g., staphylococci,wereoftenmissedby theDNA‐basedmethodsduetohavinglowcountsinthefecalsamples.Thestudiespresented in this thesis indicate that cultivation and molecular‐basedassaysarecomplementary ingeneratinganoverallpictureofthecomplexgutmicrobiota.

In theALLERGYFLORAcohortstudy,T‐RFLPwasusedtoanalyzethesalivarymicrobiotain4‐month‐oldinfantswhoseparentshadthehabitof“cleaning”theirpacifierbysuckingonit,andofcontrolchildren whose parents did not have this habit. Sharing of thepacifier between parent and infant was associated with reducedrisk of the child developing an allergy and altered salivarymicrobiota in the child. We hypothesize that the oral bacteriatransmittedfromtheparentsstimulatethechild'simmunesysteminsuchawaythatallergydevelopmentisavoided.

Samplesof theduodenal fluidsof childrenwithnewlydiagnosedanduntreatedinflammatoryboweldiseases(ulcerativecolitisandCrohn'sdisease)andcontrols (having functionalboweldisorderswithoutsignsofintestinalinflammation)wereanalyzedbycultureand pyrosequencing. The microbiota of children with ulcerativecolitisdisplayed lowerbacterialdiversity thanthatof thecontrolchildren, and certain bacterial groupswere less abundant in theformergroup.

Taken together, the studies presented in this thesis suggest thatthe compositions of the commensalmicrobiota in the oral cavityandsmallintestineaffecttheriskofdevelopingimmunoregulatorydiseases,suchasallergiesandinflammatoryboweldiseases.

Keywords: microbiota, alimentary tract, children, allergy,duodenum, inflammatory bowel disease, culture, T‐RFLP,pyrosequencing

ISBN:978‐91‐628‐9090‐2

SAMMANFATTNING PÅ SVENSKA

Allergi är den vanligaste kroniska sjukdomen hos barn ochungdomar i Sverige och är förknippadmedhög boendestandard,västerländsklivsstilochgodhygien.Enligthygienhypotesenberorallergipåatt immunsystemet inteutsätts föradekvatstimuleringav mikrober under uppväxten och därför inte mognar på ettkorrekt sätt. Även inflammatorisk tarmsjukdom är vanligare ivästländer än i fattiga länder med dålig hygien. Sjukdomen harökatkraftigtochdebuteraralltoftareundertidigbarndom,vilketkan tyda på att bristandemikrobiell stimulering är en riskfaktoräven för denna sjukdom. Västerländska spädbarn koloniserassenaremedvanligatarmbakterieränbarnifattigaländerochvårhypotes är att barnets bakterieflora kan tänkas påverka båderiskenförallergiochinflammatorisktarmsjukdomsenareilivet.

DNA‐baserade metoder såsom pyrosekvensering och Terminal‐Restriction Fragment Length Polymorphism (T‐RFLP) användsofta för att karakterisera komplexa bakteriella ekosystem menstudier av metodernas känslighet saknas. I avhandlingen harodling‐ och DNA‐baserade metoder jämförts med avseende påförmågan att detektera and identifiera olika grupper avtarmbakterier. En databas utvecklades för att kunna identifierabakterier utifrån fragmentens storlek i T‐RFLP‐analys. Resultatetvisade att bakterier med odlingstal högre än 106/ g faces, iallmänhet kunde detekterasmedDNA‐baserademetoder,medanfakultativa bakterier som fanns i lågt antal ofta missades. Bådeodlings‐ och DNA‐baserade analyser har sina begränsningar ochbådakanbehövasförattfåenhelhetsbildavdenkomplexafloran.

I Florastudien, en prospektiv kohortstudie, användes T‐RFLP föratt karaktärisera munflora hos spädbarn vars föräldrar sög påderasnappföratt”rengöra”den,ochfrånspädbarnvarsföräldrarintehadedennavana.Denmikrobiellasammansättningenisalivenskildesigmellanspädbarnenidebådagruppernavid4månadersålderochbarnvarsföräldrarsögpåderasnapphadelägreriskattsenare utveckla allergi. Vår hypotes är att bakterier från

föräldrarnasmunfloraöverförs tillbarnetochattdessabakterierstimulerarbarnetsimmunsystemsåattallergiutvecklingundviks.

Vi har också kartlagd tunntarmens bakteriefloramed odling ochpyrosekvensering hos barn med nydebuterad och obehandladinflammatorisk tarmsjukdom. Resultatet visade atttunntarmsfloran hos patienter med den inflammatoriskatarmsjukdomen ulcerös kolit uppvisade en lägre bakteriellmångfald jämfört med floran hos kontrollbarn, alltså barn medfunktionellatarmbesvärochnormaltarmslemhinna.

Vårastudiertyderpåattsammansättningenavbakteriefloraimunoch tunntarm kan påverka risken att utveckla immunreglerings‐sjukdomarsomallergiochinflammatorisktarmsjukdom.

FeiSjöberg

1

LIST OF PAPERS

Thisthesisisbasedonthefollowingpapers,referredtointhetextbytheirRomannumerals.

I. HesselmarB.,SjöbergF.,SaalmanR.,ÅbergN.,AdlerberthI.,WoldAE.PacifierCleaningPracticesandRiskofAllergyDevelopment.Pediatrics2013;131:1‐9

II. SjöbergF.,NowrouzianF.,RangelI.,HannounC.,MooreE.,AdlerberthI.,WoldAE.Comparisonbetweenterminal‐restrictionfragmentlengthpolymorphism(T‐RFLP)andquantitativecultureforanalysisofinfants'gutmicrobiota.JMicrobiolMethods2013;94:37‐46

III. SjöbergF.,Barkman C., Östman S.,Nookaew I.,AdlerberthI.,SaalmanR.,WoldAE.Alteredcompositionofduodenalmicrobiotaofchildrenwithnewlydiagnosedinflammatoryboweldisease.Inmanuscript.

IV. SjöbergF.,Nookaew I.,AdlerberthI.,WoldAE.Comparativeanalysisofinfants’gutmicrobiotabynextgenerationsequencingandquantitativeculture.Inmanuscript.

PapersIandIIarereprintedwithpermissionfromtheAmericanAcademyofPediatricsandElsevierLimited,respectively.

Microbiotaofthealimentarytractofchildren

2

TABLE OF CONTENTS

ABBREVIATIONS..........................................................................................................4

INTRODUCTION............................................................................................................6

Bacterialclassification‐taxonomy...............................................................8

Microbiotacompositioninthealimentarytract................................11

Oralcavity.....................................................................................................11

Stomach..........................................................................................................11

Smallintestine.............................................................................................12

Largeintestine.............................................................................................12

Microbialcolonizationofnewborns..................................................15

Methodsforstudyingthemicrobiota.....................................................20

Culture............................................................................................................21

DNA‐basedapproaches...........................................................................26

CultureversusDNA‐basedmethod....................................................39

Influenceonthehostofthemicrobiota.................................................41

Infection.........................................................................................................41

Microbiotainteractionswiththeimmunesystem......................41

Gutmicrobiotaandhostmetabolism................................................45

MicrobiotaandAllergy..................................................................................46

Allergy–“Themodernplague”............................................................46

Thehygienehypothesis...........................................................................49

Neonatal microbial exposure and the risk of allergydevelopment................................................................................................50

Oralfloraandallergydevelopment...................................................51

Microbiotaandinflammatoryboweldiseases....................................52

AIM.............................................................................................................................55

MATERIALANDMETHODS......................................................................................56

Generaloverviewofthefourpapers.................................................56

FeiSjöberg

3

ChildrenandSamples.....................................................................................58

TheAllergyFlorabirthcohort(PapersI,IIandIV)......................58

Inflammatoryboweldiseasestudy(PaperIII)..............................60

Methods................................................................................................................61

Quantitativeculture(PapersII,III,andIV).....................................61

DNA‐basedmethods(PapersI–IV).....................................................63

Statisticalanalysis......................................................................................66

RESULTSANDCOMMENTS........................................................................................67

Pacifiercleaningpracticesandallergydevelopmentininfants...67

DevelopmentofadatabasefortheidentificationofT‐RFs............74

Infantmicrobiotadevelopment.................................................................76

Methodologicconsiderations‐Advantagesanddrawbacks...........82

Detectionlimitsofthethreemethods...............................................82

Comparison of the performances of pyrosequencing foranalyzingsmallandlargebowelmicrobiota..................................89

DuodenalmicrobiotaandIBD.....................................................................92

DISCUSSION................................................................................................................93

ACKNOWLEDGEMENTS..........................................................................................106

REFERENCES...........................................................................................................108


4

ABBREVIATIONS

IBD

CFU

InflammatoryBowelDisease

ColonyFormingUnit

OPLS

O2PLS‐DA

OTU

OrthogonalPartialLeastSquares

Orthogonal2PartialLeastSquares‐DiscriminantAnalysis

OperationalTaxonomicUnit

T‐RFs Terminal‐RestrictionFragments

T‐RFLP

Terminal‐RestrictionFragmentLengthPolymorphism

FeiSjöberg

5


6

INTRODUCTION

Thealimentarytractisanapproximately8‐mlongtubularpassagethat extends from the mouth to the anus, through which foodpassesandbecomesdigested.Thealimentarytractiscolonizedbydiverse microbial communities, which are collectively known asthemicrobiota.

Themicrobiotaofthehumangastrointestinaltractiscomposedofhundredsof speciesandup to1014bacterial cells.Thismicrobialsociety contains potential pathogens and the mucosal immunesystem faces a challenging task in protecting the host frominvasive pathogens, while, at the same time, tolerating harmlessdietary antigens and not reacting excessively to the bulk ofcommensalbacteriaandtheirinflammatogenicproducts.

There isgrowing interest in thephysiologic interactionsbetweenthemicrobiotaandthehost.Ithasbeensuggestedthatthehumanmicrobiota contributes to health and disease through providingnutrients, stimulating the immune system, and exertingcolonization resistance against pathogens. Most of the existingbodyofknowledgeregardingmicrobiotacompositioncomesfromculture‐based studies, although massive nucleic acid sequencingstudiesofthemicrobiotaatvariousbodilysitesareunderway,e.g.,intheHumanMicrobiomeProject.Fewstudieshavecomparedtheperformanceofculture‐basedandsequence‐basedmethodologiesforcharacterizationofthecommensalmicrobiota.

FeiSjöberg

7

The figure exemplifies some of the major bacterial groupscolonizingdifferentpartsofthealimentarytract.

FemalefigurebyAgnesWold


8

Bacterial classification-taxonomy Livingorganismsareclassifiedintoacommontaxonomicsystem,the basic level of which is the species. Each newly discoveredspecies must be assigned to a genus in a binary nomenclatureestablishedbyCarlvonLinné.Agenusisamemberofsuccessivelyhigher ranks: subfamily, family, suborder, order, subclass, class,phylum(ordivision),anddomain(orempire),andaspecificsuffixis added to a taxon between genus and class (Table 1). A taxon(plural, taxa) isageneraldefinition inmicrobiology;a taxonomiccategory or group, such as a phylum, order, family, genus, orspecies.

Table1. Bacterialtaxonomy

Taxonomic rank LPSN# Suffix Example

Subspecies 416 No

P. freudenreichii subsp. freudenreichii P. freudenreichii subsp. shermanii

Species 10 599 No Propionibacterium freudenreichii

Genus 2001 No Propionibacterium

Family 271 ‐aceae Propionibacteriaceae

Suborder 21 ‐ineae Propionibacterineae

Order 126 ‐ales Actinomycetales

Subclass 6 ‐ida* Actinobacteridae

Class 97 ‐ia* Actinobacteria

Phylum 30 No Actinobacteria #Total validly published since the Approved Lists of Bacterial Names up to Auguest. 2013 [1]. *Proposed suffix

Bacterial nomenclature (a system of names) is covered by therulesof theBacteriologicalCode[2]. Severalclassification systems exist, including those proposed by Pace [3], Ludwig et al. [4], Hugenholtz [5], and Bergey'sManualofSystematicBacteriology[6],the latterbeing themostwidely accepted authority on this topicamongmicrobiologists.

FeiSjöberg

9

AccordingtoBergey’sManual,thebacteriadomainisdividedinto30 phyla, and as of November 2013, 15,974 taxa have beenidentified; these are listed in the List of ProkaryoticNameswithStandinginNomenclature(LPSN)[1].LPSNisadatabasethatliststhenamesofprokaryotes (Bacteria andArchaea) thathavebeenvalidly published in the International Journal of Systematic andEvolutionary Microbiology, under the Rules of the InternationalCodeofNomenclatureofBacteria.

The bacterial species Aspeciesisdefinedas“amonophyleticandgenomicallycoherentclusterofindividualorganismsthatshowahighdegreeofoverallsimilarityinmanyindependentcharacteristics,andisdiagnosableby a discriminative phenotypic property” [7]. However, forbacteria, this definition is ambiguous, which is known as the‘speciesproblem’[8].Theoriginaltaxonomicschemeswerebasedonisolationbycultureandclassificationaccordingtomacroscopicand microscopic appearances, metabolic behavior (mostly theability to ferment various carbohydrates), and the presence ofsurfaceantigensthatwerereactivetospecificantibodiesproducedfor this purpose (serology). Since DNA methods started to beappliedforspeciesdefinition,separatespecieshavebeenfoundtobe genetically very similar. Thus, Escherichia coli and the fourspecies of the genus Shigella, whichwere previously categorizedbased onmorphology, serotyping, and biochemical tests [9], arenow assigned to a single species based on DNA relatedness.However, they are still considered to be separate species, asdistinguishing E. coli from the highly pathogenic Shigella isimportantintheclinicalsetting.

Today,anovelspeciesistypicallyidentifiedthroughacombinationoftraditionaltests,mainlybiochemicalanalysesandlipidprofiling,aswell ason thebasisofDNArelatedness.Thenamingofanewspecies requires that its similarity to anyother knownspecies is<98.5% for the 16S rRNA gene and <70% for DNA‐DNAhybridizationof theentiregenomes[10][11].Recently,basedonthe available bacterial whole‐genome sequences, Kim et al [12]showedthattheaveragenucleotideidentity(ANI)thresholdrangeof 95%–96% has taxonomic potential for species differentiation.


10

ThisANIrangecorrespondsto16SrRNAgenesequencesimilarityof98.65%.ANIcouldserveasasubstituteforthelabor‐intensiveDNA–DNAhybridizationtechnique.

SubspeciesandstrainsSome species are subdivided into subspecies (abbreviated as‘subsp.’).Therearenoclearrulesforthis; itdependsonpracticalinterest in division below the species level. Propionibacteriumfreudenreichii is a Gram‐positive bacterium that plays animportant role in the fermentation of Swiss Emmental cheese.Table2showstheexampleofthetwosubspeciesP.freudenreichiisubsp.freudenreichiiandP.freudenreichiisubsp.Shermanii,whichare differentiated on the basis of nitrate reduction and lactosefermentation[13],factorsofimportanceincheeseproduction.

Table2. Propionibacteriumfreudenreichiisubspeciesdifferentiation.

Strain refers to a colony of genetically uniform microorganismsthatderive froma commonancestor, e.g., a clonegrowingout incultureora cloneofE.coli that inhabits the large intestineofanindividual for a period of weeks or months. Strains within aspecies may be separated by serotyping, typing of enzymeelectrophoretic mobility (due to point mutations), orcharacterization of the outer membrane protein profiles (forGram‐negative bacteria). A clinically important application ofstrain/clone typing is the tracing of the source of a diseaseoutbreak.

P. freudenreichii subsp. NO3

reductionLactose

fermentation

freudenreichii + ‐

shermanii ‐ +

FeiSjöberg

11

Microbiota composition in the alimentary tract

Oral cavity Theoralcavityisamajorgatewayforbacteriatoenterthehumanbodyandprovidesanumberofbacterialcolonizationhabitats,e.g.,the surfaces of the teeth, tongue, and tonsils, and the gingivalcrevices. Biofilms, which are composed of multilayeredmicrocolonies,mayformonseveraloftheseoralsurfaces[14].Thesaliva is rich in bacteria, usually containing around 109 colony‐formingunits(CFU)/mL.

Theoralmicrobiotaisestimatedtocontainmorethan600species,distributed in 13 phyla. The sixmajor phyla contain 96% of thetaxa, i.e., Firmicutes (37%), Bacteroidetes (17%), Proteobacteria(17%),Actinobacteria(12%),Spirochaetes(8%),andFusobacteria(5%)[15].Notably,themajority(65%)ofthebacterialtaxaintheoralcavityhavebeensuccessfullycultured.

At the genus level, Streptococcus, Actinomyces, Veillonella,Prevotella, Haemophilis, Fusobacterium, and Porphyromonas arefrequentlydetectedintheoralcavity[16‐18].

Stomach The stomach has a sparse flora due to the low pH of the lumen.Using culture, Lactobacillus, Streptococcus, Clostridia, Veillonella,coliforms, and yeasts are found at population levels that varyaccording to thediet and geographic locationof theperson [19].The composition of the gastric microbiota varies considerablybetweenindividuals,asrevealedbybothculture‐basedandDNA‐based methods. Indeed, the microbes isolated from gastriccontentsmay be transients, either having passed down from thenasopharynxororalcavity,orhavingbeingintroducediningestedfood. Helicobacter pylori is a true long‐term colonizer of thestomach. It ismorereadilydetectedbyDNA‐basedmethods thanbyculturing[19,20].


12

Small intestine The small intestine is divided into the duodenum, jejunum, andileum.

Duodenum The stomach acid kills most of the ingested bacteria, and theexcretion of digestive enzymes and bile acids into the duodenallumenalsoexertsastrongselectivepressureonthenumbersandtypes of transiting bacteria. Bacterial counts are thereforegenerallylowinsamplestakenfromtheduodenum,typically102–103 CFU/mL of contents. Only a few studies have examined thecomposition of the duodenal microbiota. Two duodenal biopsystudies have shown that the genera Streptococcus andNeisseria(bothbelongingtotheFirmicutesphylum)predominate,althoughVeillonella,Gemella,ClostridiumclusterXIVa,andHaemophilusarealso detected. In contrast to the large intestinal microbiota,membersoftheBacteroidetesandActinobacteriaphylaarescarcein the duodenalmicrobiota.When present, they are representedbythegeneraPrevotella,Bacteroides,Actinomyces,andRothia[21][22].

Jejunum and ileum Thenumbers of bacteria increase along the small intestine, from102‐4bacteria/mLintheproximaljejunumto106–107bacteria/mLin the terminal ileum [23]. The milieu is rather rich in oxygen.Culture‐based studies have indicated that Lactobacillus,Streptococcus,Clostridia,Veillonella,Bacteroides,andcoliformsarethe dominant bacterial groups, and that yeasts, mainly Candida,arealsopresent[19].

Large intestine Inthecolon,thebacterialdensityisveryhigh,oftenexceeding1011CFU/g colonic content. As there is approximately 1 kg of coloniccontents, an adult individual harbors 1014 bacteria in the colon[24]. Most studies of the large intestinal microbiota have beenbased on analyses of fecal samples because of their readyavailability. However, differences in community compositionbetweenstoolandmucosalsampleshavebeenreported[25].The

FeiSjöberg

13

colonic microbiota of an individual is rather stable over time,although inter‐individual variations can be substantial [26].Geographic region, diet, and age play significant roles in shapingthecompositionofthecolonicflora[27][28].

Strictlyanaerobicbacteriapredominate in the large intestinedueto the low level of oxygen in the lumen. The number of speciesharbored by a given individual is a subject for debate. Classicalculture‐based studies performed in the 1970’s estimated thatapproximately400speciescouldbepresentinanindividual[29].In one PCR‐based Next generation sequencing study, the V4region of the 16S rRNA gene was sequenced after PCRamplification. Numerous reads (1.8±0.6 million/sample), wereobtained from 531 individuals from the Amazonas of Venezuela,rural Malawi, and US metropolitan areas. The mean number ofoperational taxonomic units (OTUs)/sample for each area were1600, 1400, and 1200, respectively [27]. The shot‐gunmetagenomics sequence study of the Human Intestinal Tract(MetaHIT) project, which involved the examination of 124Europeansubjects,showedthateach individualharboredat least160 bacterial species [30]. The MetaHIT study concluded that“giventhe largeoverlapbetweenmicrobialsequences in thisandprevious studies, we suggest that the number of abundantintestinal bacterial species may be not much higher than thatobservedinourcohort”,i.e.,afewhundred,whichisclearlymuchlower than the number revealed by the study using16S rRNAsequencing. Using a mock community, shotgun metagenomicssequenceing showed more accurate species determination thanPCR‐based16SrRNAgenenextgenetationsequencing[31].

DNA‐based studies show that up to 80% of the microbialsequences identified in fecal samples are novel and representuncultivated bacteria [25, 32]. At the phylum level, the fecalmicrobiota is dominated by Bacteroidetes and Firmicutes, whichcomprise 60%–80% of the fecal bacteria. Actinobacteria,Proteobacteria,Verrucomicrobia,andFusobacteriaarealsofound,albeit in lower numbers (Table 3) [33, 34]. Table 3 shows theproportionsofbacteriainthecolonicmicrobiota,asdemonstratedbybothDNA‐basedtechnique[33].


14

Table3. Colonicmicrobiotaincompostion

Taxon Share of the

microbiota (%) Genus/Species

Firmicutes 40–65

Clostridial cluster

IV 25 Clostridium leptum

Faecalibacterium prausnitzii

Ruminococcus

IX 7 Megasphaera

Veillonella

Selenomonas

Megamonas

XIVa 25 Clostridium

Anaerostipes

Ruminococcus

Roseburia

Eubacterium

Coprococcus

XI Clostridium bartlettii

XV Anaerofustis stercorihominis

Lactobacillales Lactobacillus

Bacteroidetes 25

Bacteroides

Actinobacteria 3–5

Bifidobacterium

Collinsella

Proteobacteria 0‐3

E. coli

Verrucomicrobia 1‐2

Akkermansia mucinophila

Victivallis vadensis

Archaea Methanobrevibacter smithii

Methanosphaera stadtmanae

Data adapted from Duncan et al., Appl. Microbiol, 2007 [33]

FeiSjöberg

15

Metagenomicstudieshaveshownthat99%ofthephylogeneticallyassigned genes of the large intestinalmicrobiota are of bacterialorigin, although Archaea (e.g., Methanobrevibacter), Eukaryotes(e.g., yeasts), and viruses (e.g., bacteriophages) are also present[30][35].

Microbial colonization of newborns Microbial colonizationof theneonatebegins atbirth as the fetusleaves the normally sterile environment of the womb. Theestablishment of the microbiota during infancy proceeds in asequentialmannerover several yearsuntil hundredsof differentbacterialspeciesareestablishedinboththeoralcavityandthegut.Thesourcesof thecolonizingbacteriaare:parents, siblings,pets,foods, and the air. Given the different exposure conditions, thecomposition of the microbiota may vary with delivery mode,feedingregime,hygienemeasures,and livingstandard,aswellastheextentofsocialcontacts.

Oral cavity colonization Relatively little is known regarding the establishment of the oralmicrobiota in infants. There is a paucity of longitudinal studies,and most of the studies that have been conducted have beenculture‐based. More recent studies employing DNA‐basedmethodologies have focused on oral pathogens, such ascolonization with Streptococcusmutans in relation to the risk ofdevelopingdentalcaries.

Teethappearatafewmonthsofageininfants.Beforeeruptionofthe teeth, various streptococci act as the first colonizers, e.g., S.salivarius, S.mitis, and S. oralis [36]. For example, in one study,streptococci were detected 8 hours after birth, whileStaphylococcusandNeisseriawereisolatedlateronduringthefirstday, and anaerobic species of Veillonella and Bifidobacteriumappeared from the second day [37]. All these bacteria were

consistentlypresentthroughoutthefirstyearof life,althoughthebacterialcountsofstreptococcidecreasedovertime[38].Anotherstudy showed that the most frequently isolated early anaerobiccolonizers of the oral cavity were Prevotella melaninogeniea,


16

Fusobacterium nucleatum, and non‐pigmented Prevotella spp.[39].

After emergence of the teeth, the bacterial diversity in the oralcavity increases as the numbers of retention sites and potentialniches increase. It is unclear at what age the oral microbiotabecomes adult‐like. Bacteroides melaninogenicus was found in<40%of five‐year‐old children, but in essentially all 13–16‐year‐oldadolescents[40].

In the oral cavity, bacteriamay adhere to each other, forming abiofilmonavailablesurfaces,withinwhichtheyliveinasymbioticpartnership [36]. Fastidious anaerobic bacteria growth even inedentulousmouthscanbeexplainedbytheformationofbiofilms.Fusobacterium nucleaturn may have a central role in thematurationoforalbiofilmcommunities[41]

Gut colonization The neonatal gut is initially an aerobic environment, andfacultative anaerobic bacteria are usually the first colonizers.Initially,manyofthebacteriaaretransientcolonizers[42].Thegutcolonizationfluctuatesmarkedlyfromhourtohourintheneonatalperiod [43]. In the early period of life, the gut microbialcompositionvariessignificantlyfrombabytobaby[42,44,45].

As shown in twin and kinship studies, common environmentalexposuresplaygreaterrolesinshapinggutmicrobialecologythandoes genetic relatedness [27]. Vaginally delivered infants arecolonized first by the cervical and fecal flora of themother [46][47].BabiesdeliveredbyCesareansectionareexposedinitiallytobacteria from the hospital environment and healthcare workers[47, 48]. Caesarean section‐delivered infants less often harbortypicalbacteriaoffecalorigin,suchasEscherichiacoli,Bacteroides,andBifidobacterium species,while other bacteria encountered intheenvironmentaremorefrequent,e.g.,KlebsiellaandClostridiumspecies [44, 45, 49]. One study has reported a lower level ofmicrobial diversity in the gut microbiota of Cesarean section‐delivered babies, as compared with that in babies deliveredvaginally[42].

FeiSjöberg

17

Milkformula‐fedinfantshaveamorediversemicrobialcommunitythan breast‐fed infants [44]. While this might reflect greaterexposure tobacteria, it is equally likely that breastmilk exerts astrongselectivepressureon thegutbacteriapopulationand thatonly a few bacterial groups are able to withstand this selectionpressure.

Figure 1 shows the gut microbial colonization patterns of 300childrenfromthreeEuropeanbirthcohorts,followedfrom3daysof age to 1 year of age using culture‐based analyses of fecalsamples [49]. The left panel of Figure 1 shows the cumulativecolonization frequency, i.e., theproportionof infantswhoharbororhaveharboreda certainbacterial genus,while the rightpanelshows the population levels of the same genera in infants whoharbor the genus in question. It is clear that coagulase‐negativestaphylococci (CoNS) and enterococci are the first colonizersamong the facultative bacteria, and that the frequencies ofcolonization with E. coli and S. aureus are approximately equal.Colonization by Klebsiella and other members of theEnterobacteriaceae family is also relatively common in infancy(panel A). Regarding population levels, it is evident that thepopulation levelsofE.coliareconsistentlyretainedoverthefirstyear of life (panel B). Other members of the EnterobacteriaceaefamilyandenterococciarepresentinlowerlevelthanE.coliatoneyear of age. In sharp contrast, the population levels ofstaphylococci,bothcoagulase‐negativestaphylococciandS.aureus,decreasedramaticallyoverthefirstyearof life,andin1‐year‐oldchildren,theirpopulationlevelsaverage104–105CFU/gfeces.Theearly microbiota contains few species and the competition fornutrients and space is limited, with the consequence that mostbacteriagrowtoquitehighpopulation levels. Inparallelwiththeincreaseinthecomplexityofthemicrobiotaduringthefirstyearoflife, the population levels of bacteria that cannot survive thecompetition decrease. Staphylococci, which are typical skincolonizers, are examples of bacteria that are not well suited topersistinginamicrobiotathathasreachedfullcomplexity.

PanelsCandDinFigure1depictthecolonizationpatternsofsomegroups of strictly anaerobic bacteria. Bifidobacteria are the first


18

colonizers, followed by a variety of Clostridium spp. [49].Bacteroides,which is a Gram‐negative commensal gut bacterium,colonizesonlyaminorityoftheseinfantsduringthefirstmonths,although itscolonization frequency increasesovertheentire firstyear of life. Lactobacilli are not a major colonizer of the infantmicrobiota,althoughtheybecomemoreprevalenttowardstheendofthefirstyearoflife(panelC).

Intestinalcolonizationpatternduringthefirstyearoflife.TheresultsareFigure 1.presentedastheproportionofinfantsevercolonizedbyeachtimepointandthemeanlog10countforcolonizedinfantsonlyateachtimepointforfacultativeanaerobic(A and B) and strict anaerobic bacteria (C and D). CoNS, Coagulase‐negative staphylococci.ReproducedwithpermissionfromAdlerberthetal.Datashownarebasedontheresultsofcultivationoffecalsamplsfrom300Europeaninfants[49]

Regarding population levels, infants who are colonized bybifidobacteria or Bacteroides harbor consistently high levels ofthese bacteria, i.e., >109 CFU/g feces (panel D). The lactobacilluspopulation levels in colonized infants are generally much lower

FeiSjöberg

19

(panel D). Clostridia are detected and enumerated by alcohol‐treatment of the samples, which kills living bacteria but leavesspores intact. Thus, the population levels shown in panel Drepresent clostridial spore counts in the feces of culture‐positiveinfants.

Conversiontoanadult‐typemicrobiotaThe complexity of the microbiota increases as more and moreanaerobic speciesbecomeestablished in the infant gut [50] [27].Fromthetimeofweaning,diversificationofthedietmaypromotereplacement of pioneer gut colonizers with memebers of themicrobiotathatwillpersistintoadulthood[51]. At1yearofage,although still distinct, the microbiota starts to converge towardthatofanadult[42].However,itisnotuntil2–3yearsofagethattheyoungchild’sgutmicrobiotaresemblestheadultflora[27,52].

With decreasing oxygen tension, the numbers of facultativebacteria decline. Thus, the ratio of facultatives to anaerobeschanges from roughly 1:1 in newborn infants to 1:500 in adults[53].

Ameta‐analysis that combinedNGS data on infant and adult gutmicrobiota from three separate studies revealedanagegradient,with a transition from Enterococcaceae, Enterobacteraceae,Streptococcaceae, Lactobacillaceae, Clostridiaceae, andBifidobacteraceae at an early age to communities in adults thatwere enriched for Lachnospiraceae, Ruminococcaceae,Bacteroidaceae,Prevotellaceaeandothers[54].


20

Methods for studying the microbiota Muchofourknowledgeregardingthehumanmicrobiotahasbeenderived from culture‐based studies. Cultured bacteria aresubculturedforpurityandmaybeidentifiedtothefamily,genus,speciesorevenstrainlevel.Variousmethodsareavailablefortheidentification of bacterial isolates. The culture‐based andmolecular techniques that are commonly used for microbialidentification are listed inTable4. Allof thesesmethods requirepurecultureisolatesasstartingpoint.

Table4. Culture‐dependenttaxonomicresolutionlevelsbasedoncurrentbacterialidentificationtechniques.

Method Taxonomic resolution level

Family Genus Species Strain

Non‐DNA‐based

Biotyping X X X (X)

Fatty acids (VFA/FAME)a X X X

MALDI‐TOFb X X X

DNA‐based

Sequencing non‐dependent

PCR (specific‐primer) X X X

DNA‐DNA hybridization X

PFGEc X

MLVAd X

Sequencing‐dependent

16S rRNA sequencing X X (X)

Whole‐genome sequencing X X X X

Multilocus sequence typing X

(X) denotes occasionally achievable aVolatile Fatty Acid/Fatty Acid Methyl Esters bMatrix‐Assisted Laser Desorption‐Ionization‐Time‐Of‐Flight c Pulsed‐Field Gel Electrophoresis dMultiple‐Locus Variable‐number Tandem Repeat

FeiSjöberg

21

Inrecentyears,moleculartechniquesthatpermittheidentificationofpreviouslynon‐culturableanaerobeshavebeendeveloped,andthese are now widely used to describe microbial ecosystems(Table5).

Table5. Non‐culture‐dependentmethodsusedtostudybacterialcommunities

Method Taxonomic resolution level

Family Genus Species

Sequencing‐non‐dependent

PCR (specific‐primer) X X X

Microarray X X (X)

T‐RFLP, D/TGGE* X X (X)

Sequencing‐dependent

Cloning and sequencing X X (X)

NGS (PCR‐based)a X X (X)

NGS (DNA‐based) X X (X)

Single‐cell sequencing X X X

(X)denotes occasionally achievable. aNext generation sequencing

*Terminal Restriction Fragment Length Polymorphism

*Denaturing/Temperature Gradient Gel Electrophoresis

Culture Microbialcultureisessentialfortheisolationofapurecultureofabacterium. Pure cultures arerequired to assess antibioticsusceptibilitypatterns.Bacterialisolatesinpureculturesmayalsobesubtypedtothestrainlevel,whichallowsthespreadofclonestobemonitoredinepidemiologicstudies.Methods,whichincludeserotyping, PFGE (Pulsed‐Field Gel Electrophoresis), MLVA(Multiple‐locus variable‐number Tandem‐repeat), andMLST(Multilocus sequence typing), can be used to distinguishstrains within a particaulr species, although all these methodsrequirepurecultureisolatesasthestartingpoint[55].


22

Culture conditions Bacteria can be classified based on dependence on oxygen into:aerobic,anaerobic,andfacultativeanaerobic.Thus,amilieuthatisrichorpoorinoxygencanbeappliedasaselectiveforcetoenabletheselectionoffacultativeorstrictlyanaerobicbacteria.

Aerobicbacteriaderiveenergyfromoxidativeprocesses,whereasobligate anaerobic bacteria use fermentation or anaerobicrespirationastheirenergysource.Obligateanaerobicbacteriaarealso vulnerable to oxygen exposure, as they die in air. Obligateanaerobes show extensive differences with respect to oxygensensitivity. For example, Roseburia intestinalis survives for lessthan2minuteswhenexposedtoaironanagarsurface,whileotheranaerobic bacteria, here exemplified by Anaerostipes caccae,survivesquitewellinambientairforsometime(Figure2).

Figure2.AerotoleranceofsomeClostridium‐relatedFirmicutebacteriafromthehumangut.ReproducedwithpermissionfromJohnWileyandSons.EnvironmentalMicrobiology,2007[56]

Facultative anaerobic bacteria can survive under anaerobicconditions but they grow better in the presence of oxygen.Culturing in air can be applied as a selective pressure forfacultative bacteria that are present in relatively low numbersamongfarhighernumbersofstrictanaerobes.

Thenutritionalrequirementsvarybetweenbacteria.Non‐selectivemediameetthenutrientrequirementsofmanydifferentbacteria,anexamplebeingBrain‐HeartInfusionagar/broth.Selectivemediaareusedtoisolatespecificbacteriafromamixedpopulation.They

FeiSjöberg

23

permit growth only of the targeted microbe or a few types ofmicrobes.Forexample,MacConkeyagarcontainsbilesalts,whichpermitthegrowthofGram‐negativebacteriaofthetypesfoundinthe intestinal flora, i.e., the Enterobacteriaceae family, whileinhibiting the growth of most Gram‐positive bacteria. InScandinavia, Drigalski agar is used for the same purpose. It isslightlylessselectivethanMacConkey agar, permitting scant growth of Gram-positive enterococci. Staphylococci are able to grow in the presence of 10% NaCl, and media that are selective for staphylococci are based on this principle. As all other gut-colonizing bacteria die in this heavy salt environment, staphylococci that are present at very low concentrations, e.g., 103 CFU/g, can be isolated from a mixed population of 1011 fecal bacteria.

The number of bacteria is determined by quantitative culturing.Serial dilutions of a sample are plated onto solid medium. Thedilution that yields 10–100 colonies is selected. The colonies areenumerated and multiplied by the dilution factor, yielding thebacterialcountsascolony‐formingunits(CFUs).

Gram‐staining was developed by the Danish pathologist Hans‐Christian Gram [57]. First, crystal violet is added to bacteriasmeared on a glass slide. After fixationwith iodide solution, theslide is destained with ethanol or acetone and then counter‐stainedwith saffranin.Gram‐positivebacteria stainblue, as theirtightlymeshedandthickcellwallretainscomplexesofcytoplasmicproteins and crystal violet. The thin cell wall of Gram‐negativebacteriacannotretainthecrystalviolet‐proteincomplexes,whichleak out of the cell, so that it loses the color. Gram‐negativebacteriaarestainedredbythesaffraninappliedattheendofthestainingprocess.

Isolate based identification methods Atrainedmicrobiologistcan,at times, identifyabacterialspeciessolelybaseduponcolonymorphologyonselectiveornon‐selectivemedia and smell, e.g., Pseudomonas aeruginosa, E. coli, and S.aureus. Gram‐staining followed by observation under themicroscope adds information regarding the presence of cocci(round bacteria), bacilli (rod‐shaped bacteria), and endospores


24

(bacilli,clostridiae). Inaclinical laboratory,manyoftheclinicallyrelevantbacteriamayberelativelyeasilyidentifiedinthisway.

BiochemicalreactionsBiochemical tests are commonly used to identify bacteria. Thepresence or absence of certain bacterial enzymesmay be highlyinformative. Catalase converts hydrogen peroxide to water andoxygen, and all staphylococci have this enzyme. Coagulaseconverts fibrinogen to fibrin anda clot is formed.Staphylococcusaureus is coagulase‐positive, while many other staphylococcalspecies,suchasS.epidermidisandS.haemolyticus,arecoagulase‐negative and are grouped together as the coagulase‐negativestaphylococci.Coagulase‐positiveS.aureusisapathogen,inthatitcauses sepsis, osteomyelitis, endocarditis, pneumonia, and localskin abscesses. In addition, its production of toxins, expecially“superantigens”,maycausefoodpoisoningandshock.Incontrast,thecoagulase‐negativestaphylococciseldomcausedisease,exceptin immunocompromised individuals and newborn infants. Low‐gradeandlong‐terminfectionsofprostheticdevicesare,however,causedbycoagulase‐negativestaphylococci.

Biotyping, which entails enrichment of a pure culture of thebacteriatofermentdifferenttypesofcarbohydrates,was,priortothe advent of DNA‐based methods and matrix‐assisted laserdesorptionionization–timeofflight(MALDI‐TOF),themajorwayto achieve bacterial speciation. Commercial biotyping kits areavailable, such as API 20E for theEnterobacteriaceae family andRapid ID 32A for anaerobic bacteria, e.g., Bacteroides andClostridium.

FattyacidcharacterizationThere are two types of fatty acid analysis methods: Fatty AcidMethyl Esters (FAME); and Volatile Fatty Acids (VFA). FAMEanalyzes long (>12) fatty acids that are part of the bacterial cellwall, whereas VFA analyzes the short‐chain fatty acids (SCFA),which are produced by anaerobic bacteria through thefermentation of carbohydrates in the culture medium. SCFAinclude acetic acid, lactic acid, propionic acid, and butyric acid.Differentgroupsofbacteriaproducedifferentproportionsoffatty

FeiSjöberg

25

acids,which are separated anddetected by gas chromatography,withtheresultingpatternindicatingtheidentityofthebacterium[58]. For example, bifidobacteria mainly produce acetate, whilesomeclostridialspeciesarebutyrateproducers.

MALDI‐TOFSince 2005,matrix‐assisted laser desorption ionization – time offlight (MALDI‐TOF) has become the standard method for rapididentification of bacteria in the clinical laboratory [59]. Speciesidentification by MALDI‐TOF is more rapid, accurate, andinexpensive than other procedures based on molecular orbiochemicaltests.Acolonyofthemicrobeinquestionissmeareddirectlyontoametalsurface,whichisthenbombardedbyalaser,andthemacromoleculesspreadfromthecolony,i.e.,nucleicacids,proteins, and carbohydrates, are analyzed bymass spectroscopy.Thespectrumgeneratedbyaspecificbacteriumisanalyzedusingdedicated software, and identification is achieved by comparisonwithstoredprofiles.


26

DNA-based approaches During the last decades, many DNA‐based methods have beendevelopedtoclassifybacteria.Thesemethodologiesaregenerallyreferred to as “molecular” methods. They all depend on thedetectionofdifferencesinDNAsequencesbetweenbacteria.Theycan be basically divided into sequencing‐dependent andsequencing‐non‐dependentmethods.

Sequencing-non-dependent

PolymerasechainreactionPolymerasechainreaction(PCR)wasdeveloped in1983byKaryMullis [60] and is now a widely used technique in medical andresearchlaboratories.Species‐specificPCRcanbeusedtoidentifybacteria down to species level. The primers used comprisesynthetic small fragments of DNA that hybridize to the targetedDNA sequence of interest. During the PCR, the target DNAsequenceisamplifiedmanytimes,enablingitsdetectionbasedonsizeorbysequencing.

16SrRNAgeneIn the 1980s, CarlWoese described an rRNA‐based phylogenetictaxonomyfortheclassificationofbacteria[61].The16SrRNAgenecontainshighlyconservedregionsinwhichtheDNAsequencesareverysimilarbetweendifferentbacteria.Interspersedamongtheseconservedregionsarehypervaribleregions,knownasVariable(V)regions, the sequences of which differ among bacterial taxa.Conserved regions are suitable for universal PCR primerconstruction, while the variable regions can be used for taxonidentification.However, the conserved regionsarenotabsolutelyidenticalbetweenallbacteria.Therefore,primersarealwaysmoreor less selective, favoring certain groups and disfavoring others[62]. Another problem is that, depending on the size of thesequencedfragment, itmaynotbepossibletoidentifyabacterialtaxon, since there are too few sites within that region thatunequivocallydefinespecifictaxa.

Figure3showsa representationof theentire1500‐bp16SrRNAgene,withitsnineVregions,V1‐V9,whichrangeinlengthfrom50

FeiSjöberg

27

to100nucleotides[63].The figureshowsthechanceofcorrectlyclassifyingabacteriumtothephylum,class,order,familyorgenuslevel, based on the utilization of a 100‐nt that encompassesdifferent parts of the 16S rRNA gene. It is evident that themostpotentdiscriminationisachievedusingaprimerthatencompasseseithertheV2regionortheV4region.However,itisalsoclearthatclassification to thegenus level isnear impossibleusinga100‐ntsequence, as themaximal accuracy that can be achieved is 80%.Using such relatively short sequences, bacteria can be reliablyclassifiedonlytothephylumorclasslevel(Figure3).

Figure3.(A)ClassificationaccuracyratefortheBergeycorpuswithsequencesegmentsof100bases,moving25basesatime.Theaverageclassificationaccuracyrateatthegenuslevelwas70%overall100‐baseregions.(B)Averagebootstrapconfidence.ReproducedwiththepermissionoftheAmericanSocietyforMicrobiology.Wangetal.ApplEnvironMicrobiol,2007.

Figure4 shows the chanceof correctly assigninga genus, family,order, class or phylum to a 16S rRNA gene segment of differentlength.Asshown inthe figure,with400‐bpsequencesof the16SrRNA gene, there is an 88.7% chance of correctly identifying thegenusofthebacteriuminquestion,a94.6%chanceofassigningittothecorrectfamily,anda>99%chanceofcorrectlydeterminingitsclassorphylum.


28

Despitethisgeneralrelationship,somespeciescannotberesolvedtothespecieslevel,evenbysequencingthefull‐length16SDNA.Asthis is true for streptococcal species, highly pathogenicpneumococcicannotbedistinguished fromawiderangeofotherstreptococcalspecies,collectivelyreferredtoastheα‐streptococcior“viridans”streptococci.Thesestreptococciarenumerousintheoral cavity flora and upper small intestinal flora. TheEnterobacteriaceae family (E. coli, Klebsiella etc.) is alsonotoriouslydifficulttoseparatesolelybasedon16SrRNAanalysis,due to the close genetic relationships. Enterobacteriaceae,,particularlyE.coli,areimportantmembersofthegutflora.

Figure4.Overallclassificationaccuracybyquerysize.Numbersarepercentagesoftestscorrectlyclassified.FigureadoptedwiththepermissionfromtheAmericanSocietyforMicrobiologyWangelal.ApplEnvironMicrobiol,2007.

Another complication is that bacteriamayhave several copies of16SrRNAgenes.Although40%ofbacteriahaveoneortwocopies,uptosevencopiesarecommon[64].

Within a single bacterium, the different copies of the 16S rRNAgenesmaydiffer in theirDNAsequences,whichmayhamper thesequenceannotation[65].

Another factot that influences the detection limit of 16S rRNAgene‐basedanalysisistheefficacyofDNAextraction.DNAismorereadilyreleasedfromGram‐negativebacteria,suchasBacteroidesandEnterobacteriaceae, while someGram‐positive bacteria, such

FeiSjöberg

29

asEnterococcusandStaphylococcus,haveverythickandsturdycellwalls,soitmaybedifficulttoaccesstheDNAofthesebacteria.

A final concern is “universal” PCR primer mismatch [66]. Inparticaulr,Gram‐positivebacteriathathaveahighGCcontent(e.g.,Actinobacteria)arenotoriouslydifficulttoanalyzewithcommonlyusedPCRprimers[67].

FingerprintingmethodsDNA fingerprinting is a technique that is used to identify individual microbes in a community based on their respective DNA profiles.Terminal Restriction Fragment Length Polymorphism (T‐RFLP),Denaturing Gradient Gel Electrophoresis (DGGE), andTemperature Gradient Gel Electrophoresis (TGGE) are DNAfingerprintingmethods.Theygenerateapatternoffragmentsthatare characteristic of a strain or an entire bacterial community.They are based on differences in theDNA sequence of a specificgene, often the 16S rRNA gene, which are revealed byelectrophoretic separation. Whereas fingerprinting techniquesprovide less information than current sequencingmethods, theyare orders of magnitude cheaper and faster to perform. Asignificant disadvantage of fingerprinting methods is that taxonidentification is not straightforward and requires furtherprocessing, e.g., the excision of DNA bands from the gel andsubsequentidentificationoftheDNAfragmentsbysequencing.

DGGEandTGGEarebasedontheGC‐contentsofthePCR‐amplifiedsequences of a microbial community. The amount of energyneededtoseparatethedouble‐strandedDNAvariesdependingonthe GC content, with a single strain producing one band and acommunity giving a series of bands in the gel. Downstreamidentificationrequiresbandexcisionandsequencing.

T‐RFLP was developed by Liu et al. in 1997 [68] and is widelyused.T-RFLP is based on differences in the 16S rRNA gene sequence between bacteria (Figure 5) After amplification of the 16S gene by PCR and cleavage of the products with restriction enzymes, each bacterial taxon (genus or species) generates a fragment of 16S rRNA gene of a certain length, while a complex bacterial community


30

generates a series of DNA fragments of different sizes. The terminal fragments are detected by the generated fluorescence, as the primer is labeled with a fluorochrome. A series of non-labeled fragments is also generated upon digestion, but only one of the fragments (the terminal one) is detected per bacterium (Figure 5).

Figure5.SchematicoftheT‐RFLPmethodologyandthefingerprintsobtainedforthefecalsamplesfromtwoinfants

As is the case with other fingerprinting methods, directinformationregardingthetaxonomicidentityofthespeciesisnotgenerated.InthecaseofT‐RFLP,databasescanbeproducedthat“translate” fragment size to genus/species [69]. In theory, it ispossible to predict the size of the terminal fragment based onknowledgeofthesequenceofthe16SrRNAgeneforthespeciesinquestion, combined with knowledge of the cleavage sites of therestriction enzymes used, in a so‐called ‘in silico analysis’.However, this does not work well in practice, due to technicalproblems with the method, such as the nonlinear relationshipbetweenthefragmentsizeandthemigrationtimeofthefragment[48,70].Furthermore,thefluorescentdyelabelcontributestotheoverall molecular mass of the fragment. The molecular mass

FeiSjöberg

31

standard is usually labeled with another fluorochrome, and inmany cases, the manufacturer does not reveal the size of theirfluorochrome. As a result, while in silico analysis can aid inidentification,itisinsufficientonitsown.

MicroarrayMicroarray analysis, which is an approach that is intermediatebetween fingerprinting and next‐generation sequencing, allowsrapidanalysis ofmultiple samples, asdo fingerprintingmethods,but it also allows identification. Microarray analysis is based onshortDNAprobesthatareattachedtoasolidsurface; theprobesusuallyconsistofpartsofthe16SrRNAgenethatarespecificfordifferentbacterial taxa.Thousandsof probes canbe synthesized,andthebacterialDNAinthesamplecanhybridizetothespecies‐specific oligonucleotide probes. Hybridization is detected byfluoresenceorchemiluminesence.Thehumanintestinaltractchip(HITChip) [71] and phyloChip [72] are two microarrays thatcontain probes that hybridize with the sequences of commonmembers of the human gut microbiota. A disadvantage of themethod is that the microarray can only identify already knownbacteria. A mosre serious problem is that sequence homologiesamong different bacteria may result in cross‐hybridization andincorrectclassification.

Sequencing dependent DNA-based methods

CloningandsequencingThe first method of sequencing to identify individual bacteriawithin a community, e.g., in a fecal sample, was the cloning of aPCR‐amplifiedfragmentofabacterialgene(usuallythe16SrRNAgene) into a vector carried by E. coli, followed by Sangersequencing analysis of the cloned PCR fragment. Thismethod isvery labor‐intensive and at most some hundreds to a thousandDNAmolecules ina samplecanbe reasonablysequenced.On theonehand,onlythemostabundantbacteria inthecommunitycanbe detected and minor populations are often missed. However,long fragments, sometimes theentire16S rRNAgeneof 1500‐nt,canbeclonedandsequenced,providinggoodtaxonomicaccuracy.


32

NextgenerationsequencingSince2005,variousnext‐generationsequencing(NGS)techniqueshavebeendevelopedandappliedinstudiesofthemicrobiota.NGSisbasedonparallelsequencingofhighnumbersofDNAmolecules.Parallel sequencing technologies overcome the limitations ofSanger sequencing (referred to as ‘first‐generation sequencing’).NGSenablesrapidanalysisoflarge‐scalesequencesinmicrobiotastudies, e.g., thehumanmicrobialproject (HMP), andatdifferentbody sites, such as the digestive tract, mouth, skin, nose, andfemaleurogenitaltract[73].

There are severalNGS techniques available. Themost frequentlyused techniques for studiesofmicrobiota involve theRoche/454and Illumina/Solexa sequencers. In general, theRoche sequencergenerates long read lengths (≈700 bases) but has lower dataoutput capacity (≈700Mb/run),whereas the Illumina sequencerproducesshorterbutpaired‐endreads(≈100–300bases)andhasagreateroutput(≈15–600Gb/run).

NGSmicrobiotastudiescanbefurtherdividedintoPCR‐basedNGSand DNA sequence‐based metagenomic NGS (Figure 6). In PCR‐basedNGS,aspecifiedregionofthe16SrRNAgeneofthestudiedcommunity is amplified, and the amplicons are sequenced. InmetagenomicNGS,alltheDNApresentinasample(includingDNAof human, viral and Archaeal origin) is recovered directly andsequenced,withoutpriorPCRamplification.Thus, thedifferencesdetected in many parts of the genome between bacterial taxacontribute information and also PCR amplification biases areavoided, which increases the ability to analyze faithfully thegenetic variationswithin a community.However, thequality andreliabilityof the sequenceassembly remain critical, since the fullDNA sequences are based on the assembly of shorter DNAsequences. This assembly is based on either known full‐genomesequences of bacteria (mapping assembly) or, for bacteria thathavenotbeen culturedand thereforehavenot generatedwhole‐genome data, algorithms (so‐called ‘de novo assembly’) [74].Naturally, such a construction of genes and genomes facesmanypotential pitfalls. Metagenomic analysis may be further dividedinto functional metagenomics, in which metabolic pathways are

FeiSjöberg

33

identified, and sequence‐based metagenomics, which addressespopulationstructureandgeneticrelatedness.

Figure6.Schematicofthenextgenerationsequencingmethodology

Single‐cellsequencingSingle‐cell sequencing is an emerging technique. Individual cellscan be isolated from microbial communities using fluorescence‐activatedcellsorting(FACS)ormicrofluidicchips.Aftercell lysis,


34

the genomic DNA is amplified by the multiple displacementamplification(MDA)technique[75].MDAisanon‐PCRmethod;itstarts with the annealing of random hexamer primers to thedenatured DNA, which is followed by strand‐displacementsynthesisatconstanttemperature.ThisgeneratesDNAproductsofhighmolecularweight(7–10kbin length)with lowererrorratesthan traditional Tag polymerase‐based PCR amplification.Porphyromonasgingivalis thatwas recovered from a biofilm in ahospitalsinkwassuccessfullyanalyzedusingMDAcombinedwithNGStechniques[76].Single‐cellsequencingisapromisingmethodthatmayovercomethelimitationsofotherNGSmethodsduetoitsabilities todetect strain variations and to obtain sequences fromtaxathatarepresentinlowabundanceinacommunity.

FeiSjöberg

35

Advantages and drawbacks of PCR‐based NGS andmetagenomicsAlthoughNGSanalysisof16SrRNAgeneamplicons isapowerfultool, there are many technological challenges involved in theanalysisoftheresultingmassivedataset. Inprinciple,all thedatafrom16S‐based analyses of the human gutmicrobiota should becomparable.However,meta‐analysesof14differentstudiesoftheWestern adult fecal microbiota using PCR‐based NGS of the 16SrRNA gene have shown that “the samples generally clustered bystudy,andthe16SrRNAtargetregion,DNAextractiontechnique,andsequencingplatformproducedsystematicbiases inobserveddiversitythatcouldobscurebiologicallymeaningfulcompositionaldifferences”[54].Thethresholdsareusedintheassignmentofthesequence to a new “operational taxonomic unit” are also factorsthatmaygeneratedifferencesintheobservedresults[77].Asthecut‐offlevelsforDNAidentitythatareusedtodefineanewspeciesare based on the sequence of full‐length 16S rRNA, there is nounambiguous definition as to when a short NGS sequencerepresentsa“new”taxon,whencomparedtoasimilar,albeitnon‐identical, short NGS sequence [78]. PCR has an inherentimperfection, misincorporation of bases by the DNA polymeraseenzymesisroughly1.2x10‐6/PCRcycle[79].Hence,mistakesearlyin the amplification cycle can multiply exponentially.Incorporation of nucleotide base during a sequencing reactionproduces a light reaction, homopolymers (single‐base repeats)could be miss recorded since the light intensities do notnecessarilyincreaselinearlywiththenumberofconsecutivebases[80].

Whichsequencealignmentprogram is chosenandhowsequencegaps are treated, whether variable positions are removed (aprocesstermed“masking”)[78](Figure7).Effectsofthedifferentgaptreatmentsinfluencegeneticdistancewhichwasalsostronglyaffectedbythevariationinsequencelengthandregionofthe16SrRNA genes. All these factors will influence the subsequentdiversity estimation as phylogenetic distance was calculated bysummingthebranchlengthfortheentiretree[78].


36

Figure 7. Schematic of effects of aligment quality, distance calculation and sequencemaskingonthesample’srichnessorphylogenticdiversity.

The Human Microbial project is a systematic project to identifyand address the methodologic problems associated with PCR‐based NGS of the 16S rRNA gene. Four different leading NGScenterswereassignedacommunityof21species, i.e.,“theMock”(Figure8).Usingdifferentsequencingplatformsandtargeted16S

FeiSjöberg

37

regions,sequenceartifactscouldbeidentifiedthatresultedinthemisclassificationofbacteria, generating falsemeasuresof speciesdiversityandcreating “new” taxa thatwereactuallynot includedin theMock i.e.,Azomonas,Cronobacter andBergeriella [66], butthat were instead the consequences of mistakes in the PCR,sequencing,andalignmentprocesses.

Figure8.Mockcommunity‐basedaccuracyofcommunityrepresentationcomparedacrosstechnologyand16Swindow. TheMCwassequencedbydifferentcentersonboth3730 and 454 platforms. Each sequencing trial is represented as a column. For 3730sequencingoftheV1–V9window,ampliconsderivedfromacommonamplificationprotocolweresequencedwithshortcapillaries(1),longcapillaries(2),andthreereadsperclone(3).454sequencingwasperformedbyfourcenters(A,B,C,andD)withthree16Swindows(V1–V3,V3–V5,andV6–V9).(A)Theobservedgenuslevelfrequencydataoverexpectedgenus‐level frequency ratio foreachof theMCmembers is shownasaheatmapusingabinarylogarithm scale.The expected frequency ratio is based on the whole genome coveragesinferredfrommappedIlluminaWGSreadstotheMCreferencegenomesequences.Generawithobserved frequenciesdifferingmorethanfour‐fold fromexpectedaremarkedwith+


38

or–forover‐orunder‐representation,respectively.(B)Thefractionofmisclassified(0.1%of the total combined data set) and unclassified (4.6% of the total combined data set)sequencesdisplayedasafrequencyheatmap.Thefrequencyvaluesaredepictedasabinarylogarithmscale[66].

ChimeraformationduringPCRFalselyidentifiednoveltaxacouldduetomethodologicalproblemssuchas formationofchimeras, introductedvariations in theDNAsequence during PCR multiplication (Figure 9). Chimeras werefoundtoreproduciblyformamongindependentamplificationsandcontributed to falseperceptionsof samplediversityand the falseidentificationofnovel taxa,with less‐abundantspeciesexhibitingchimeraratesexceeding70%[31].

Figure9.FormationofchimericsequencesduringPCR.AnabortedextensionproductfromanearliercycleofPCRcanfunctionasaprimerinasubsequentPCRcycle.Ifthisabortedextension product anneals to and primes DNA synthesis from an improper template, achimeric molecule is formed [31]. Haas et al. Genome Res. 2011, Cold spring HarborlaboratoryPress.

FeiSjöberg

39

Culture versus DNA-based method Bacteria that can be cultured in the laboratory represent only asmallfractionofthebacteriathatexistinnature.Nonetheless,itisonly through the isolation of individual bacterial species in pureculture that characterization of physiologic features and a fullassessment of virulence potential, i.e., antibiotic resistance andpathogenicity,canbeperformed.

Molecular techniques permit the identification of bacteria thathave hitherto not been cultured, mainly very strict anaerobes.However, many strictly anaerobic bacteria, such as Roseburia,Faecalibacterium, and Eubacterium, are readily cultured [33], sothe failure to culture certain gut bacteria may reflect a lack ofeffort rather than an inherent unculturability [56]. Bomar et al.[81]characterizedthebacterialgenometoidentifygrowthfactorsand to design media for cultivating previously “unculturable”organisms.Apossiblereasonforafailuretocultureabacteriumisits lackof specificmetabolicpathways.Forexample,Tropherymawhipplei lacks the genes for nine pathways of amino acidbiosynthesis;genomesequencingwasusedtoidentifythemissingpathways and to design a culture medium that supported thegrowthofT.whipplei[82].

ThereisageneralbeliefthatDNA‐basedmethods,especiallyNGS,have the ability to reveal “all” thebacteria in the gutmicrobiota.However,everymethodhasitslimitations.Itisimportanttopointout that in a complex bacterial community, Bacteria with smallpopulation size will be missed. The table below exemplifies thenumber of reads that will be genetated if 10,000 sequences areobtained from a faecal sample in which a Bacteroides strain ispresent at 1010 CFU/g, aBifidobacterium strain is present at 109CFU/g,etc(Table6).Specieswithpopulationsizedof<106willbemissede.g.Enterococcus


40

Table6. Theoreticaldetectionlimitofsequencingofacomplexemicrobiota.

FeiSjöberg

41

Influence on the host of the microbiota

Infection Thegutmicrobiotaservesasareservoirforpathogenicbacteria.E.coli, Klebsiella, and other members of the Enterobacteriaceaefamily are the major causes of urinary tract infections and areoftenthecausesofinfantsepsis.[83].Commensalbacteriaseldomcause disease at the site of colonization, although there areexceptions to this rule. While Clostridium difficile is carried bymost people without producing any symptoms, if competingbacteria are suppressed by treatment with broad‐spectrumantibiotics,C.difficilemayexpand innumbersand its toxinsmaytrigger diarrhea and mucosal inflammation [84]. The gutmicrobiotamayserveasareservoirforantibiotic‐resistantstrains,andalsoasamilieu inwhichresistancegenescanbetransferredbetweenstrains[83].

Microbiota interactions with the immune system

Basic immunology concepts The fundamental role of the immune system is to eliminate orneutralizemicroorganisms.Innateimmunityandadaptiveimmunityarethetwoarmsofthehumanimmunesystem.

Innate immunity recognizes conserved microbial structures andbecomesactivatedwithinminutesofinfection.Theinnateimmunesystem is therefore crucial for the elimination of invadingmicroorganisms. It comprises cells (neutrophils, eosinophils,basophils,naturalkillercells,mastcells,monocytes,macrophages,and dendritic cells) and soluble components (complement andcoagulation factors, acute phase proteins, and cytokines), whichsynergizetokillandeliminateinvadingorganisms.

Adaptiveimmunity(alsoknownasacquiredimmunity)recognizesspecific antigens through highly diverse antigen receptors on TandBlymphocytes.Uponactivation,naïveTcellsdifferentiateintovarious subtypes with different effector and regulatory functions.NaïveBcellsdifferentiateintoantibody‐secretingplasmacellsorinto


42

memoryBcells,whichare rapidly reactivatedwhen theyencounterthesameantigenanew.

Microbesareeliminatedthroughco‐operationbetweentheinnateandacquiredimmunesystems.Forexample,antibodies(acquiredimmunity) activate complement (innate immunity), leading toenhancedphagocytosisandactivatedTcells(acquiredimmunity),which increase the capacity of macrophages to kill ingestedbacteria.

The mucosa-associated lymphoid tissue The mucosal surfaces are protected by the mucosa‐associatedlymphoid system. Themouth, nose, and intestines are importantsites for the induction of immune responses, both mucosal andsystemic, and indeed, most foreign antigens enter the body viathesesites.

Nasopharynx‐associatedlymphoidtissueThe nasopharynx‐associated lymphoid tissue (NALT) comprisestheadenoids, thepalatine tonsils, and isolated lymphoid follicles,sometimes referred to asWaldeyer’spharyngeal ring [85]. It hasalsobeenimplicatedinthedevelopmentofimmunematuration,asintranasal exposure to a microbial extracts may induce activeimmunematuration [86] The tonsils are rich in FOXP3+ T cells,whicharethoughttoberegulatoryTcells,andinvitroactivationofhumantonsillaTcellsinducesimmunesuppressiveactivity[87].

Gut‐associatedlymphoidtissueThe gut‐associated lymphoid tissue (GALT) comprises themesenteric lymph nodes, the Peyer’s patches, isolated lymphoidfollicles in the intestinal wall, intraepithelial lymphocytes, anddiffuselyscattered lymphocytes in thegut laminapropria(Figure10) [88‐90]. The small intestine is a major inductive site foracquiredmucosalimmuneresponses[91].ThePeyer’spatchesarelocated in the small bowel, with approximately 200 of thesestructuresbeingfoundintheaverageadult.Bacteriathatresideatmucosalsitesareabletotranslocate,i.e.,remainintactandviablewhile crossing the mucosal barier [89] [90]. This results inimmune activation and it has been proposed that the gut

FeiSjöberg

43

microbiota is central to immunematuration during infancy [88].Animalsthatarerearedgerm‐freehavepoorlydevelopedPeyer’spatches and have lower serum IgG and IgA levels, as comparedwith conventionally reared animals [92]. Regulatory T cellsobtainedfromgerm‐freemicehavepoorersuppressivecapacitiesthan the regulatory T cells from mice that have an establishedmicrobiota[93].

Figure10.Gut‐associatedlymphoidtissue.Copyright@IngegerdAdlerberth


44

Oral immune system Relativelylittleisknownabouttheimmunedefensesystemoftheoralcavity.Sublingual immunotherapy,whichinvolvesplacinganallergen under the tongue, is a convenient way of reducingallergenicimmuneresponses,e.g.,tograsspollens.

Bacteria in the oral cavity have a pronounced capacity totranslocatetothecervicallymphnodes[94]andtherefore,theoralmicrobiotamay be amajor immune stimulant, albeit one that isgenerallyoverlooked.

Oral tolerance Immunologicunresponsivenesstoharmlessantigens,suchasfoodproteins, is crucial for the maintainance of health. Thisunresponsiveness is actively induced and antigen‐specific, and istermed‘oraltolerance’.Oraltoleranceisdefinedastheabilityofanorally administered antigen to suppress subsequent local andsystemicimmuneresponses[95].Despitedecadesofresearch,theexact mechanisms underlying oral tolerance have not beendefined. The small intestine is a major site of induction of oraltolerance, although the contributions of the NALT and the oralmucosaarenotwellunderstood.Oraltoleranceinvolvesboththemucosalepithelium,whichproducestolerogenicpackagedantigen,so‐called “tolerosomes” [96], and tolerogenic antigen‐presentingcells in the gut‐associated lymhoid system [95].Oral tolerance ismoreshort‐lived inanimals thatarerearedwithoutamicrobiotathaninconventionallyrearedanimals[97],andgerm‐freeanimalshaveareducedcapacitytoconvertantigensintotolerosomes[96].

FeiSjöberg

45

Gut microbiota and host metabolism

Short-chain fatty acids Themicrobiotahelpsthehosttometabolizeotherwiseindigestibledietary fibers,hostmucin,andresistantstarches intoshort‐chainfatty acids (SCFAs) [98]. SCFAs are produced through bacterialfermentation and include acetate, propionate, and butyrate,isovalericacid, andother fattyacids.Acetate is themainSCFA inthebloodstream;itisreadily absorbedandtransportedtotheliverwhere it is used in the production of blood lipids. Propionate isused by the liver to produce glucose, through a process termedglucogenogenesis [99]. Butyrate is produced by certain obligateanaerobic bacteria within the Firmicutes phylum, mainly thosebelongingtoclostridialclustersIVandXIVa[100].Butyrateservesas an importantnutrient forhost colonocytes [101, 102]. In fact,colonic epithelial cells derive their energy mainly from the gutlumen,ratherthanfromtheblood.Butyratehasbeenreportedtoexert anti‐inflammatory [103] and anti‐tumorigenic [104] effectsin the colon, andRoedigerproposed in the1980s thatulcerativecolitis, an inflammatoryboweldisease,wasdue toan inabilityofthecolonicepitheliumtoutilizebutyrateasfuel[101].


46

Microbiota and Allergy

Allergy – “The modern plague” Theterm’allergy’,meaning‘changedreactivity’,wasintroducedin1906byClemens vonPirquet[105]. Initially,allergywasequatedwith immunity. However, it was subsequently reserved forimmune reactions that resulted in adverse symptoms, i.e.,hypersensitivity. Allergy is today defined as immune-mediated hypersensitivity; in allergy, the immunesystemreacts toharmlesssubstances, termed allergens, and this reaction producessymptoms. Different immune effector mechanisms are implicated in allergy, including IgE antibodies (so-called ‘atopic allergy’), T cells (contact dermatitis), and possibly also IgG antibodies.

The prevalence of IgE‐mediated allergy, including asthma, atopiceczema, and hay fever, has increased dramatically in Westernsocietiesinparallelwithadecreaseintheprevalenceofinfectiousdiseases.Thesamepatternisseenforautoimmunediseases(suchas multiple sclerosis) and inflammatory bowel (e.g., Crohn’sdisease) (Figure 11) [106]. Today, one third of Swedish schoolchildrensufferfromIgE‐mediatedallergy[107].

Figure11.InverseRelationbetweentheIncidenceofPrototypicalInfectiousDiseases(PanelA)andtheIncidenceofImmuneDisorders(PanelB)from1950to2000.Reproducedwithpermissionfrom[106],CopyrightMassachusettsMedicalSociety.

FeiSjöberg

47

IgE-mediated allergy Allergy caused byIgEantibodiesistermedatopicallergy(fromtheGreekatopos,meaning‘outofplace’).

In IgE‐mediatedallergy, theallergensarealmostalwaysproteinsthat are commonly present in pollens, food, and animal dander.TheseantibodiesbindtotissuemastcellsviaFcεreceptorsonthecell surface.Theproductionof IgEantibodiesdirectedagainstanallergenistermed‘sensitization’.

Allergy occurs when the sensitized individual re‐encounters thesameallergenatalatertime‐point,whereupontheallergenbindsand cross‐links at least two IgE antibodies and thus, the Fcreceptors on themast cell. This inducesmast cell degranulationand the release of histamine, which rapidly increases thepermeability of the capillaries and venules. This early phase ofallergy takes place within 10minutes of the encounter with theallergen.Later,theactivatedmastcellproducesandreleasesmastcell‐specific leukotrienes (LTC4, LTD4 and LTE4), which causeprolonged vascular leakage and airway smooth muscleconstriction. Lastly, cytokines produced by mast cells,macrophages, and T cells at the allergenic focus synergize toattract eosinophils and more T cells to the site of allergenchallenge, thereby contributing to tissue swelling and airwayrigidity.

Dependingontheaffectedorgan,theinvolvedsymptoms/diseasesinclude allergic conjunctivitis (eyes), allergic rhinitis (nose)asthma(bronchi),vomitinganddiarrhea(gut),urticariaoreczema(skin).Asystemic,oftenlife‐threateningallergicreactionistermedanaphylaxis [108]. There is no straightforward relationshipbetween the route of exposure and the organ in which thesymptomsappear.Forexample, foodallergydenotesanyallergicreaction to ingested food, including gastrointestinal symptoms(vomiting, diarrhea), skinmanifestations (urticaria, eczema), andlesscommonly,bronchialconstriction.


48

The atopic march Atopic allergy usually appears during childhood and earlyadolescence. Theterm’atopicmarch‘(alsocalled‘allergicmarch’)(Figure 12) refers to the sequential progression of the differentmanifestations of atopic allergy in susceptible children. This’march’startsatinfancywithatopiceczemaand/orfoodallergyasthefirstmanifestation(s).Asthmausuallypresentsinearlyschool‐age,whileallergic rhinitis is the lastorgan‐specificmanifestationoftheatopicstate,typicallydébutinginearlyadolescence.Asthmaand allergic rhinitis usually persist throughout adult life, whilefoodallergyandatopiceczematendtovanishwithtime.However,this is not universally true and while food allergies to milk andeggsmaydisappear,allergiestofishandnutsmaybylife‐long.

Figure 12. The atopicmarch. Reprinted withpermission fromChildhood atopic eczema;Copyright BMJ PublishingGroup Ltd.

One‐third of children with eczema develop asthma during laterchildhood [109] and two‐thirds develop allergicrhinoconjunctivitis[110].

Other types of allergy Other typesofallergy thatdonot involve IgE‐mediatedreactionsare non‐IgE‐mediated food allergy and celiac disease (whichappear to be T‐cell‐mediated, perhaps with a contribution fromIgG) and contact dermatitis (T‐cell‐mediated). Non‐IgE‐mediatedfoodallergyismostcommonlytocow’smilkandsoya.Symptomsinclude diarrhea or constipation and pain. The symptoms mayappear many hours and even days after consumption of theoffending food, while IgE‐mediated food allergy may presentwithinminutesoffoodconsumption.

FeiSjöberg

49

The hygiene hypothesis Allergy is currently the most common chronic disease amongchildren in the Western world. While the cause of allergy isunknown, epidemiologic studies point to reduced exposure tomicrobes of infants and young children as a risk factor. Thus,poverty, crowded housing, large families, early contactwith petsor farm animals, and early exposure to foodbornemicrobes, areassociatedwith a reduced risk of allergydevelopment [111‐113](Figure13).The "hygienehypothesis"wasproposedbyStrachanin1989[114].Strachanobservedthatsmallfamilysizes[114]andhigh parental social class [115] were both strongly andindependentlyassociatedwithhighriskofhayfeverat16yearsofage in two large population‐based British cohorts. He proposedthat some infectious agent(s) circulate(s) in poor and largefamilies, and that exposure to this/these infectious agent(s)induced a type of immune maturation that counteracted thedevelopment of atopic disease. Twenty‐five years after theformulationof thehygienehypothesis, evidencehasaccumulatedthat an “unhygienic” lifestyle is protective against allergydevelopment. The three major protective factors are: poverty;largefamilysize;andanimalcontacts[111].Growinguponafarmwith milk production is particularly protective against allergydevelopment[113].

Figure13.Lifestylefactorsareassociatedwithprotectionagainstallergy.


50

Birth by vaginal delivery, which exposes the neonate to thecomplexmaternalmicrobiota,hasbeenassociatedwithareducedriskofallergydevelopment,ascomparedwithCesareandelivery,insome[116]butnotallstudies[117].

Factorswithsmallornoeffectsontheriskofallergydevelopmentincludegestational age atbirth, breastfeedingor formula feedingandvaccinations.

Neonatal microbial exposure and the risk of allergy development Commensal microbial colonization is the major stimulus for theimmune system of the infant and, thus, is likely to affect itsdevelopment [88]. The acquisition of commensal gut bacteria isvery rapid in infants born and living under poor sanitaryconditions in the urban slum of Lahore, Pakistan, as comparedwithSwedishinfants[118].TheturnoverofindividualstrainsofE.coli is also much higher in the former group than in the lattergroup. On average, an urban slum‐dwelling infant in Pakistancarries 8.5 differentE. coli strains over the first 6months of life[119].Astudyperformedinthe1980’sreportedanaverageof4.2E.colistrainsoveraperiodof11–18monthsinthelargebowelsofSwedishinfants[120],andintheSwedishFLORAstudy,ourgrouphasreportedanaverageof2.1E.colistrainsover12monthsinthebowelsofinfantswhowerebornaroundyear2000[121].

Manystudieshaveinvestigatedtherelationshipbetweentheearlygutcolonizationpatternandsubsequentallergydevelopment.Themost consistent observation is that the presence of a gutmicrobiotaof lowcomplexityduringtheneonatalperiodisariskfactor for allergydevelopment [122‐125]. There is little evidencethat any particular group of bacteria is particularly protective. Apossibleexception isS.aureus, a skinbacteriumthathasbecomepartof theneonatalgutmicrobiota inrecentyears,probablyasaconsequenceoflimitedcompetitionfrom“traditional”gutbacteriathathavebecome lessprevalent in thegutmicrobiota inparallelwith an increasingly hygienic lifestyle. S. aureus elaboratesstrongly immune‐stimulating toxins, so‐called ‘superantigens’,

FeiSjöberg

51

which may offer broad stimulation of the immune system,equivalent to that conferred by hundreds of strains of ‘normal’bacteria.

Oral flora and allergy development To the best of my knowledge, there have been no prospectivestudies on the relationship between early oral colonization bybacteria and allergy development. The prevalence of allergicdisease in relation to antibody responses to oral pathogens wasstudied as part of the American National Health Program [126],antibodies were analyzed in serum samples from 9385 subjectsaged ≥12 years. Individuals with high concentrations of IgGantibodies to Porphyromonas gingivalis had a significantly lowerprevalenceofphysician‐diagnosedasthmaorhayfever,aswellasself‐reported wheeze (defined as a positive response to thequestion:“Apartfromwhenyouhaveacold,doesyourchesteversound wheezy or whistling”). Patient‐reported wheeze wassignificantly negatively correlatedwith the concentrations of IgGantibodiesagainstActinobacillus actinomycetemcomitans[126]. Inanimalstudies,oraladministrationofbacterialextractsprotectedmice from allergic airway inflammation via the recruitment ofregulatory T cells to the airway [127]. Mice challenged with P.gingivalis showed reduced ovalbumin‐induced allergic airwayinflammation and airway responsiveness, concomitant withreduced levels of the Th2‐type cytokines IL‐4, IL‐5, and IL‐13[128].


52

Microbiota and inflammatory bowel diseases Inflammatory bowel disease (IBD) comprises several diseases, ofwhich ulcerative colitis and Crohn’s disease are the mostprevalent. Severe inflammation of the intestinal wall is thehallmark of these diseases. In ulcerative colitis, inflammation isconfinedtothecolonandtothemucosallayer,whiledeeperpartsof thegutwall are spared. In contrast,Crohn’sdisease canaffectany part of the gastrointestinal tract and the resultinginflammation may span the entire gut wall. Furthermore,granuloma formation is a hallmark of Crohn’s disease but is notseen in ulcerative colitis. Granulomas are aggregates of activatedmacrophages that are surrounded by scattered lymphocytes andfibrosis. Sometimes, the macrophages merge, formingmultinucleatedgiantcells.

The etiology of IBD is unknown; interactions between the hostimmune system and an unbalanced intestinal microbiota aresuggestedtobeofkeyimportance[129].TheprevalenceofIBDishighest in Northern Europe and North America at 20/100,000persons.Theincidenceisincreasingnotonlyintraditionallyhigh‐incidence regions, but also in low‐incidence areas, such assouthernEuropeandAsia.

One‐fifthofallcasesof IBDoccur inchildhood,andtheincidenceofpediatricIBDinSwedenhasincreaseddramaticallyoverthelastdecades [130].Familyhistory isawell‐knownrisk factor for IBDdevelopment. Several IBD‐associated genes have been identified,themostwell‐knownbeingNOD2,whichisarisk‐associatedgenefor Crohn’s disease [129]. NOD2 is the intracellular receptor forbacterialcellwallpeptidoglycan,afactthatlinksIBDwithimmunereactionstothemicrobiota.However,geneticsusceptibilityalonedoesnotexplain theobserved increases in theprevalenceof IBDoverthepastdecades.Inaddition,individualswhohavemigratedfromacountrywithalowprevalenceofIBDtoahigh‐prevalencecountrydevelop thediseaseat the samerateas thatobserved inthecountrytowhichtheyhavemigrated[131,132].

FeiSjöberg

53

Similartoallergy,IBDhasbeenlinkedtoaWesternlifestylewithgood sanitary conditions [133, 134]. Numerous studies haveinvestgated themicrobiota inpatientswith IBD. Inmostof thosestudies, the investigatedpatientshad already received treatmentfortheirdiseaseandmayhavehadthediseaseformanyyears,inwhich case the observed differences may be secondary to thedisease process and/or the administered medication.Investigations of the microbiota in pediatric cases of new‐onsetIBDare summarized inTable7. In summary, the resultsof these11studiesshowthatchildrenwithIBDhave:1)increasedlevelsofEnterobacteriaceae and other facultative anaerobic bacteria; and2) decreased level of Firmicutes, particularly Clostridiales(Roseburia,Ruminococcus andEubacterium spp.). The results forBacteroidetesandFaecalibacteriumprausnitziiareinconsistent.Acommon finding is that themucosa‐associated colonicmicrofloraofpatientswithIBDislessdiversethanthatofhealthyindividuals.These studies have all investigated the colonic or distal ilealmicrobiota.

Two reasons can be put forward to explain why an alteredmicrobiota triggers IBD: 1) that the IBD microbiota contains ahigher proportion of inflammatogenic bacteria, i.e., facultativeGram‐negative bacteria that live in proximity to themucosa andhave the capacity to translocate themucosal barrier; and2) thatbacteriawithprotectivepropertieshavebeenlost.

Regardingthefirstpossibility,colitisdoesnotoccurinagerm‐freeenvironment in animal models of IBD [135], and CD4+ T cellsactivatedbycommensalbacterialantigenscantransferthedisease[136]. Monocolonization by Bacteroides fragilis could initiatechronic inflammatory lesions in susceptible mice, while E. coliinduced acute‐type inflammatory reactions [137]. Regarding thesecond possibility, i.e., the potential loss of protective bacteria,butyrateisproducedbybacteriaintheFirmicutesphylum,mainlythose belonging to clostridial clusters IV and XIVa [100], anddecreased numbers of the members of these cluster have beenfound in patients with IBD [138]. This observation has led tosuggestions that butyric acid‐producing anaerobic bacteria couldbeusedasanovelprobiotictreatmentforIBD[139].


54

Table7. Gutmicrobiotainnew‐onsetIBD

Age(year), subject & sample

Methods Major finding Ref

Mean age=12, feces CD=27,UC=16, Ctrl=IBS=30

16S rRNA PCR. Cloning & sequencing

Correlation to age Young CD ↑E. coli & ↓Bacteroidetes Older UC ↓ E. coli & ↑ Bacteroidetes

[140]

Age=16‐60, feces CD=30, UC=33 Ctrl=IBS=34

16S PCR Cloning & sequencing

↑Escherichia/Shigella in CD ↓Faecalibacterium in CD vs. Ctrl ↓Faecalibacterium in CD vs. UC

[141]

Age=8‐13, colonoscopy biopsy, CD=12, UC=7, Ctrl=GI symptoms=7

Culture 16S rRNA RT‐PCR

↑Aerobic & faculta ve bacteria in IBD vs. Ctrl ↓Bacteroides vulgatus in IBD vs. Ctrl

[142]

Age=12‐73, colonoscopy biopsy, CD=20, UC=15 Ctrl=family history of colon cancer =15

DGGE. 16S Clone libraries.

DGGE band: UC≠CD Bacteria association: UC>CD(=Ctrl) Bacteroidetes: CD>UC

[143]

Age=11‐18, colonoscopy biopsy, CD=15 Ctrl=GI symptoms=26

NGS 16S, V1‐V3. Fungal 18S PCR. Mycobacterium PCR.

↓Roseburia, Ruminococcus, Eubacterium, and Subdoligranulum in CD vs. Ctrl

[144]

Age=6‐16, colonoscopy biopsy, CD=13, UC=12 Ctrl= GI symptoms=6

NGS 16S, V3‐V6. RT‐PCR

↑Faecalibacterium prausnitzii in CD vs. Ctrl. ↓Diversity in CD vs. Ctrl NO disease‐specific PCA clustering

[145]

Age=6‐16, colonoscopy biopsy, CD=29, UC=13 Ctrl= GI symptoms =42

Culture and PCR of microaerophilic Gram‐negatives

No significant difference in microaerophilic Gram‐negatives bacteria

[146]

Age=3‐18, Ileum biosy CD=32, Ctrl= GI symptoms =36

Pseudomonas 16S PCR

↑Pseudomonas in CD vs. Ctrl Pseudomonas species were less diverse in CD

[147]

Age=5‐18, Colonic biopsy, CD=62, UC=26 Ctrl= GI symptoms =54

Mycobacterium avium subsp. Paratuberculosis PCR

↑Mycobacterium avium subsp. paratuberculosis in CD vs. Ctrl

[148]

Age=8‐16, colonoscopy biopsy, CD=9, UC=6, Ctrl=GI symptom =8, Healthy ctrl=16 (faecal)

End‐point PCR. qPCR

↓Bdellovibrio bacteriovorus in CD vs. Ctrl

[149]

Age=3‐17, colonoscopy biopsy, CD= 447 Ctrl=IBS= 221

NGS 16S, V4. NGS shotgun and metagenomics

↑Enterobacteriaceae, Pasteurellacaea, Veillonellaceae & Fusobacteriaceae in CD vs. Ctrl. ↓Erysipelotrichales, Bacteroidales and Clostridiales in CD vs. Ctrl. Dysbiosis reflected in biopsy but weakly in stool, Antibiotic treatment =more dysbiosis

[150]

Abbreviations: CD= Crohn’s disease, UC=Ulcerative colitis, Ctrl=Control, GI=Gastrointestinal, IBS=Irritable bowel disease, ↑=Increased level of, ↓=Decreased level of,

FeiSjöberg

55

AIM

Themainpurposeofthisthesiswastostudytheimportanceofthecommensal microbiota of children in health andimmunoregulatory diseases, such as allergy and inflammatoryboweldisease.Tothisend,wehaveestablisheddifferentbacterialidentification methods and compared the suitability of thesemethodsforcharacterizationoftheinfantmicrobiota.

Papers I and III ‐Whether or not themicrobiota plays a role inimmunedisorders

Paper I: To investigatewhether different pacifier cleaning habitsaffectoralmicrobiotacompositionandallergyoutcomesininfants.

Paper III: To compare the duodenal microbiota at the onset ofpediatric IBD in patients with that of controls using high‐throughput pyrosequencing and quantitative culturing, beforemedicationandintheabsenceofanyantibiotictreatment.

Papers II and IV – Infant gut flora development and comparativeanalysisofthemethodsusedforstudyingthemicrobiota

Paper II: To adapt T‐RFLP to the characterization of infant fecalmicrobiota, including the building of a database for theidentification of T‐RFLP fragments. Furthermore, to compare T‐RFLPwithquantitativeculturing for theanalysisof the intestinalmicrobiota in infants with respect to sensitivity for detectingdifferentgroupsofgutbacteria.

Paper IV: To compare pyrosequencing with T‐RFLP andquantitative culturing for the detection of various groups ofcommensalbacteriaintheintestinalmicrobiotaofinfants.


56

MATERIAL AND METHODS

General overview of the four papers I.PacifierCleaningPracticesandRiskofAllergyDevelopment

II. Comparison between terminal‐restriction fragment lengthpolymorphism (T‐RFLP) and quantitative culture for analysis ofinfants'gutmicrobiota

FeiSjöberg

57

III. Altered composition of duodenalmicrobiota of childrenwithnewlydiagnosedinflammatoryboweldisease

IV. Comparative analysis of infants’ gut microbiota by nextgenerationsequencingandquantitativeculture


58

Children and Samples Detaileddescriptionsofthematerialsandmethodsusedaregivenineachpaper,andonlythoseofparticularrelevancetotheresultsaredescribedbelow.

The AllergyFlora birth cohort (Papers I, II and IV) Pregnant women were recruited into the AllergyFlora study atMölndal Hospital, in the greater Gothenburg area of South‐WestSweden. A total of 187 children born in the period 1998–2002were enrolled at 1–3 days of age to examine the relationshipbetween theearly intestinal colonizationpatternandsubsequentallergy development. In this cohort, families with at least oneallergic parent were mainly targeted, so as to achieve a highproportion of children with allergic disease. Upon inclusion, astructured interview was conducted, which focused on thepregnancyanddelivery, andon the family structureandhousingconditions.Diariescoveringtheinfant’sfirstyearoflifewerekeptby the parents, who were asked to record food introduction,weaning, diseases andmedications, and other significant events.This information was reported during structured telephoneinterviewswiththeparentswhenthechildrenwere6monthsold.Duringthisinterview,theparentswerealsoasked:“Doesthechilduseapacifier?”and“Isitcleanedbyboiling,rinsingintapwater,orby the parents sucking on it?” (more than one option could beselected).Theexclusioncriteriawerepretermbirth(<38weeksofgestation), neonatal intensive care, and presumed inability tocomplywithaquitedemandingstudyprotocol,e.g.,difficulties inunderstandingtheSwedishlanguage.

Fecal samples were collected from the 184 infants. From infantnumber117onwards,asalivasamplewasobtainedat4monthsofageandbloodsampleswereobtainedat4,18,and36monthsofage.Theaimwastomonitortheimmunematurationoftheinfantsin relation to the intestinal colonization pattern and allergydevelopment.

A pediatric allergologist examined the children, reviewed theirmedical charts, and performed a structured interview with the

FeiSjöberg

59

parents when the children were 18 and 36 months old, andwhenever symptoms suggestive of the commencement of allergywere noted. Venous blood was analysed for eosinophilicgranulocytecounts,which increase inpatientswithallergies, andforallergen‐specificIgE(“sensitization”,seebelow).

For clinical diagnosis, the examining pediatrician used thefollowingcriteria:

Eczema: Diagnosed according to Williams criteria for “atopicdermatitis”.“Eczemaat18months”denoteddiagnosisatanytimebeforeorat18monthsofage;“Eczemaat36months”requiredthepresenceofsymptomsafter24monthsofage.

Asthma: Persistent wheezing for ≥4 weeks or ≥3 episodes ofwheezing combinedwith≥1minor criterion (symptomsbetweencolds, eczema, allergic rhinoconjunctivitis or food allergy). For“asthmaat36months”,≥1wheezingepisodehadtohaveoccurredafter24monthsofage,andresponsestoinhaledglucocorticoidsorleukotrieneantagonistswereincludedamongtheminorcriteria.

Sensitization: Presence of specific IgE against inhalant allergens(birch,timothygrass,mugwort,cat,dog,horse,D.pteronyssinus,D.farinae, and mold; Phadiatop®; Phadia AB, Uppsala, Sweden) oragainst food antigens (milk, egg, soy, fish, wheat, and peanut;ImmunoCAP® food‐mix test; Phadia AB). A positive reactionwasdefinedwhenthesumoftheallergen‐specificIgEswas≥0.35kU/Lineitherassay.

Parentalhistoryofallergy:Physiciandiagnosisofasthma,allergicrhinoconjunctivitisoreczema,asreportedintheinterviews.

InPaperI,allergicmanifestationat18monthsofagewasrelatedto the parents’ habits of cleaning their baby’s pacifiers in theALLERGYFLORA cohort. Saliva samples collected at 4 months ofagewereavailablefor33children.

In Papers II and IV, the fecal samples from six infants wereanalyzed in each study. Infants were selected from the


60

ALLERGYFLORAcohortstudy.FourinfantswereanalyzedinbothPaperIandPaperII.Sampleswerecollectedbytheparentsonsixoccasionsduringthefirstyearoflife(at1,2,and4weeksandat2,6, and12monthsof age). InPaper I, all36 samples from the sixinfants were analyzed. In Paper IV, a total of 20 samples wereanalyzed,duetothefactthatforfourinfants,onlythe1‐weekand1‐yearsampleswereanalyzed.

Inflammatory bowel disease study (Paper III) ThestudywasdesignedtocharacterizethemicrobiotaofchildrenwithIBDasearlyaspossible,i.e.,beforediagnosisandinitiationoftreatment. Childrenwhowere referred for suspected IBD to theGastroenterology Unit at the University Hospital in Gothenburg,Sweden, were included. The exclusion criteria were any priortreatment with anti‐inflammatory, disease‐modifying drugs orantibiotic treatment in theprevious3months.Thechildrenwerehospitalizedforclinicalwork‐upandtheirintestinalbiopsieswereexaminedbyanexperiencedpathologist.Diagnosisofthepatientswas based on clinical presentation, endoscopic and histologicfeatures,andradiologicappearanceaccordingtothePortocriteria.The following diagnostic groupswere included:Ulcerative colitis(N=8);Crohn’sdisease(N=5);andControls(N=8).Thelattergrouprepresentedthosechildreninwhomtheintestinalbiopsyshowedno signs of inflammation; their symptoms were deemed to becaused by a functional intestinal disorder. Symptomatic childrenwith the following diagnoses were not included in the study:allergiccolitis(N=2);celiacdisease(N=2);infectiouscolitis(N=1);unspecificcolitis(N=4);andindeterminatecolitis(N=1).

Duodenalcontentsweresampledinconnectionwithgastroscopy.Theduodenalfluidwasimmediatelytransportedtothelaboratoryforquantitativeculturing(seebelow)andsubsequentassessmentbypyrosequencing(seebelow).

FeiSjöberg

61

Methods

Quantitative culture (Papers II, III, and IV) Samples (of feces or duodenal fluids)were transported in a gas‐tight sachet in which an anaerobic milieu was created using ananaerobic generator (Benex limited, County Clare, Ireland).Anaerobiosis was confirmed upon arrival to the laboratory bycheckingtheanaerobicindicator.Culturingwasperformedwithin24 h. The detection limit was 300 (102.5) colony‐forming units(CFU)/gforfecesand33(101.52)CFU/mLforduodenalcontents.

Inbrief,acalibratedspoonfulof feceswasseriallydiluted inten‐fold steps in sterile peptone water and cultivated on two non‐selectivemediaandeightselectivemedia(Table1).Forduodenalfluids,100μlwereseriallydilutedandcultivatedonnon‐selectivemedia aerobically (Columbia blood agar) and anaerobically(Brucella blood agar). Bacterial colonies with differentmorphologies were enumerated on appropriate media,subcultured for purity, and speciated using a combination ofbiochemicaltestsandgenus‐orspecies‐specificPCR,aspreviouslydescribed [49]. Pure‐culture isolateswere stored frozen at‐80°C.Anaerobic spore formers (“clostridia”) were quantified byanaerobic culture of alcohol‐treated samples. Alcohol treatmentkills vegetative cells, whereas bacterial spores survive. Thus,spore‐forming bacteria can be detected by this method. Spore‐faming bacteria include Bacillus, which are aerobic, andClostridium, which are anaerobic. The group “Clostridia”, definedbyculturedoesnotconformtothegenusClostridia,asdefinedbyDNA‐based taxonomy. The latter comprises 19 groups (I–XIX),which include both spore‐formers and non‐spore‐formingbacteria. A portion of a 1:10 dilution of stool samplewasmixedwith an equal volume of 99% ethanol on a shaker at roomtemperature for 30 min. The sample was diluted and plated onagar.


62

Table1.Cultureconditionsandspeciesidentificationmethods

Culture condition, agar & duration Species identification

Facultative bacteria Aerobically at 37°C Days Morphology Bacterial typing

Enterobacteriaceae Drigalski 1 G‐ bacilli API20Ea

Staphylococcus Staphylococcus 2 G+ cocci Catalase+, Coagulase±

Enterococcus Enterococcosel 1 G+ cocci Esculine hydrolysis

Non‐selective Colombia blood 1 Total CFU counts

Anaerobic bacteria Anaerobically at 37°C

Bacteroides Bacteroides bile esculin 3 G‐ bacilli Rapid ID 32Aa

Clostridium (sporeforming) Brucella blood, alcohol treatment 3 G+ bacilli Rapid ID 32Aa

Clostridium difficile CCFAb 3 G+, G± bacilli Rapid ID 32Aa

Bifidobacterium Beerens 3 G+ bacilli Species‐specific PCR

Lactobacillus Rogosa 3 G+ bacilli Species‐specific PCR

Non‐selective Brucella blood 3 Total CFU countsc

aBiotyping,bioMérieux.bCycloserinecefoxitinfructoseeggyolk.cFacultativebacteriagrowingontheplateweresubtracted

G+=Gram‐positiv,G‐=Gram‐negative,G±=Gram‐variable

FeiSjöberg

63

DNA-based methods (Papers I–IV)

DNA extraction for T-RFLP and pyrosequencing BacterialDNAwasextractedfrom200μlofthawedduodenalfluidor 160 mg (wet weight) of frozen feces using the QIAamp DNAStool Mini Kit (Qiagen, Hildens, Germany). The DNA yield wasincreased by adding additional steps [151], in which four glassbeads(3.0mmindiameter)and0.5gofzirconiabeads(0.1mmindiameter)wereaddedtothesamplesuspendedinASLbuffer.Thesuspensionwashomogenizedat5m/sectwicefor30secusingaFastPrepCelldisrupter (Savant Instruments,NY,USA), incubatedat 95°C for 5min, and shaken at 1,200 rpm (IKAVibrax Shaker;Labortechnik, Germany) for 30 min at 5°C. DNA was quantifiedspectrophotometrically(NanoDropTechnologies,Wilmington,DE,USA).

T-RFLP (Papers I, II, and IV) Terminal‐Restriction Fragment Length Polymorphism (T‐RFLP)wasperformedonsalivasamplesandfecalsamples.

PCRamplificationofbacterial16SrRNAgenesBacterial 16S rRNA genes were amplified using the universalprimersENV1(5′‐6‐FAM‐AGAGTTTGATIITGGCTCAG‐3′,E.colint8–27)andENV2(5′‐CGGITACCTTGTTACGACTT‐3′,E.colint1511–1492) [152]. The forward primer was labeled withfluorescent6‐FAMatthe5′‐end.ThePCRmixturecontained50ngDNA (25 ng and 75 ng were also tested; 25 ng DNA generatedsmallerT‐RFpeaks), 25μl ofHot StartTaqMasterMix (Qiagen),0.3μMofprimer,1mMMgCl2,andwateraddedtoafinalvolumeof50μl.Forpure‐culturestrains, the finalvolumewas25μlandthe reagent concentrationswere50%of those used for the fecalsamples. The PCR was performed in an Eppendorf Mastercyclergradient (Eppendorf, Hamburg, Germany) using the followingprogram: an initial activation of Taq polymerase at 95°C for 15min, 25 cycles (28 and 32 cycleswere also tested) of 94°C for1min, 50°C for 45 s, and 72°C for 2min,with a final extension at72°C for 7 min. Negative controls that contained all the DNAextraction reagents but that lacked bacteria were included


64

throughoutthewholeanalysis,todetectDNAcontaminationfromthereagentsorinstruments.Foreachsample,threePCRswereruninparallel.

RestrictionenzymedigestionPurifiedPCRproductsfromfecalsamples(100ng)weredigestedwith 16 U (4, 10, and 20 U were also tested) of the MspIendonuclease(NewEnglandBiolabs,Ipswich,USA)at37°Cfor5hin a final volumeof5μl. Forpure‐culturebacteria,30ngofPCRproductsweredigestedwith16UMspI inafinalvolumeof10μl.Thereactionwasstoppedbyheatingat65°Cfor20min.

FragmentanalysisFluorescently labeled T‐RFs were separated and detected usingthe ABI PRISM 310 genetic analyzer (Applied Biosystems,California, USA)with the Pop 4 gel. The resolution (number andquality of peaks) was compared using different volumes ofdigestedPCR amplicons (0.5, 0.7, 1, 1.5, 2.5, 3.8, and4.8 μl), runvoltages(9.9,12.2,and15kV), injectiontimes(5sand10s)andsize standards (ROX 500, ROX1000, combined ROX 500 andROX1000, LIZ 600, and LIZ 1200; Applied Biosystems).Reproducibilitywasinvestigatedbyrunningsamplesinduplicate.

The fragment lengths were analyzed using GeneMapper ver. 4(AppliedBiosystems)andtheLocalSouthernMethod.’True’peakswereseparatedfrom‘noise’withinafragmentlengthrangeof28–1000 bp, according to Abdo et al. [153]. Briefly, the data werestandardizedbydividing the areaof eachpeakby the total peakareaof theparticular sample.The standarddeviation (SD)of thedatasetwasthencomputedassumingthatthetruemeanwaszero.Peaks with an area three standard deviations greater than themeanwereconsideredtobetruesignals.Theywerecollectedandremoved;thereafter,theprocesswasiterateduntilnomore‘true’peakswereidentified.T‐RFsthatdifferedbynomorethan±1bpindifferentanalyseswereconsideredtobeidentical.

AllthefecalsampleswereanalyzedinthreeT‐RFLPanalyses.Onlythose fragments that were represented in two out of three runswereconsidered,andtheirpeaksizesandareaswereaveraged.

FeiSjöberg

65

Pyrosequencing (Papers III and IV) DNAextractionsof the fecal samples andduodenal fluid sampleswere pyrosequenced on a 454 Life Sciences Genome SequencerFLX instrument (GATC Biotech AG, Konstanz, Germany). Aproofreading polymerase was used for PCR amplification of theV1‐V3 region of the bacterial 16S rRNA genes using the forwardprimer27F(5′‐AGAGTTTGATCCTGGCTCAG‐3′)andthereverseprimer534R(5′‐ATTACCGCGGCTGCTGG‐3′).ThePCRprogramusedwasas follows:98°C for30s, followedby25cyclesof98°Cfor10s,56°Cfor30s,and72°Cfor10s,andafinalextensionat72°Cfor1min.ThePCRproductswerecolumn‐purified.AsecondPCRwasperformedtoaddthesequencingadaptorsandmultiplexidentifiers.MultiplexidentifiersareshortDNAsequencesthatareused to identify the different samples. The conditions for thesecondPCRwere thesameas thosedescribedabove,except thatsteps2–4wereonlyperformed for fivecycles.ThePCRproductswere column‐purified, quantified using the Qubit system(Invitrogen, CA, USA), and all the samples were pooled. Prior tosequencingontheGSFLX,thepoolofampliconswasgel‐purifiedand the DNA was recovered with a standard PCR and gelpurification kit. After quantification with the Qubit system, thelibrarywassequencedontheRocheGSFLXfollowingthestandardprotocolsfromRoche.

SequenceanalysisTherawsequencereadswereanalyzedusingthesoftwarepackageQuantitative Insights into Microbial Ecology (QIIME), employingdefault parameters for each step [154], except where specified.The raw reads were assigned to samples according to theirbarcodes. The singletons and possible chimeric sequences wereremovedbefore the subsequent analysis.Only highquality readsthathadaminimumqualityscoreof25andareadlengthof200–500baseswereretainedforfurtheranalysis.

Operational Taxonomic Units (OTUs) were chosen based onrepetitive sequences derived by sequence clustering using theUCLUSTsoftware[155],with aminimumpairwiseidentityof97%.Therepresentativesequence(readsofthesameDNAsequence)ofeachOTUwaschosenbasedonabundanceandlength,andaligned


66

usingPyNAST(apython‐basedimplementationofNASTinQIIME).The taxonomy of each representative OTU was subsequentlyassigned using the Ribosomal Database Project (RDP) classifiernomenclaturemethod [63]. Finally, an abundance table for eachOTUwasgeneratedfromthefrequencyreadcount.

Statistical analysis OrthogonalPartialLeastSquares(OPLS,SIMCAP+V13;UmetricsAB,Umeå,Sweden)wasused toobtain theoverallpatternof themicrobiota.OPLSisapredictionandregressionvarietyofprincipalcomponentanalysiswhich finds information in theXdata that isrelated toYvariables.ThepresenceorabsenceofeachOTUwasusedastheX‐variables,ratherthanthenumberofreadsperOTU.Y represents a diagnostic group, e.g., children with ulcerativecolitisorCrohn’sdisease,thediseasedcontrolsinPaperIII,orthechildrenwhoseparents“clean”theirpacifierbysuckingonitvis‐à‐vischildrenwhoseparentsdidnothavethishabit(PaperI).

TheT‐RFLPanalysisofthefecalsamplesfromsixinfantscollectedover the first year of life was depicted by principal componentanalysis.

Univariate analyses were performed to confirm the unequaldistributionof theOTUs that showed the strongest contributionsin the respective OPLS models used to compare the diagnosticgroups. Univariate analyses were performed by using Fisher’sexacttests(SPSS;IBMCorp.,NY,USA)inPaperIII.

One‐wayANOVAwasperformedtocompare theoveralldiversityat thephylumandgenus levelsbetween thediagnosticgroups inPaperIII(GraphpadPrism5;SanDiego,USA).

FeiSjöberg

67

RESULTS AND COMMENTS

Pacifier cleaning practices and allergy development in infants In Paper I, the effects of pacifier cleaning practices on thecompositionoftheinfantoralmicrobiotaandallergydevelopmentwere studied. In total, 184 children in the ALLERGYFLORA birthcohortwerestudied.Overall,80%ofthechildrenhadatleastoneallergicparent,andtheseinfantswerethereforeconsidered"high‐risk"with respect to allergy development. Almost all the infantswere breast‐fed for at least 4months, and 74% (N=136) used apacifierduringtheirfirst6monthsoflife.

All the infantswereexamined forsensitization(IgEantibodies tocommonallergens)andclinicalallergyat18and36monthsofage.At 18 months of age , 25% had eczema and 5% had asthma.Sensitizationtofoodantigensoccurredin15%oftheinfants,whilesensitization to inhaled antigens was uncommon (see Table 3,PaperI).

When the child was 6months old, the parents participated in atelephone interview that was designed to find out how theycleaned the pacifier used by the child. Several alternativeresponseswereallowed.

AsshowninTable1,almostalltheparentscleanedthepacifierbyrinsingitintapwater,althoughapproximately50%oftheparentsalso boiled the pacifier and almost 50% reported ‘cleaning’ thepacifierbysuckingonitbeforegivingitbacktotheinfant.


68

Table 1: Pacifier cleaning practices among the 136 infants who used a pacifier

during the first 6 months of life.

Technique No. of children

%

Rinsing in tap water 113 83

Boiling 74 54

Parental sucking of the pacifier 65 48

Not reported

Total

13

136

10

100

Techniques for cleaning the pacifier reported in a telephone interview when the child was 6 months of age.

Pacifiercleaningpracticesandallergyat18monthsofageThepresenceofeczema,asthmaoratopyat18monthsofagewasrelated to pacifier use and pacifier cleaning practices. Figure 1shows the risks of developing eczema, asthma, and sensitization.AnOddsRatio(OR)of1denotesthattheriskisashighasthatforinfants who did not use a pacifier. Pacifier use per se was notsignificantly associated with clinical allergy or sensitization(Figure 1). Boiling the pacifierwas associatedwith an increasedprevalence of asthma, although the effect was not significant.However, parental ‘cleaning’ of the pacifier by sucking on itwasstronglyassociatedwithreducedrisksofeczemaandasthma.

Figure1:Therelationshipbetweenatopyorallergy inchildrenat18monthsofageand inrelationtopacifieruseanddifferentpacifiercleaningpractices.OddsRatiosaregiven forpacifiervsnopacifieruse.Forchildrenusingapacifier the

FeiSjöberg

69

way of “cleaning” the pacifier was compared to those not using that specificcleaningmethod.ns=notsignificant.

Bloodsamples,whichweretakeninconjunctionwiththeclinicalinvestigationat18monthsofage,wereanalyzedforIgEantibodiesto common environmental allergens (sensitization) and bloodeosinphilic granulocytes, a cell type whose blood concentrationmaybeelevatedinallergicchildren.Sensitizationtendedtobelessprevalent among infantswhoseparents "cleaned" thepacifier bysucking on it (Figure 1), and these infants also had significantlyfewer blood eosinophils at 18months of age than children whoused a pacifier but whose parents did not have the habit ofcleaningitbysuckingonit(Figure2).

Figure 2: Blood eosinophilcountsat18monthsofageforpacifier‐using children whoseparents ‘cleaned’ theirpacifierby sucking it compared tothose whose parents did notsuckthepacifier

It isdifficult todistinguishallergic asthma fromwheezingdue tothe frequency of respiratory infections in young childen. Weanalyzedwhether respiratory infections in the first 6months ofchildhood were related to pacifier cleaning practices; infectionswererecordedby theparents inadiaryandreportedduring thetelephoneinterviewcoductedat6monthsofage.Theprevalence


70

ofrespiratoryinfectionsinthefirst6monthsoflifedidnotdifferbetween children whose parents sucked their pacifier (1.4) andotherpacifier‐usingchildren(1.5;ANOVA,p=0.75).

AnalysisofconfoundingfactorsConfounders, such asparental educationordeliverymode, couldbe responsible for the observed associations between pacifiercleaningpracticeandallergydevelopment.Theparents'habitsofcleaningthebaby'spacifierwereanalyzedinrelationtoarangeoflifestyle factors. Parents of vaginally delivered infants weresignificantlymorelikelytosuckthechild’spacifierthanparentsofbabiesdeliveredbyCesareansection(Fisher’sexacttest,p=0.02).Conversely,motherswith a University level degree tended to belesslikelytosuckthechild’spacifier,althoughthisassociationwasnotsignificant(seeTable4,PaperI).

Logistic regression analysiswasperformed to adjust fordeliverymode andmother’s occupation. The protective effect of parentalpacifiersuckingoneczemadevelopmentduringthechild’sfirst18monthsofliferemained(OR0.27,95%confidenceinterval0.086–0.819; p=0.02). The protective effect on asthma could not beanalysedinthiswayduetothelownumberofcases.

DeliverymodeandparentalpacifiercleaningDeliveryvia the vaginal route exposes the infant to thematernalvaginal microbiota, while parent‐infant pacifier‐sharing exposesthe child to the parental oral microbiota,We examinedwhetherthese exposures conferred additive protective effects bystratification of the cohort into three groups: (1) vaginallydelivered infants with pacifier‐sucking parents; (2) Cesarean‐delivered infantswhoseparents did not suck their pacifiers; and(3) infantswhowereeithervaginallydeliveredorexposedtotheparents’ oral microbiota by pacifier sharing. The prevalence ofeczemaininfantsat18monthsofagewascomparedbetweenthethreegroups (Figure3).Thegroup thatwasexposed toboth thematernal vaginal and parental oral microbiota had the lowestprevalence of eczema (20%), whereas the infants who wereexposed to neither the maternal vaginal nor parental oralmicrobiotahadthehighestprevalenceofeczema(54%).

FeiSjöberg

71

Thegroupofchildrenwhowereeithervaginallydeliveredorwhoseparentssuckedtheirpacifiershadanintermediateprevalenceofeczema(31%)(Figure3).

Figure3:Prevalencesofeczemaat18monthsofageinthecohortstratifiedas:(1)vaginallydeliveredinfantswithpacifier‐suckingparents(blackbar,N=59);(2)infantswhowereeithervaginallydeliveredorexposedtotheparents’oralmicrobiotabypacifiersharing(hatchedbar,N=48);and(3)Cesarean‐deliveredinfantswhoseparentsdidnotsuckontheirpacifier(openbar,N=13)

Parentalpacifiersuckingandallergyduringthechild’sfirst36monthsoflifeThechildreninthecohortwerealsofollowedupat36monthsofage.Kaplan–Meiercurveswerecalculatedforeczema,asthma,andsensitization(seeFigure4PaperI).Developmentofeczemaupto36 months of age was significantly less likely in those infantswhoseparentssuckedontheirpacifiersduringtheirfirst6monthsof life, as compared with other pacifier‐using children (p=0.04).The protective effect of parental pacifier sucking on asthma andsensitizationwasnotstatisticallysignificant.

The children in the cohort also underwent clinical examinationsfor allergy at 8 years of age, and the results have recently beensummarized. The preliminary results suggest that infants whoseparents"cleaned"theirpacifierbysuckingonitduringthechild'sfirst 6 months of life have a significantly lower prevalence ofasthma and reduced bronchial hyperreactivity (measured by themetacholine test) than infants who used a pacifier that was not"cleaned"inthisway(BillHesselmar,personalcommunication).


72

Parentalpacifiersuckingandsalivarymicrobiotaofchildrenat4monthsofageIf theparents’ salivarymicrobes are transferred to the infantviathepacifier,thismayaffectthedevelopmentoftheinfant’ssalivarymicrobiota.

Salivary samples had been obtained from 64 of the infants at4months of age and had been stored frozen.DNAwas extractedfrom these samples andanalyzedbyT‐RFLP toobtain anoverallpicture of the microbiota. Salivary samples from Cesarean‐delivered infants were excluded, since it is known that deliverymodeaffectsoralmicrobiotacomposition [156]and, furthermore,delivery mode and pacifier cleaning mode were related in ourcohort (see above). After exclusion of Cesarean‐delivered infantsand non‐pacifier users, 33 infants remained, 21 of whom hadparentswhosuckedontheirpacifier.

Figure 4 shows an O‐PLS (orthogonal projection onto latentstructures) analysis of the T‐RFLP results. O‐PLS is a regressionvariety of principal component analysis that can be used toseparate two groups (in this case, pacifier‐using infants whoseparentsdid,ordidnot,"clean"thepacifierbysuckingonit)basedonaverylargesetofvariables.ThepresenceorabsenceofeachT‐RF(usuallyrepresentingabacterialgenus,seePaperII)wasusedas the set of X‐variables. In all, 845 T‐RFswere identified in thegroup of infants. As seen in the figure, the two groups of infantscould be separated based on the salivary T‐RFLP patterns. Oneinfant (outlier, bottom left) had a microbiota that was clearlydistinctfromtheothers.AsidentificationofT‐RFLPpeaksrequiresa substantial amount of additional work, we do not, at present,knowtheidentityofthemicrobesthataremoreorlessprevalentinthesalivasamplesofthetwogroupsofinfants.

FeiSjöberg

73

Figure 4: The oral microbiota of children at 4 months of age in relation toparents'habitofcleaningthepacifier.Eachsymbolrepresentsone infantand itspositionisbasedonthetotalityofbacterialtaxa(genus/species)inthesaliva,asgenerated by T‐RFLP (digestion of total bacterial DNA and separation of thefragments).Thedatawereanalyzedbyorthogonalprojectiontolatentstructures,which is a variation of principal component regression analysis,with parentalpacifier sucking as the Y‐variable, and the totality of bacterial taxa as the X‐variables.The firstcomponent,representing the largestvariationsamong theX‐variables, is represented along the horizontal axis. The second component,depictedalong theverticalaxis, represents the second largestvariationsamongtheX‐variables.


74

Development of a database for the identification of T-RFs In theory, it should be possible to deduce the identity of abacteriumthatgeneratesafragmentofacertainsize,i.e.,aT‐RFintheT‐RFLPmethod,sincetherestrictionsitesareknownanditispossible to calculate the sizes of the fragments generated afterdigestionofthePCRproductsfromamplificationofthe16SrRNAgeneofthatbacterialtaxon.

However, this has proven to be practically impossible.Wenotedthat the relationship between fragment size and migration timewasnot linear(seeFiguresupplementS1 inPaper II),andthis isalso evident from Table 2, in which the theoretical "in silico"digestionresultsarecomparedwiththeobservedapparentsizes.Thelargerthefragment,thegreaterthedeviation,whichmakesitimpossible to introduce any automatic correction factor thatwouldmakeitpossibleto"translate"directlythemigrationtimetoT‐RFsize.

Table 2: Type‐strain in silico digestion and T‐RFLP analysis for comparison of the fragment size discrepancies between true and observed T‐RFs.

Type strain In silico Observed Discrepancy

T‐RF size (nucleotide)

Lactobacillus vaginalis NCTC 12197T 29 30 1

Bacteroides uniformis ATCC 8492T 97 91 ‐6

Clostridium lituseburense NCIMB

10716T 195 188 ‐7

Clostridium ramosum ATCC 25582T 295 286 ‐9

Clostridium difficile NCTC 12197T 457 448 ‐9

Citrobacter braakii CDC 80‐58T 496 485 ‐11

Lactobacillus fermentum ATCC 25582T 581 570 ‐11

Furthermore, to judge the sizes of the T‐RFs, a molecular sizestandardhastoberuninparallel.Severalmolecularweight(MW)standardswere tried, and Table 3 shows that, depending on the

FeiSjöberg

75

MW standard, the T‐RFs generated by digestion of two pure‐culturebacterial strains (Bacteroides andE.coli )anda longPCRfragmenthaddifferentapparentsizes.

Table 3: T‐RFLP results for a sample with three fragments run in combination with different size standards.

Theseresultsclearlydemonstratethat it is inpracticeimpossibleto identifythespecies/genusbasedon insilicodigestion. Instead,theactual apparent size (in relative to the chosenMWstandard)must be investigated in real life, using pure‐culture isolates thatare run, one at a time, and subsequent examination of thegeneratedT‐RFs.

To construct a database,we analyzed a large set of pure‐culturestrainsbyT‐RFLP.AdistinctadvantageofthisexercisewasthatitallowedidentificationoftheconditionsthatgenerateasingleT‐RFafter digestion. A problemwith themethod is the generation of"pseudofragments", i.e., fragmentsthatappearifacleavagesiteisnot used by the restriction enzyme, generating a longer thanexpectedfragment.Iftoolittlerestrictionenzymeisused,thiscanposeasignificantproblem.

Despitetheseprecautions,weobservedpseudofragmentsonsomeoccasions.ForsamplesinwhichBacteroideswerepresentinveryhigh numbers, fragments were visualized that could not beattributed to any taxon, and we concluded that they werepseudofragmentscreatedbyincompletedigestion(PaperII).

Internal size standard used Apparent T‐RF size of the

bacterial DNA (nucleotides)

Rox 500 93 487 0

Rox 1000 93 486 858

Rox 500 & 1000 94 487 857

Liz 1200 92 484 878


76

The databasewas developed by using pure‐culture isolates fromall the groups of gut bacteria that were isolated by culture themicrobiota of the ALLERGYFLORA cohort were used,complemented by strains obtained from the culture collection ofUniversityofGothenburg(CCUG)straincollection(42).Lastly,weperformedcloningandsequencingon9fecalsamples,whichwerefound to contain a high number of T‐RFs that could not beidentified.

Using thismethod, a database that consisted of 130 specieswasconstructed,whichenabledidentificationofthebacteriabasedontheapparentsizesoftheirT‐RFs.Throughtheuseofthisdatabase,approximately 60% of the peaks found in infant fecal samplescouldbedetermined.T‐RFLPidentifiedbacteriatothegenuslevelandalsoprovidedsomesub‐genusdiscrimination.

Infant microbiota development InPaper II, longitudinally sampled fecal samples fromsix infantswere analyzed by T‐RFLP. It is well known that infants acquiresuccessively more and more bacterial species, and that thecomplexity of themicrobiota increaseswith time [50]. However,whenweanalyzedtheearlyinfantfaecalsamplesusingT‐RFLP,aninterestingpictureemerged. Inmostof the infants(Figure5),weobserved rather high bacterial diversity, as characterized in thefecal samples at 1 week and 2 weeks of age, while microbiotacomplexity decreased thereafter, to reach a nadir at around 1month of age. Thereafter, complexity increased markedly onceagain, with the highest level of complexity detected in the lastsamplecollectedat12monthsofage.Thispatternwasalsoseeninthetwoinfantanalyzedwithpyrosequencing(PaperIV).

Thehighmicrobiotalcomplexityofthesamplesobtainedat1and2weeksofagewasunexpected.Figure6showsanexampleoftheT‐RFLPpatternofasingleinfantwhowasfollowedusingrepeatedsamplingofthefeces.Asshowninthefigure,severaloftheT‐RFsthatwerefoundintheearlyfecalsampleswerenotpresentinthemicrobiota after the first weeks of life and did not reappearsubsequently.

FeiSjöberg

77

Figure5.Microbiotacomplexityasafunctionofageinsixchildren. Thenumberoftaxadetectedineachsampleisdepictedbyabar;eachinfantisrepresentedbysixbars,oneforeachsamplingoccasion(at1,2and4weeksandat2,6and12monthsofage).

Figure6:T‐RFLPanalysisof faecalmicrobiotacomposition inan infant followedduringthefirstyearof life.BacterialDNAwasextractedfromfaecalsamplesobtainedat1and2weeks of age and at 1, 2, 6 and 12months of age and analyzed by T‐RFLP. Each peakrepresents a terminal restriction fragment of a specific length that corresponds to abacterialphylotype,usuallyagenus.Adatabasewasconstructed fromT‐RFLPanalysisofknown bacterial species or cloned and sequenced bacterial 16s rRNA genes, and thisdatabasewasused to identified individualT‐RFs in faecal samples.T‐RFno.denotes thenumberofT‐RFs identified in the sample.a thenumbersofT‐RF thatcouldbe identifiedusingthedatabase.TheinfantdepictedcorrespondstoinfantFinpaperII.


78

Anoverallpictureof themicrobiotalcomposition,asanalyzedbyT‐RFLP,wasobtainedbyprincipalcomponentanalysis(Figure7).Here,eachsamplewaspositionedbasedontheT‐RFsfoundinthatsample. In general, samples thatwere obtained between 1weekand 6months of age (samples 1–5) showed clustering by infant,suggesting thateach infanthadadistinctmicrobiota. In contrast,allthe12‐monthsamples(A6–F6)andthe6‐monthsamplesfromtwochildren(A5andF5)appearedinvicinityofoneanotherbutfar from the earlier samples. Thus, the later samplesweremoresimilar tooneanother than theywere to theearly samples fromtherespectiveinfants.Thisconvergenceagreeswiththeresultsofapreviousstudy[157].

Figure7:Microbiotadevelopmentover the firstyearof life forsix infants.Fecalsamples from six infants (A–F)obtainedon six separateoccasionsover the firstyear of life were analyzed by T‐RFLP, and the T‐RF pattern was analyzed byprincipal component analysis. Sampling schedule: 1=1 week; 2=2 weeks; 3=1month;4=2months;5=6months;and6=12months.ThepositionofthesampleisdictatedbythetotalityoftheT‐RFspresentinthesample.

FeiSjöberg

79

In Paper IV, we examined by pyrosequencing fecal samples thatwere obtained longitudinally between 1week and 1 year of agefrom six infants. Four of these infants were the same as thoseanalyzedinPaperII,butduetoinsufficientremainingmaterial,wehad to select 2 additional infants (table 4). For inclusion, theselected infants had to have not been consumpted of anyantibioticsduringthefirstyearoflife.

Table 4: Infants and sample in papers II and IV.

Infant Analyzed & called Delivery Breast feeding

Paper II Paper IV mode Exclucive Partial

1 A A Vaginal 4 8

2 Ba Vaginal 0 0

3 C D Vaginal 4 12

4 D E Vaginal 4 0

5 E B Vaginal 4 4

6 F Vaginal 6 7

7 C Vaginal 5 10

8 F Cesarean 4 9 Bold denotes faecal samples were analyzed at age 1 & 2 weeks, 1, 2, 6 & 12 months.

Italicdenotes faecal samples were analyzed at age 1 week and 1 year aDeveloped food allergy

Themostprevalent(foundin4/6infants)bacterialtaxoninfecalsamples at one week and one year by culturing, T‐RFLP, andpyrosequencingareshowedintable5.

At one week of age anaerobic bacteria were less prevalent.However,at1yearofage,manyanaerobicbacteriacouldbefoundin the fecal samples from the infants such as Roseburia,Eubacterium,Faecalibacterium,VeillonellaandRuminococcus.Thisis in accordancewith the results of previous studies, e.g., that ofPalmer and colleagues, who reported that the infant microbiotastartstoconvergetowardanadultprofileby1yearofage[42].


80

Table 5: Most prevalent bacterial taxon in the fecal sample of infant at one week and one year as determined by culture, T‐RFLP and pyrosequencing.

One week One year

Culture T‐RFLP Pyro Culture T‐RFLP Pyro

Bacteroidetes

Bacteroides X X X X X X Parabacteroides X X Bacteroidales X

Porphyromonadaceae X

Alistipes X

Proteobacteria

Enterobacteriaceae X X X X X X

Ralstonia X

Acidovorax X X

Haemophilus X X

Pasteurellaceae X X

Sutterella X X

Acinetobacter X

Actinobacteria

Bifidobacterium X X X X X X

Propionibacterium X

Eggerthella X

Firmicutes

Co‐neg Staph* X X X X

S. aureus X

Gemella X

Clostridium X X X X X

Clostridiales X

Enterococcus X X

Streptococcus X X X

Lactobacillus X X

Lactobacillales X

Coprobacillus X X

Ruminococcus X

Veillonella X X X X

Veillonellaceae X X

FeiSjöberg

81

Selenomonas X

Peptostreptococcaceae X

Eubacterium X

Lachnospiraceae X

Roseburia X

Anaerotruncus X

Blautia X

Butyricicoccus X

Dorea X

Faecalibacterium X

Oscillibacter X

Dialister X

Clostridiaceae X

Verrucomicrobia

Akkermansia X

*Coagulase‐negative staphylococci, based on coagulase test


82

Methodologic considerations-Advantages and drawbacks In total, 40 fecal samples collected from 8 infants who werefollowedfrom1weekto1yearof lifewereanalyzed.Allofthesesamples had previously been analyzed by quantitative culturing,andthesamplesfromsixinfantswereanalyzedusingbothT‐RFLPandpyrosequencing.Thispermitsustocomparetheperformanceofthethreemethodsfortheanalysisofthelargebowelmicrobiotaofinfants.

Detection limits of the three methods In the 1‐week samples, twice as many OTUs were identified bypyrosequencing as were detected by T‐RFLP, which in turnidentifiedtwiceasmanyOTUsasdidquantitativeculturing(Figure7).WithT‐RFLPandpyrosequencing, an increased complexity ofthemicrobiotaoverthefirstyearoflifewasnoted,aphenomenonthatwasnot seenusingquantitative culturing.This supports thenotion that exclusively oxygen‐sensitive anaerobic bacteria settleinthecommensalmicrobiotamainlyduringthesecondhalfofthefirstyearoflife;theculturingconditionsappliedwerenotadaptedto isolateand identify thesebacteria(Table5),as themain focuswas on the early microbiota and on bacterial groups that havesignificant impacts on immune maturation in the infant. It isknown that obligate anaerobes elicit very meager immuneresponses, supposedly because they do not translocate themucosal barrier[90]. At 1 year of life, 45–70 individual generawere identified in each infant by pyrosequencing (Figure 7). Itshouldbenotedthatwhileallsixinfantscontributedfecalsamplesat1weekand1yearofage,weonlyanalyzedthosesamplesthatwereobtainedfromtwoinfantsintheintermediateperiod(2and4weeksand2and6monthsofage).

FeiSjöberg

83

Figure 7: Numbers of taxa inindividual fecal samples frominfants, as identified byquantitativeculturing,T‐RFLP,and pyrosequencing, anddifferences in thecompositionsof the early (1week) and late(1 year) microbiota, asrevealedbypyrosequencing

Quantitative culture (Figure 8a) showed that Bacteroides werepresentinhighnumbers,>109CFU/gfeces,bothinthe1‐weekand1‐year samples. By culturing, it was clearly evident thatEscherichiawas thedominant genus among theProtebacteria (E.colibeingitssinglespecies).However,sequencingofthe16SrRNAgene cannot adequately resolve the different genera of theEnterobacteriaceae family (Escherichia, Klebsiella, Proteus etc.),which makes it impossible to use this type of DNA‐basedmethodology to distinguish clinically important members of theinfant microbiota. In young infants, Klebsiella is initially ascommon as E. coli in the commensal microbiota, although thisgenusandEnterobacteriaceaemembersotherthanE.colibecomeless prevalentwhen themicrobiota "matures" toward the endofthe first year of life. The inability of sequence‐basedmethods todistinguishbetweendifferentmembersof theEnterobacteriaceaeis a distinct disadvantage given the different ecologic niches ofthese bacteria. The primary habitat of E. coli is in thegastrointestinal tracts of humans and many warm‐bloodedanimals wheras Klebsiella, Enterobacter, and non‐E. coliEnterobacteriaceae, are commonly found in nature. Vaginallydelivered infants are, therefore, colonizedbyE.colimuch earlierthan Cesarean‐delivered infants, while the latter more oftenacquire other types of enterobacteria. These differences canreadily be revealed by culturing [49, 118, 119], though not bysequencingmethods.


84

AsshowninFigure8a,quantitativeculturerevealsthatthecountsof Enterobacteriaceae decrease approximately ten‐fold (one logunit)between1weekand1yearofage,whichisaconsequenceofincreased competition caused by the successively more complexmicrobiota and the increasingly anaerobic milieu. In contrast,obligate anaerobes do not decrease in number:bifidobacteria and Bacteroides reamin numerous at 1 week and 1 year of age in infants who harbour the genus in question in their gut microbiota.

Interestingly, Bacteroides were detected equally well by all threemethods. Thus, all the fecal samples that contained Bacteroidesspp., as determined by culturing,were also found to contain theBacteroides genus by pyrosequencing and T‐RFLP. Notably,samples that were positive for Bacteroides contained highpopulation counts of this genus, i.e., >107.9 CFU/g feces. In sharpcontrast,Bifidobacteriumspp.weredetectedbypyrosequencinginallthesamplesthatcontained>106CFUbifidobacteria/gfecesbutwere missed in the three samples with lower bifidobacterialpopulation counts (Figure 8b). T‐RFLP also failed to detectbifidobacteriainthesesamples,aswellasinasamplewithahighcountofbifidobacteria(Figure8c).

FeiSjöberg

85

Figure8:Populationlevelsofthemajortaxainearlyandlatefecalsamplesfrominfants,asrevealedbyquantitativeculturing(a)andtheprobabilitesofdetectingormissingthesetaxainpyrosequencing(b)andT‐RFLP(c)analyses,respectively.


86

We determined the number of CFUs (in theory, each CFUrepresentsonelivecultivatablebacterium)thathadtobepresentina fecalsample for it togenerateonesequenceread(Figure9).We found that for Enterobacteriaceae and Bacteroides, 105–106CFU/gfeceswereneededtogenerateoneread.Thisisclosetothetheoreticalcalculation.

Bacte

roide

s

Bifidob

acte

rium

Enter

obac

teria

ceae

101

102

103

104

105

106

107

108

109

Figure9:NumbersofCFUperreadinsamplesthatweredemedpositivebybothcultureandpyrosequencing

However, the sensitivity for detection of bifidobacteriawas 100‐timeslower:108CFU/ghadtobepresentinordertogenerateonepyrosequencing read. This very low level of sensitivity wasmasked by the fact that the infants' feces contained very highnumbersofthisgenus.It isplausiblethatbifidobacteriaareoftenmissed in the fecal samples from adults, where they are alsopresent in high numbers (109/g) [50], although they are seldomreportedinpapersthatusesequence‐basedmethodologies.

Samples that contained Enterococcus, Staphylococcus andLactobacillus spp (all Gram‐positive bacteria) were oftenmissedby pyrosequencing. This is partly due to the low populations ofthesebacteria,butalsothefactthattheywerefrequentlymissedinthose samples in which they were present in realtively high

FeiSjöberg

87

number. Thiswas evenmore evidentwhenT‐RFLPwasused. Incontrast, all the samples that contained Gram‐negativeEnterobacteriaceae,asdetectedbyculturing,werealsofoundtobepositivebypyrosequencing,eveniftheywerepresentinnumbersthatwerenohigherthanthoseoftheGram‐positivebacteria.

Table 6 combines the data from the two methodology papers(PapersIIandIV),andcomparesthethreeDNA‐basedmethods(T‐RFLP,cloning‐sequencing,andpyrosequencing)withculture.Thesensitivityofcultureisarbitrarilysetat100%,i.e.,amethodthatdetects the genus less often than culturing is assigned a value<100%, while a method that is more sensitive than culture isassigned a value >100%. Some factors that may influence thedetection limitarealso included,suchasthenumberofcopiesofthe16SrRNAgenecarriedbyasinglebacterialcellandtheGram‐stainingstatus.

Table6ComparisonsofthedetectionsensitivitiesofDNA‐basedmethodsrelativetothesensitivityofculturing(setat100%)

Cloning and sequencing showed very low sensitivity for theinvestigated genera. This is understandable, considering the few

Phylum T‐RFLP Pyro Cloning CFU 16S copy no. Gram‐

Taxon %b

median median staining

Bacteroidetes

Bacteroides 100 105 40 9.9 6 G‐ bacil l i

Proteobacteria

Enterobacteriaceae 74 100 10 8.6 7 G‐ bacil l i

Actinobacteria

Bifidobacterium 79 84 0.2 9.9 4 G+ bacill i

Firmicutes

Clostridium 106 38 3 6.0* 9 G+ bacill i

Lactobacillus 67 80 0.1 6.8 6 G+ bacill i

Enterococcus 6 31 0 5.3 4 G+ cocci

Staphylococcus 30 40 0 4.7 5 G+ cocci

b based on the 834 clones obtianed from 9 faecal samles in paper II

* alcohol‐treated faecal sample

detectability (%)a

a compared to the result obtained by culture, 100 %= as good as culture.


88

reads per sample that this method generates, usually in thehundreds. If themicrobiota contains 5×1011 bacteria/g and 500sequencesareobtained,theywill,inprinciple,allbederivedfrombacteriathatarepresentinpopulationlevelsofatleast1010/g,i.e.,somedominant obligately anaerobicbacteria.However, it shouldbenotedthatbifidobacteria,whichclearlyarepresentatsuchhighlevels,arestillnotdetected,onceagaindemonstratingtheinferiorcapacitiesofDNA‐basedmethodstodetectthisgroupofbacteria.Three reasons can be put forward to explain the insensitivity tobifidobacteria: 1) they are Gram‐positive (it is much easier todisrupt the Gram‐negative bacterial cell wall when extractingDNA); 2) they carry relatively few copies of the 16S rRNA gene;and3)PCRprimermismatch.

PCR primer mismatch has an important negative impact ondetection sensitivity [66]. Bifidobacteria belong to theActinobacteriaphylum,previouslyconsideredtobeGram‐positivebacteriawithahighGCcontent[67].Thecommonlyusedforwardprimer8–27(E.coli16Snucletoidepositions)hasbeenshowntocontain threeprimermismatches [62]and is thus inadequate foramplifyingtheDNAofbifidobacteria,leadingtolowsensitivityforthis quantitatively dominant bacterial genus in the gut.Degenerative primers might increase the primer affinity duringPCRby incorporationof inosine (I), as in thecaseof theprimersweusedforT‐RFLP:ENV‐1,5´‐AGAGTTTGATIITGGCTCAG‐3´;and ENV‐2, 5´‐CGG ITA CCTTGT TAC GAC TT‐3´. Normally, thenucleotides in DNA bind as follows: adenine (A) binds only tothymine(T);andcytosine(C)bindsonlytoguanine(G).However,inosinecanbindtoanyoftheabove‐mentionednucleoties. i.e.,A,T,CorG.Analysisof the16SrRNAgenes fromdifferentbacteriahas revealed themost common primermismatch positions [67].Thus, inosine is incrporated at these nucleotide positions in theprimers.Despite thismodification,ourPCRprimercouldstillnotamplify the bifidobacteria DNA with satisfactory sensitivity. Thefact that they were still identified in most samples can beattributedtotheirexceedinglyhighcountsininfantfecalsamples.

AsDNAismoreeasilyreleasedfromGram‐negativebacteria,suchas Bacteroides and Enterobacteriaceae, than from Gram‐positive

FeiSjöberg

89

bacteria, such as Enterococcus and Staphylococcus, we combinedenzymatic lysis with an additional mechanical bead‐beatingmethod to break the bacterial cell wall (Paper IV). However,despite this extra measure to improve the efficacy of DNAextraction,thedetectionsensitivityforGram‐positivebacteriadidnot improve. Collectively, our results suggest that Gram‐negativebacteria (Bacteroides and Enterobacteriaceae) are highlyoverestimated when DNA‐based methodologies are used toanalyzecomplexmicrobiota.

Comparison of the performances of pyrosequencing for analyzing small and large bowel microbiota Pyrosequencingwas used to analyze the duodenal fluid samplesfrom childrenwith IBD and from control children (Paper III), aswellasthefecalsamplesfrominfants(PaperIV).Thesegroupsofsamples were analyzed at the same time and on the samesequencing plate, which permitted direct comparisons of theperformanceofpyrosequencingwiththesedifferentsampletypes.Themeannumbersof sequencereadsdidnotdifferbetween thetwo cohorts: the duodenal samples gave32±1.2 k reads, and thefecalsamplesgave33±7kreads(Figure10a).

Culturingrevealedthattheduodenalsamplescontainedmaximally<106 CFU/g feces, as comparedwith >1011 for the fecal samples.Despitethis,theduodenalfluidgeneratedanaverageof104OTUspersample,ascomparedwithanaverageof43OTUsforthefecalsamples (Figure10b)! It seemshighlyunlikely that theduodenalmicrobiota contains 200‐times more bacterial genera than doesthecolonicmicrobiota,evenifthefecalsampleswerefromyounginfants and the duodenal microbiota was from olderchildren/adolescents whose microbiota could be moremature/complex. Indeed, the colonic microbiota containedmoreT‐RFs than the duodenal microbiota when they were bothanalyzed by T‐RFLP (Figure 10c).Moreover, pyrosequencing didnot show significantly higher numbers ofOTUs at 1week of age(averageof38±8OTUs)thanat1yearofage(54±9OTUs),whichrepresentanincreaseofonly16OTUsintheaverageinfant.


90

Figure 11. Duodenal fluid samples and fecal samples analyzed usingpyrosequencing,T‐RFLP,andquantitativeculturing.

Themajordifferencenotedbetweenthetwosampletypeswasthevast difference in the density of bacteria, with the fecal sampleshaving 5 log units higher counts, asmeasured by culturing, thanthe duodenal samples (medians: 1010 vs. 104.7) (Figure 10d). Ifanything, the colonic counts should be underestimated ascompared to the duodenal counts, as the colonic microbiota ismuch more anaerobic and is thus more difficult to culture. InFigure10e,wehavecalculatedhowmanyCFUcountsareneededto generate one sequence read. In general, the fecal samplesrequired 106 CFU to generate one sequence read, while theduodenalsamplescouldgenerateonereadalreadyat101CFU(10bacteria).ThemostlikelyexplanationforthelowerOTUsdetected

FeiSjöberg

91

in the fecal samples is that the high population levels of certainfecal bacteria hamper the detection of subdominant genera,therebylimitingthedepthofpyrosequencing.

Inthisrespect,T‐RFLPhasacertainadvantage.Despitebeinglesssensitive overall than pyrosequencing, it is more robust whenthere is an unequal distribution of population sizes, since thefragments are separated according to size before they aredetected. Therefore, the detection of, for example, lactobacilli isnothamperedbythepresenceof10,000‐timesmoreBacteroides.


92

Duodenal microbiota and IBD The duodenal fluids of childrenwith newly diagnosed IBDwerecharacterized using high‐throughput sequencing and culture.Wecompared diversity and composition in children with IBD withsymptomatic controls, i.e., children with symptoms compatiblewith IBD and, hence, referred to a pediatric specialist clinic fordiagnosis, but whose gut mucosa displayed no signs ofinflammation (thereby excluding a diagnosis IBD). To control forany effects on microbiota composition, we excluded all childrenwhohadreceiveddisease‐modifyingdrugsorwhohadbeentakingany antibiotics in the previous 3 months. Duodenal fluids wereanalyzedbyquantitativeculturingandpyrosequencing(averageof32,000 reads/sample, average read of 500 nt). Quantitativeculturingrevealedbacterial levelsof>105CFU/mL in26%of thechildren(twopatientswithulcerativecolitisandthreecontrols).

Pyrosequencingshowedthat theduodenalmicrobiotaofchildrenwith ulcerative colitis was less diverse, i.e., contained fewerOperational Taxonomic Units (OTUs), than those of the controls.Although the microbiota of children with Crohn’s disease alsotendedtobelesscomplexthanthoseofcontrols,therewasgreatervariabilityintheCrohn’sdiseasegroupandthedifferencedidnotreachstatisticalsignificance.Thelowerdiversityofthemicrobiotain children with ulcerative colitis was not restricted to anyparticularbacterialgroup,butwasseenacrossthreemajorphyla:Actinobacteria, Bacteroidetes and Firmicutes, included bacterialgenera, e.g., Collinsella (P<.001), Bacillus (P=.007), Lactobacillus(P=.007), and unclassified Sphingobacteriales (P=.007), wereisolated less frequently from the children with ulcerative colitisthan from the controls. Collinsella was also less common inpatientswithCrohn’sdiseasethanincontrolsubjects(P=.007).

Comparedtocontrolsubjects,theduodenalmicrobiotaofchildrenwithulcerativecolitis is lessdiverseandseveralbacterialgroupsarepresentatreducedfrequencies.Whetherthesechangesinthemicrobiotacontributetothediseaseprocessorareaconsequenceoftheinflammatorydiseaseremainstobedetermined.

FeiSjöberg

93

DISCUSSION

The commensal microbiota of the alimentary tract containsenormous amounts of antigens and inflammatogenic substances.Commensal bacteria and fungi translocate across the mucosalmembranes, whereby they come in contact with the immunesystem,which is therebyactivated.The commensalmicrobiota isan important stimulus for the immune system and has beensuggestedtoplayrolesinseveralimmune‐mediateddiseases.Thishas spurred interest in the composition and diversity of themicrobiotaduringinfancyandchildhood,inasearchformicrobesthat contributeeithernegativelyorpositively to theetiologyandpathogenesis of allergy and inflammatory bowel disease.Traditionally, quantitative culture on selective and non‐selectivemediahasbeenusedtomapthemicrobiota.However, in the lastcouple of decades, DNA‐based methodologies have becomestandard.Thesemethodspermitanalysesoffrozensamples,sincetheydonotrequirethatthebacteriaareviable.However,therearefew studies that directly compare culture‐based and DNA‐basedmethodologies in terms of their sensitivity and accuracy incharacterizingcomplexmicrobiota,suchasthatofthealimentarytract. In this thesis, we have compared quantitative culturing offecal samples with the DNA‐based methods of T‐RFLP,pyrosequencing,andcloning‐sequencing.Anunderstandingoftheadvantages and limitations of each method will enable moreaccurate analyses of the microbiota and, thus, a betterunderstanding of its role in allergy and inflammatory boweldisease.

Terminal‐Restriction Fragment Polymorphism (T‐RFLP) exploitsvariationswithin the16S rRNAgenes forbacterial identification.Theubiquitousbacterial16S rRNAgene is amplifiedbyPCRandcleavedwitharestrictionenzyme(oracombinationofenzymesinsomeapplicationsofthemethod).Eachbacterialtaxongeneratesafragment, termed the T‐RF, of defined size. When applied to acomplex microbiota, T‐RFLP provides rapid fingerprinting of abacterial community and themethod is well suited to analyzingbacterial diversity in different environments, such as soil, water,


94

andthegutsofhumansandanimals[158,159].AdisadvantageofT‐RFLPisthatitisnotimmediatelyevidentwhichbacterialtaxongenerates a particular T‐RF. Several web tools are available toidentify T‐RFs [160]. These databases allow calculations of thesizesofT‐RFsfrom16SrRNAgenesoncetheuserentersthePCRprimersandrestrictionenzyme(“insilicodigestion”).However,inpractice, the utilization of theweb‐based tools requires accuratefragment size determination. We show here, as previouslydemonstrated[161],thattherearediscrepanciesbetweenthetrueand observed T‐RF sizes that are inherent to the method. Wedemonstrate here that one source of this discrepancy is thatdifferent fluorescein dyes are used to label the sample and themolecular weight standard, which has also been pointed outpreviously [70]. Moreover, the differences between thetheoretically calculated and apparent fragment sizes vary acrossbacterial taxa,whichmean thata correctionalgorithmcannotbeappliedtocompensateforthediscrepancy.Kaplanetal.[162]havesuggested that thepurinecontentof the16S fragment influencestheT‐RFLPruns,whichmaybeonesourceofthevariationinT‐RFmigration time observed between bacterial taxa. Commonlyavailable T‐RFLP analysis programs do not take all these factorsinto consideration, and our results suggest that this is, indeed,impossible.Therefore,theuseofautomatedT‐RFLPprogramsmayresultinfaultyidentificationofthebacterialtaxainacommunity.To identify the T‐RFs generated from the analysis of the infantfecalmicrobiota,wedevelopedourowndatabase.T‐RFLPanalysesofpure‐culture isolatesofbacteria from infant fecal samples andfromtypestrainsofgutbacteriaformedthebasisfortheanalysis,which was complemented by the cloning and sequencing of alimitednumberofisolatesfromtheinfantfecalsamples.AlthoughweselectedsampleswithmanyunidentifiedT‐RFsforcloningandsequencing, this procedure mostly generated a large set ofBacteroides sequences. The limited sensitivity of the cloning andsequencing procedure means that only quantitatively dominantbacterialpopulationscanbedetected(seebelow).

TheT‐RFLPdatabaseallowedustoidentifydirectlyapproximately60% of the peaks detected in the infant fecal samples. Thus,despiteextensiveefforts, a substantialproportionof the infantile

FeiSjöberg

95

gutmicrobiota still consisted of unidentified bacteria,which is aclear limitation to using the T‐RFLP method for gut microbiotasamples.

Using our database, we showed that T‐RFLP permittedidentification at the genus level inmost cases. Several genera, infact,generatedmorethanonedistinctT‐RF,whichisinaccordancewiththeresultsofapreviousstudy[163].Thus,T‐RFLPhasalmostthe same taxonomic resolution potential as culturing, which bydefinition permits identification of the species, provided thatvarioustypingmethodsareappliedtothepure‐cultureisolates.

Pyrosequencing of moderately long fragments of the 16S rRNAgene (average: 500 nt) andwith an average number of reads of30,000 per sample (today termed ’ultra‐deep sequencing’) wascomparedwithculturingandT‐RFLPfortheanalysisofinfantfecalsamples. As pyrosequencing does not use information on theentire16SrRNAgene,itisimpossibletodeterminewithaccuracythe species of a bacterial taxon. Even identification to the genuslevelmaybeambiguous.Wangetal.calculatedthatwhenusinga400‐ntprimer the chancesof correctly assigning a genus, family,order, class, and phylumwere 88.7%, 94.6%, 97.7%, 99.2% and99.8%, respectively [63]. This fact is seldom taken intoconsiderationinstudiesthatusesequence‐basedmethodsfortheanalysis of complex bacterial communities. As the sequence isaligned to previously published sequences, and since thesesequencesareattributed to species, it is tempting tobelieve thatthe sequence does contain enough information to determineaccurately the species fromwhich it is derived. However, this isnot the case. We chose to use the genus level as the taxonomic"cut‐off" for our pyrosequencing method. However, despite thisconservativeapproach,thereisanapproximately11.3%riskthatasequencewillbeassignedtothewronggenus[63].

Sequencing the entire 16S rRNA gene, as in the cloning andsequencingmethod,inmostcasesgeneratessufficientinformationto assign a species to the sequence in question. However, thismethod is of limited use in analyses of very dense bacterialcommunities, such as the fecal microbiota. As it is immensely


96

labour‐intensive (each DNA fragment must be cloned into abacterium,whichmustsubsequentlybepropagatedandlysedandits DNA subjected to sequencing), the number of DNA fragmentsthat can be analyzed per sample is fairly limited (usually <1000per samples). It is clear that analysis of a random collection of1000bacteriaamongapopulationof1011gives little informationabout the overall community, as only the most quantitativelydominant bacteria are likely to be analysed, for purely statisticalreasons. This was evident in our attempts to identify non‐cultivatablebacteria ina few infant fecal samplesby cloning andsequencing; few bacterial taxa were identified and most of thesequenceswerederivedfromBacteroides.

In some instances, not even sequencing of the entire 16S rRNAgene yields the genus, the reason being the close geneticrelatedness of these bacterial genera. Examples include thestreptococci (which include pneumococci, which are highlyvirulentbacteria,aswellasnumerousspeciesthatarecollectivelyreferredtoasthealpha‐streptococci),andtheEnterobacteriaceaewhich containsmanyof themost clinically relevant gut bacteria,suchasE.coliandKlebsiella,whicharealsohighlyprevalentintheinfant microbiota. DNA‐based differentiation of theEnterobacteriaceaefamilyusuallyrequiressequencingofboththe16Sgeneandadditionalhousekeepinggenes,e.g.,gyrB [164]anddnaJ [165]. In contrast, Escherichia, Klebsiella, Enterobacter,Citrobacter, and Proteus are easily distinguished using culturingandsimplebiotypingtests.Thesespeciesareofclinicalimportanceinthataremajorcausesofurinarytract infectionsandalsoquitecommon causes of sepsis and local infections that emanate fromthegastrointestinaltract.

For some applications, identification to the genus level isinadequate. For example, the clinical and immunologic effects ofbeing colonized by S. aureus are profoundly different from theeffects of being colonized by any of the species collectivelyreferredtoascoagulase‐negativestaphylococci.Whiletheformeris one of themost common invasive pathogens, the latter causeinfections only in immunocompromised hosts, includingpremature infants. Furthermore, S. aureus produces a range of

FeiSjöberg

97

extremely potent T‐cell‐stimulating toxins, the superantigenictoxins, as well as the B‐cell‐superantigenic protein A. We haveshown that early gut colonizationbyS.aureus is associatedwithdecreased risk of developing food allergy within the first 18months of life, and that gut colonization with superantigen‐producing strains is associated with higher serum levels of IgA[166]andsolubleCD14[167]inthecolonizedinfants.

In conclusion, the taxonomic accuracies of T‐RFLP andpyrosequencing are similar, i.e., they permit, in most cases,taxonomicidentificationtothegenuslevel.Whenamoredetailedtaxonomicdistinctionisrequired,thesemethodsareinadequate.

Apartfromtheidentificationofspecificbacterialtaxainacomplexmicrobial community, it is often of interest to determine thegeneral complexity of the microbial community, i.e., how maybacterial taxa it contains. Thus, the feature that is most oftenassociatedwithprotectionagainstallergydevelopmentistohavemany different bacterial groups making up the early microbialmicrobiota [122, 124, 125, 156, 168].The determination ofmicrobiotal complexity of is by no means straightforward. It isoftenreportedthatthousandsofindividualtaxaarefoundinfecalsamplesanalyzedbyDNA‐basedmethods,suchaspyrosequencing.However, in our hands, pyrosequencing of infant fecal samplesdetected on average only 43 OTUs per sample and a mere 126uniqueOTUs inall20 infant fecalsamplescombined,despiteour"ultra‐deep" sequencing rate of 30,000 reads per sample andaveragelengthof500nucleotides.

Althoughthemicrobiotaofaninfantislesscomplexthanthatofanadult, it is difficult to reconcile our resultswith thosepreviouslyobtained,whichsuggeststhatthousandsofOTUsaregeneratedbysequencing one microbiota from one individual. A distinctpossibility is that previousDNA‐based studies overestimated thenumbersofdifferenttaxa[30]andfalsely identifiednoveltaxa inthe gut flora due to methodologic errors [54, 66]. A number ofpitfallsarelinkedtopyrosequencing,andfaultsmaybeintroducedat several steps, i.e., the PCR step, sequencing step, andalingement/annotationstep.PCRhasaninherentimperfection, in


98

that roughly 1.2x10‐6/PCR cycle of the bases is incorrectlyincorporated [79].AnotherproblemwithPCR is the formationofchimeras, which occurs when two DNA sequences from twodifferent bacteria become amplified to generate a single PCRproduct,which isachimeraof thetwobacterialsequencesbut isinterpreted as being from “a new species”, the chimera ratewasmore than 70% of the less‐abundant specieswere chimera [31].Thus,extremecaremustbetakentoavoidtheappearanceoffalsesequencesthatmaycreateafalseimpressionofthewidediversityof a bacterial community. Classical culture‐based studiesemployingthebestavailablestrictanaerobicculturemethodologyin the 1970s revealed that about 200 distinct species could beidentifiedinoneindividual[169].Asmicroscopicevaluationofthefecalsamplerevealedanumberofbacterialcelltypesthathadnotappeared in the culture‐based analysis, it was assumed that themeticulous culturingapproachhad succeeded in culturinghalfofthe bacterial species present, and the estimate of 400 species inthe intestinal microbiota of an adult individual was proposed[169]. It should be noted that this culture‐based approachidentified facultative bacteria that are normally missed by DNA‐basedmethodologies,suchasenterococci.Thenumberoftaxathatmay be identified by sequencing alone is likely to be lower than400.

A very interesting observation was made by comparing thenumber of OTUs detected by pyrosequencing of the duodenalmicrobiota(fromchildrenwithIBDandage‐matchedcontrols),ascomparedwiththemicrobiotaofthefecalsamples(frominfants).Quiteremarkably,anaverageof100OTUswasfoundperduodenalsample,whichismorethantwiceasmanyaswereidentifiedinthecolonicmicrobiota.Itisveryunlikelythattheduodenumcontainsagreatervarietyofbacterial taxa than thecolon, the latterbeingregardedasoneofthemostcomplexmicrobialecosystemsknown.Indeed, T‐RFLP identified 2.5‐times more T‐RFs in the averagefecalsamplethanintheaverageduodenalfluidsample.Themostlikely explanation for this is that the lownumbers of bacteria inthe duodenal fluidmeant that theDNA from almost every singlebacterialcellcouldbeanalyzed(30,000readswouldcover3×104bacteria, which is a typical population level in the duodenum).

FeiSjöberg

99

These findings provide a clear indication that pyrosequencing,despitebeinglabeledas“ultra‐deep”,onlyskimsthesurfaceofthecomplex colonic microbiota, identifying quantitatively thedominant groups, while missing quite a high number of sub‐dominanttaxa.

Themodest number of different taxa detected in our studymayreflect the fact that the gut microbiota of Swedish infants is,indeed, quitemeagre, due to prolonged exclusive breast‐feeding,small family size, and high standards of hygiene. We havepreviously reported delayed colonization of Swedish infants, ascompared with Pakistani infants, by typical commensal gutbacteria[118],andsluggishturnoverofindividualE.colistrainsinthe gut microbiota [119]. These findings suggest that infants inaffluentWesterncountriesareexposedtolowmicrobialloads,andthat this contributes to low microbiota complexity. To compareadequately theperformanceofpyrosequencing inelucidating thecomplexityof theduodenalandcolonicmicrobiota,samples fromthesameindividualsmustbeanalyzed.Suchstudiesareunderway,as we have corresponding fecal samples for all the duodenalsamplesfromchildrenwithIBDandtheircontrols.

We and other groups have shown that culture is substantiallymoresensitivethanDNA‐basedmethodsforthedetectionofbothobligateandfacultativeanaerobicbacteria,providedthatselectivemedia and appropriate culturing conditions are available [170][69].Culture‐basedanalysiscandetectcultivatablebacteriadownto population levels of 102.5 bacteria/g feces, while our studysuggests that the current sequencing techniques detect fecalbacteria when they are present at ≥106 CFU/g sample, asapproximately 1 in 105–106 fecal bacteria yielded onepyrosequencingread.Thisisinaccordancewithapreviousreport[171] and the theoretical calculations showing that at least100,000 sequences must be analyzed to obtain one read of abacteriumpresentatalevelof106CFU/gwithinatotalpopulationof1011CFU/g.Thus,thehighpopulationlevelsofcertainbacteria,which are readily detected by PCR and sequencing, e.g.,Bacteroides, hamper the detection of subdominant less‐dominantgenera.Thus,muchofthedepthofpyrosequencingis“wasted”on


100

the masses of sequences obtained from quantitatively dominantgenera,suchasBacteroides.

The fact that bacterial genera that are present in relatively lownumbersaremissedbypyrosequencingwasreadilydemonstratedbycomparingtheresultsfromtheanalysesoffecalsamplestakenfrom 1‐week‐old and 1‐year‐old infants. OPLS analysis basedsolely on the pyrosequencing data showed that Staphylococcus,Enterococcus,andLactobacillusweredetectedmore frequently inthe 1‐week samples than in the 1‐year samples. In fact,staphylococci and enterococci were present at 1 year of age inpracticallyalltheinfants,andlactobacilliweremoreprevalentat1year than at 1week of age. However, as their population countsdecreased, concomitant with increased competition from themicrobiotaoverthefirstyearoflife,thenumbersofthesebacteriafellbelowthedetectionlimitofpyrosequencingand“disappeared”fromthemicrobiotalanalysis.

ThereisageneralbeliefthatDNA‐basedmethods,especiallynext‐generationsequencing,areable to reveal “all” thebacteria in thegutmicrobiota. As shown in our studies and as discussed above,this is a misconception; all methods have certain limitations.Culture is, by definition, substantially more sensitive than anyother method, provided that selective media and appropriateculture conditions are available. This includes all facultatives,which may easily be isolated from among billions of obligateanaerobes using the simple technique of culturing the sampleunder aerobic conditions, which few of the gut anaerobes cantolerate.Asmanyfacultativebacteriahaveimportantfunctionsasimmunemodulatorsorpathogens,webelievethat it is importanttopointoutthatsuchbacteriaoftenescapedetectionwhenDNA‐based methods are employed. Therefore, in future studies,combining both culture‐based and DNA‐based methods withgreater sequencing depth may pave the way for a betterunderstandingofboththebacterialcompositionandphysiologyofthemicrobiotalivingwithinus.

Our interest in microbiota analysis derives from the hypothesisthattheinfantilemicrobiotaisamajorstimulusforthedeveloping

FeiSjöberg

101

immunesystemandthatmicrobiotalcomplexityandcompositionmay affect the risk of allergy development. Indeed, a lowcomplexityof theearlymicrobiota isarisk factor forsubsequentallergydevelopment.InPaperII,analysisoftheT‐RFLPpatternsofinfant fecal samples yielded an interesting picture, which, to thebest of our knowledge, has not been reported before. Inmost ofthe infants, a quite high bacterial diversity characterized the 1‐weekand2‐week fecal samples, althoughmicrobiotal complexitydecreased thereafter to reach a nadir at around 1month of age.Thereafter, the complexity of themicrobiota increasedmarkedlyagain, being highest in the last sample collected at 12months ofage. It is well known thatmicrobiotal complexity increases overthefirstyearoflifeasmoreandmoreanaerobicbacterialspeciesbecome established in the gut [50]. However, the high bacterialcomplexityofthesamplesobtainedat1and2weeksofagewasanunexpected finding. Several bacterial species that were found intheearlyfecalsamplesdisappearedfromthemicrobiotaafterthefirstweeks and did not reappear. Some of thesemay have comefrom the birth canal, as all the infants examined here werevaginally delivered, or from the environment. Such non‐professional gut colonizersmay have a chance to expand duringthefirstweeksbuttheyarenotabletowithstandthecompetitionas more “professional” gut bacteria are acquired. Furthermore,theymaybemoresensitivetotheantibacterialpeptidesproducedin themucosa in response to colonization.Previous reportshaveshownthatbacteriathatoriginatefromthematernalvaginalfloracanbeculturedfromthebaby'sgastriccontentsimmediatelyafterbirth [47], although many of these bacteria cannot colonize theinfant'sgut.

InPaperI,wedemonstratethatacommonparentalpractice‐oralcleaning of their infant’s pacifier before it is given back to theinfant, is associated with protection against early eczemadevelopment and asthma symptoms. Early acquisition of acomplex intestinal microbiota has been identified as providingprotection against allergy development, which suggests thatcommensalbacteriaprovide thedeveloping immunesystemwithcrucialsignalsforitspropermaturation[122,123,168][125].Wepropose that the underlying mechanism involves immune


102

stimulationbymicrobes,whicharetransferredtothebabyviathepacifier. Undoubtedly, this habit allows for close oral contactbetween the parents and child, facilitating bacterial transfer at avery young age, before the child starts to use feeding utensils.Parentswhodonot suck their infant’s pacifiermayhave amorecautious style of livingwith respect to bacterial transfer to theirchildren. Using T‐RFLP to characterize the infant’s salivarybacteria,we have gathered evidence that this practice influencesthecompositionofthe infant’soralmicrobiota.Atpresent,wedonot know the identities of the microbes that are more or lessprevalentinthesalivaofthetwogroupsofinfants,asdevelopmentofanoralbacteriadatabasefortheidentificationofthesemicrobesrequires substantial additional work (cloning and sequencing).TheT‐RFLPpeakidentificationdatabasedevelopedinPaperIIwasdesignedprimarilytoidentifybacteriainfecalsamplesandisnotsuitable for oralmicrobiota identification.We have attempted toconstruct an oral bacteria database, but this database at presentcontainsonly50speciesandthisisclearlynotsufficienttoidentifymorethanasmallfractionofthevastnumberofbacterialgenerapresentintheoralcavity.

Oral tolerance isactively induced immunologicunresponsivenesstoharmlessproteins,and ithasbeenknownfordecadesthat thepresence of commensalmicrobiota improves oral tolerance [97].Hitherto, attention has focused almost exclusively on the smallintestineasthesiteofinductionoforaltoleranceandtheinfluenceofthegutmicrobiotaonthehandlingofdietaryantigens.However,the oral cavity is actually the first site of encounters betweenforeign antigens and the lymphoid system, and it allows thelymphoidsystemtosampleantigensthathavenotbeendenaturedby acid or degraded by digestive enzymes. The oropharynx issurroundedbydense lymphoid tissues, i.e., the adenoids and thepalatine and pharyngeal tonsils. Similar to the gut‐associatedlymphoid tissues, thesearecoveredwithMcells,whichspecializein antigen uptake and delivery to antigen‐presenting cells andlymphocytesbeneaththeepithelium[172].Furthermore,dendriticcells capture antigens in the oral mucosa and migrate to thecervico‐mandibular lymph nodes [173], similar to the way inwhichdietaryantigensarecarriedtothemesentericlymphnodes.

FeiSjöberg

103

The tonsils are rich in T cells that have a regulatory phenotype[87],andapplicationsofcontactallergenstotheoralmucosaleadto active tolerance induction [174]. Thus, there is no reason tobelievethatactiveoraltoleranceisnotinducedalreadyintheoralcavity. Exposure of the infant to parental salivamight acceleratethedevelopmentofacomplexoral/pharyngealmicrobiota,which,similar toa complexgutmicrobiota [122,123,168] [125],mightbeneficially affect tolerogenic handling of antigens by theoral/pharyngeal lymphoid tissues. Moreover, oral bacteria areswallowed, thereby also affecting the composition of themicrobiota in the small intestine, which may in turn regulatetolerance development in the gut. Further studies are nowrequired to establish if parental pacifier sucking represents asimpleandsafemethod toreduceallergydevelopment in infantsandyoungchildren,asourstudysuggests.

In Paper III, the duodenal fluids of childrenwith new‐onset IBDwere characterized using pyrosequencing and culturing. Thesechildrenhadnotreceivedeitherantibioticsordisease‐modulatingdrugs.TheduodenalmicrobiotaofchildrenwithIBDwasdistinctlydifferent from that of diseased control children. A potentialweaknessofourstudyisthelimitednumberofcollectedsamples.However, thispermittedin‐depthsequenceanalysis,withameannumber of 32,000 sequence reads per sample. Thus, the presentstudy represents the most comprehensive examination of IBD‐associatedmicrobiotaDNAsequencesperformedtodate,includingprevious studies of the colonic/fecal flora in patients with IBD[150]

Inparticular,a lower levelofdiversitywas found tocharacterizetheduodenalmicrobiotaofchildrenwithIBD.Astherehavebeenno previous studies of the duodenal microbiota of patients withIBD, comparisons are difficult. In dogs with spontaneous IBD,duodenal biopsies contained fewer species than those collectedfrom healthy dogs, and the percentages of Fusobacteria,Bacteroidaceae, Prevotellaceae, and Clostridiales were lower.Instead, their duodenal microbiota contained more members ofthe Proteobacteria phylum, e.g., Enterobacteriaceae,Diaphorobacter,andAcinetobacter[175][176].


104

The lowercomplexityof themicrobiota inchildrenwithIBDalsomeant that a number of bacterial genera (e.g., Collinsella,Sphingobacteriales,Enterococcus, andLactobacillus)werepresentatlowerfrequenciesintheirduodenalfluids.Collinsella,whichisaGram‐positive anaerobe that belongs to the Actinobacteriumphylum,waspresentinreducednumbersinboththechildrenwithulcerativecolitisandthechildrenwithCrohn’sdisease.Thisgenusconsistsoffourspecies,withC.aerofaciensbeingfrequentlyfoundin the human large bowel [177] [178], albeit at lower levels inpatientswithsevereirritablebowelsyndrome[179].Thenumbersof Collinsella have previously been reported to be lower in thebowelmicrobiota of persons in theWestern hemisphere than inpeople from Asia and Africa [180] [181], and Collinsella is alsofound at lower frequencies in populations at high risk for coloncancer than in low‐risk populations [182]. It remains to beelucidatedwhetherCollinsellahasadirectprotectiveeffectonguthealth and whether this genus is particularly sensitive to thealteredconditionsprevailinginpatientswithIBD.

AnOTUthatbelongstotheSphingobacterialesorderofthephylumBacteroideteswasalsodetectedlessfrequentlyintheIBDpatients,particularlyinthosewithulcerativecolitis.ThisOTUcouldnotbeassigned to any genus, family or class. The nameSphingobacteriales alludes to the production of sphingolipids byspecieswithin thephylumBacteroidetes [183]. SphingolipidsarerecognizedbyinvariantnaturalkillerT(iNKT)cellsandmayexertimmunomodulatoryeffects[184].InvariantNKTcellsareimportant regulators of immune system. Animal studies have shown aprotective effect of inhibitory sphingolipids against experimentaliNKTcell‐mediated,oxazolone‐inducedcolitis [185].Lactobacilluswas also detected more frequently in the diseased controls.Lactobacilliarewidelypresentinfoodproducts,suchasyoghurts,and various probiotic lactobacilli have been shown to haveimmune‐stimulating properties [186]. These bacteria are alsocommonly found in the oral [15], small intestinal, and colonicmicrobiota.

Adrawbackofourstudyisthatwelackedduodenalfluidsamplesfrom healthy children. Esophagogastro‐duodenoscopy is a

FeiSjöberg

105

challenging intervention that cannot be performed on healthychildren for purely research purposes. In addition, in a study ofIBDconductedindogs,ethicalconcernsprecludedtheinclusionofhealthy control animals [176]. The few studies of the humanduodenal microbiota that have been carried out have focusedprimarily on specific bacterial groups in a limited number ofdiseases,i.e.,smallintestinalbacterialovergrowth,irritablebowelsyndrome, and celiac disease. These studies have been reviewedbyWang [187],who concludes that little is known regarding theduodenalmicrobiota, either in healthy individuals or in patients.Ourdiseasedcontrolsshowedtheanticipatedtypicalhealthysmallintestinalflora,whichisdominatedbythegeneraofLactobacillus,Enterococcus,andStreptococcaceaeandcomprisesmembersofallfourdominantintestinalphyla.Withthesequence‐basedanalysis,we demonstrate that the microbiota of control children showgreater diversity across all phyla and contain more OTUs thatcannot be assigned to any known bacterial phylum or that caneven be classified as bacterial in origin, than the microbiota ofchildrenwithulcerativecolitis.

Hopefully, our studies canbeused to improvemethodologies forthe analysis of the intestinal microbiota of infants and youngchildren,astheseadvancesareneededtohighlighttheimportanceofmicrobiotalcomplexityandcompositionforthedevelopmentofimmunoregulatory dieases. such as allergy and inflammatorybowel disease. Although the hygiene hypothesis fits with allavailable epidemiologic data,we still do not know exactlywhichtypesofmicrobialstimuliarerequiredtoprovidetheappropriatematuration signals to the developing immune system so that itdevelopsfunctionaltolerancetoharmlessantigens.


106

ACKNOWLEDGEMENTS

”Nog finnsdetmålochmening ivår färd ‐mendetärvägen,somärmödanvärd”DennadiktavKarinBoyesstämmersåvälmed den lyxiga tiden som doktorand. Jag har vuxit enormt sompersonochhaftroligtunderhelamindoktorandperiod,aldrighaftenendamåndagsångest.Mångaharbidragittilldennaarbetsglädjeochhjälptmiggenomföradettaarbete! TUSEN TACK till er alla!

Särskilttacktillallafamiljer,barn, läkareochsjuksköterskorsomdeltog i forskningstudieernapåMölndalsSjukhus resp.DrottningSilviasBarnsjukhus.

Ia ochAgnes för att ni skyddarmigunder era vingaroch guidatmig genomhela arbetet! Ia för din visdom, snällhet och rättvisa,dindörräralltidöppenförmig.Agnesdetfinnsingenhandledaresomdu,duharalltiduppmuntratmigochfårmigattsesakerfrånandra perspektiv. Jag har fått fria händer att jobbameddet somintresseratmig.

Foroughsomfickmigattdoktoreraoch lärdemigalltdukunde.EdforthethousandtimesIbargedintoyourofficewithhundredsofquestions,everytimeyoutaughtmesomethingnew.Vincent,Ireallyappreciateyourprofessionalhelpineditingmythesis.

DamernaEva, Ingela, Jolanta ‐ tack förattni finnsutanerhadedet inte gått! Eva för din klokhet, Ingela för din positiva livssyn,Jolanta (min psykolog) för alla komplimanger och personligaterapisamtal.

Sofiadetfinnssåmångaordmendeträckernogmed:jagtyckersåmycket om dig. Anna‐Carin for alla fina komplimanger, menegentligen så jag tycker du är en ”Natural Beauty”. Hardis förtrevligsällskapiTaos.ErikaLsomalltidärgladochpositiv.NahidKaramiforalladiskussionerochplasmidsamarbete.DulärdemigNanodropp även under ditt avhandlingstid, det säger så mycketom dig som person. Eva J,Merna och Nahid Kondori för allatrevligaluncherochsällskapunderåren.TackaallaklokaPost‐doc.Anna och Jenny.Madeleine för din prisbelönta avhandling och

107

andra förstklassiga föreläsningar som du delat med dig.Åsa fördinagedignakunskaper iPrism. Ignacio chefen!Föralla trevligasamarbeten, alla mina dumma frågor och dina nyanserade svar.Tidigaredoktorander&kollegorAnnaÖ,AnnaS,Cillan,EricaB.Ni är underbara som förgyllde vardagen, för all hjälp och alladiskussionerunderåren.

Liselott för all hjälp under åren, allt från lån av kemikalier tilllöneförhandling. Personalen på CCUG, Kent, Kristel, Maria,SusannochElisabethförallastammarsomjagfickochdetkännsalltidvälkomnadeattkommaned till er.Sofia som lärdemigalltom sekvensringskonst.GunnarDahlen ochLisbeth tack för alltsamarbete,nilärdemigsåmycketommunflora.RobertSaalmanförallaintressantadiskussioneromIBD.

Alla trevliga och snälla läkare på våning 6,Annika, Susann ochBrynja.Lisaförkurssällskapetochdinomtänksamhet,Bodil föralla trevliga svampplockningsturer, fina blombilder och allabotaniska diskussioner. Inger Mattsby för alla diskussionermellanhimlenoch jord samtditt intresse förkinesiskkulturochmat.ChristineWenneråsförattdualltidharvaritsåtrevligsamtdinafinafärgersomtröstatminaögon.

Alladoktorandkollegor,Johannasomharfraktatprovianttillmigända från Kina. Jag kan skriva flera rader om hur snäll, trevlig,hjälpsamochklokJohannaär,menframföralltärhonenängelmotsinapatienter.Kicki förduär så roligochklok, Johan förattdustårutmedmina”lille”trakasserier,egentligenvilljagbarafådigatt le. Maria Bankvall för allt samarbete med RAS, för dintålmodighet,noggrannhetochdinstarkaviljaatt läradigT‐RFLP.HuameiochLiföralltkinesisktsnackochtrevligtlunchsällskap.

谢谢父母的养育之恩！妈谢谢您让我上最好的英文学校。您一直都是我的好榜

样！

MinafinabarnBellaochFeliciasomärtröstochglädjeimittliv.MintryggemanMikael,somvillkorslöstställtuppochstöttatmiggenomhelaresan,utandighadejagaldrigkommitsåhärlångt!


108

REFERENCES

1. ParteAC:LPSN‐‐listofprokaryoticnameswithstandingin nomenclature. Nucleic Acids Res 2014, 42(Databaseissue):D613‐616.

2. Book. In: International Code ofNomenclature of Bacteria:Bacteriological Code, 1990 Revision. Edited by Lapage SP,Sneath PHA, Lessel EF, Skerman VBD, Seeliger HPR, ClarkWA.Washington(DC);1992.

3. Pace NR: Amolecular view ofmicrobial diversity andthebiosphere.Science1997,276(5313):734‐740.

4. Ludwig W, Strunk O, Westram R, Richter L, Meier H,Yadhukumar,BuchnerA,LaiT,SteppiS,JobbGetal:ARB:asoftware environment for sequencedata.NucleicAcidsRes2004,32(4):1363‐1371.

5. Hugenholtz P: Exploring prokaryotic diversity in thegenomicera.Genomebiology2002,3(2):REVIEWS0003.

6. OrenA,GarrityGM:Thenandnow:asystematicreviewof the systematicsofprokaryotes in the last80years.AntonieVanLeeuwenhoek2014,106(1):43‐56.

7. Rossello‐Mora R, Amann R: The species concept forprokaryotes. FEMSmicrobiology reviews 2001, 25(1):39‐67.

8. Fraser C, Alm EJ, Polz MF, Spratt BG, Hanage WP: Thebacterial species challenge: making sense of geneticand ecological diversity. Science 2009, 323(5915):741‐746.

9. Johnson JR: Shigella and Escherichia coli at thecrossroads:machiavellianmasqueradersortaxonomictreachery?JMedMicrobiol2000,49(7):583‐585.

10. Murray RG, Stackebrandt E: Taxonomic note:implementation of the provisional status Candidatusfor incompletely described procaryotes. Internationaljournalofsystematicbacteriology1995,45(1):186‐187.

11. Stackebrandt E EJ: Taxonomic parameters revisited:tarnishedgoldstandards.MicrobiolToday2006,33:152–155.

109

12. Kim M, Oh HS, Park SC, Chun J: Towards a taxonomiccoherence between average nucleotide identity and16S rRNA gene sequence similarity for speciesdemarcation of prokaryotes. International journal ofsystematic and evolutionary microbiology 2014, 64(Pt2):346‐351.

13. Bergey's Manual of Systematic Bacteriology, vol. 2:Williams&Wilkins;1986.

14. Lemon KP, Earl AM, Vlamakis HC, Aguilar C, Kolter R:Biofilm development with an emphasis on Bacillussubtilis. Current topics in microbiology and immunology2008,322:1‐16.

15. DewhirstFE,ChenT,IzardJ,PasterBJ,TannerAC,YuWH,LakshmananA,WadeWG:Thehumanoralmicrobiome.JBacteriol2010,192(19):5002‐5017.

16. JenkinsonHF,LamontRJ:Oralmicrobialcommunitiesinsickness and in health. Trends in microbiology 2005,13(12):589‐595.

17. Human Microbiome Project C: Structure, function anddiversity of the healthy human microbiome. Nature2012,486(7402):207‐214.

18. AasJA,PasterBJ,StokesLN,OlsenI,DewhirstFE:Definingthenormalbacterial floraof theoral cavity. Journalofclinicalmicrobiology2005,43(11):5721‐5732.

19. Savage DC: Microbial ecology of the gastrointestinaltract.AnnuRevMicrobiol1977,31:107‐133.

20. Bik EM, Eckburg PB, Gill SR, Nelson KE, Purdom EA,Francois F, Perez‐Perez G, Blaser MJ, Relman DA:Molecular analysis of the bacterialmicrobiota in thehuman stomach. Proc Natl Acad Sci U S A 2006,103(3):732‐737.

21. OuG,HedbergM,HorstedtP,BaranovV,ForsbergG,DrobniM,SandstromO,WaiSN,JohanssonI,HammarstromMLetal: Proximal small intestinal microbiota andidentification of rod‐shaped bacteria associated withchildhood celiac disease. The American journal ofgastroenterology2009,104(12):3058‐3067.


110

22. ChengJ,KalliomakiM,HeiligHG,PalvaA,LahteenojaH,deVos WM, Salojarvi J, Satokari R: Duodenal microbiotacomposition and mucosal homeostasis in pediatricceliacdisease.BMCgastroenterology2013,13:113.

23. Finegold SM: Normal human intestinal flora. Annalidell'Istitutosuperioredisanita1986,22(3):731‐737.

24. O'Hara AM, Shanahan F: The gut flora as a forgottenorgan.EMBOreports2006,7(7):688‐693.

25. EckburgPB,BikEM,BernsteinCN,PurdomE,DethlefsenL,SargentM,GillSR,NelsonKE,RelmanDA:Diversityofthehuman intestinal microbial flora. Science 2005,308(5728):1635‐1638.

26. Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI,KnightR:Bacterialcommunityvariationinhumanbodyhabitats across space and time. Science 2009,326(5960):1694‐1697.

27. Yatsunenko T, Rey FE, Manary MJ, Trehan I, Dominguez‐BelloMG,ContrerasM,MagrisM,HidalgoG,BaldassanoRN,AnokhinAPetal:Humangutmicrobiomeviewedacrossageandgeography.Nature2012,486(7402):222‐227.

28. Finegold SM, Attebery HR, Sutter VL: Effect of diet onhuman fecal flora: comparison of Japanese andAmericandiets.TheAmerican journalof clinicalnutrition1974,27(12):1456‐1469.

29. Finegold SM SV, Mathisen GE.: Human intestinalmicroflora in health and disease‐normal indigenousintestinalflora,vol.3–31.London:AcademicPress;1983.

30. QinJ,LiR,RaesJ,ArumugamM,BurgdorfKS,ManichanhC,NielsenT,PonsN,LevenezF,YamadaTetal:Ahumangutmicrobialgene catalogue establishedbymetagenomicsequencing.Nature2010,464(7285):59‐65.

31. Haas BJ, Gevers D, Earl AM, Feldgarden M, Ward DV,GiannoukosG,CiullaD,TabbaaD,HighlanderSK,SodergrenE et al: Chimeric 16S rRNA sequence formation anddetection in Sanger and 454‐pyrosequenced PCRamplicons.GenomeRes2011,21(3):494‐504.

32. HayashiH,SakamotoM,BennoY:Phylogeneticanalysisofthe human gut microbiota using 16S rDNA clone

111

librariesandstrictlyanaerobicculture‐basedmethods.MicrobiolImmunol2002,46(8):535‐548.

33. Duncan SH, Louis P, Flint HJ: Cultivable bacterialdiversityfromthehumancolon.LettApplMicrobiol2007,44(4):343‐350.

34. Zoetendal EG, Rajilic‐Stojanovic M, de Vos WM: High‐throughput diversity and functionality analysis of thegastrointestinal tract microbiota. Gut 2008,57(11):1605‐1615.

35. ReyesA,HaynesM,HansonN,AnglyFE,HeathAC,RohwerF, Gordon JI: Viruses in the faecal microbiota ofmonozygotic twins and their mothers. Nature 2010,466(7304):334‐338.

36. KononenE:Developmentoforalbacterialflorainyoungchildren.AnnMed2000,32(2):107‐112.

37. RotimiVO,DuerdenBI:Thedevelopmentofthebacterialflorainnormalneonates.JMedMicrobiol1981,14(1):51‐62.

38. McCarthyC, SnyderML, ParkerRB:The IndigenousOralFloraofMan. I.TheNewborn to the1‐Year‐Old Infant.Archivesoforalbiology1965,10:61‐70.

39. Kononen E, Asikainen S, Jousimies‐Somer H: The earlycolonization of gram‐negative anaerobic bacteria inedentulous infants. Oral microbiology and immunology1992,7(1):28‐31.

40. Socransky SS, Manganiello SD: The oral microbiota ofmanfrombirthtosenility.Journalofperiodontology1971,42(8):485‐496.

41. Bradshaw DJ, Marsh PD, Watson GK, Allison C: Role ofFusobacterium nucleatum and coaggregation inanaerobe survival in planktonic and biofilm oralmicrobial communities during aeration. Infect Immun1998,66(10):4729‐4732.

42. Palmer C, Bik EM, DiGiulio DB, Relman DA, Brown PO:Developmentofthehumaninfantintestinalmicrobiota.PLoSBiol2007,5(7):e177.


112

43. TapiainenT,YlitaloS,EerolaE,UhariM:Dynamicsofgutcolonization and source of intestinal flora in healthynewborninfants.Apmis2006,114(11):812‐817.

44. AzadMB,KonyaT,MaughanH,GuttmanDS,FieldCJ,ChariRS, SearsMR,BeckerAB, Scott JA,KozyrskyjALetal:Gutmicrobiota of healthy Canadian infants: profiles bymode of delivery and infant diet at 4months. CMAJ :Canadian Medical Association journal = journal del'Associationmedicalecanadienne2013,185(5):385‐394.

45. BiasucciG,RubiniM,RiboniS,MorelliL,BessiE,RetetangosC:Modeofdeliveryaffects thebacterialcommunity inthe newborn gut. Early human development 2010, 86Suppl1:13‐15.

46. BrookI,BarrettCT,BrinkmanCR,3rd,MartinWJ,FinegoldSM: Aerobic and anaerobic bacterial flora of thematernal cervix and newborn gastric fluid andconjunctiva: a prospective study. Pediatrics 1979,63(3):451‐455.

47. Dominguez‐BelloMG, Costello EK, ContrerasM,MagrisM,HidalgoG,FiererN,KnightR:Deliverymodeshapes theacquisition and structure of the initial microbiotaacrossmultiple body habitats in newborns. Proc NatlAcadSciUSA2010,107(26):11971‐11975.

48. PendersJ,ThijsC,VinkC,StelmaFF,SnijdersB,KummelingI,vandenBrandtPA,StobberinghEE:Factorsinfluencingthe composition of the intestinalmicrobiota in earlyinfancy.Pediatrics2006,118(2):511‐521.

49. AdlerberthI,StrachanDP,MatricardiPM,AhrneS,OrfeiL,AbergN,PerkinMR,TripodiS,HesselmarB,SaalmanRetal:Gutmicrobiotaanddevelopmentofatopiceczemain3 European birth cohorts. J Allergy Clin Immunol 2007,120(2):343‐350.

50. Adlerberth I, Wold AE: Establishment of the gutmicrobiota in Western infants. Acta Paediatr 2009,98(2):229‐238.

51. El Aidy S, Van den Abbeele P, Van de Wiele T, Louis P,Kleerebezem M: Intestinal colonization: how keymicrobialplayersbecomeestablished in thisdynamic

113

process: microbial metabolic activities and theinterplay between the host andmicrobes. BioEssays :news and reviews inmolecular, cellular and developmentalbiology2013,35(10):913‐923.

52. Ellis‐PeglerRB,CrabtreeC,LambertHP:Thefaecalfloraofchildren in the United Kingdom. J Hyg (Lond) 1975,75(1):135‐142.

53. Adlerberth I. HLÅ, Wold AE., : Development of thegastrointestinal tract chapter 17 Ontogeny of theintestinal flora. Ontario, Canada: B. C. decker inc.Hamilton,Ontario;1999.

54. Lozupone CA, Stombaugh J, Gonzalez A, Ackermann G,WendelD,Vazquez‐BaezaY,JanssonJK,GordonJI,KnightR:Meta‐analyses of studies of the human microbiota.GenomeRes2013,23(10):1704‐1714.

55. ElberseKE,NunesS,Sa‐LeaoR,vanderHeideHG,SchoulsLM: Multiple‐locus variable number tandem repeatanalysis for Streptococcus pneumoniae: comparisonwithPFGEandMLST.PLoSOne2011,6(5):e19668.

56. Flint HJ, Duncan SH, Scott KP, Louis P: Interactions andcompetition within the microbial community of thehuman colon: links between diet and health. EnvironMicrobiol2007,9(5):1101‐1111.

57. AdamsE:Studiesingramstaining.Staintechnology1975,50(4):227‐231.

58. Morrison WR, Smith LM: Preparation of Fatty AcidMethyl Esters and Dimethylacetals from Lipids withBoronFluoride‐‐Methanol.Journalof lipidresearch1964,5:600‐608.

59. Seng P, Drancourt M, Gouriet F, La Scola B, Fournier PE,RolainJM,RaoultD:Ongoingrevolutioninbacteriology:routine identification of bacteria by matrix‐assistedlaser desorption ionization time‐of‐flight massspectrometry.ClinInfectDis2009,49(4):543‐551.

60. Mullis K, Faloona F, Scharf S, Saiki R, Horn G, Erlich H:Specific enzymatic amplification of DNA in vitro: thepolymerasechainreaction.ColdSpringHarbor symposiaonquantitativebiology1986,51Pt1:263‐273.


114

61. Woese CR: Bacterial evolution. Microbiological reviews1987,51(2):221‐271.

62. Frank JA, Reich CI, Sharma S, Weisbaum JS, Wilson BA,Olsen GJ:Criticalevaluationof twoprimers commonlyusedforamplificationofbacterial16SrRNAgenes.ApplEnvironMicrobiol2008,74(8):2461‐2470.

63. Wang Q, Garrity GM, Tiedje JM, Cole JR:Naive Bayesianclassifier forrapidassignmentofrRNAsequences intothenewbacterialtaxonomy.ApplEnvironMicrobiol2007,73(16):5261‐5267.

64. Acinas SG, Marcelino LA, Klepac‐Ceraj V, Polz MF:Divergenceandredundancyof16SrRNAsequences ingenomes with multiple rrn operons. J Bacteriol 2004,186(9):2629‐2635.

65. UedaK,SekiT,KudoT,YoshidaT,KataokaM:Twodistinctmechanisms cause heterogeneity of 16S rRNA. JBacteriol1999,181(1):78‐82.

66. Jumpstart Consortium Human Microbiome Project DataGeneration Working G: Evaluation of 16S rDNA‐basedcommunityprofiling forhumanmicrobiome research.PLoSOne2012,7(6):e39315.

67. Farris MH, Olson JB: Detection of Actinobacteriacultivated fromenvironmentalsamplesrevealsbias inuniversal primers. Lett ApplMicrobiol 2007, 45(4):376‐381.

68. LiuWT,MarshTL,ChengH,ForneyLJ:Characterizationofmicrobialdiversitybydeterminingterminalrestrictionfragmentlengthpolymorphismsofgenesencoding16SrRNA.ApplEnvironMicrobiol1997,63(11):4516‐4522.

69. Sjoberg F, Nowrouzian F, Rangel I, Hannoun C, Moore E,Adlerberth I, Wold AE: Comparison between terminal‐restriction fragment length polymorphism (T‐RFLP)and quantitative culture for analysis of infants' gutmicrobiota.JMicrobiolMethods2013,94(1):37‐46.

70. Hahn M, Wilhelm J, Pingoud A: Influence of fluorophordye labels on the migration behavior of polymerasechainreaction‐‐amplifiedshorttandemrepeatsduring

115

denaturing capillary electrophoresis. Electrophoresis2001,22(13):2691‐2700.

71. Rajilic‐Stojanovic M, Heilig HG, Molenaar D, Kajander K,Surakka A, Smidt H, de Vos WM: Development andapplication of the human intestinal tract chip, aphylogenetic microarray: analysis of universallyconserved phylotypes in the abundant microbiota ofyoung and elderly adults. Environ Microbiol 2009,11(7):1736‐1751.

72. Wilson KH, Wilson WJ, Radosevich JL, DeSantis TZ,Viswanathan VS, Kuczmarski TA, Andersen GL: High‐density microarray of small‐subunit ribosomal DNAprobes.ApplEnvironMicrobiol2002,68(5):2535‐2541.

73. GroupNHW, Peterson J, Garges S, GiovanniM,McInnes P,WangL,SchlossJA,BonazziV,McEwenJE,WetterstrandKAet al:TheNIHHumanMicrobiomeProject.GenomeRes2009,19(12):2317‐2323.

74. Walker AW, Duncan SH, Louis P, Flint HJ: Phylogeny,culturing, and metagenomics of the human gutmicrobiota.Trendsinmicrobiology2014,22(5):267‐274.

75. Spits C, Le Caignec C, De Rycke M, Van Haute L, VanSteirteghem A, Liebaers I, Sermon K: Optimization andevaluation of single‐cell whole‐genome multipledisplacement amplification. Human mutation 2006,27(5):496‐503.

76. McLean JS, Lombardo MJ, Ziegler MG, Novotny M, Yee‐GreenbaumJ,BadgerJH,TeslerG,NurkS,LesinV,BramiDet al: Genome of the pathogen Porphyromonasgingivalis recovered from a biofilm in a hospital sinkusingahigh‐throughputsingle‐cellgenomicsplatform.GenomeRes2013,23(5):867‐877.

77. RoeschLF,FulthorpeRR,RivaA,CasellaG,HadwinAK,KentAD, Daroub SH, Camargo FA, Farmerie WG, Triplett EW:Pyrosequencing enumerates and contrasts soilmicrobialdiversity.ISMEJ2007,1(4):283‐290.

78. Schloss PD: The effects of alignment quality, distancecalculationmethod, sequence filtering, and region on


116

the analysis of 16S rRNA gene‐based studies. PLoSComputBiol2010,6(7):e1000844.

79. PatinNV,KuninV,LidstromU,AshbyMN:EffectsofOTUclustering and PCR artifacts on microbial diversityestimates.Microbialecology2013,65(3):709‐719.

80. QuinceC, LanzenA, CurtisTP,DavenportRJ,HallN,HeadIM, Read LF, Sloan WT: Accurate determination ofmicrobialdiversityfrom454pyrosequencingdata.NatMethods2009,6(9):639‐641.

81. BomarL,MaltzM,ColstonS,Graf J:Directedculturingofmicroorganismsusingmetatranscriptomics.mBio2011,2(2):e00012‐00011.

82. Renesto P, Crapoulet N, Ogata H, La Scola B, Vestris G,ClaverieJM,RaoultD:Genome‐baseddesignofacell‐freeculturemediumforTropherymawhipplei.Lancet2003,362(9382):447‐449.

83. KaramiN,MartnerA,EnneVI,SwerkerssonS,AdlerberthI,Wold AE: Transfer of an ampicillin resistance genebetween two Escherichia coli strains in the bowelmicrobiota of an infant treated with antibiotics. TheJournal of antimicrobial chemotherapy 2007, 60(5):1142‐1145.

84. Adlerberth I,HuangH, Lindberg E, AbergN,HesselmarB,Saalman R, Nord CE, Wold AE, Weintraub A: Toxin‐producingClostridiumdifficilestrainsaslong‐termgutcolonizers in healthy infants. J Clin Microbiol 2014,52(1):173‐179.

85. Brandtzaeg P:Functionofmucosa‐associated lymphoidtissue in antibody formation. Immunologicalinvestigations2010,39(4‐5):303‐355.

86. Kasahara K, Matsumura Y, Ui K, Kasahara K, Komatsu Y,Mikasa K, Kita E: Intranasal priming of newbornmicewithmicrobial extracts increases opsonic factors andmatureCD11c+ cells in theairway.American journal ofphysiology Lung cellular and molecular physiology 2012,303(9):L834‐843.

87. Siegmund K, Ruckert B, Ouaked N, Burgler S, Speiser A,Akdis CA, Schmidt‐Weber CB: Unique phenotype of

117

human tonsillar and in vitro‐induced FOXP3+CD8+ Tcells. Journal of immunology (Baltimore,Md :1950)2009,182(4):2124‐2130.

88. Crabbe PA, Bazin H, EyssenH, Heremans JF:ThenormalmicrobialfloraasamajorstimulusforproliferationofplasmacellssynthesizingIgAinthegut.Thegerm‐freeintestinal tract. Int Arch Allergy Appl Immunol 1968,34(4):362‐375.

89. Steffen EK, Berg RD, Deitch EA: Comparison oftranslocationratesofvariousindigenousbacteriafromthegastrointestinaltracttothemesentericlymphnode.JInfectDis1988,157(5):1032‐1038.

90. Berg RD: Bacterial translocation from thegastrointestinaltract.AdvExpMedBiol1999,473:11‐30.

91. Macdonald TT, Monteleone G: Immunity, inflammation,and allergy in the gut. Science 2005, 307(5717):1920‐1925.

92. Carter PB, Pollard M: Host responses to "normal"microbial flora in germ‐free mice. Journal of theReticuloendothelialSociety1971,9(6):580‐587.

93. Ostman S, Rask C, Wold AE, Hultkrantz S, Telemo E:ImpairedregulatoryTcell function ingerm‐freemice.Europeanjournalofimmunology2006,36(9):2336‐2346.

94. Von Bultzingslowen I, Adlerberth I, Wold AE, Dahlen G,Jontell M: Oral and intestinal microflora in 5‐fluorouracil treated rats, translocation to cervicalandmesenteric lymph nodes and effects of probioticbacteria. Oral microbiology and immunology 2003,18(5):278‐284.

95. Pabst O, Mowat AM: Oral tolerance to food protein.Mucosalimmunology2012,5(3):232‐239.

96. Karlsson M, Lundin S, Dahlgren U, Kahu H, Pettersson I,Telemo E: "Tolerosomes" are produced by intestinalepithelialcells.EurJImmunol2001,31(10):2892‐2900.

97. Moreau MC, Corthier G: Effect of the gastrointestinalmicroflora on induction and maintenance of oraltolerance toovalbumin inC3H/HeJmice. Infect Immun1988,56(10):2766‐2768.


118

98. Koropatkin NM, Cameron EA, Martens EC: How glycanmetabolism shapes the human gutmicrobiota.NaturereviewsMicrobiology2012,10(5):323‐335.

99. Wong JM, de Souza R, Kendall CW, Emam A, Jenkins DJ:Colonic health: fermentation and short chain fattyacids.Journalofclinicalgastroenterology2006,40(3):235‐243.

100. Louis P, Flint HJ: Diversity,metabolism andmicrobialecologyofbutyrate‐producingbacteriafromthehumanlargeintestine.FEMSmicrobiologyletters2009,294(1):1‐8.

101. Roediger WE: Role of anaerobic bacteria in themetabolic welfare of the colonicmucosa inman. Gut1980,21(9):793‐798.

102. VelazquezOC,LedererHM,RombeauJL:Butyrateandthecolonocyte. Production, absorption, metabolism, andtherapeutic implications. Adv Exp Med Biol 1997,427:123‐134.

103. SegainJP,RaingearddelaBletiereD,BourreilleA,LerayV,Gervois N, Rosales C, Ferrier L, Bonnet C, Blottiere HM,Galmiche JP:Butyrate inhibits inflammatory responsesthroughNFkappaB inhibition: implications forCrohn'sdisease.Gut2000,47(3):397‐403.

104. Avivi‐Green C, Polak‐Charcon S, Madar Z, Schwartz B:Apoptosis cascade proteins are regulated in vivo byhigh intracolonic butyrate concentration: correlationwith colon cancer inhibition. Oncology research 2000,12(2):83‐95.

105. KayAB:100yearsof 'Allergy': canvonPirquet'swordberescued?ClinExpAllergy2006,36(5):555‐559.

106. Bach JF: The effect of infections on susceptibility toautoimmune and allergic diseases. The New Englandjournalofmedicine2002,347(12):911‐920.

107. Aberg N, Hesselmar B, Aberg B, Eriksson B: Increase ofasthma, allergic rhinitis and eczema in Swedishschoolchildrenbetween1979and1991.ClinExpAllergy1995,25(9):815‐819.

119

108. Johansson SG, Bieber T,Dahl R, FriedmannPS, LanierBQ,LockeyRF,MotalaC,OrtegaMartellJA,Platts‐MillsTA,RingJetal:Revisednomenclatureforallergyforglobaluse:Reportof theNomenclatureReviewCommitteeof theWorldAllergyOrganization,October2003.JAllergyClinImmunol2004,113(5):832‐836.

109. van der Hulst AE, Klip H, Brand PL: Risk of developingasthma in young children with atopic eczema: asystematic review. J Allergy Clin Immunol 2007,120(3):565‐569.

110. Spergel JM: Epidemiology of atopic dermatitis andatopicmarch inchildren. Immunologyandallergy clinicsofNorthAmerica2010,30(3):269‐280.

111. Hesselmar B, Aberg N, Aberg B, Eriksson B, Bjorksten B:Doesearlyexposuretocatordogprotectagainstlaterallergy development? Clin Exp Allergy 1999, 29(5):611‐617.

112. MatricardiPM,RosminiF,RiondinoS,FortiniM,FerrignoL,Rapicetta M, Bonini S: Exposure to foodborne andorofecalmicrobes versus airborne viruses in relationto atopy and allergic asthma: epidemiological study.BMJ2000,320(7232):412‐417.

113. Riedler J,Braun‐FahrlanderC,EderW,SchreuerM,WaserM, Maisch S, Carr D, Schierl R, Nowak D, von Mutius E:Exposure to farming in early life and development ofasthma and allergy: a cross‐sectional survey. Lancet2001,358(9288):1129‐1133.

114. StrachanDP:Hayfever,hygiene,andhouseholdsize.Bmj1989,299(6710):1259‐1260.

115. Strachan DP:Family size, infectionandatopy: the firstdecade of the "hygiene hypothesis". Thorax 2000, 55Suppl1:S2‐10.

116. BagerP,MelbyeM,RostgaardK,BennCS,WestergaardT:Mode of delivery and risk of allergic rhinitis andasthma.JAllergyClinImmunol2003,111(1):51‐56.

117. McKeever TM, Lewis SA, Smith C, Hubbard R: Mode ofdeliveryandriskofdevelopingallergicdisease.JAllergyClinImmunol2002,109(5):800‐802.


120

118. AdlerberthI,CarlssonB,deManP,JalilF,KhanSR,LarssonP,MellanderL,SvanborgC,WoldAE,HansonLA:IntestinalcolonizationwithEnterobacteriaceae inPakistaniandSwedish hospital‐delivered infants. Acta paediatricaScandinavica1991,80(6‐7):602‐610.

119. Adlerberth I, Jalil F, Carlsson B, Mellander L, Hanson LA,Larsson P, Khalil K, Wold AE: High turnover rate ofEscherichiacolistrainsintheintestinalfloraofinfantsinPakistan.EpidemiolInfect1998,121(3):587‐598.

120. KuhnI,TullusK,MollbyR:Colonizationandpersistenceof Escherichia coli phenotypes in the intestines ofchildrenaged0to18months.Infection1986,14(1):7‐12.

121. Nowrouzian F, Hesselmar B, Saalman R, Strannegard IL,Aberg N, Wold AE, Adlerberth I: Escherichia coli ininfants' intestinalmicroflora: colonization rate, strainturnover,andvirulencegenecarriage.PediatrRes2003,54(1):8‐14.

122. Wang M, Karlsson C, Olsson C, Adlerberth I, Wold AE,StrachanDP,MartricardiPM,AbergN,PerkinMR,TripodiSetal:Reduceddiversityintheearlyfecalmicrobiotaofinfantswithatopiceczema. JAllergyClin Immunol2008,121(1):129‐134.

123. Bisgaard H, Li N, Bonnelykke K, Chawes BL, Skov T,Paludan‐Muller G, Stokholm J, Smith B, Krogfelt KA:Reduced diversity of the intestinalmicrobiota duringinfancy is associated with increased risk of allergicdisease at school age. J Allergy Clin Immunol 2011,128(3):646‐652e641‐645.

124. Ismail IH, Oppedisano F, Joseph SJ, Boyle RJ, Licciardi PV,Robins‐Browne RM, Tang ML: Reduced gut microbialdiversity in early life is associated with laterdevelopment of eczema but not atopy in high‐riskinfants. Pediatric allergy and immunology : officialpublicationoftheEuropeanSocietyofPediatricAllergyandImmunology2012,23(7):674‐681.

125. Abrahamsson TR, Jakobsson HE, Andersson AF, BjorkstenB, Engstrand L, Jenmalm MC: Low gut microbiota

121

diversity in early infancy precedes asthma at schoolage.ClinExpAllergy2014,44(6):842‐850.

126. ArbesSJ,Jr.,SeverML,VaughnB,CohenEA,ZeldinDC:Oralpathogensandallergicdisease:results from theThirdNational Health and Nutrition Examination Survey. JAllergyClinImmunol2006,118(5):1169‐1175.

127. Navarro S, Cossalter G, Chiavaroli C, Kanda A, Fleury S,Lazzari A, Cazareth J, Sparwasser T, Dombrowicz D,GlaichenhausNetal:Theoraladministrationofbacterialextracts prevents asthma via the recruitment ofregulatory T cells to the airways.Mucosal immunology2011,4(1):53‐65.

128. Card JW, Carey MA, Voltz JW, Bradbury JA, Ferguson CD,CohenEA,SchwartzS,FlakeGP,MorganDL,ArbesSJ,Jr.etal:Modulation of allergic airway inflammation by theoralpathogenPorphyromonas gingivalis. Infect Immun2010,78(6):2488‐2496.

129. KhorB,GardetA,XavierRJ:Geneticsandpathogenesisofinflammatory bowel disease. Nature 2011,474(7351):307‐317.

130. HildebrandH,FinkelY,GrahnquistL,LindholmJ,EkbomA,Askling J:Changingpatternofpaediatric inflammatorybowel disease in northern Stockholm 1990‐2001. Gut2003,52(10):1432‐1434.

131. Tsironi E, Feakins RM, Probert CS, Rampton DS, Phil D:Incidenceof inflammatoryboweldisease isrisingandabdominal tuberculosis is falling in Bangladeshis inEast London, United Kingdom. The American journal ofgastroenterology2004,99(9):1749‐1755.

132. Yang SK, Loftus EV, Jr., Sandborn WJ: Epidemiology ofinflammatoryboweldiseaseinAsia.Inflammatoryboweldiseases2001,7(3):260‐270.

133. Feillet H, Bach JF: Increased incidenceof inflammatorybowel disease: the price of the decline of infectiousburden? Current opinion in gastroenterology 2004,20(6):560‐564.

134. RadonK,WindstetterD,PoludaAL,MuellerB,vonMutiusE,Koletzko S:Contactwith farmanimals inearly lifeand


122

juvenile inflammatory bowel disease: a case‐controlstudy.Pediatrics2007,120(2):354‐361.

135. Taurog JD,Richardson JA, Croft JT, SimmonsWA, ZhouM,Fernandez‐Sueiro JL,BalishE,HammerRE:Thegermfreestate prevents development of gut and jointinflammatorydisease inHLA‐B27 transgenic rats. TheJournalofexperimentalmedicine1994,180(6):2359‐2364.

136. Aranda R, Sydora BC, McAllister PL, Binder SW, Yang HY,Targan SR, Kronenberg M: Analysis of intestinallymphocytes inmouse colitismediated by transfer ofCD4+,CD45RBhighTcellstoSCIDrecipients. Journalofimmunology (Baltimore, Md : 1950) 1997, 158(7):3464‐3473.

137. MedinaC,SantanaA,LlopisM,Paz‐CabreraMC,AntolinM,MourelleM,GuarnerF,VilasecaJ,GonzalezC,SalasAetal:Induction of colonic transmural inflammation byBacteroides fragilis: implication of matrixmetalloproteinases. Inflammatory bowel diseases 2005,11(2):99‐105.

138. Sokol H, Pigneur B, Watterlot L, Lakhdari O, Bermudez‐HumaranLG,Gratadoux JJ,BlugeonS,BridonneauC,FuretJP, Corthier G et al: Faecalibacterium prausnitzii is ananti‐inflammatory commensalbacterium identifiedbygutmicrobiotaanalysisofCrohndiseasepatients.ProcNatlAcadSciUSA2008,105(43):16731‐16736.

139. Van Immerseel F, Ducatelle R, De VosM, Boon N, Van DeWiele T, Verbeke K, Rutgeerts P, Sas B, Louis P, Flint HJ:Butyric acid‐producing anaerobic bacteria as a novelprobiotic treatmentapproach for inflammatoryboweldisease.JMedMicrobiol2010,59(Pt2):141‐143.

140. Nwosu FC, Thorkildsen LT, Avershina E, Ricanek P,Perminow G, Brackmann S, Vatn MH, Rudi K: Age‐dependent fecalbacterial correlation to inflammatoryboweldiseasefornewlydiagnoseduntreatedchildren.Gastroenterologyresearchandpractice2013,2013:302398.

141. Thorkildsen LT, Nwosu FC, Avershina E, Ricanek P,Perminow G, Brackmann S, Vatn MH, Rudi K: Dominantfecal microbiota in newly diagnosed untreated

123

inflammatory bowel disease patients. Gastroenterologyresearchandpractice2013,2013:636785.

142. ConteMP,SchippaS,ZamboniI,PentaM,ChiariniF,SegantiL, Osborn J, Falconieri P, Borrelli O, Cucchiara S: Gut‐associated bacterialmicrobiota in paediatric patientswith inflammatory bowel disease. Gut 2006,55(12):1760‐1767.

143. Bibiloni R, Mangold M, Madsen KL, Fedorak RN, TannockGW:Thebacteriologyofbiopsiesdiffersbetweennewlydiagnosed, untreated, Crohn's disease and ulcerativecolitispatients.JMedMicrobiol2006,55(Pt8):1141‐1149.

144. Kellermayer R, Mir SA, Nagy‐Szakal D, Cox SB, Dowd SE,Kaplan JL, Sun Y, Reddy S, Bronsky J, Winter HS:MicrobiotaseparationandC‐reactiveproteinelevationin treatment‐naive pediatric granulomatous Crohndisease.Journalofpediatricgastroenterologyandnutrition2012,55(3):243‐250.

145. HansenR,RussellRK,ReiffC,LouisP,McIntoshF,BerrySH,Mukhopadhya I, Bisset WM, Barclay AR, Bishop J et al:Microbiota of de‐novo pediatric IBD: increasedFaecalibacterium prausnitzii and reduced bacterialdiversity in Crohn's but not in ulcerative colitis. TheAmerican journalofgastroenterology2012,107(12):1913‐1922.

146. Hansen R, Berry SH, Mukhopadhya I, Thomson JM,SaundersKA,NichollCE,BissetWM,LoganathanS,MahdiG,Kastner‐ColeDetal:Themicroaerophilicmicrobiotaofde‐novo paediatric inflammatory bowel disease: theBISCUITstudy.PLoSOne2013,8(3):e58825.

147. WagnerJ,ShortK,Catto‐SmithAG,CameronDJ,BishopRF,Kirkwood CD: Identification and characterisation ofPseudomonas16SribosomalDNAfromilealbiopsiesofchildren with Crohn's disease. PLoS One 2008,3(10):e3578.

148. Kirkwood CD,Wagner J, Boniface K, Vaughan J, MichalskiWP, Catto‐Smith AG, Cameron DJ, Bishop RF:Mycobacterium avium subspecies paratuberculosis in


124

childrenwithearly‐onsetCrohn'sdisease.Inflammatoryboweldiseases2009,15(11):1643‐1655.

149. IebbaV,SantangeloF,TotinoV,NicolettiM,GagliardiA,DeBiase RV, Cucchiara S, Nencioni L, Conte MP, Schippa S:Higher prevalence and abundance of Bdellovibriobacteriovorus in the human gut of healthy subjects.PLoSOne2013,8(4):e61608.

150. GeversD,KugathasanS,DensonLA,Vazquez‐BaezaY,VanTreurenW,RenB,SchwagerE,KnightsD,SongSJ,YassourMetal:The treatment‐naivemicrobiome innew‐onsetCrohn'sdisease.Cellhost&microbe2014,15(3):382‐392.

151. SalonenA,NikkilaJ,Jalanka‐TuovinenJ,ImmonenO,Rajilic‐Stojanovic M, Kekkonen RA, Palva A, de Vos WM:Comparativeanalysisof fecalDNAextractionmethodswith phylogenetic microarray: effective recovery ofbacterialandarchaealDNAusingmechanicalcelllysis.JMicrobiolMethods2010,81(2):127‐134.

152. Gutell RR: Collection of small subunit (16S‐ and 16S‐like)ribosomalRNAstructures:1994.NucleicAcidsRes1994,22(17):3502‐3507.

153. AbdoZ,SchuetteUM,BentSJ,WilliamsCJ,ForneyLJ, JoyceP: Statistical methods for characterizing diversity ofmicrobial communities by analysis of terminalrestriction fragment length polymorphisms of 16SrRNAgenes.EnvironMicrobiol2006,8(5):929‐938.

154. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K,Bushman FD, Costello EK, FiererN, PenaAG, Goodrich JK,GordonJIetal:QIIMEallowsanalysisofhigh‐throughputcommunity sequencing data. Nat Methods 2010,7(5):335‐336.

155. Edgar RC: Search and clustering orders ofmagnitudefaster than BLAST. Bioinformatics 2010, 26(19):2460‐2461.

156. Nelun Barfod M, Magnusson K, Lexner MO, Blomqvist S,Dahlen G, Twetman S: Oral microflora in infantsdelivered vaginally and by caesarean section. Int JPaediatrDent2011,21(6):401‐406.

125

157. KurokawaK,ItohT,KuwaharaT,OshimaK,TohH,ToyodaA, Takami H, Morita H, Sharma VK, Srivastava TP et al:Comparative metagenomics revealed commonlyenrichedgenesetsinhumangutmicrobiomes.DNARes2007,14(4):169‐181.

158. Kibe R, Sakamoto M, Hayashi H, Yokota H, Benno Y:Maturationofthemurinececalmicrobiotaasrevealedbyterminalrestriction fragment lengthpolymorphismand 16S rRNA gene clone libraries. FEMSmicrobiologyletters2004,235(1):139‐146.

159. Ozutsumi Y,HayashiH, SakamotoM, ItabashiH, BennoY:Culture‐independent analysis of fecal microbiota incattle.BiosciBiotechnolBiochem2005,69(9):1793‐1797.

160. Kent AD, Smith DJ, Benson BJ, Triplett EW:Web‐basedphylogenetic assignment tool for analysis of terminalrestriction fragment length polymorphism profiles ofmicrobial communities. Appl Environ Microbiol 2003,69(11):6768‐6776.

161. Pandey J, Ganesan K, Jain RK: Variations in T‐RFLPprofileswithdiffering chemistriesof fluorescentdyesused for labeling the PCR primers. JMicrobiolMethods2007,68(3):633‐638.

162. Kaplan CW, Kitts CL: Variation between observed andtrue Terminal Restriction Fragment length isdependent on true TRF length and purine content. JMicrobiolMethods2003,54(1):121‐125.

163. NieminenTT,VihavainenE,PalorantaA,Lehto J,PaulinL,AuvinenP, SolismaaM,BjorkrothKJ:Characterizationofpsychrotrophic bacterial communities in modifiedatmosphere‐packed meat with terminal restrictionfragment length polymorphism. Int J Food Microbiol2011,144(3):360‐366.

164. Fukushima M, Kakinuma K, Kawaguchi R: Phylogeneticanalysis of Salmonella, Shigella, and Escherichia colistrainsonthebasisofthegyrBgenesequence.Journalofclinicalmicrobiology2002,40(8):2779‐2785.

165. Pham HN, Ohkusu K, Mishima N, NodaM, Monir ShahM,Sun X, Hayashi M, Ezaki T: Phylogeny and species


126

identification of the family Enterobacteriaceae basedondnaJsequences.Diagnosticmicrobiologyand infectiousdisease2007,58(2):153‐161.

166. LundellAC,HesselmarB,NordstromI,SaalmanR,KarlssonH, Lindberg E, Aberg N, Adlerberth I, Wold AE, Rudin A:HighcirculatingimmunoglobulinAlevelsininfantsareassociated with intestinal toxigenic Staphylococcusaureusandalowerfrequencyofeczema.ClinExpAllergy2009,39(5):662‐670.

167. LundellAC,AdlerberthI,LindbergE,KarlssonH,EkbergS,AbergN,SaalmanR,HockB,SteinkassererA,HesselmarBet al: Increased levels of circulating soluble CD14 butnotCD83ininfantsareassociatedwithearlyintestinalcolonization with Staphylococcus aureus. Clin ExpAllergy2007,37(1):62‐71.

168. Abrahamsson TR, Jakobsson HE, Andersson AF, BjorkstenB, Engstrand L, Jenmalm MC: Low diversity of the gutmicrobiota in infantswithatopiceczema. JAllergyClinImmunol2012,129(2):434‐440e432.

169. MooreWE,HoldemanLV:Humanfecalflora:thenormalflora of 20 Japanese‐Hawaiians. Appl Microbiol 1974,27(5):961‐979.

170. Lagier JC, Armougom F, Million M, Hugon P, Pagnier I,RobertC,BittarF,FournousG,GimenezG,MaraninchiMetal:Microbialculturomics:paradigmshift inthehumangut microbiome study. Clin Microbiol Infect 2012,18(12):1185‐1193.

171. Lagier JC, Million M, Hugon P, Armougom F, Raoult D:Human gut microbiota: repertoire and variations.Frontiersincellularandinfectionmicrobiology2012,2:136.

172. CutlerCW,JotwaniR:Dendriticcellsattheoralmucosalinterface.JDentRes2006,85(8):678‐689.

173. Eriksson K, Ahlfors E, George‐Chandy A, Kaiserlian D,Czerkinsky C:Antigen presentation in themurine oralepithelium.Immunology1996,88(1):147‐152.

174. VanHoogstraten IM, Boos C, BodenD, Von BlombergME,ScheperRJ,KraalG:Oralinductionoftolerancetonickel

127

sensitization in mice. The Journal of investigativedermatology1993,101(1):26‐31.

175. Xenoulis PG, Palculict B, Allenspach K, Steiner JM, VanHouse AM, Suchodolski JS: Molecular‐phylogeneticcharacterizationofmicrobialcommunities imbalancesinthesmallintestineofdogswithinflammatoryboweldisease.FEMSMicrobiolEcol2008,66(3):579‐589.

176. Suchodolski JS, Dowd SE,Wilke V, Steiner JM, JergensAE:16S rRNA gene pyrosequencing reveals bacterialdysbiosis in the duodenum of dogs with idiopathicinflammatory bowel disease. PLoS One 2012,7(6):e39333.

177. GosslingJ,SlackJM:Predominantgram‐positivebacteriain human feces: numbers, variety, and persistence.InfectImmun1974,9(4):719‐729.

178. ReyFE,GonzalezMD,ChengJ,WuM,AhernPP,GordonJI:Metabolicnicheofaprominentsulfate‐reducinghumangut bacterium. Proc Natl Acad Sci U S A 2013,110(33):13582‐13587.

179. MalinenE,Krogius‐KurikkaL,LyraA,NikkilaJ,JaaskelainenA,RinttilaT,Vilpponen‐SalmelaT,vonWrightAJ,PalvaA:Association of symptoms with gastrointestinalmicrobiotainirritablebowelsyndrome.Worldjournalofgastroenterology:WJG2010,16(36):4532‐4540.

180. Peach S, Fernandez F, Johnson K, Drasar BS: The non‐sporing anaerobic bacteria in human faeces. J MedMicrobiol1974,7(2):213‐221.

181. Benno Y, Suzuki K, Suzuki K, Narisawa K, Bruce WR,MitsuokaT:ComparisonofthefecalmicroflorainruralJapaneseandurbanCanadians.MicrobiolImmunol1986,30(6):521‐532.

182. MooreWE,CatoEP,HoldemanLV:Somecurrentconceptsinintestinalbacteriology.TheAmericanjournalofclinicalnutrition1978,31(10Suppl):S33‐42.

183. Kato M, Muto Y, Tanaka‐Bandoh K, Watanabe K, Ueno K:Sphingolipid composition in Bacteroides species.Anaerobe1995,1(2):135‐139.


128

184. WuL,GabrielCL,ParekhVV,VanKaerL:Invariantnaturalkiller T cells: innate‐like T cells with potentimmunomodulatory activities. Tissue antigens 2009,73(6):535‐545.

185. AnD,OhSF,OlszakT,NevesJF,AvciFY,Erturk‐HasdemirD,Lu X, Zeissig S, Blumberg RS, Kasper DL: SphingolipidsfromasymbioticmicroberegulatehomeostasisofhostintestinalnaturalkillerTcells.Cell2014,156(1‐2):123‐133.

186. van Baarlen P, Wells JM, Kleerebezem M: Regulation ofintestinal homeostasis and immunity with probioticlactobacilli.Trendsinimmunology2013,34(5):208‐215.

187. Wang ZK, Yang YS: Upper gastrointestinal microbiotaanddigestivediseases.Worldjournalofgastroenterology:WJG2013,19(10):1541‐1550.

microbiota of the alimentary tract of children - … · microbiota of the alimentary tract of ......

Documents