&get_box_var;ORIGINAL ARTICLE

The Impact of Breastfeeding on Nasopharyngeal MicrobialCommunities in InfantsGiske Biesbroek1, Astrid A. T. M. Bosch1,2, Xinhui Wang1, Bart J. F. Keijser3, Reinier H. Veenhoven2†,Elisabeth A. M. Sanders1, and Debby Bogaert1

1Department of Pediatric Immunology and Infectious Diseases, Wilhelmina Children’s Hospital, University Medical Center Utrecht,Utrecht, The Netherlands; 2Research Center Linnaeus Institute, Spaarne Hospital, Hoofddorp, The Netherlands; and 3Research GroupMicrobiology and Systems Biology, TNO Earth, Environmental and Life Sciences, Zeist, The Netherlands

Abstract

Rationale: Breastfeeding elicits significant protection againstrespiratory tract infections in infancy. Modulation of respiratorymicrobiota might be part of the natural mechanisms of protectionagainst respiratory diseases induced by breastfeeding.

Objectives: To study the association between breastfeeding andnasopharyngeal microbial communities, including all cultivable andnoncultivable bacteria.

Methods: In this observational study, we analyzed the microbiotaof infants that had received exclusive breastfeeding (n = 101) andexclusive formula feeding (n = 101) at age 6 weeks and 6 months by16S-based GS-FLX-titanium-pyrosequencing.

Measurements and Main Results: At 6 weeks of age the overallbacterial community compositionwas significantly different betweenbreastfed and formula-fed children (nonmetric multidimensionalscaling,P=0.001). Breastfed children showed increased presence andabundance of the lactic acid bacterium Dolosigranulum (relativeeffect size [RES], 2.61; P = 0.005) and Corynebacterium (RES, 1.98;

P = 0.039) and decreased abundance of Staphylococcus (RES, 0.48;P 0.03) and anaerobic bacteria, such as Prevotella (RES, 0.25; P,0.001) and Veillonella (RES, 0.33; P, 0.001). Predominance (.50%of themicrobial profile) ofCorynebacterium andDolosigranulumwasobserved in 45 (44.6%) breastfed infants compared with 19 (18.8%)formula-fed infants (relative risk, 2.37; P = 0.006). Dolosigranulumabundance was inversely associated with consecutive symptomsof wheezing and number of mild respiratory tract infectionsexperienced. At 6 months of age associations between breastfeedingand nasopharyngeal microbiota composition had disappeared.

Conclusions: Our data suggest a strong association betweenbreastfeeding and microbial community composition in the upperrespiratory tract of 6-week-old infants. Observed differences inmicrobial community profile may contribute to the protective effectof breastfeeding on respiratory infections and wheezing in earlyinfancy.Clinical trial registered with www.clinicaltrials.gov (NCT 00189020).

Keywords: breastfeeding; microbiota; microbial communities;respiratory tract; infants

Breastfeeding has been associated witha protective effect on frequency (1, 2),duration (3), and severity (4) of respiratory

tract infections (RTIs) in children at leastup to 6 months of age in developingcountries (1, 2, 5) and high-income

countries (3, 6). Several componentsof breast milk may contribute to thisprotective effect, such as secretory IgA

(Received in original form January 14, 2014; accepted in final form June 10, 2014 )†Deceased.

Supported by The Netherlands Organization for Scientific Research through NWO-VENI grant (91610121) and ZonMw grant (91209010). The randomizedcontrolled trial was financed by the Dutch Ministry of Health. The sponsors had no role in study design, data collection, analysis and interpretation, or writing ofthe report. The corresponding author had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of thedata analysis.

Author Contributions: Conceived and designed the experiments, G.B., B.J.F.K., and D.B. Performed the clinical study, R.H.V. and E.A.M.S. Performed theexperiments, G.B., A.A.T.M.B., B.J.F.K., and D.B. Analyzed the data, G.B., A.A.T.M.B., X.W., B.J.F.K., and D.B. Wrote the paper, G.B., A.A.T.M.B., B.J.F.K.,E.A.M.S., and D.B. Critical review of the paper, all authors.

Correspondence and requests for reprints should be addressed to Debby Bogaert, M.D., Ph.D., Department of Pediatric Immunology and InfectiousDiseases, Wilhelmina Children’s Hospital, University Medical Center Utrecht, PO Box 85090, Room KC.03.068.0, 3584 EA Utrecht, The Netherlands.E-mail: dbogaert@umcutrecht.nl

This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org

Am J Respir Crit Care Med Vol 190, Iss 3, pp 298–308, Aug 1, 2014

Copyright © 2014 by the American Thoracic Society

Originally Published in Press as DOI: 10.1164/rccm.201401-0073OC on June 12, 2014

Internet address: www.atsjournals.org

298 American Journal of Respiratory and Critical Care Medicine Volume 190 Number 3 | August 1 2014

antibodies; oligosaccharides; and severalmicrobicidal and/or immune stimulatorymilk-derived factors, including lactoferrin,cytokines, and growth factors (4, 7). Mainbacterial respiratory pathogens, such asStreptococcus pneumoniae, Haemophilusinfluenzae, and Staphylococcus aureus,are common and usually asymptomaticresidents of the upper respiratory tract,in particular the nasopharyngeal niche, inhealthy children. From here these bacteriamay start behaving as pathogens andspread throughout the respiratory tract,or even invade the bloodstream to causemucosal and systemic disease (8).

Because many of the causativepathobionts of RTIs are commonly carriedin the nasopharynx of children, it is possiblethat breastfeeding influences risk of diseasethrough a direct effect on nasopharyngealcolonization with potential pathogens.However, until now conflicting results havebeen found regarding a possible correlationbetween breastfeeding and pathogencolonization using conventional culturemethods (9, 10). Nevertheless, recentlystrong associations between breastfeedingand the microbiota composition of the guthave been observed, with more pathobiontsand less “beneficial” commensal bacteria

observed in the feces of formula-fed childrencompared with breastfed infants (11, 12).Commensal bacterial communities maythereby act as first-line defense againstacquisition, overgrowth, and invasionof pathobionts and may consequentlyinfluence susceptibility to infections. Effectsof feeding type on the total commensalbacterial populations of the respiratory tracthave not been studied yet, althoughinfluences on the commensal microbiotamight be similar as observed in the gut.We therefore analyzed nasopharyngealmicrobiota in 101 healthy breastfedinfants and 101 formula-fed infants atthe consecutive ages of 6 weeks and6 months using deep sequencing techniques.

Some of the results of these studies havebeen previously reported in the form ofabstracts (13–15).

Methods

Sample Collection and StorageNasopharyngeal samples were obtained fromchildren that participated in a randomizedcontrolled pneumococcal conjugate vaccinationtrial performed in the Netherlands between2005 and 2008 and previously described (16).Written informed consent was obtained fromboth parents before enrolment of the infants.Deep transnasally obtained nasopharyngealswabs were collected during home visits andimmediately inoculated in Transwab (modifiedAmies) transport medium and stored within24 hours at 2808C for further analyses. Forthe current study, nasopharyngeal sampleswere selected from 6-week-old exclusivebreastfed and formula-fed children that weresampled between August 2005 and February2006 and between December 2005 and June2006 at age 6 months. All samples werecultured for Moraxella catarrhalis, H.influenzae, S. aureus, and S. pneumonia asreported previously (16–18).

Questionnaires on breastfeeding,crowding conditions, antibiotics use, healthstatus, and other relevant environmentalfactors were obtained by research nurses atboth sampling moments. All children wereconsidered nonfebrile and healthy at themoment of sampling.

Ethical StatementThis randomized controlled trial wasapproved by an acknowledged DutchNational Ethics Committee (StichtingTherapeutische Evaluatie Geneesmiddelen,

http://www.stegmetc.org) and the trial wasundertaken in accordance with EuropeanStatements for Good Clinical Practice,which included the provisions of theDeclaration of Helsinki of 1989.

Construction of thePhylogenetic LibraryNasopharyngeal samples were processedfor 454 GS-FLX-titanium sequencing of the16S small subunit ribosomal DNA gene:a conserved gene with variable regions amongbacteria. Preparation of the 16S ampliconlibrary has been described previously anddetails are provided in theMETHODS section ofthe online supplement (19). In short, DNAwas extracted using phenol/bead beating incombination with the AGOWA Mag miniDNA extraction kit (AGOWA, Berlin,Germany). Subsequently, an ampliconlibrary was generated by amplification of theV5-V7 hypervariable region of this gene (19).The purified and equimolar pooled ampliconswere sequenced unidirectionally in the GS-FLX-Titanium Sequencer (Life Sciences,Roche, Branford, CT).

Sequences were processed using modulesimplemented in the Mothur 1.20.0 softwareplatform (20). Sequences were denoised andchecked for quality and chimeras (usingChimera Slayer). High-quality alignedsequences were classified using the RDP-IInaive Bayesian Classifier and clustered intospecies-based phylotypes (operationaltaxonomic units [OTUs], defined by 97%similarity). For all samples rarefaction curveswere plotted and community diversity indices(Shannon diversity, Chao1, and Simpsonindex) calculated. Sequence data were subjectedto weighted UniFrac analysis using theUniFrac module implemented in Mothur andvisualized with iTOL software. The WeightedUniFrac algorithm calculates the distancebetween microbial communities based onthe phylogenetic relatedness of lineages andrelative abundance in each sample.

Statistical AnalysisData analyses were performed in R version 2.7(R Foundation for Statistical Computing,Vienna, Austria), Microsoft Excel 2011(Redmond, WA), and SPSS version 20 (IBMCorporation, Armonk, NY) using severalstatistical methods (i.e., Pearson correlation,chi-square tests, and univariate andmultivariate linear models). To correct forpotential confounders, we used a stepwiseapproach starting with univariate regression,followed by a classical stepwise multivariate

At a Glance Commentary

Scientific Knowledge on theSubject: Breastfeeding has beenassociated with a protective effecton respiratory infections and atopicdiseases in infants and children.Although several immune stimulatoryand antimicrobial components of breastmilk are considered to contribute to itsprotective effect, there is conflictingevidence regarding an associationbetween breastfeeding and bacterialcolonization in the respiratory tract.

What This Study Adds to theField: This study shows for the firsttime that breastfeeding has an effect onthe complete microbial population(microbiota) of the respiratory tract.The increased presence, abundance,and dominance of the lactic acidbacteria Dolosigranulum andCorynebacterium might contribute tothe protective effect of breastfeedingagainst respiratory infections.

ORIGINAL ARTICLE

Biesbroek, Bosch, Wang, et al.: Breastfeeding and Infant Nasopharyngeal Microbiota 299

regression model including all variables thatwere significantly associated with microbiotamembers or showed a trend toward anassociation with a P value less than 0.15, asdescribed by Wang and coworkers (21).Because of the high number of comparisons,associations were considered statisticallysignificant after correction for multiple testingby false discovery rate (FDR) (q value = 0.2).We log-transformed the relative abundancesof OTUs because of the skewed distributionin our population. We calculated relativeeffect sizes (RESs) of significant associationsfrom the standardized regression coefficientsof the multivariate model to indicate themagnitude of change between groups. Weused the taxonomy-independent clusteringmethod nonmetric multidimensional scaling(nMDS) to compare microbiota profiles fordissimilarities. To obtain insight in OTU

network structures, OTUs were hierarchicallyclustered using complete linkage and Pearsoncorrelation, and displayed using CytoscapeV2.8.2 (The Cytoscape Consortium, SanDiego, CA). The optimal number of clusterswas calculated using the Silhouette indexscreening several different distance andlinkage measures. Highly correlating clustersand OTUs (correlation coefficient . 0.6)were visualized by connecting red (negativelycorrelating) or green (positively correlating)lines between nodes.

Results

Characteristics of theStudy PopulationWe analyzed nasopharyngeal samples of 202infants by 16s rDNA pyrosequencing at both

6 weeks and 6 months of age. Of these infants,101 received formula feeding and 87 infantsreceived breastfeeding up to the age of 6months. Fourteen children got breastfeedingup to around 3 months of age (range, 11–21wk) (Table 1) and were not included in theanalyses of 6-month samples. Breastfed andformula-fed children differed in populationcharacteristics. Formula-fed childrenhad significantly more often one or moresiblings and were more frequently exposedto indoor smoking compared with breastfedchildren. Although not significant, morechildren were delivered by caesareansection in the group of formula-fedcompared with breastfed children.

Culture ResultsOver time colonization rates for S. aureusdeclined significantly and H. influenzae,

Table 1. Baseline Characteristics of 6-Week- and 6-Month-Old Children

6 Weeks 6 Months

Breastfeeding*FormulaFeeding†

P ValueBreastfeeding*

FormulaFeeding†

P Value(N = 101) (N = 101) (N = 87) (N = 101)

Male sex 50 (49.5) 53 (52.5) NS 44 (50.6) 53 (52.5) NSCaesarian section‡ 8 (8.8) 14 (14.6) NS 6 (7.8) 14 (14.6) NSPneumococcal conjugate vaccination-7

immunizationx64 (63.4) 65 (64.4) 55 (63.2) 65 (64.4)

Presence of siblings in household 63 (62.3) 78 (77.2) 0.02 52 (59.8) 77 (76.2) 0.02Daycare attendancejj 0 0 NS 46 (52.9) 49 (48.5) NS

Environmental tobacco exposure 3 (3.0) 8 (7.9) NS 1 (1.1) 11 (10.9) ,0.01Average number of cigarettes/daywhen exposed (SD)

13 (6.1) 12.9 (6.6) 20 (NA) 9.6 (8.3)

Month of sampling Aug–Feb Aug–Feb Dec–June Dec–JuneAntibiotics consumption (,1 mo)¶ 2 (2.0) 3 (3.0) NS 3 (3.4) 4 (4.0) NS

Upper respiratory tract symptomsCurrent cold** 17 (16.8) 21 (20.8) NS 21 (24.1) 24 (23.8) NSCurrent otitis media — — 1 (1.1) 1 (1.0) NSHistory of wheezing — — 6 (6.9) 16 (15.8) 0.07Inhaled bronchodilators — — 9 (10.3) 13 (12.9) NSOral or inhaled steroids — — 2 (2.3) 4 (4.0) NSNumber of previous upper respiratorytract infections††

— — 27 (31.0) 35 (34.6) NS

Bacterial colonizationStreptococcus pneumonia 27 (26.7) 19 (18.8) NS 58 (66.7) 60 (59.4) NSMoraxella catarrhalis 35 (34.7) 42 (41.6) NS 69 (79.3) 70 (69.3) NSHaemophilus influenza 19 (18.8) 23 (22.8) NS 43 (49.4) 49 (48.5) NSStaphylococcus aureus 53 (52.5) 58 (57.4) NS 10 (11.5) 19 (18.8) NS

Definition of abbreviations: NA = not applicable; NS = nonsignificant.Data are shown as n (%) unless otherwise noted. P values were calculated using the chi-square test. A P value of greater than 0.05 was considerednonsignificant. P values were depicted when there was a trend (P value between 0.05 and 0.01) or a significant difference (P value < 0.01).*Children were classified as being breastfed if exclusive breastfeeding was given at the age of 6 weeks and 6 months.†Children were classified as formula fed if children were exclusively formula fed at 6 weeks and 6 months of age.‡Information about mode of delivery of 15 infants (10 from the breastfed group and 5 from the formula-fed group) is missing.xOver half of the children within this cohort received pneumococcal conjugate vaccination in a routine vaccination schedule (i.e., at 2, 4, and 11 months of age).jjDaycare is defined as at least 4 hours per week of daycare with more than one child from a different family.¶Defined as use of an antibiotic within 1 month before sampling.**Presence of mild symptoms of a respiratory tract infection at time of sampling as reported by parents.††Number of parental-reported upper respiratory tract infections in the previous 6 months.

ORIGINAL ARTICLE

300 American Journal of Respiratory and Critical Care Medicine Volume 190 Number 3 | August 1 2014

S. pneumoniae, and M. catarrhalis increasedsignificantly (chi-square, P, 0.05) (16–18).In the present study, colonization rates didnot significantly differ between breastfedand formula-fed infants at both ages. Toidentify these pathobionts in the sequencedatabase, we correlated the cultureresults to the corresponding species-levelphylotypes (OTUs, based on >97%similarity) of Moraxella, Streptococcus,Haemophilus, and Staphylococcus. Weobserved a strong correlation betweenthe culture results and the highest rankingOTU per correlating genus (see Table E1in the online supplement).

Overall Microbial DiversityThrough barcoded 16s rDNApyrosequencing, we obtained 1,138,617high-quality sequences with on average2,798 6 1,008 (SD) per sample. Sequencedepth was sufficient to obtain high degreeof sequence coverage for all samples (mean,0.996; median, 0.997; range, 0.977–1.000).Sequencing of the nasopharyngealmicrobiota of both age groups resultedin the identification of 13 phyla and 1,031bacterial OTUs (.3 sequences), witha mean of 19.4 OTUs (95% confidenceinterval (CI), 18.4–20.4) per sample. Thenumber of OTUs per sample significantlyincreased over time (mean, 19.4 OTUs persample, 95% CI, 17.8–21.0 at 6 wk; mean,23.5 OTUs per sample, 95% CI, 21.6–25.3at 6 mo), whereas the average bacterialdiversity as measured by Shannon diversityindex significantly decreased from 1.0(mean value; 95% CI, 0.96–1.1) at 6 weeksto 0.86 (mean value; 95% CI, 0.79–0.92)at 6 months of age. Together thesedata suggest a less uniform distribution(i.e., decreased evenness) of OTUs at6 months compared with 6 weeks ofage. Similarly, a significantly highernumber of OTUs was found in formula-fed (mean, 19.2; 95% CI, 17.2–21.3)compared with breastfed children (mean,16; 95% CI, 14.3–17.8; t test P = 0.02) at6 weeks, whereas the Shannon diversityindex was not different between feedinggroups.

Sample Clustering Based onTaxonomy-Dependent and-Independent Clustering MethodsWe used the taxonomy-dependentWeighted UniFrac analyses to evaluateif overall microbial profiles differedsignificantly between formula-fed and

breastfed infants (Figure 1). The UniFracclustering showed distinct microbiotaprofiles dominated by either Moraxella(red), Streptococcus (dark blue), Haemophilus(green), Staphylococcus (light blue), ora combination of Corynebacterium andDolosigranulum (yellow/pink). Breastfedinfants were significantly distant in theWeighted UniFrac from formula-fedinfants at 6 weeks of age (distance-basedamova, P = 0.002). At 6 months of age, nophylogenetic differences were observedbetween groups (amova, P = 0.83).

To confirm the differences at 6 weeksand the absence of differences at 6 months,we used the taxonomy-independentclustering method nMDS (Figure 2). Usingthis technique we observed a clear shift inmicrobiota profiles between breastfed andformula-fed infants (Figure 2A; geomeans,F test P = 0.001) at 6 weeks of age, whichhad disappeared at 6 months of age(Figure 2B), confirming the WeightedUniFrac analyses.

Microbiota Differences Underlyingthe Observed Infant ClusteringPatternsTo identify bacterial phylotypes that differedin relative abundance between breastfedand formula-fed infants, we comparedmicrobiota composition between bothgroups and performed multivariateregression analyses.

Breastfed children wereoverrepresented in the Corynebacterium–and Dolosigranulum–dominated microbiotaprofiles at 6 weeks of age (Figure 1A): adominant profile (>50% of sequences in anindividual sample) of these species wasobserved in 45 (44.6%) breastfed comparedwith 19 (18.8%) formula-fed infants (P =0.006). The association of Dolosigranulumwith breastfeeding remained positive in themultivariate regression analyses at genuslevel (RES, 2.61; 95% CI, 1.4–5.0; P = 0.005)and at OTU level (RES, 2.61; P = 0.005)(Table 2). The sequences in this latter OTUshowed 100% homology to Dolosigranulumpigrum when blasted against the NationalCenter for Biotechnology Information(NCBI) GenBank and SILVA 16S referencedatabases (see http://www.arb-silva.de/contact/).

Two different CorynebacteriumOTUs were also positively correlated withbreastfeeding in multivariate analyses(RES, 1.98 and 1.93) (Figure 3, Table 2).The most abundant Corynebacterium

OTU (Corynebacterium 1) showedhigh homology (100%) with the speciesC. pseudodiphteriticum and C.propinquum. The other breastfeeding-related Corynebacterium was 97%homologous to either C. accolens, C.fastidiosum, or C. segmentosum,respectively (see Figure E1 and Table E2).Intriguingly, Dolosigranulum andCorynebacterium OTUs seemed to cooccursignificantly (Pearson chi-square test, P ,0.001), suggesting their interrelatedness orinterdependency.

By contrast, in formula-fed childrenat 6 weeks of age we observed anoverrepresentation of a Staphylococcus–dominated cluster (Figure 1A); a dominantStaphylococcus profile was observedin 23.7% of formula-fed and 17.8%of breastfed children (P = 0.05). OnOTU level, only the most dominantStaphylococcus OTU, resembling S. aureusin culture, was negatively associated withbreastfeeding (RES, 0.48; P = 0.032)(Table 2). Other bacteria that werenegatively correlated with breastfeedingwere OTUs belonging to the generaActinomyces, Gemella, Veillonella,Prevotella, Rothia, and Granulicatella, ofwhich OTUs belonging to the latter fourgenera remained significantly different afterFDR correction. These differences werereflected in presence and absence as well asin relative abundance of the associatedOTU (Table 2, Figure 3).

At 6 months of age differences betweengroups had diminished, although, notsignificantly, Granulicatella (RES, 0.43; 95%CI, 0.22–0.83; P = 0.013) was still negativelyassociated with breastfeeding.

Microbial Inference NetworkBecause the presence or abundance ofindividual bacteria most likely affectsthe presence of others due to ecologicinteractions, insight into the communitystructure of the microbiota helpsinterpretation of the observed increasesor decreases in individual species. To thispurpose we calculated the cooccurrenceand coexclusion of bacteria and createda bacterial association network fromsamples taken at 6 weeks of age (Figure 4).OTUs in this network are representedby nodes colored according to the clusterthey belong to, and depicted with theirPearson correlations (positive andnegative correlations in green and redlines, respectively). The most abundant

ORIGINAL ARTICLE

Biesbroek, Bosch, Wang, et al.: Breastfeeding and Infant Nasopharyngeal Microbiota 301

OTUs Haemophilus, Moraxella, andStaphylococcus matching the cultureresults for H. influenzae, M. catarrhalis,and S. aureus, respectively, clusteredseparately from one another and from theremaining bacterial OTUs. Moraxella andStaphylococcus were also negatively

associated with the cluster containing theCorynebacterium and DolosigranulumOTUs(cluster 3) This latter cluster also interactednegatively with two large and closely relatedclusters of OTUs containing mostly oralbacteria: the first one (cluster 1) includedamong others the most abundant

Streptococcus OTU (matching S.pneumoniae culture results), the anaerobesVeillonella, Rothia, Granulicatella,Actinomyces, and several nonpneumococcalstreptococcal OTUs, whereas the secondcluster (cluster 6) also included anaerobicbacteria, such as several different Prevotella

Figure 1. Weighted UniFrac of nasopharyngeal samples taken at 6 weeks and 6 months of age. Clustering of samples based on evolutionary(phylogenetic) relatedness using Weighted UniFrac analyses. The clustering is shown in a circle dendrogram. Each branch represents a sample and eachadjacent histogram the relative abundance of the six most abundant genera found in that sample. Note that multiple operational taxonomic units cancorrespond to the depicted genus. The differences in length of the branches among samples reflect their distance (i.e., dissimilarity) to each other. TheUniFrac dendrogram with the corresponding metadata was displayed using iTol software. To clarify the differences, we calculated separate dendogramsfor breastfed and formula-fed children at 6 weeks and 6 months of age. Overall, samples were clustered mostly on dominance of either Moraxella,Streptococcus, Haemophilus, Staphylococcus, or Corynebacterium. At 6 weeks of age breastfed children were overrepresented in the Corynebacterium

and Dolosigranulum clade and underrepresented in Streptococcus– and Staphylococcus–dominated clades. At 6 months of age the Moraxella

clade expanded, whereas the Corynebacterium and Staphylococcus clades almost diminished. Differences according to feeding type were less clearat this age.

ORIGINAL ARTICLE

302 American Journal of Respiratory and Critical Care Medicine Volume 190 Number 3 | August 1 2014

Figure 2. Nonmetric multidimensional scaling (nMDS) of microbiota profiles of breastfed and formula-fed children at 6 weeks and 6 months of age.Microbiota profiles were compared between groups by nMDS to find dissimilarities between samples. Each dot in the figure is the representation of themicrobiota profile of a single sample. Shown in both figures are the plots of breastfed (b, red) and formula-fed (f, blue) children. The boxes representthe geometric mean of both groups where the length of the line between sample (dot) and geomean (box) represents the distance of that sample to thegeometric mean of the group. Longer lines represent higher distances of samples (i.e., higher variability between sample compositions). (A) nMDS plots ofbreastfed (red) and formula-fed (blue) children at 6 weeks of age. The geomean of microbiota profiles differed significantly (F test, P , 0.01) betweenbreastfed and formula-fed children. (B) nMDS plots of breastfed (red) and formula-fed (blue) children at 6 months of age, showing no differences ingeomean of microbiota profiles between both groups.

ORIGINAL ARTICLE

Biesbroek, Bosch, Wang, et al.: Breastfeeding and Infant Nasopharyngeal Microbiota 303

OTUs, Gemella, Leptotrichia, andFusobacterium, and Klebsiella. The OTUspositively associated with breastfeedingwere all in the Corynebacterium and

Dolosigranulum cluster, whereas the OTUsassociated with formula feeding were all,except for Staphylococcus, observed in the“oral” clusters 1 and 6.

Respiratory Health CharacteristicsAt all sample moments, we asked parentshow many (mild) upper RTIs and episodesof wheezing had occurred since the previous

Table 2. Relative Effect Sizes and 95% CIs of OTUs Significantly Related to Feeding Type by Multivariate Analyses at 6 Weeksof Age

OTU

Relative Abundance (%) Presence (% of Individuals)

BF FF RES 95% CI P Value BF FF P Value

Dolosigranulum 8.21 5.12 2.61 1.35–5.04 0.005 73 52 0.004Corynebacterium (1) 17.6 8.96 1.98 1.04–3.76 0.031 82 69 NSCorynebacterium (11) 0.09 0.02 1.93 1.02–3.85 0.011 15 5 0.032Staphylococcus 17.69 21.96 0.48 0.25–0.93 0.015 72 81 NSActinobacillus 0.23 0.31 0.41 0.21–0.79 0.001 10 23 0.021Actinomyces 0.003 0.02 0.36 0.19–0.69 0.002 2 13 0.006Gemella 0.37 1.14 0.38 0.20–0.73 0.003 28 46 0.013Veillonella (1) 0.05 0.07 0.33 0.17–0.63 0.003 11 30 0.001Rothia 0.001 0.26 0.31 0.16–0.59 0.001 6 24 <0.001Bacteroidetes unclassified 0.07 0.43 0.29 0.15–0.55 0.001 3 18 <0.001Veillonella (2) 0.000 0.07 0.28 0.15–0.54 0.001 3 20 <0.001Prevotella 0.20 1.06 0.25 0.13–0.48 <0.001 5 27 <0.001Granulicatella 0.008 0.19 0.21 0.11–0.41 <0.001 4 26 <0.001

Definition of abbreviations: BF = breastfeeding; CI = confidence interval; FF = formula feeding; NS = not significant; OTU = operational taxonomic units;RES = relative effect size.Overview of RES, 95% CI, P values, average relative abundance, and presence of OTUs that were significant or showed a trend toward an associationwith feeding type at 6 weeks of age in the multivariate analyses correcting for delivery mode, the presence of siblings, and environmental tobaccoexposure. Relative effect sizes that remained significant after false discovery rate correction are shown in bold. In the multivariate analyses remained in themodel next to breastfeeding. Significance in presence of individuals is calculated with a two-tailed Fisher exact test. The number in parentheses behind theOTU name is the hierarchical number of that genus assigned OTU when ranked on average relative abundance.

Figure 3. Relative effect sizes of operational taxonomic units that were significantly associated with feeding type at 6 weeks of age. Relative effect sizeswith their 95% confidence intervals of operational taxonomic units that were significantly decreased or increased in breastfed infants compared withformula-fed infants in multivariate analyses.

ORIGINAL ARTICLE

304 American Journal of Respiratory and Critical Care Medicine Volume 190 Number 3 | August 1 2014

visit (Table 1). In a post hoc analysis wecorrelated respiratory symptoms thatoccurred between 6 weeks and 6 monthsto the OTUs significantly associated(positively and negatively) with breastfeedingat 6 weeks of age (Table 2). Weobserved a negative correlation betweenDolosigranulum and wheezing (averagerelative abundance of 1.5% [95% CI,0.2–3.7%] and 6.8% [95% CI, 5.0–8.8%]in children with and without symptomsof wheezing, respectively; P = 0.02),even when correcting for feeding typein the model (P = 0.05). Similarly,Dolosigranulum was negativelyassociated with number of parental-reported RTIs with a mean relativeabundance of 5.2% (95% CI, 2.3–8.3%)and 6.7% (95% CI, 4.6–8.8%; P = 0.04) inchildren with compared with childrenwithout RTIs, respectively. Gemella waspositively associated with the number ofRTIs with relative abundance of 1.2%(95% CI, 0.5–1.8%) and 0.4% (95% CI,0.2–0.8%; P = 0.03) in children with

compared with children without RTIs,respectively. These significantassociations remained present even aftercorrecting for breastfeeding in themultivariate model (P = 0.05 andP = 0.04 for consecutive RTIs andbreastfeeding, respectively).

Discussion

Breastfeeding is highly advocated for itsbeneficial impact on infant health, forexample because of its protective effectsagainst development of infectious and atopicdiseases (2, 22). To our knowledge, thepresent deep sequencing study is the firstto describe the correlation between infantfeeding and the composition of the upperrespiratory microbiota, and was executed ina large group of infants. Results suggesta major impact of breastfeeding on thecomposition of the respiratory microbiotain the nasopharynx, especially in theimportant first weeks of life when

susceptibility to (severe) infections ishigh (23).

We observed an increase in presenceand abundance of Dolosigranulum andseveral Corynebacterium spp., commoncommensals of the respiratory tractin children (24). Not much is knownabout Dolosigranulum and only recentlythe first genome was mapped by wholegenome sequencing (Dolosigranulumpigrum ATCC 51524, Broad Instituteof Harvard and MIT, http://www.broadinstitute.org/). Dolosigranulumare members of the Carnobacteriaceaefamily, which are gram-positive lacticacid bacteria. Lactic acid bacteria aregenerally considered to be contributors toa healthy microbiota (25) and the attributedprobiotic features of several membersof this class of bacteria are interesting inlight of our observed positive correlationbetween Dolosigranulum and breastfeeding.Corynebacterium is a heterogeneousgram-positive clade of the Actinobacteria,and a common colonizer of skin and

Figure 4. Operational taxonomic units (OTUs) association network. To identify patterns of cooccurrence between the 100 most abundant OTUs we usedhierarchical clustering with average linkage and Pearson correlation distance. The optimal number of clusters and the optimal distance and linkagemeasures were identified by the Silhouette index. The 14 identified clusters of OTUs are discriminated by different node colors, which are described in thelegend. The average relative abundance of the OTU is reflected in the node size (based on a log2 scale). Positive correlation (correlation coefficient, >0.6)between individual OTUs is depicted by green lines. OTUs that are positively associated with breastfeeding in multivariate analysis are depicted byblue circles. OTUs negatively associated with breastfeeding are depicted by red circles around the nodes. For visualization purposes we did not depictthe names of all nodes. Also note that the graphical distance between nodes and clusters is not an actual reflection of the mathematical distance.

ORIGINAL ARTICLE

Biesbroek, Bosch, Wang, et al.: Breastfeeding and Infant Nasopharyngeal Microbiota 305

mucosa in humans. Interestingly,the species pseudodiphteriticum andpropinquum, both highly homologousto the Corynebacterium OTUs that arepositively associated with breastfeeding, areasaccharolytic (not capable of fermentingoligosaccharide). Moreover the specieC. accolens, which also resembles ourbreastfeeding-associated OTUs, is highlylipophilic. Because we observed a strongcooccurrence between Dolosigranulum andCorynebacterium, the presence of onespecies might favor the presence ofthe other, for example by metabolicinterdependency or by the fact that theyboth use different components of breastmilk itself.

Next to the potential metabolicfeatures selecting for this complementarycolonization profile, the increase inCorynebacterium in breastfed children mayalso be a consequence of their presence onthe human skin or in human milk itself.Hunt and coworkers (26) identified in all47 human milk samples the presence ofCorynebacterium spp. indicating this genusmight be a member of the core microbiotadetected in human milk. In light of theirhigh abundance in the respiratory tract ofespecially breastfed infants, and theirpresence in breast milk itself, it seemsworthwhile to further explore theirmetabolic features, interrelatedness withother bacteria, and potential beneficialeffect on newborn health. A potentialbeneficial role in respiratory health of theabove species is further supported by theobserved inversed association betweenDolosigranulum abundance and parentalreported symptoms of wheezing andRTIs in the consecutive months. Althoughour study was not designed for clinicaloutcomes measures, recent data of Lauferand coworkers (27, 28) also showed that thepresence of both Corynebacterium andDolosigranulum in respiratory microbiotawas associated with a reduced risk of acuteotitis media. Because our data and previousstudies using conventional culture showno apparent differences in colonization ofthe most abundant pathobiontsHaemophilus and Streptococcus, it might bethat protective effects of breastfeeding aremediated through enhancement ofprotective microbiota, as was hypothesizedfor the gut microbiota (12). In the gut ofbreastfed children, microbiota were alsopredominated by nonpathogenic residents,such as Bifidobacteria and Lactobacillus

species, whereas the microbiota of formula-fed infants seem to be colonized bya more diverse range of bacteria includingpathobionts.

At 6 weeks of age we found in formula-fed infants an increase in abundance of severalanaerobes, such as Veillonella and Prevotella,several streptococcal OTUs, Rothia, Gemella,and Granulicatella, compared with breastfedchildren. Similar results were observed byHolgerson and coworkers (29) who observedincreased carriage rates of Prevotella,Granulicatella, and Veillonella in the oralmicrobiota of 3-month-old children that wereformula fed. This increases the likelihood thatour observations are truly related to infantfeeding regimens. Although the bacteria thatwe observed to be negatively associated withbreastfeeding are not clearly described aspathogens, Actinomyces, Veillonella, Rothia,Gemella, and Prevotella were recently foundto be associated with an increased risk ofotitis media (27). We also observed a positivecorrelation between Gemella and consecutiveRTIs supporting a possible unbeneficial rolein respiratory health.

Despite the fact that breastfeeding isgenerally associated with short- and long-termhealth benefits (1, 2, 6, 22), we observed earlybut no clear late differences (6 mo) inmicrobial profiles of breastfed compared withformula-fed children. Nevertheless, earlymicrobiota differences may be importanthealth determinants for later in life, as wasillustrated by the observed associationsbetween colonization of Dolosigranulum at6 weeks of age and respiratory healthcharacteristics in the approximatelyconsecutive 4 months in this study. Microbiotadifferences may be lost or more subtle after theintroduction of solid foods generally startingaround 4 months of life, because theintroduction of solid foods was shown to havea major impact on the gut microbiota ininfants (30, 31). Increased daycare attendanceat 6 months compared with 6 weeks of agemight also contribute to the diminished effectof breastfeeding on microbiota composition at6 months of age (32).

The higher levels of Staphylococcus andCorynebacterium, regardless of feedingtype, in the nasopharynx at 6 weekscompared with 6 months of age might beexplained by environmental encounters.Both species are common colonizers ofthe adult human skin (33) and, for thebreastfed infants, breast milk also containssignificant amounts of Corynebacterium(26). We have no information about the

time between completion of breastfeedingand nasopharyngeal sampling and wetherefore cannot rule out that breast milkitself might have “contaminated” oursamples. Nevertheless, we are confidentthat our findings truly reflect thenasopharyngeal and not breast milkmicrobiota for the following reasons: breastmilk contains approximately 103–104 cfu/ml(34), whereas CFU counts are 10–100 timeshigher in the nasopharyngeal samples (19).Furthermore, breast milk contains, nextto Corynebacteria, more staphylococci,whereas staphylococcal carriage wasdecreased in breastfed infants. If thedifferences we observed in breastfedcompared with formula-fed childrenwere biased by contamination of breastmilk, we would have found an increase inStaphylococcus species, not the reverse.This finding also underlines there isa distinction between bacterial passageand true colonization. On the contrary,Dolosigranulum was significantly enhancedin breastfed children and is not a memberof the bacterial communities in breast milk.Nonetheless, it is very interesting to relatebacterial compositions in breast milk of themother to microbiota compositions inupper respiratory and gut sites in newprospective studies.

To appreciate the results of our study,several limitations of our study should betaken into account. Our study included onlytwo sampling points in time. Longitudinalstudies with shorter time intervals aredesired to further elucidate the successionand dynamics of the respiratorymicrobial ecosystem in relation to feeding.Furthermore, this study was not primarilydesigned to investigate feeding-type–relatedeffects and we therefore did not obtaininformation on the exact momentum ofintroduction of supplementary feedingin these children. Moreover, informationon respiratory symptoms was retrievedretrospectively from the parents, whichmight have led to a recall bias. Similarly,antibiotic usage was also retrospectivelyretrieved. Nevertheless, most of thechildren received no antibiotics in thefirst 6 weeks and 6 months of life andthe number of prescriptions was almostequally distributed between feeding groups.Moreover, they match prescriptionfrequencies in the Netherlands, whichare still relatively low compared withworldwide prescription frequencies (35).Finally, we observed that other subject

ORIGINAL ARTICLE

306 American Journal of Respiratory and Critical Care Medicine Volume 190 Number 3 | August 1 2014

characteristics were slightly skewedbetween groups (i.e., frequency of indoorsmoking, the presence of siblings, anddelivery mode). Although we performedstringent multivariate modeling correctingour primary outcome measures forpotential confounders, new prospectivestudies should be conducted to verify theobserved correlations.

In conclusion, our data strongly suggesta difference of breastfeeding on bacterialcolonization of the upper respiratory tract of

6-week-old infants, which may contribute tothe protective effect of breastfeeding onrespiratory infections and wheezing inearly infancy. More studies are warrantedto understand the implications of thesefindings for respiratory health and disease.n

Author disclosures are available with the textof this article at www.atsjournals.org.

Acknowledgment: The authors are thankful toElske J. M. van Gils, Gerwin D. Rodenburg, and

the research team who performed therandomized controlled trial. Furthermore theyacknowledge the laboratory logisticsorganized and supervised by Jacob Bruin ofthe Streeklaboratorium Haarlem. They arealso thankful to Jordy Coolen and MartienP. M. Caspers for their assistance in Mothurand other post-sequencing analyses.Overall, they gratefully acknowledge allstudy and laboratory staff and collaboratinginstitutes for their dedication to thisproject. Most of all they thank the childrenand their families that have made this studypossible.

References

1. Tarrant M, Kwok M-K, Lam T-H, Leung GM, Schooling CM. Breast-feeding and childhood hospitalizations for infections. Epidemiology2010;21:847–854.

2. Duijts L, Jaddoe VWV, Hofman A, Moll HA. Prolonged and exclusivebreastfeeding reduces the risk of infectious diseases in infancy.Pediatrics 2010;126:e18–e25.

3. Lopez-Alarcon M, Villalpando S, Fajardo A. Breast-feeding lowers thefrequency and duration of acute respiratory infection and diarrheain infants under six months of age. J Nutr 1997;127:436–443.

4. Paricio Talayero JM, Lizan-Garcıa M, Otero Puime A, BenllochMuncharaz MJ, Beseler Soto B, Sanchez-Palomares M, SantosSerrano L, Rivera LL. Full breastfeeding and hospitalization as a resultof infections in the first year of life. Pediatrics 2006;118:e92–e99.

5. WHO Collaborative Study Team on the Role of Breastfeeding on thePrevention of Infant Mortality. Effect of breastfeeding on infant andchild mortality due to infectious diseases in less developed countries:a pooled analysis. Lancet 2000;355:451–455.

6. Duijts L, Ramadhani MK, Moll HA. Breastfeeding protects againstinfectious diseases during infancy in industrialized countries.A systematic review. Matern Child Nutr 2009;5:199–210.

7. Labbok MH, Clark D, Goldman AS. Breastfeeding: maintaining anirreplaceable immunological resource. Nat Rev Immunol 2004;4:565–572.

8. Bogaert D, De Groot R, Hermans PWM. Streptococcus pneumoniaecolonisation: the key to pneumococcal disease. Lancet Infect Dis2004;4:144–154.

9. Hokama T, Hamamoto I, Takenaka S, Hirayama K, Yara A, Adjei A.Throat microflora in breastfed and formula-fed infants. J Trop Pediatr1996;42:324–326.

10. Kaleida PH, Nativio DG, Chao HP, Cowden SN. Prevalence of bacterialrespiratory pathogens in the nasopharynx in breast-fed versusformula-fed infants. J Clin Microbiol 1993;31:2674–2678.

11. Penders J, Thijs C, Vink C, Stelma FF, Snijders B, Kummeling I, van denBrandt PA, Stobberingh EE. Factors influencing the composition of theintestinal microbiota in early infancy. Pediatrics 2006;118:511–521.

12. Azad MB, Konya T, Maughan H, Guttman DS, Field CJ, Chari RS,Sears MR, Becker AB, Scott JA, Kozyrskyj AL; CHILD StudyInvestigators. Gut microbiota of healthy Canadian infants: profiles bymode of delivery and infant diet at 4 months. CMAJ 2013;185:385–394 10.1503/cmaj.121189.

13. Biesbroek G, Bosch AATM, Keijser BJF, Wang X, van Gils EJM,Veenhoven RH, Sanders EAM, Bogaert D. Effect of feeding type onthe composition of the nasopharyngeal microbiota in infants.Presented at the Dutch Society of Microbiology annual meeting.April 16, 2013, Arnhem, The Netherlands. Poster presentation.

14. Biesbroek G, Bosch AATM, Keijser BJF, Wang X, van Gils EJM,Veenhoven RH, Sanders EAM, Bogaert D. Effect of feeding type onthe composition of the nasopharyngeal microbiota in infants.Presented at the European Society for Pediatric Infectious Diseasesannual meeting. May 30, 2013, Milan, Italy. Oral presentation.

15. Biesbroek G, Bosch AATM, Keijser BJF, Wang X, Veenhoven RH,Sanders EAM, Bogaert D. Effect of feeding type on the composition

and dynamics of nasopharyngeal microbiota in infants. Presented atthe International Human Microbiome Congress. March 19, 2012,Paris, France. Poster presentation.

16. van Gils EJM, Veenhoven RH, Hak E, Rodenburg GD, Bogaert D,Ijzerman EPF, Bruin JP, van Alphen L, Sanders EAM. Effect ofreduced-dose schedules with 7-valent pneumococcal conjugatevaccine on nasopharyngeal pneumococcal carriage in children:a randomized controlled trial. JAMA 2009;302:159–167.

17. van Gils EJM, Hak E, Veenhoven RH, Rodenburg GD, Bogaert D, Bruin JP,van Alphen L, Sanders EAM. Effect of seven-valent pneumococcalconjugate vaccine on Staphylococcus aureus colonisation in arandomised controlled trial. PLoS ONE 2011;6:e20229.

18. van Gils EJM, Veenhoven RH, Rodenburg GD, Hak E, SandersEAM. Effect of 7-valent pneumococcal conjugate vaccine onnasopharyngeal carriage with Haemophilus influenzae and Moraxellacatarrhalis in a randomized controlled trial. Vaccine 2011;29:7595–7598.

19. Biesbroek G, Sanders EAM, Roeselers G, Wang X, Caspers MPM,Trzciński K, Bogaert D, Keijser BJF. Deep sequencing analyses oflow density microbial communities: working at the boundary ofaccurate microbiota detection. PLoS ONE 2012;7:e32942.

20. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB,Lesniewski RA, Oakley BB, Parks DH, Robinson CJ, et al.Introducing Mothur: open-source, platform-independent,community-supported software for describing and comparingmicrobial communities. Appl Environ Microbiol 2009;75:7537–7541.

21. Wang X, Eijkemans MJC, Wallinga J, Biesbroek G, Trzciński K, SandersEAM, Bogaert D. Multivariate approach for studying interactionsbetween environmental variables and microbial communities.PLoS ONE 2012;7:e50267.

22. Verhasselt V. Neonatal tolerance under breastfeeding influence. CurrOpin Immunol 2010;22:623–630.

23. Nair H, Simões EA, Rudan I, Gessner BD, Azziz-Baumgartner E, ZhangJS, Feikin DR, Mackenzie GA, Moısi JC, Roca A, et al.; Severe AcuteLower Respiratory Infections Working Group. Global and regionalburden of hospital admissions for severe acute lower respiratoryinfections in young children in 2010: a systematic analysis. Lancet2013;381:1380–1390.

24. Bogaert D, Keijser B, Huse S, Rossen J, Veenhoven R, van Gils E, BruinJ, Montijn R, Bonten M, Sanders E. Variability and diversity ofnasopharyngeal microbiota in children: a metagenomic analysis.PLoS ONE 2011;6:e17035.

25. Gorbach SL. Lactic acid bacteria and human health. Ann Med 1990;22:37–41.

26. Hunt KM, Foster JA, Forney LJ, Schutte UME, Beck DL, Abdo Z, FoxLK, Williams JE, McGuire MK, McGuire MA. Characterization of thediversity and temporal stability of bacterial communities in humanmilk. PLoS ONE 2011;6:e21313.

27. Laufer AS, Metlay JP, Gent JF, Fennie KP, Kong Y, Pettigrew MM.Microbial communities of the upper respiratory tract and otitis mediain children. MBio 2010;2:e00245–10–e00245–10.

28. Hilty M, Qi W, Brugger SD, Frei L, Agyeman P, Frey PM, Aebi S,Muhlemann K. Nasopharyngeal microbiota in infants with acute otitismedia. J Infect Dis 2012;205:1048–1055.

ORIGINAL ARTICLE

Biesbroek, Bosch, Wang, et al.: Breastfeeding and Infant Nasopharyngeal Microbiota 307

29. Holgerson PL, Vestman NR, Claesson R, Ohman C, Domellof M,Tanner ACR, Hernell O, Johansson I. Oral microbial profilediscriminates breast-fed from formula-fed infants. J PediatrGastroenterol Nutr 2013;56:127–136.

30. Koenig JE, Spor A, Scalfone N, Fricker AD, Stombaugh J, Knight R,Angenent LT, Ley RE. Succession of microbial consortia in thedeveloping infant gut microbiome. Proc Natl Acad Sci USA 2011;108:4578–4585.

31. Spor A, Koren O, Ley R. Unravelling the effects of the environment andhost genotype on the gut microbiome. Nat Rev Microbiol 2011;9:279–290.

32. Xu Q, Almudervar A, Casey JR, Pichichero ME. Nasopharyngeal bacterialinteractions in children. Emerg Infect Dis 2012;18:1738–1745.

33. Grice EA, Segre JA. The skin microbiome. Nat Rev Microbiol 2011;9:244–253.

34. Jeurink PV, van Bergenhenegouwen J, Jimenez E, Knippels LMJ,Fernandez L, Garssen J, Knol J, Rodrıguez JM, Martın R. Human milk:a source of more life than we imagine. Benef Microbes 2013;4:17–30.

35. de Bont EGPM, van Loo IHM, Dukers-Muijrers NHTM, Hoebe CJPA,Bruggeman CA, Dinant G-J, Cals JWL. Oral and topical antibioticprescriptions for children in general practice. Arch Dis Child 2013;98:228–231.

ORIGINAL ARTICLE

308 American Journal of Respiratory and Critical Care Medicine Volume 190 Number 3 | August 1 2014