Expression of the blood-group-related glycosyltransferase B4galnt2 influences the intestinal microbiota in mice.

0.62 MB·English

by  Staubach, Fabian

TheISMEJournal(2012)6,1345–1355
&2012InternationalSocietyforMicrobialEcology Allrightsreserved1751-7362/12
www.nature.com/ismej
ORIGINAL ARTICLE
Expression of the blood-group-related
glycosyltransferase B4galnt2 influences
the intestinal microbiota in mice
Fabian Staubach1,2, Sven Ku¨nzel1, Andrea C Baines3, Andrew Yee4, Beth M McGee5,
Fredrik Ba¨ckhed6, John F Baines1,7,10 and Jill M Johnsen8,9,10
1Department of Evolutionary Genetics, Max Planck Institute for Evolutionary Biology, Plo¨n, Germany;
2DepartmentofBiology,StanfordUniversity,Stanford,CA,USA;3DepartmentofHumanGenetics,University
of Michigan, Ann Arbor, MI, USA; 4Life Sciences Institute, University of Michigan, Ann Arbor, MI, USA;
5Howard Hughes Medical Institute, Ann Arbor, MI, USA; 6Department of Molecular and Clinical Medicine,
Sahlgrenska Center for Cardiovascular and Metabolic Research/Wallenberg Laboratory, University of
Gothenburg, Gothenburg, Sweden; 7Institute for Experimental Medicine, Christian-Albrechts-University
of Kiel, Kiel, Germany; 8Department of Medicine, Research Institute, Puget Sound Blood Center, Seattle,
WA, USA and 9Department of Medicine, University of Washington, Seattle, WA, USA
Glycans on mucosal surfaces have an important role in host–microbe interactions. The locus
encoding the blood-group-related glycosyltransferase b-1,4-N-acetylgalactosaminyltransferase 2
(B4galnt2) is subject to strong selective forces in natural house-mouse populations that contain a
common allelic variant that confers loss of B4galnt2 gene expression in the gastrointestinal (GI)
tract.Wereasonedthatalteredglycan-dependentintestinalhost–microbeinteractionsmayunderlie
these signatures of selection. To determine whether B4galnt2 influences the intestinal microbial
ecology, we profiled the microbiota of wild-type and B4galnt2-deficient siblings throughout the GI
tract using 16S rRNA gene pyrosequencing. This revealed both distinct communities at different
anatomic sites and significant changes in composition with respect to genotype, indicating a
previously unappreciated role of B4galnt2 in host–microbial homeostasis. Among the numerous
B4galnt2-dependent differences identified in the abundance of specific bacterial taxa, we
unexpectedlydetectedadifferenceinthepathogenicgenus,Helicobacter,suggestingHelicobacter
spp.alsointeractwithB4galnt2glycans.Incontrasttootherglycosyltransferases,wefoundthatthe
hostintestinalB4galnt2expressionisnotdependentonpresenceofthemicrobiota.Giventhelong-
term maintenance of alleles influencing B4galnt2 expression by natural selection and the GI
phenotypes presented here, we suggest that variation in B4galnt2 GI expression may alter
susceptibility toGI diseasessuch asinfectious gastroenteritis.
TheISME Journal(2012)6, 1345–1355;doi:10.1038/ismej.2011.204; published online26January 2012
SubjectCategory: microbe–microbe and microbe–host interactions
Keywords: B4galnt2; blood group; glycosylation; intestine; microbiota; mouse
Introduction polymorphic blood group antigen of unknown
function expressed on the red cells of 90% of
Host glycans have been shown to participate
human populations (Conte and Serafini-Cessi,
directly in specific host–microbe interactions
1991)anddetectableinseveralotherhumantissues,
(Bishop and Gagneux, 2007), and signatures of
including the intestinal mucosa and kidney in 98%
selection observed at the loci of carbohydrate blood
of humans (Morton et al., 1988). The glycosyltrans-
group-relatedgenesinhumans(Calafelletal.,2008;
ferase b-1,4-N-acetylgalactosaminyltransferase 2
Ferrer-Admetlla et al., 2009) are most likely the
(B4galnt2) is responsible for catalyzing the last step
result of host–pathogen interactions (Anstee, 2010).
in the biosynthesis of the Sd(a)/Cad antigen by the
The Sd(a)/Cad carbohydrate determinant is a
addition of an N-acetylgalactosamine (GalNAc)
residueviaab-1,4linkagetoasubterminalgalactose
residue substituted with an a-2,3-linked sialic acid
Correspondence:JMJohnsen,DepartmentofMedicine,Research (Lo et al., 2003; Montiel et al., 2003). In the
Institute,PugetSoundBloodCenter,Seattle,WA98104,USA. intestine, there is evidence of a gradient in GalNAc
E-mail:[email protected]
residues conferred by B4galnt2 from the ileum to
10Theseauthorscontributedequallytothiswork.
the colon (Robbe et al., 2003). The character of
Received16May2011;revised22September2011;accepted30
November2011;publishedonline26January2012 intestinal mucus itself also changes with location,
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1346
both in variation in the types of mucins present Stecher and Hardt, 2010). We performed high-
(Corfield et al., 2001) and in the balance of throughput 16S rRNA gene profiling at multiple
glycolipids and glycoproteins that may be capable intestinal sites in mouse sibling pairs that differed
of acting as B4galnt2 substrates (Dohi et al., 1996; only in the presence or absence of B4galnt2 expres-
Kawamura et al., 2005).Thus, distinct carbohydrate sion. In addition, we analyzed B4galnt2 expression
mucosal phenotypes are likely to result from patterns in germ-free versus conventional mice.
B4galnt2 expression at different intestinal anatomic Here we describe this detailed characterization of
sites. mouse GI bacterial communities in seven distinct
Intestinal expression of B4galnt2 is observed locations from the duodenum to the colon and
throughout vertebrates from fish (Stuckenholz provide evidence for a significant effect of B4galnt2
et al., 2009) to humans, including mice. Although expression on intestinal bacterial populations.
conservation across species is evidence for a func-
tional constraint preventing the loss of intestinal
expression, mice genetically deficient in B4galnt2 Materials and methods
(B4galnt2(cid:2)/(cid:2)) display no obvious phenotype, gastro-
intestinal (GI) or otherwise, under specified patho- Animal material and tissue sampling
gen-free (SPF) conditions in the laboratory (Mohlke All animal protocols (except for the germ-free
et al., 1999; Johnsen et al., 2008). protocols, see below) were approved by the
The absence of intestinal B4galnt2 expression is University of Michigan University Committee on
also observed in mice homozygous for a sponta- the Use and Care of Animals (UCUCA). C57BL6/J
neous cis-regulatory mutation termed Modifier of animals were purchased from the Jackson Labora-
von Willebrand Factor-1 (Mvwf1). Mvwf1 specifi- tory (Bar Harbor, ME, USA). B4galnt2 knockout
cally turns off B4galnt2 expression in the intestinal animals first engineered by Dr John Lowe were
epithelium and instead directs B4galnt2 expression provided with permission and courtesy of Dr David
in vascular endothelium. Endothelial cell expres- Ginsburg. Genetic background, maternal effect,
sionofB4galnt2resultsinaberrantposttranslational housingconditions,genderanddietwereaccounted
modification of the blood-clotting glycoprotein, von for in the following mating scheme: B4galnt2(cid:2)/
Willebrand factor (VWF), resulting in accelerated (cid:2)animals bred for more than 20 generations to a
VWF clearance and low circulating VWF levels C57BL6/J background were mated to C57BL6/J
(Mohlkeetal.,1999),similartothecommonhuman animals to generate heterozygous B4galnt2þ/
bleedingdisorder,vonWillebranddisease(Sweeney (cid:2)parents. B4galnt2þ/(cid:2) parents were then inter-
et al., 1990). As in the B4galnt2 knockout mouse, crossed to generate B4galnt2þ/þ and B4galnt2(cid:2)/(cid:2)
loss of intestinal B4galnt2 expression due to Mvwf1 malesiblingoffspring.Four‘sibpair’cagescontaining
does not result in an obvious GI phenotype in the atleastone B4galnt2þ/þ andone B4galnt2(cid:2)/(cid:2) sibling
laboratory. Mvwf1 is common among laboratory were raised and housed together with standard
mouse strains (Johnsen et al., 2008), and is present mouse chow in the same room in a specified
at intermediate frequencies in wild Mus musculus pathogen-free(SPF)animalfacilityatthesametime.
domesticus(housemouse)populations,wherethere Studies were performed at 10 weeks of age to
is strong evidence of both recent and long-term ensure adequate time for the colonization and
selectionattheB4galnt2locus(Johnsenetal.,2009). development of mature, stable microbiota (Rehman
These observations seem paradoxical: Mvwf1 et al., 2011). Tissues were harvested by fresh
confers what would be expected to be a detrimental dissection of 1cm of bowel at each of the following
phenotype in M. musculus (a mild bleeding dia- defined anatomic sites: mid-duodenum (D2), mid-
thesis), yet the Mvwf1 allele is common in wild jejunum, terminal ileum (ending at the cecal valve),
mice. The prevalence of natural murine B4galnt2 cecum and mid-descending colon. With the excep-
variants isfurtherillustratedinourrecent surveyof tion of the cecum, luminal contents were separated
multiple wild-mouse populations, in which we frommucosalspecimensby flayingopenthebowel,
found loss of B4galnt2 intestinal glycans conferred washing twice with 0.7ml ice-cold RNALater
by the Mvwf1 allele class to be frequent, but not (Ambion, Carlsbad, CA, USA) and the 1.4-ml
always accompanied by the gain of B4galnt2 effluent containing stool pooled and marked as
vascular expression (Linnenbrink et al., 2011). ‘luminal contents’. The bowel was then placed in
We hypothesized that loss of intestinal B4galnt2 1.4ml ice-cold RNALater and marked as ‘mucosa’.
expression is at least in part responsible for the For the cecal samples, the cecal pouch was
strikingsignaturesofselectionattheB4galnt2locus dissected and flayed open, the entire specimen
in wild house mice. In this study, we sought to placedina15-mltubewith4mlice-coldRNALater,
determine the influence of B4galnt2 expression on andthespecimengentlyshakentoremovethececal
the resident microbiota throughout the GI tract as contents. The mucosa was retrieved with forceps
evidence of an intestinal B4galnt2 phenotype, as and rinsed with an additional 0.7ml of ice-cold
shifts in the intestinal microbiota are known to RNALater. This rinse was added to the 4-ml stool
affect host susceptibility to pathogens and disease solution and marked ‘luminal contents’. The rinsed
(Bishop and Gagneux, 2007; Sekirov et al., 2010; cecal tissue was then placed in 1.4ml ice-cold
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1347
RNALater and marked ‘mucosa’. Owing to varia- amountsofpurifiedPCRproductfromsevenknock-
bility in the presence of luminal contents in the out and six wild-type individuals were pooled to
sections of the small intestine, only the contents of generatelibrariesforeachanatomiclocation.Pooled
the cecum and descending colon were analyzed. amplicon libraries were further purified using
Instruments were cleaned between each anatomic Ampure Beads (Agencourt, Bernried, Germany). A
site to avoid cross-contamination. sample of each library was run on an Agilent
(Waldbronn, Germany) bioanalyzer as described in
the Roche Titanium Amplicon Sequencing protocol
Germ-free animals beforeenteringemulsionPCRandsequencing.Each
C57BL/6J animals were raised under conven- library was sequenced on one-eighth of a picotiter
tional or germ-free conditions at the University of plate on a 454 GS-FLX (Roche).
Gothenberg. All animal protocols were approved
by the Research Animal Ethics Committee in
Gothenburg. At 6 weeks of age, small bowel was
harvested for Dolichos biflorus agglutinin (DBA) Sequence analysis
lectin histochemistry as described below. Rawreadsobtainedfromthesequencerwerefiltered
using a perl script according to the following
criteria: average quality X25, no ambiguous bases,
DBA lectin histochemistry length between 110 and 200bp, excluding MID tags
DBA lectin is reactive with terminal N-acetyl-D- and bacterial primers, and a perfect match to the
galactosamine (GalNAc) residues and specifically MID and bacterial primer. Primerand tag sequences
detects the Mvwf1 switch in B4galnt2 expression were removed using alignment with the Needle-
fromtheintestinetothebloodvessel(Mohlkeetal., man–Wunsch algorithm (Needleman and Wunsch,
1999). Fresh small bowel harvested from C57BL6/J 1970). Sequences were sorted by individuals to
andB4galnt2(cid:2)/(cid:2)animalswasfixedinZ-fix(Anatech produce group files for analysis in MOTHUR
Ltd, Battle Creek, MI, USA) at room temperature v.1.12.3(Schloss, 2009). Classificationinto bacterial
overnight, then paraffin embedded. DBA lectin phyla was performed with the Ribosomal Database
histochemicalstainingwasperformedonembedded Project classifier tool (Wang et al., 2007) and the
tissues using horseradish peroxidase-conjugated RDP taxonomy database as implemented in
DBA (EY Laboratories, Inc., San Mateo, CA, USA) MOTHUR. Reads were aligned with the kmer
as previously described (Mohlke et al., 1999). algorithm available under the align.seqs command
(Schloss, 2009) in MOTHUR to the silva reference
database (Pruesse et al., 2007). Sequences that did
DNA extraction not match the reference alignment in the expected
Approximately100mg oftissue orluminal contents positions were removed.
was transferred to a 2-ml screwcap tube containing
glass beads 0.1, 0.5 and 1mm in size, each 50mg
(BioSpec Products, Bartlesville, OK, USA), and
1.4ml lysis buffer argininosuccinate lyase from Statistical analysis
the QIAmp DNA stool mini kit (Qiagen, Hilden, Aligned sequences were used to build a distance
Germany).After beadbeatinginaPrecellys (Peqlab, matrix and group sequences into operational taxo-
Erlangen, Germany) bead beater (3(cid:3)15s at nomic units (OTUs) in MOTHUR. Species richness
6500r.p.m.), samples were heated to 951C for estimates and collector’s curves were generated
10min and constantly shaken in a thermomixer based on these OTUs and drawn using the R
(Eppendorf). Bacterial DNA was then extracted statistics package v.2.11.1 (R Development Core
using the QIAmp DNA stool mini kit (Qiagen) Team, 2010). Phylogenetic trees were built based
following the manufacturer’s instructions. on the MOTHUR-derived distance matrices with
FastTree v. 2.1.3 (Price et al., 2009, 2010) and
submitted to the Fast Unifrac online tool for
PCR and pyrosequencing principal coordinate analysis (PCoA) of Unifrac
Universal bacterial primers for the V3 16S rRNA distances(Hamadyetal.,2009).Linearmodelswere
variable region were used to amplify bacterial 16S fitted to the data using the ‘lm’ function implemen-
RNA as described by Dethlefsen et al. (2008). PCR ted in R. Two mixed effect models were made with
ampliconprimerswerefusedtomolecularidentifier the ‘lme’ function contained in the nlme package v.
(MID)tagsand454sequencingadaptersaccordingto 3.1-96(Pinheiroetal.,2009)toaccountforpotential
the manufacturer’s instructions (Roche, Basel, Swit- maternal (cage of origin) effects (models 6 and 7 in
zerland).AllPCR reactions were performed in50-ml Table 1). To compare species richness between
duplicates and combined after PCR. The products different sections of the gut, we subsampled the
were extracted with the Qiagen MinElute Gel same number of reads per individual and section
Extraction Kit and quantified with the Quant-iT (in silico capping), calculated Chao’s richness
(Invitrogen, Karlsruhe, Germany) dsDNA BR Assay estimator (Chao, 1984) and compared the number
Kit on a Nanodrop 3300 fluorometer. Equimolar of OTUs observed in the subsamples.
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1348
Table1 LinearmodelcomparisonsfortheeffectofgenotypeonUnifracPCos
Model Dependentvariable Explanatoryvariables AIC P-value RSS Rsq
1 UnifracPCo1 Nullmodel 16.66 5.66
2 UnifracPCo1 Gutsection (cid:2)145.86 1.20E(cid:2)32 0.69 0.88
3 UnifracPCo2 Nullmodel (cid:2)115.81 1.15
4 UnifracPCo2 Gutsection (cid:2)213.42 6.87E(cid:2)20 0.31 0.73
5 UnifracPCo2 Gutsection+genotype (cid:2)217.864 0.016a 0.28 0.07a
6b UnifracPCo2 Fixed¼genotype NA 0.017 NA NA
Random¼section/cage
7b UnifracPCo2 Fixed¼genotype NA 0.019 NA NA
Random¼section/cage/genotype
Abbreviations:Cage,cageoforigin(maternaleffect);NA,notapplicableinthatcontext;section,locationalongtheintestine.
aANOVAincomparisonwithmodel4.
bLinearmixedeffectsmodel.
P-valuesbyANOVA.
Candidate bacterial species distinguishing the
genotypes were determinedby calculatingthepoint
biseral correlation between species-level (97%
sequence identity) OTUs and genotype with the
‘multipatt’ function as part of the ‘indicspecies’
R package v. 1.5.1 (De Caceres and Legendre, 2009).
This approach considers not only the difference in
read number assigned to an OTU between geno-
types, but also the number of individuals of each
genotype in which the OTU is observed, thus Figure 1 Small-bowel DBA lectin staining. (a) Wild-type
drawing statistical power from biological replicates. (C57BL6/J) intestine exhibits robust intestinal epithelial DBA
The number of sequence reads per sample was lectin staining (brown). Mucus (arrows indicate mucus-contain-
ing goblet cells) is strongly DBA lectin positive. (b) B4galnt2(cid:2)/(cid:2)
capped in silico by random subsampling to obtain
bowel demonstrates complete loss of DBA lectin staining of the
equalsamplesizesforspeciesrichness,Unifracand
epitheliumandmucus.
for detection of candidate species. To provide an
initial overview of the sequence reads associated
with each region of the GI tract, we classified breeding scheme (see Materials and methods)
sequences by aligning them to the SILVA reference yielded a total of six B4galnt2þ/þ and seven
database (Pruesse et al., 2007) and obtained taxo- B4galnt2(cid:2)/(cid:2) individuals. We surveyed the bacterial
nomicinformationfromRDPClassifier(Wangetal., communities associated with the mucus layer of
2007). five major functional compartments of the murine
intestine (duodenum, jejunum, ileum, cecum and
descending colon) and the luminal contents of the
Results
cecum and colon. The 16S rRNA V3 region was
successfully amplified from all individuals at all
DBA lectin staining
sites, with the exception of the duodenum of one
DBA lectin staining of the small bowel from
B4galnt2(cid:2)/(cid:2) individual (see Supplementary Table 1
C57BL6/J animals demonstrates strong DBA reacti-
fordistributionofreadsperGItractlocation).Atotal
vityinintestinalepithelialcells,themucusmaterial
of337449readsofthehypervariableregionV3were
contained in goblet cells, and the lining of the
analyzed afterqualityfiltering.AsingleB4galnt2(cid:2)/(cid:2)
intestinalluminalsurface.NoDBAreactivityisseen
inthesmallbowelofB4galnt2(cid:2)/(cid:2)animals(Figure1). individualdisplayedastronglydeviantpatternwith
respect to the relative abundance of the major
ThesedatashowthatDBAlectinhistochemistryisa
bacterial phyla compared with all 12 other mice in
specific method for detection of B4galnt2-carbohy-
the small intestine (Supplementary Figure 1),
drates in the bowel, which are evident on the
despite no notable phenotype at the time of dissec-
intestinal epithelium and mucus of B4galnt2-suffi-
cient animals but absent in B4galnt2(cid:2)/(cid:2) animals. tion. This individual was housed with two other
siblings, one B4galnt2þ/þ and one B4galnt2(cid:2)/(cid:2),
which did not exhibit this pattern. Although we
High-throughput sequencing of the B4galnt2þ/þ and cannotexcludethatthisaberrationisaconsequence
B4galnt2(cid:2)/(cid:2) intestinal microbiota oflossofB4galnt2expression,wespeculatethatthis
To determine the influence of B4galnt2 expression animal was subclinically ill or otherwise compro-
on the resident microbiota throughout the GI tract, mised, resulting in a shift distinct from other
we performed high-throughput pyrosequencing genotypes. Thus, we removed this individual from
(454) of the bacterial 16S rRNA gene. Our sibpair further analyses.
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1349
Bacterial phyla of the GI tract by anatomic site locationsof the GI tract and in the luminalcontents
The communities within thedistinct sectionsof the of the cecum and colon (Supplementary Figures 2
GI tract differ largely in their composition and and 3). For this analysis, sequences were grouped
proportion of the major bacterial phyla (Figure 2). intospecies-level(97%similarity)OTUsandChao’s
The mucosa of the small intestine (duodenum, species richness estimator was calculated.
jejunum and ileum) is dominated by Firmicutes Although we detected no significant difference in
(77%, 78% and 91%, respectively). As previously species richness with respect to B4galnt2 genotype
reported by Ley et al. (2005), we also detected in any of the sections, several interesting patterns
sequences of Cyanobacteria origin in the small are apparent with regard to the individual GI tract
intestine (1.5% in duodenum and 2.1% in the locations.Speciesrichnessishighestintheluminal
jejunum). The mucosa of the cecum and colon samples and on average lower in the mucus layer
contained a comparatively lower proportion of (335 versus 187 species, Chao’s species richness
Firmicutes (40% and 25%, respectively) and higher estimator),consistentwiththeprotectivefunctionof
proportion of Bacteroidetes (19% and 15%, respec- the mucus layer and a more controlled interaction
tively) and Proteobacteria (37% and 59%, respec- between host and microbes closer to the intestinal
tively). The cecum samples also contained epithelium. The high richness observed in the
Deferribacteres (2.5%). cecum may relate to its role as a biofermenter
The differences between the mucosa and the offering stable conditions for the growth of diverse
adjacent luminal contents of the cecum and colon species (Savage, 1977). Interestingly, we also
are particularly striking. Cecum and colon mucosal observe a progressive decline of bacterial species
tissues contain 19% and 15% Bacteroidetes, com- richness along the small intestine (201, 168 and 82
pared with 32% and 64%, respectively, in the for duodenum, jejunum and ileum, respectively,
lumen. The opposite pattern is found for Proteo- Chao’s species richness estimator), mainly due to a
bacteria, with 37% and 59%, respectively, in the much lower diversity in the ileum.
cecum and colon mucosal tissues, compared with
8% and 3% in the lumen. Furthermore, the Species composition differs between B4galnt2þ/þ and
Tenericutes are more abundant in the lumen. This B4galnt2(cid:2)/(cid:2) mice
remarkable distinction between the mucosal and To test whether the composition of bacterial com-
luminal communities might be expected in light of munities is influenced by B4galnt2 expression,
previousfindingsthatbacterialpopulationsoccupy- we compared bacterial communities between
ing human colonic mucosa are distinct from those B4galnt2þ/þ and B4galnt2(cid:2)/(cid:2) mice using the phylo-
present in fecal samples (Eckburg et al., 2005; geny, Bernried, Germany-based beta-diversity mea-
Willing et al., 2010). Our results provide direct sure UniFrac. This metric represents the distance
support for the presence of distinct communities between bacterial communities by comparing the
inhabiting the niches found along the GI tract, shared branch length between phylogenetic trees
including those associated with the tissue versus underlying two communities. A matrix of UniFrac
lumen at a given location. distancesbetweentheindividualmicewasanalyzed
forprincipalcoordinates(PCo)topartitionvariation
among samples into the most important indepen-
Bacterial species richness differs between sections of dentcomponents(Figure3a).Theresultingvariance
the GI tract along the principal coordinates was analyzed in a
Tocharacterizethediversityofthesecommunitiesat linear model framework with respect to the experi-
lower taxonomic levels, we tested for differences in mental setup (Figure 3b, Table 1).
bacterial species richness in the B4galtn2þ/þ and Our model identifies B4galnt2 genotype as a
B4galnt2(cid:2)/(cid:2) mice at the five mucosal anatomic significantdeterminantofPCo2,andhencebacterial
duodenum jeiunum ileum cecum colon cec.cont col.cont
100
80
60
40 Firmicutes
Bacteroidetes
Proteobacteria
Cyanobactera
20 Actinobacteria
Deferribacteres
TM7
Tenericutes
0 others
+/+ -/- +/+ -/- +/+ -/- +/+ -/- +/+ -/- +/+ -/- +/+ -/-
Figure2 Abundanceofmajorbacterialphylabyanatomicsite.Averageabundanceofsequencingreadsfrommajorbacteriaphylain þ/þ
(B4galnt2þ/þ)and(cid:2)/(cid:2)(B4galnt2(cid:2)/(cid:2))micealongtheintestineinpercentoftotalbacterialsequencesobtained.cec.cont,cecumcontent;
col.cont,coloncontent.
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1350
duodenum jejunum ileum
0.15 0.00
0.25
0.3 0.20 0.10
duodenum -0.10
jejunum 0.15 0.05
ileum
0.2 cceeccuumm con. 00..0150 0.00 -0.20
colon
colon con. +/+ -/- +/+ -/- +/+ -/-
0.1 cecum cecum content
% 0.10
90 0.10
o2 5. 0.0 0.05 0.05
C
P 0.00 0.00
-0.1 -0.05 -0.05
+/+ -/- +/+ -/-
colon colon content
-0.2
0.00 -0.05
-0.05 -0.10
-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4
-0.10
PCo2 29.12% -0.15
-0.15
-0.20
+/+ -/- +/+ -/-
Figure 3 Principal coordinate analysis (PCoA) of Unifrac distances across anatomic sites and genotypes. (a) PCoA of unweighted
Unifrac distances for the whole dataset (all anatomic sites sampled). Colors by section, filled symbols¼þ/þ; open symbols¼(cid:2)/(cid:2);
symbolshapebycageoforigin.(b)BoxplotofPCo2froma,boldhorizontallinesrepresentthemedian,theboxedgesmarktheupperand
lowerquartiles,whiskersextendtowardsthemaximumandminimum;colorcodeasina.
community composition. Both a linear model
including section and genotype as additive effects
(model 5, Table 1) and the application of mixed-
effects models to account for maternal (cage of
origin) effects(models6and7)assign genotype asa
significant explanatory effect of community compo-
sition. The effect is visualized in Figure 3b:
ThevaluealongPCo2assignedtoB4galnt2þ/þ mice
is higher for all portions of the GI tract except
the colon, indicating a systematic effect of B4galnt2
expression on bacterial communities along the
GI tract.
This analysis also reveals the striking effect of
anatomic location on bacterial composition along
the GI tract. The individual locations of the gut
analyzed in our study differ strongly in the compo-
sition of their associated bacteria, accounting for
88%and73%,respectively,ofthevarianceobserved
along the first two PCos.
DifferencesinthebacterialcommunityofB4galnt2þ/þ Figure 4 Unweighted Unifrac PCoA of the ileal mucosal
and B4galnt2(cid:2)/(cid:2) mice by location bacterial community. Solid symbols¼B4galnt2þ/þ, empty
Because distinct locations along the GI tract differ symbols¼B4galnt2(cid:2)/(cid:2).Symbolshapebycageoforigin.
largely in their bacterial composition, genotype-
dependent differences between the bacterial com-
munities of B4galnt2þ/þ and B4galnt2(cid:2)/(cid:2) mice are Bonferronicorrectionfortestingallsectionsandthe
likely to be site-specific. Thus, to analyze the first two PCos, analysis of variance).
potential influence of B4galnt2 on bacterial com-
munitiesindependentofvariationbetweensections,
we applied UniFrac to each anatomic location Individual bacterial species associated with
separately. We find the largest genotype effect on B4galnt2þ/þ and B4galnt2(cid:2)/(cid:2) genotypes
thebacterialcommunitiesassociatedwiththeileum Toshedlightontheidentityofbacterialspeciesthat
(Figure 4, P¼7.49(cid:3)10(cid:2)5, R2¼0.81, P¼0.001 after make up some of the differences between the
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1351
intestinal bacterial communities associated with DBA lectin staining was performed on the small
B4galnt2þ/þ and B4galnt2(cid:2)/(cid:2) mice, we applied a bowels of mice housed under conventional and
common ecological measure of species habitat germ-free conditions. Intestinal DBA lectin staining
association (Dufre´ne and Legendre, 1997) to OTU was present and appeared similar under both
clusters (97% sequence identity). OTUs identified conventional and germ-free conditions (data not
asindicatorsofB4galnt2expression‘habitat’(thatis, shown), indicating that B4galnt2 expression occurs
those consistently present or more abundant in in the absence of intestinal bacteria.
one genotype compared with the other) were
classified using the Ribosomal Database Project tool
(Wangetal.,2007).Numerousinterestingcandidate Discussion
taxa belonging to the three major phyla Bacteroi-
detes,FirmicutesandProteobacteriawereidentified At birth the intestinal tract is sterile, but is rapidly
bythismethod(Table2).MembersoftheFirmicutes colonized by a diverse spectrum of bacteria, in
appear to be the most widely influenced. A total of addition to archaea and eukaryotes. These organ-
11 indicator OTUs belonging to the classes Clostri- isms are provided a nutrient-rich and largely stable
diales or Lactobacillales were identified from this environment by the host. In turn, the host relies on
group, with at least one member differentiating the microbiota for a variety of metabolic processes
B4galnt2 expression habitats in all seven sampled and their presence is required for normal host
locations of the GI tract. The Bacteroidetes con- intestinaldevelopment,mucosalintegrityandmain-
tained eight OTUs distributed across all jejunum, tenance of immunologic balance (Backhed et al.,
cecum and the luminal contents of the cecum and 2005; O’Hara and Shanahan, 2006; Artis, 2008;
colon, five of which belonged to the genus Barne- Fraser et al., 2009). Furthermore, these complex
siella. Four OTUs belonging to the Proteobacteria communities protect the host from pathogenic
wereidentifiedintheduodenum,colonandluminal organisms in several ways: by occupying microbial
contentsofthecolon,threeofwhichbelongedtothe niches resulting in displacement, by production of
genus Helicobacter. antimicrobialfactorsandbycompetingfornutrients
and receptors (O’Hara and Shanahan, 2006). By
systematically profiling multiple locations through-
Intestinal expression of B4galnt2 does not require out the murine GI tract, we have characterized
the presence of bacteria distinctmicrobialcommunitiesatdiscreteanatomic
Expression of intestinal glycosyltransferases, such sites throughout the intestine and found striking
as the Fut2 glycosyltransferase, has been shown to differences between sites and between the mucosa
be influenced by the microbiota present (Bry et al., and adjacent luminal contents. Although this is
1996;Mengetal.,2007).Thisregulatorymechanism seeminglyincontrasttoapreviousreportthatfound
likely has an important role in the host’s ability to no difference between the populations of the
alterthemucosalsurfaceinresponsetotheenviron- mucosa and lumen in a humanized mouse model
ment. To determine whether B4galnt2 expression (Turnbaugh et al., 2009), this apparent discrepancy
similarlyrequiresthepresenceofintestinalbacteria, maybeattributabletothecurrentstudy beingbased
Table2 CandidatebacterialspeciesasdeterminedbyOTUgenotypecorrelation
Section Genotype OTU r P Seq Taxonomy
Duodenum +/+ 154 0.76 0.02 6 Campylobacterales(100);Helicobacteraceae(100);Helicobacter(100)
+/+ 134 0.712 0.03 14 Lactobacillales(100);Streptococcaceae(100);Lactococcus(100)
(cid:2)/(cid:2) 37 0.844 0.003 15 Lactobacillales(100);Streptococcaceae(100);Streptococcus(80)
(cid:2)/(cid:2) 63 0.392 0.037 76 Pseudomonadales(100);Moraxellaceae(100);Acinetobacter(100)
Jejunum +/+ 135 0.579 0.039 171 Clostridiales(100);Clostridiaceae(100)
(cid:2)/(cid:2) 48 0.496 0.048 22 Bacteroidales(100);Porphyromonadaceae(100);Barnesiella(100)
Ileum (cid:2)/(cid:2) 62 0.707 0.015 8 Clostridiales(100);Clostridiaceae(100)
Cecum +/+ 595 0.768 0.018 7 Clostridiales(100);Lachnospiraceae(100);Moryella(86)
(cid:2)/(cid:2) 37 0.622 0.029 59 Clostridiales(100);Lachnospiraceae(100);Moryella(100)
(cid:2)/(cid:2) 103 0.61 0.024 41 Bacteroidales(100);Porphyromonadaceae(100)
(cid:2)/(cid:2) 110 0.481 0.048 70 Bacteroidales(100);Porphyromonadaceae(100);Barnesiella(100)
Colon +/+ 123 0.704 0.046 9 Lactobacillales(100)
+/+ 30 0.597 0.044 15 Campylobacterales(100);Helicobacteraceae(100);Helicobacter(100)
Cecumcont. +/+ 331 0.763 0.012 44 Clostridiales(100)
+/+ 499 0.742 0.017 9 Clostridiales(100);Lachnospiraceae(100)
(cid:2)/(cid:2) 192 0.671 0.048 8 Clostridiales(100);Lachnospiraceae(100);Syntrophococcus(88)
(cid:2)/(cid:2) 105 0.632 0.047 13 Bacteroidales(100);Porphyromonadaceae(100);Barnesiella(100)
Coloncont. +/+ 23 0.632 0.027 17 Bacteroidales(100);Porphyromonadaceae(100);Barnesiella(100)
(cid:2)/(cid:2) 192 0.667 0.046 12 Bacteroidales(100);Porphyromonadaceae(100);Barnesiella(100)
Abbreviation:OTU,uniqueidentifierfortherespectiveOTU(artificialspeciesname);sequencenumbercalculationsarebasedontaxonomy;
class,family,genusarelistedaccordingtoRDPtaxonomy;bootstrapvaluesareshowninbrackets.
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1352
on sampling of unperturbed native bacterial diseased mice in an IL10 knockout model of IBD
communities. (Ye et al., 2008). Interestingly, Lachnospiracaea are
Wefindthatvariationintheexpressionofasingle major determinants of the recently reported ‘enter-
glycosyltransferase, B4galnt2, is associated with otypes’ described in a large study of human fecal
significantshiftsinthecompositionoftheintestinal metagenomes (Arumugam et al., 2011). Ye et al.
microbiota. These differences are consistent with (2008) also identified differences in their murine
host carbohydrate-specific selection on colonizing IBD model in the genus Barnesiella, which we too
microbial populations, as B4galnt2(cid:2)/(cid:2) animals and foundtobeinfluencedbyB4galnt2genotype.Others
their control B4galnt2þ/þ littermates shared the have associated Barnesiella with CD8þ T-cell
samemicrobeexposuresovertheirlifetime.Thefact function in mice (Presley et al., 2010), suggesting
thattheileumdisplayedthemostclearseparationin more than a bystander role for this genus in
overallcompositionasmeasuredbytheunweighted intestinal inflammation.
UniFrac metric also suggests B4galnt2-dependent Intriguingly, we found that the indicator OTUs for
immune activity, as strong antibacterial activity is B4galnt2þ/þ and B4galnt2(cid:2)/(cid:2) mice at the same
presentinthistissue(Petnicki-Ocwiejaetal.,2009). location of the gut often were members of the same
In addition to overall changes in composition, we bacterial family or genus (Figure 5). Streptococcus
identify a number of specific bacterial lineages and Lactococcus were indicative of B4galnt2þ/þ and
influenced by host B4galnt2 expression. Although B4galnt2(cid:2)/(cid:2) mucosal communities in the duodenum,
it could be argued that many bacterial orders found as were two different Moryella OTUs in the cecum.
inthebowelhaveatleastonememberthathasbeen Likewise, closely related indicator OTUs distin-
shown to be important for gut health, nearly all the guished by B4galnt2 genotype were found in the
candidatelineagesweidentifiedasbeinginfluenced luminal content of cecum and colon (Lachnospi-
by B4galnt2 expression have been previously iden- raceae and Barnesiella, respectively). Thus, similar
tified as significant to intestinal communities. For species appear to substitute each other depending
example,wediscovereddifferencesinseveralOTUs on B4galnt2 genotype, indicating that many closely
belonging to the order Clostridiales in the jejunum, related species have the potential to occupy distinct
ileum, cecum and cecum content, similar to two B4galnt2 glycan-defined niches in the mucosa.
recentreportsthatfounddistinctClostridialesOTUs Unexpectedly, among the bacteria identified were
in inflammatory bowel disease (IBD) samples com- several OTUs belonging to the genus Helicobacter.
pared with the controls (Willing et al., 2010; Frank Helicobacter spp. are known to naturally infect and
etal.,2011).Furthermore,weobservedifferencesin cause disease in laboratory mice, most commonly
the family Lachnospiraceae (Order: Clostridiales), H.hepaticus,H.bilisandH.typhlonius(Fengetal.,
which was identified to be decreased in patients 2005). Although a difference in Helicobacter abun-
with Crohn’s disease localized to the ileum com- dancewould be expected toresultinanaltered risk
pared with controls (Willing et al., 2010) and also for enteritis or other hepatobiliary inflammatory
reported to be disparate between healthy and processes, in a protected laboratory environment
Duodenal Mucosa Cecal Mucosa
00..001168 O(LTaUct o1c3o4c.1cuMs) 00..000078 O(MToUry 5e9ll5a.)5M
Abundancen 000...0000.11002481 O(STtrUe p3t7o.c1oMccus) Abundancen000...000000564 O(MToUry 3e7ll.a5)M
Relative 000...000000246 Relative 000...000000123
0 0
ko1ko2ko3ko4ko5wt1wt2wt3wt4wt5wt6 ko1ko2ko3ko4ko5ko6wt1wt2wt3wt4wt5 wt6
Colon contents Cecal Contents
0.018 0.006 OTU 499.5S
0.016 OTU 23.6S (Lachnospiraceae)
Relative Abundance 00000.....000000.100010446821 (O(BBTaaUrrnn 1ee9ssi2iee.ll6llaaS)) Relative Abundance00000.....000000000051234 O(LTaUch 1n9o2s.p5irSaceae)
0.002
0 0
ko1ko2ko3ko4ko5ko6wt1wt2wt3wt4wt5wt6 ko1ko2ko3ko4ko5ko6wt1wt2wt3wt4wt5wt6
Figure 5 Individual bacterial species associated with B4galnt2 genotype. OTUs of closely related species appear to substitute by
B4galnt2 genotype (B4galnt2þ/þ¼wt, B4galnt2(cid:2)/(cid:2)¼ko). (a) Duodenal mucosa: OTU 134.1M (Lactococcus sp.), OTU 37.1M
(Streptococcussp.),(b)cecalmucosa:OTU595.5M(Moryellasp.),OTU37.5M(Moryellasp.),(c)coloncontents:OTU23.6S(Barnesiella
sp.),OTU192.6S(Barnesiellasp.),(d)cecalcontents:OTU499.5S(Lachnospiraceae),OTU192.5S(Lachnospiraceae).
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1353
animalswithHelicobactermayexhibitamoresubtle Helicobacter spp.
or subclinical phenotype. In humans, although 0.009
Duodenal mucosa
H. hepaticus and H. bilis are both associated with 0.008 OTU 154M
hepatoenteric disease (Goldman and Mitchell, (Helicobacter spp.)
0.007
2010), the most prevalent and best-studied patho- e Colonic mucosa
c
genic Helicobacter is H. pylori. dan 0.006 O(HTeUlic 3o0bMacter spp.)
H.pyloripredominantlyinhabitsthegastricmucosa n
u 0.005
and is associated with a spectrum of human b
A
intestinal diseases ranging from gastritis to malig- e 0.004
v
nancy(TimothyandMartin,2009).MostH.pyloriis ati 0.003
locatedwithintheoverlyinggastricmucuslayerand el
R 0.002
does not interact with the underlying epithelium.
However, under some conditions H. pylori can 0.001
adhere to the gastric mucosa, triggering virulence
0
factors in the bacterium and an inflammatory
1 2 3 4 5 6 1 2 3 4 5 6
response from the host (Timothy and Martin, ko ko ko ko ko ko wt wt wt wt wt wt
*
2009). Thus, the mere presence of Helicobacter
may result in a bystander effect by which the Figure6 AbundanceofHelicobacterspp.byB4galnt2genotype.
nearby-resident microbiota may be influenced not Relative abundance of OTUs of two Helicobacter spp. in the
duodenum and colon segregated by B4galnt2 genotype (indivi-
only by niche competition with Helicobacter but
duals are the same as in Figure 5, B4galnt2þ/þ¼wt and
also by more general environmental changes due to
B4galnt2(cid:2)/(cid:2)¼ko). *ko6 duodenal mucosal sample was not
mucosal inflammation and upregulation of host availableforanalysis(seeResults).
defenses.
Adhesion to host glycans presented on the gastric disease-causing organisms and/or change specific
mucosa, notably carbohydrate antigens of the ABH binding targets for GI pathogens. For example, even
and Lewis blood group systems, is a critical step in in this study intended to characterize the symbiotic
thepathogenesisofH.pylori(Kobayashietal.,2009). microbiota, we detected a difference in the patho-
Furthermore, H. pylori is known to be capable of genicgenusHelicobacter,whichisknowntoadhere
specificallybindingadditionalcarbohydratemoieties, to other host blood group carbohydrate structures
including terminal b-1,3-GalNAc residues (Miller- during pathogenesis.
Podraza et al., 2005), while other murine entero- In summary, we have found that B4galnt2 expres-
hepatic Helicobacter spp. isolates also demonstrate sioninfluencestheintestinalmicrobiota.Helicobacter
evidence of carbohydrate-specific adhesion (Hynes and other enteric pathogens known to interact with
etal.,2003).Intheduodenalandcolonicmucosa,two GalNAc residues, including protozoans (Entamoeba
indicator Helicobacter spp. were detected in the histolytica (Frederick and Petri, 2005)), viruses
B4galnt2þ/þ animals, yet these Helicobacter were (Norovirus (Shirato et al., 2008)) and a spectrum of
largely undetectable in B4galnt2(cid:2)/(cid:2) individuals bacteria (Krivan et al., 1988; Karlsson, 1995), are
(Figure 6). Thus, we postulate a novel direct interac- excellent candidates to underlie the striking signa-
tion between Helicobacter and host mucosal turesofselectionattheB4glant2locusinwild-mouse
B4galnt2-derivedb-1,4-GalNAcresiduestobealikely populations.
mechanismresponsibleforthesignificantincreasein
abundance of these Helicobacter species observed in
the B4galnt2þ/þ mice.
Acknowledgements
The conservation of intestinal B4galnt2 expres-
sionacrossspeciessuggestsanimportantfunctional
We wish to thank David Ginsburg for his assistance with
role for B4galnt2 in the GI tract. This hypothesis is the B4galnt2(cid:2)/(cid:2) mice, Philipp Rausch, Ateequr Rehman
supported by our finding that the loss of B4galnt2 and Sri Girish Srivinasa Murthy for helpful discussions
expression significantly impacts the composition of and assistance in data analysis, and Silke Carstensen for
the resident microbiota. We propose that the com- excellent technical assistance. This work was supported
plex signatures of natural selection observed in by the Puget Sound Blood Center (JMJ), the National
house mice (Johnsen et al., 2009) are at least in part Institutes of Health RO1-HL039693 (JMJ), the Deutsche
Forschungsgemeinschaft(DFG)ExcellenceCluster‘Inflam-
duetothevariationofB4galnt2-GalNAcresidueson
mation at Interfaces’ (JFB) and DFG grants BA 2863/2-1
intestinal mucosal surfaces (the loss of glycans
(JFB)andSTA1154/1-1(FS)andanintramuralgrantfrom
during evolution is discussed by Bishop and
theCAU KielMedical Faculty (JFB).
Gagneux (2007)). Taken together, these data support
a scenario in which the loss of intestinal B4galnt2
expression offers a significant fitness advantage in
the face of pathogens reliant on the presence of the
References
otherwise ubiquitous intestinal mucosal B4galnt2-
derived GalNAc. Host B4galnt2 glycans may AnsteeDJ.(2010).Therelationshipbetweenbloodgroups
serve as a carbon source for both symbiotic and anddisease. Blood 115:4635–4643.
TheISMEJournal
B4galnt2influencestheintestinalmicrobiota
FStaubachetal
1354
Artis D. (2008). Epithelial-cell recognition of commensal Hamady M, Lozupone C, Knight R. (2009). Fast UniFrac:
bacteria and maintenance of immune homeostasis in facilitating high-throughput phylogenetic analyses of
the gut.NatRev Immunol8: 411–420. microbial communities including analysis of pyrose-
ArumugamM,RaesJ,PelletierE,LePaslierD,YamadaT, quencingandPhyloChip data.ISME J 4:17–27.
MendeDRetal.(2011).Enterotypesofthehumangut Hynes SO, Teneberg S, Roche N, Wadstrom T. (2003).
microbiome.Nature473:174–180.(Erratumin:Nature Glycoconjugate binding of gastric and enterohepatic
2011;474:666). Helicobacterspp.Infect Immun71:2976–2980.
Backhed F, Ley RE, Sonnenburg JL, Peterson DA, Gordon JohnsenJM,LevyGG,WestrickRJ,TuckerPK,GinsburgD.
JI. (2005). Host-bacterial mutualism in the human (2008). The endothelial-specific regulatory mutation,
intestine.Science 307:1915–1920. Mvwf1, is a common mouse founder allele. Mamm
Bishop JR, Gagneux P. (2007). Evolution of carbohydrate Genome 19: 32–40.
antigens—microbial forces shaping host glycomes? Johnsen JM, Teschke M, Pavlidis P, McGee BM, Tautz D,
Glycobiology 17:23R–34R. Ginsburg D et al. (2009). Selection on cis-regulatory
BryL,FalkPG,MidtvedtT,GordonJI.(1996).Amodelof variationatB4galnt2anditsinfluenceonvonWillebrand
host-microbial interactions in an open mammalian factorinhousemice.MolBiolEvol26:567–578.
ecosystem.Science 273:1380–1383. Karlsson KA. (1995). Microbial recognition of target-cell
Calafell F, Roubinet F, Ramı´rez-Soriano A, Saitou N, glycoconjugates.Curr OpinStructBiol 5:622–635.
BertranpetitJ,BlancherA.(2008).Evolutionarydynamics Kawamura YI, Kawashima R, Fukunaga R, Hirai K,
ofthehumanABOgene.HumGenet124:123–135. Toyama-Sorimachi N, Tokuhara M et al. (2005).
Chao A. (1984). Nonparametric estimation of the number Introduction of Sd(a) carbohydrate antigen in gastro-
of classes ina population. ScandJ Stat11:265–270. intestinal cancercells eliminates selectin ligands and
ConteR,Serafini-CessiF.(1991).Comparisonbetweenthe inhibits metastasis. CancerRes 65:6220–6227.
erythrocyte and urinary Sda antigen distribution in a KobayashiM,LeeH,NakayamaJ,FukudaM.(2009).Roles
large number of individuals from Emilia-Romagna, a of gastric mucin-type O-glycans in the pathogenesis of
regionof northernItaly.Transfus Med1:47–49. Helicobacterpyloriinfection.Glycobiology19:453–461.
CorfieldAP,CarrollD,MyerscoughN,ProbertCS.(2001). KrivanHC,RobertsDD,GinsburgV.(1988).Manypulmon-
Mucins in the gastrointestinal tract in health and ary pathogenic bacteria bind specifically to the carbo-
disease. Front Biosci6:D1321–D1357. hydrate sequence GalNAc beta 1-4Gal found in some
De Caceres M, Legendre P. (2009). Associations between glycolipids.ProcNatlAcadSciUSA85:6157–6161.
species and groups of sites: indices and statistical Ley RE, Ba¨ckhed F, Turnbaugh P, Lozupone CA, Knight
inference.Ecology90:3566–3574. RD, Gordon JI. (2005). Obesity alters gut microbial
DethlefsenL,HuseS,SoginML,RelmanDA.(2008). The ecology.ProcNatl Acad SciUSA 102:11070–11075.
pervasive effects of an antibiotic on the human gut LinnenbrinkM,JohnsenJM,MonteroI,BrzezinskiCR,Harr
microbiota,asrevealedbydeep16SrRNAsequencing. B, Baines JF. (2011). Long-term balancing selection at
PLoS Biol 6:e280. thebloodgroup-relatedgeneB4galnt2inthegenusMus
DohiT,YuyamaY,NatoriY,SmithPL,LoweJB,OshimaM. (Rodentia;Muridae).MolBiolEvol28:2999–3003.
(1996). Detection of N-acetylgalactosaminyltransferase Lo PL, Cabuy E, Chiricolo M, Dall’Olio F. (2003).
mRNAwhichdeterminesexpressionofSdabloodgroup Molecular cloning of the human beta1,4 N-acetylga-
carbohydrate structure in human gastrointestinal lactosaminyltransferase responsible for the biosyn-
mucosaandcancer.IntJCancer67:626–631. thesis of the Sd(a) histo-blood group antigen: the
Dufre´ne M, Legendre P. (1997). Species assemblages and sequence predicts a very long cytoplasmic domain.
indicatorspecies:theneedforaflexibleasymmetrical JBiochem 134:675–682.
approach.Ecol Monogr67:345–366. Meng D, Newburg DS, Young C, Baker A, Tonkonogy SL,
EckburgPB,BikEM,BernsteinCN,PurdomE,Dethlefsen Sartor RB et al. (2007). Bacterial symbionts induce a
L, Sargent M et al. (2005). Diversity of the human FUT2-dependentfucosylatednicheoncolonicepithe-
intestinal microbialflora. Science 308:1635–1638. lium via ERK and JNK signaling. Am J Physiol
Feng S, Ku K, Hodzic E, Lorenzana E, Freet K, Barthold Gastrointest Liver Physiol293:G780–G787.
SW. (2005). Differential detection of five mouse- Miller-PodrazaH, LanneB, Angstro¨mJ,TenebergS,Milh
infecting helicobacter species by multiplex PCR. Clin MA,JovallPAetal.(2005).Novelbindingepitopefor
DiagnLab Immunol12:531–536. Helicobacter pylori found in Neolacto carbohydrate
Ferrer-AdmetllaA,SikoraM,LaayouniH,EsteveA,Roubinet chains.JBiol Chem 280:19695–19703.
F, Blancher A et al. (2009). A natural history of FUT2 Mohlke KL, Purkayastha AA, Westrick RJ, Smith PL,
polymorphisminhumans.MolBiolEvol26:1993–2003. PetryniakB, LoweJB et al. (1999). Mvwf, a dominant
FrankDN,RobertsonCE,HammCM,KpadehZ,ZhangT, modifierofmurinevonWillebrandfactor,resultsfrom
ChenHetal.(2011).Diseasephenotypeandgenotype altered lineage-specific expression of a glycosyltrans-
are associated with shifts in intestinal-associated ferase. Cell96:111–120.
microbiota in inflammatory bowel diseases. Inflamm Montiel MD, Krzewinski-Recchi MA, Delannoy P,
Bowel Dis17:179–184. Harduin-Lepers A. (2003). Molecular cloning, gene
Fraser C, Alm EJ, Polz MF, Spratt BG, Hanage WP. (2009). organization and expression of the human UDP-
Thebacterialspecieschallenge:makingsenseofgenetic GalNAc:Neu5Acalpha2-3Galbeta-R beta1,4-N-acetyl-
andecologicaldiversity.Science323:741–746. galactosaminyltransferase responsible for the bio-
FrederickJR,PetriJrWA.(2005).Rolesforthegalactose-/ synthesis of the blood group Sda/Cad antigen: evi-
N-acetylgalactosamine-bindinglectinofEntamoebain dence for an unusual extended cytoplasmic domain.
parasite virulence and differentiation. Glycobiology Biochem J373:369–379.
15:53R–59R. Morton JA, Pickles MM, Vanhegan RI. (1988). The Sda
Goldman CG, Mitchell HM. (2010). Helicobacter spp. antigen in the human kidney and colon. Immunol
otherthanHelicobacterpylori.Helicobacter15:69–75. Invest17:217–224.
TheISMEJournal