Content uploaded by Diego Mora
Author content
All content in this area was uploaded by Diego Mora on Nov 19, 2015
Content may be subject to copyright.
1
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
www.nature.com/scientificreports
Bidobacteria exhibit social
behavior through carbohydrate
resource sharing in the gut
Christian Milani
1
, Gabriele Andrea Lugli
1
, Sabrina Duranti
1
, Francesca Turroni
2
,
Leonardo Mancabelli
1
, Chiara Ferrario
1
, Marta Mangifesta
1
, Arancha Hevia
3
,
Alice Viappiani
1
, Matthias Scholz
4
, Stefania Arioli
1
, Borja Sanchez
3,†
, Jonathan Lane
5
,
Doyle V. Ward
6
, Rita Hickey
5
, Diego Mora
7
, Nicola Segata
4
, Abelardo Margolles
3
,
Douwe van Sinderen
2
& Marco Ventura
1
Bidobacteria are common and frequently dominant members of the gut microbiota of many
animals, including mammals and insects. Carbohydrates are considered key carbon sources for the
gut microbiota, imposing strong selective pressure on the complex microbial consortium of the gut.
Despite its importance, the genetic traits that facilitate carbohydrate utilization by gut microbiota
members are still poorly characterized. Here, genome analyses of 47 representative Bidobacterium
(sub)species revealed the genes predicted to be required for the degradation and internalization of
a wide range of carbohydrates, outnumbering those found in many other gut microbiota members.
The glycan-degrading abilities of bidobacteria are believed to reect available carbon sources in
the mammalian gut. Furthermore, transcriptome proling of bidobacterial genomes supported
the involvement of various chromosomal loci in glycan metabolism. The widespread occurrence of
bidobacterial saccharolytic features is in line with metagenomic and metatranscriptomic datasets
obtained from human adult/infant faecal samples, thereby supporting the notion that bidobacteria
expand the human glycobiome. This study also underscores the hypothesis of saccharidic resource
sharing among bidobacteria through species-specic metabolic specialization and cross feeding,
thereby forging trophic relationships between members of the gut microbiota.
Bidobacteria are Gram positive bacteria with high GC-content genomes that typically reside in the
gastro intestinal (GIT) tract of various animal species, including warm blooded mammals and social
insects
1,2
. Noteworthy, bidobacteria reach a very high relative abundance as part of the infant gut micro-
biota
3-5
, and this early life prevalence supports their purported role as modulators of various metabolic
and immune activities of their host
1
. Non-digestible carbohydrates derived from the diet, together with
host-produced glycans found in the mammalian gut represent critical energy sources believed to be
responsible for the survival and proliferation of many microbial components of the gut microbiota,
including bidobacteria
6
. It is therefore important to know what molecular strategies are employed
by members of the gut microbiota to harvest and metabolize complex glycans in order to understand
how they underpin their ecological tness in, and adaptation to, the GIT environment. Furthermore,
1
Laboratory of Probiogenomics, Department of Life Sciences, University of Parma, Italy.
2
APC Microbiome Institute
and School of Microbiology, Bioscience Institute, National University of Ireland, Cork, Ireland.
3
Departamento de
Microbiologia y Bioquimica de Productos Lacteos, IPLA – CSIC, Villaviciosa, Asturias, Spain.
4
Centre for Integrative
Biology, University of Trento, Trento, Italy.
5
Teagasc Food Research Centre, Moorepark, Cork, Ireland.
6
Broad
Institute, 415 Main St., Cambridge, USA.
7
Department of Food Environmental and Nutritional Sciences, University
of Milan, Italy.
†
Present address: Department of Analytical Chemistry and Food Science, Faculty of Food Science
and Technology, University of Vigo, Spain. Correspondence and requests for materials should be addressed to M.V.
(email: marco.ventura@unipr.it)
Received: 27 May 2015
Accepted: 02 October 2015
Published: 28 October 2015
OPEN
www.nature.com/scientificreports/
2
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
carbohydrate metabolism in the mammalian gut may occur in a concerted manner by various members
of the microbiota, including bidobacteria
7
. us, carbohydrates are presumed to be partly responsible
for microbiota dynamics in this changeable environment. Genomics has been crucial in revealing the
complex interactions and molecular dialogue between a host and its resident (bido)bacteria, and in
unraveling bidobacterial gut colonization strategies
8
. Recently, all 47 currently recognized bidobacte-
rial sub/species have been genomically decoded
9,10
, thus providing the necessary genetic background for
this group of bacteria. Here, we describe an analysis of the genomes and transcriptomes of representa-
tives of all 47 (sub)species that are currently assigned to the Bidobacterium genus, predicting bidobac-
terial genome-based strategies for carbohydrate metabolism that impact on the overall glycobiome of the
gut microbiota and host, while also being pivotal for the establishment of trophic relationships between
members of gut microbiota.
Results and Discussion
Genomes of the Bidobacterium genus and carbohydrate metabolism. Genome sequences
from the type strain of each of the 47 Bidobacterium (sub)species (Table S1) were used to assess the
contribution of carbohydrate metabolism to the corresponding pan-genome, core genome and vari-
ome, determined as described previously
9
. In total, 18,181 and 551 BifCOGs (Bidobacterium-specic
Clusters of Orthologous Genes) represent the pan-genome (pan BifCOGs) and the core of bidobacte-
rial genomic coding sequences (core BifCOGs) of the Bidobacterium genus, respectively. Functional
annotation of the core BifCOGs, based on a recently updated EggNog database
11
, indicates, as expected,
that a large part of the conserved core genes encode housekeeping functions and functions involved in
the adaptation to/interaction with a particular environment, including carbohydrate metabolism (Fig.
S1). Interestingly, about 5.5% of the core BifCOGs is associated with carbohydrate metabolism (Fig. S1),
whereas the carbohydrate metabolism functional family is the most represented COG family of the pan
BifCOGs (13.7%) (Fig. S1). is nding suggests a strong selective pressure towards the acquisition and
retention of accessory genes for carbohydrate utilization by bidobacteria in order to be competitive in
a particular ecological niche. e commitment of bidobacteria towards the utilization of a wide array
of simple and complex carbohydrates is further underlined by EggNog proling of adult human’s faecal
metagenomes sequenced as part of the Human Microbiome Project (HMP). Results showed that the
average abundance of the carbohydrate metabolism functional family in HMP metagenomic datasets is
8.0%, thus being 58% lower than the abundance detected in the bidobacterial pan-genome, i.e. 13.7%
(Fig. S1). Notably, the carbohydrate metabolism functional family is the most abundantly represented
COG family within the Bidobacterium pan-genome (Fig. S1). e pan-genome analysis also allowed the
identication of the Truly Unique Genes (TUG), consisting of genes present in just one of the examined
bidobacterial genomes (Fig. S1). Interestingly, no functional annotation can be made for the majority
of TUGs (54.1%) (Fig. S1). However, 13.2% of the identied TUG can be attributed to a COG fam-
ily representing proteins involved in carbohydrate metabolism, including glycosyl hydrolases (GH) and
carriers for carbohydrate uptake. is data supports the notion that carbohydrate metabolism supports
adaptation and specialisation processes, and consequently speciation within the Bidobacterium genus.
Carbohydrate-active enzymes encoded by the genus Bidobacterium. Considering the impor-
tance of carbohydrate metabolism and energy conversion systems in gut inhabitants such as bidobacte-
ria, we decided to investigate the presence of genes predicted to encode GHs, polysaccharide lyases (PLs),
carbohydrate esterases (CE), glycosyl transferases (GT) and carbohydrate binding modules (CBM), and
other carbohydrate metabolic pathway components using the predicted core- and pan-genome of the
Bidobacterium genus. Genomic data showed that B. scardovii and B. biavatii have a signicantly larger
set of genes involved in carbohydrate metabolism, also termed glycobiome
12
, as compared to other
bidobacterial species (Fig.1). Normalization of GH counts against genome size (Mbp), generating a GH
index, provided an insight into the relative extent of adaptive events related to carbohydrate metabolism
found in individual bidobacterial species. Notably, four species (B. scardovii, B. biavatii, Bidobacterium
saeculare and Bidobacterium dentium) possess a GH index that is approximately 50% higher than the
bidobacterial average (Fig. 1). Classication according to the Carbohydrate Active Enzymes (CAZy)
system
13
revealed that the pan-genome of the analysed Bidobacterium representatives includes 3,385
genes encoding predicted carbohydrate-active enzymes, including members of 57 GH, 13 GT and seven
CE families (Fig.1, Fig. S2 and Fig. S3), while no putative Polysaccharide Lyases (PL)-encoding genes
were detected. Genes encoding members of CAZy family GH13 are most commonly found in bidobac-
terial genomes, especially for those bidobacteria isolated from the mammalian gut (24.2% of the total
predicted bidobacterial GH repertoire, Fig.1). Enzymes belonging to this family are widespread in bac-
teria and are characterized by their degradative abilities towards a wide range of carbohydrates, including
plant-derived polysaccharides, such as starch and related substrates (amylose and amylopectin and/or
(cyclo)maltodextrins), and trehalose. Furthermore, stachyose, ranose and melibiose may also represent
a target for members of the GH13 family
14
and their complete breakdown is achieved with the involve-
ment of GH36 enzymes, which were shown to be abundant in the analysed bidobacterial genomes
(Figure S2). Such carbon sources indeed represent very common glycans found in the adult mammalian
(omnivore and herbivore) diet
6
. We also observed distinct dierences between predicted glycobiomes
of Bidobacterium species with regards to their (predicted) ability to degrade plant polysaccharides as
www.nature.com/scientificreports/
3
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
opposed to host-derived glycans, such as N- or O-linked glycoproteins, human milk oligosaccharides
(HMOs) and glycosaminoglycans
1
. e bidobacterial glycobiome was shown to contain members of
GH families that are known to be involved in host-glycan degradation, such as GH33, encompassing
exo-sialidases, GH29 and GH95, which represent fucosidases, GH20, which include hexosaminidase
and lacto-N-biosidase activities, GH112, representing lacto-N-biosidases, GH38 and GH125, involv-
ing α -mannosidases as well as GH101 and GH129, which include α -N-acetylgalactosaminidases.
Interestingly, members of the B. scardovii, B. longum subsp. infantis and Bidobacterium bidum species
possess the most extensive set of such host-glycan-degrading GH families (Fig. 1). Such glycobiome
specialization probably reects species-specic adaptation to a particular ecological niche and illustrates
strict co-evolution with their mammalian host. Clustering of bidobacterial species based on their pre-
dicted GH and carbohydrate-degradation pathway repertoire (Fig.1) allows the identication of a cluster,
designated as GHP/A, representing species with a considerable array of predicted GH43 family enzymes
involved in degradation of complex plant polysaccharides, such as (arabino)xylan, which plays an impor-
tant structural role as a main constituent of the plant cell wall
15
, and as such represents a substantial
component of plant cell wall-derived dietary ber
16
. is nding suggests that bidobacteria rich in
GH43 member-specifyings genes are adapted to hosts with a vegetarian or omnivore diet. A subgroup
of the cluster GHP/A, named GHP/A1, appears to possess a wider array of GHs (Fig.1). is subgroup
encodes a large number of GH3 family enzymes that are predicted to be involved in the degradation
of an extensive range of plant-derived polysaccharides (e.g., cellodextrin, (arabino)xylan and (arabino)
galactan), as well as involved in bacterial cell wall biosynthesis and turnover. Notably, bidobacterial
species isolated from honey/bumblebees specify a peculiar GHP/C cluster, whose members possess a
Figure 1. Predicted glycobiome of the Bidobacterium genus and some additional members of the
Bidobacteriaceae family. GH families and carbohydrate-utilization pathways proles, based on CAZy
database and Pathway-tools soware, respectively, were used to construct a hierarchical clustering of all
tested species of the Bidobacterium genus and additional members of the Bidobacteriaceae family. is
clustering highlights the presence of three distinct clusters named GHP/A, GHP/B and GHP/C that display a
dierent repertoire of GHs as well as a dierent repertoire of plant carbohydrate degradation pathways. GH
arsenal prediction for every analyzed Bidobacteriaceae species is represented by a bar plot. e presence of
pathways for degradation of simple or complex carbohydrates is represented by the red color in the heat map
and the GH index (the number of GHs predicted in each genome normalized by genome size expressed as
Mbp) is illustrated as an orange bar plot. Pathways denominations are indicated as follows: 1 Bidobacterium
shunt, 2 galactose degradation I (Leloir pathway), 3 melibiose degradation, 4 ribose degradation, 5 lactose
degradation III, 6 glycogen degradation I, 7 glycogen degradation II, 8 sucrose degradation IV, 9 L-arabinose
degradation I, 10 xylose degradation I, 11 D-mannose degradation, 12 (1,4)-ß-xylan degradation, 13 starch
degradation V, 14 chitin degradation (chitinase), 15 trehalose degradation IV, 16 Pectin (homogalacturonan)
degradation, 17 2'-deoxy-α-D-ribose 1-phosphate degradation, 18 trehalose degradation I (low osmolarity),
19 L-rhamnose degradation II.
www.nature.com/scientificreports/
4
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
discrete set of GH43 and GH3 enzymes, but in addition these particular bidobacterial genomes are
predicted to encode a very limited repertoire of GH13 representatives, in contrast to all other bidobac-
teria. is is in accordance with the absence or paucity of carbohydrates with α -glucosidic linkages in the
honey/bumblebee diet
17
. e remaining bidobacterial species, not tting in clusters GHP/A or GHP/C,
are included in cluster GHP/B, which is characterized by an under-representation of GH43 and GH3.
In order to further evaluate the importance of the bidobacterial contribution to degradation of
complex carbohydrates found in the human gut, we compared the bidobacterial GH repertoire with all
the so far sequenced members of the human gut microbiome
18,19
. For this purpose, we used GH data of
2,674 genomes available in the CAZy database
13
with the addition of data obtained from the bidobac-
terial genomes analysed in this study. Interestingly, the Bidobacterium genus was shown to specify one
of the largest arsenal of GH13, GH43, GH3 and GH51 family members (2.0 fold-, 2.6 fold-, 5.8 fold-,
7.0 fold-, respectively, more with respect to the average GH arsenal of the gut microbiome), along with
Bacteroides spp. (Bacteroidales family) and Clostridiales family. Furthermore, abundance of these GH
families has been shown by a small number of other members of the gut microbiota such as Paenibacillus
spp. (Bacillales family) and Streptomyces spp. (Actinomycetales family) (Fig.2). ese ndings corroborate
the very substantial contribution to glycan-breakdown potential by bidobacteria in the mammalian gut.
Secreted glycosyl hydrolases of bidobacteria. In order to allow GH enzymes produced by dif-
ferent organisms to act in concert and to access glycans with a degree of polymerization exceeding that
of the corresponding uptake system, certain GH enzymes are expected to be located on the cell surface
or secreted into the environment. Notably, even though the majority of the identied GH enzymes from
the bidobacterial pan-genome is predicted to be intracellular, 10.9% of the deduced bidobacterial
GH pool is predicted to be extracellular, of which 32.9% are members of the GH13 family and anno-
tated as pullulanases and α -amylases, 24% are members of the GH43 family and predicted to act as
β -xylosidases and α -L-arabinofuranosidases, while 12% are members of the GH51 family and classied
as α -L-arabinofuranosidases. ese data suggest that bidobacteria secrete a relevant pool of GH for the
degradation of plant polysaccharides, which would accordingly represent an important resource for the
host in the context of obtaining access to dietary bres. e secreted bidobacterial pan-genome also
encompasses GHs involved in host-glycan degradation (see above), being classied as families GH29,
GH33, GH95 and GH101, and constituting 0.4%, 3.5%, 0.8% and 2.7% of this secreted pan-genome,
respectively. Predicted extracellular GHs were identied in 43 species of the genus Bidobacterium, with
a particularly high prevalence in B. biavatii (endowed with 17 secreted GHs, including 4 GH43 and 4
Figure 2. Comparative analysis of bidobacterial GHs against other gut bacteria. e central heat map
shows GH prediction data of 2721 sequenced bacterial strains belonging to bacterial orders residing in the
human gut, identied by dierent color codes as explained in the underlying table. e four heat map rows,
situated above the main heat map, represent an enlarged view of the GH51, GH3, GH43 and GH13 content.
Data regarding Bidobacteriales are highlighted in blue. Data regarding Clostridiales and Bacteroidales are
highlighted in green.
www.nature.com/scientificreports/
5
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
GH13 members), B. scardovii (endowed with 11 secreted GHs, including 3 GH43 and 3 GH51, annotated
as α -L-arabinofuranosidases) and B. bidum (endowed with 11 secreted GHs, including two GH84 and
two GH33 members, predicted to act as N-acetyl β -glucosaminidases and sialidases, respectively). e
remaining bidobacterial genomes are predicted to encode seven or less secreted GHs. e two secreted
N-acetyl β -glucosaminidases and the two secreted sialidases found in B. bidum are crucial for the uti-
lization of HMOs and intestinal glycoconjugates such as mucin, thus further supporting the notion that
this microorganism is highly adaptated to colonize and persist in the mammalian gut
20,21
. Nevertheless,
B. bidum is unable to use sialic acid as its sole carbon source and the activity of sialidases appears to be
important only to allow access to other carbohydrates associated to sialylated host-encoded glycans
20,21
.
Furthermore, the released sialic acid can be utilized by other bidobacteria, such as B. breve, resulting in
cross-feeding between bidobacterial species that share the same ecological niche
20,21
.
Saccharolytic pathways of bidobacteria. Predictions of complete pathways for the degradation of
simple (di/trisaccharides) and complex sugars through Pathway Tools soware
22
showed that B. biavatii
species the largest number of such pathways (14 complete pathways) (Fig.1), while the predicted rep-
ertoire for carbohydrate degradation of B. bombi, Bidobacterium crudilactis, B. longum subsp. infantis,
B. minimum and Bidobacterium ruminantium is limited to just nine pathways (Fig.1). Notably, when
assessing presence/absence data of pathways involved in the breakdown of simple and complex car-
bohydrates shown in the GHP clustering prole (Fig. 1), bidobacterial species isolated from honey/
bumblebees, constituting cluster GHP/C (Fig.1), lack genes encoding glycogen degradation I and gly-
cogen degradation II pathways. Based on recent discoveries, glycogen metabolic pathways are present
in bacterial species able to face diverse environments and showing a exible lifestyle
23
. More recently, it
was reported that the ability of Lactobacillus acidophilus to synthesize and store energy in the form of
glycogen, either prior to or during its transit through the host, potentially confers a competitive advan-
tage in the GI tract
24
. is is supported by the fact that glycogen storage is ubiquitous among enteric
bacteria, possibly due to the necessity to support rapid growth in the intestinal environment where
there is intense trophic competition
25
. On the other hand, it was hypothesized
26
that the loss of glycogen
pathways is a strong indication of genome degradation associated with parasitic or symbiotic behaviour
of bacteria. ough the original study of Henrissat et al. was based on the analysis and comparison
of just 55 fully sequenced bacterial genomes, their ndings have more recently been substantiated by
others involving 1202 bacterial proteomes
23
. Among Bidobacteriaceae, it is noteworthy that B. actinoco-
loniiforme, B. asteroides, B. bohemicum, B. bombi, B. coryneforme and B. indicum, species isolated from
insects and constituting the GHP/C cluster, show a smaller genome size compared to all other bido-
bacterial species isolated from mammals. We therefore speculate that Bidobacterium members linked to
insects have enjoyed a longer adaptation history to their hosts, in evolutionary terms, because mammals
appeared on earth in the late Paleocene (between 65 and 23 mya), whereas insects in the Devonian period
(between 459 and 359 mya) or even before according to a recent phylogenomics study
27
. is hypothesis
is conrmed by comparison of the phylogenetic supertree of all known bidobacterial species and that
obtained for their corresponding hosts, revealing a co-evolution host-microbe prole (Fig. S4).
e genomes of B. asteroides, B. actinocoloniiforme, B. indicum, B. coryneforme, B. bombi and B.
bohemicum possess a complete trehalose degradation IV pathway, which is absent in the majority of the
other bidobacteria and in all other examined members of the family Bidobacteriaceae. e acquisition
of this pathway, which is apparently specic for bidobacterial species isolated from the insect gut, may
be related to the fact that trehalose is used as carbohydrate storage and blood-sugar by many insects
including bees
28
. Remarkably, a large majority of bidobacterial genomes included in cluster GHP/B do
not encompass pathways for L-arabinose and/or xylose metabolism (Fig.1). ese genomes encode rela-
tively few GH43 enzymes compared to the GHP/A and GHP/C clusters, thus conrming a strict genetic
adaptation to ecological niches where these plant carbohydrates are not available. e ecological origin
of most of the GHP/B members that do not possess L-arabinose and/or xylose degradation pathways is
either raw/fermented milk or faecal material/gastrointestinal tract of suckling animals (Table S1), where
a milk-based diet represents the main nutrient retrieval opportunity. In this context, B. breve, B. bidum
and B. longum subsp. infantis have been isolated from infants, B. thermacidophilum subsp. porcinum
and B. choerinum from piglet, B. crudilactis from raw bovine milk and B. mongoliense from fermented
mare’s milk (Table S1), thus suggesting that these species have evolved to focus solely or predominantly
on the degradation of carbon sources present in milk and have lost or did not acquire the ability to use
(certain) plant polysaccharides. Fermented milk is a strict anthropogenic environment and these bacteria
are unlikely to have evolved in fermented milk per se. In fact, the natural ecological environment of these
species is still expected to be the gut from various (suckling) animals. In this environment such bidobac-
terial taxa have enjoyed the presence of a rich reservoir of milk-based carbohydrates (oligosaccharides/
lactose), which thus caused their genomes to acquire a genetic arsenal that allowed them to access these
carbon sources. In contrast, B. ruminantium and B. boum taxa have been isolated from the bovine rumen
(Table S1), an environment rich in plant polysaccharides, although they lack GH43-encoding genes and
consequently cluster in GHP/B. Other rumen-derived bidobacteria do encode GH43 enzymes, perhaps
indicating that certain bidobacteria rely on cross-feeding, which is in line with bidobacteria being a
minor component of the rumen microbiota, although their functional role is largely unknown. B. breve
deserves a special mention as this species seems to have adopted a non-specialist strategy of acquiring
www.nature.com/scientificreports/
6
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
constituent elements of both plant- and host-derived carbohydrates (consistent with its isolation from
both infants and adults), yet is lacking the ability to directly access (many) HMOs, or mucin, nor possess-
ing the ability to metabolize xylose/arabinose-containing carbohydrates, though it can metabolize a wide
range of α /β -glucose- and α /β -galactose-containing sugars
14,20
. us, it may, perhaps co-operatively, rely
on other (bido)bacterial species like B. bidum or B. longum subsp. infantis in order to sustain growth
on the above mentioned complex sugars
29
.
Carbohydrate utilization patterns of Bidobacterium. Fermentation proles of the 47 sequenced
Bidobacterium strains revealed that all strains are able to ferment a common set of sugars, which include
glucose, sucrose and ranose (Fig. S5, panel a). In contrast, fermentation capabilities for other sugars,
such as lactose, galactose, maltose, melibiose, fructose, lactulose, maltodextrins, turanose, β -gentibiose
and xylose, were shown to be variable for the majority of the strains tested (Fig. S5, panel a). Notably,
we identied bidobacterial taxa such as B. cuniculi displaying an ability to grow on a wide range of
simple and complex carbohydrates (Fig. S5, panel a), thus suggesting metabolic expansion of its car-
bohydrate acquisition abilities perhaps to enhance competitiveness in one or more ecological niches.
In contrast, other bidobacterial species, such as Bidobacterium animalis subsp. animalis, only utilize
a relatively small number of the carbohydrates assayed here (Fig. S5, panel a), an observation which,
together with the very limited number of predicted GHs/carbohydrate pathways encoded by this taxon
(Fig. 1), underlines a rather high level of genetic adaptation to an ecological niche. Growth studies
highlighted the existence of dierent carbon sources that are dierentially utilized by bidobacteria,
in a manner consistent with our predictions. In this context growth experiments involving most mem-
bers of the B. asteroides phylogenetic group showed that they do not exhibit any appreciable growth
on glycogen, an observation which is consistent with in silico pathway predictions. Cultivation trials
performed on plant-derived carbohydrates such as arabinose or xylose revealed that these sugars are
utilized by a majority of the bidobacteria tested except for those taxa that form the GHP/B cluster
(Fig.1), which are not predicted to encode enzymes of the L-arabinose degradation I and xylose deg-
radation I pathways. e ubiquitous monosaccharide mannose that is found in both plant and ani-
mal glycans, is easily shunted into the glycolytic pathway via isomerization of mannose-6-phosphate to
fructose-6-phosphate
30
. Growth on mannose is, however, not a wide-spread property among the tested
bidobacterial taxa, a nding that is consistent with genomic data (Fig. 1), which revealed a variable
distribution of genes supporting bidobacterial mannose utilization such as genes encoding predicted
mannosidases and mannose-6-phosphate isomerases. Interestingly, when mannosidase-encoding genes
were detected, they were in the large majority of cases shown to be located within a putative N-glycan
(host-derived) degradation cluster. Lactose constitutes a typical glycan that specically occurs in the
mammalian diet, though normally limited to the early stages of life
31
. In silico analyses involving the
genomes of 47 bidobacterial taxa revealed a ubiquitous distribution in their chromosomes of genes
encoding β -galactosidases. is nding suggests a rather wide-spread utilization of lactose as well as
other galactose-containing glycans such as galactan, galacto-oligosaccharides, Human milk oligosaccha-
rides (HMO) and mucin, which require particular β -galactosidases for their degradation, among mem-
bers of the Bidobacterium genus
20,32
. Growth experiments involving lactose showed that, except for B.
ruminantium and B. thermacidophilum subsp. thermacidophilum, all other tested bidobacteria are able
to ferment this disaccharide. In contrast, in silico analyses suggested that L-rhamnose is rarely used by
bidobacteria due to the apparent lack of genes encoding L-rhamno-gamma-lactonase, L-rhamnoate
dehydratase and 2-keto-3-deoxy-L-rhamnoate aldolase. Notably, fermentation proles involving this car-
bon source conrmed that only B. biavatii is able to ferment L-rhamnose. Furthermore, we evaluated
the abilities of members of the genus Bidobacterium to utilize typical host glycans like mucin and
HMO. Interestingly, in addition to the currently known bidobacterial archetypes that can utilize these
host-glycans (B. bidum and B. longum subsp. infantis)
20,33
, we identied that B. biavatii, B. crudilac-
tis, B. kashiwanohense, Bidobacterium stellenboschense and Bidobacterium mongoliense are all capable
of HMO metabolism. Inspection of their genome sequences revealed the presence of genes predicted
to encode sialidases, fucosidases, N-acetyl-β -hexosaminidases, endo-α -N-acetylgalactosaminidase and
β -galactosidases, which have previously been shown to be crucial for the breakdown of these complex
carbohydrates
20,33
. is further illustrates the broad catabolic abilities of bidobacteria in general and in
particular their specialization to utilize complex carbohydrates that are commonly found in the (infant)
mammalian GIT.
In order to substantiate the notion that bidobacterial genomes contain specic genes responsible
for the utilization of key carbohydrates that are present in their ecological niches and correspond to
their saccharolytic phenotype, we investigated the transcriptome for a representative bidobacterial spe-
cies for each of the seven bidobacterial phylogenetic clusters described previously
10
grown on (where
possible) glucose, glycogen, lactose, xylose, rhamnose, mannose, trehalose or HMO as the sole carbon
source (Fig. S5, panel b). RNAseq experiments allowed the identication of the transcriptomes for each
bidobacterial strain tested in cases where growth was obtained, revealing that the carbohydrate metab-
olism COG family [COG category (G)], is one of the most represented in the transcriptomes (Fig. S5,
panel c). Manual inspection of the identied transcriptomes highlighted the existence of a large arsenal
of genes encoding GHs and other enzymes that constitute parts of carbohydrate metabolic pathways,
including suspected carbohydrate carrier systems that are expressed when bidobacteria are cultivated
www.nature.com/scientificreports/
7
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
on a carbohydrate (Fig. S6). In this context, we identied β -galactosidases of the GH2 and GH42 families
that were shown to be expressed when bidobacteria are grown on lactose, as well as MFS and ABC
systems predicted to act as carriers for lactose and/or glucose and/or galactose (Fig. S6). Cultivation of
bidobacteria on mannose, rhamnose, trehalose or xylose resulted, in cases where growth was observed,
in the transcription of genes encoding specic catabolic pathways such as D-mannose degradation,
L-rhamnose degradation II, trehalose degradation I and xylose degradation I (Fig. S6), observations that
are consistent with our in silico assignments. e transcriptomes of bidobacteria grown on glycogen
clearly showed that transcription of genes encompassing the glycogen degradation I pathway, predicted
to be indispensable for glycogen to glucose breakdown, is switched on under these circumstances (Fig.
S6). e only exceptions are represented by the transcriptomes of Bidobacterium pseudocatenulatum
and B. stellenboschense in which the genes predicted to specify amylomaltase and glucokinase enzymes
are not expressed, suggesting the existence of an alternative pathway for conversion of the intermedi-
ate carbohydrate maltose into β -D-glucose-6-phosphate. Transcriptomic data recovered from growth
of B. biavatii on the complex substrate HMO revealed up-regulation of a sizable number of dierent
GH-encoding genes, such as those predicted to belong to GH1 (predicted β -D-fucosidases), GH2 (puta-
tive β -galactosidases), GH3 (putative β -N-acetylhexosaminidases), GH29 (α -L-fucosidases), GH30 (pos-
sible β -fucosidases), GH36 (likely α -N-acetylgalactosaminidases), GH42 (putative β -galactosidases),
GH85 (endo-β -N-acetylglucosaminidases) and GH112 (predicted galacto-N-biose/lacto-N-biose phos-
phorylases) (Fig. S6). Other up-regulated GH-coding genes included those specifying members of GH13,
GH32 and GH43 families, even though their predicted enzymatic activities do not appear to be directed
against HMO.
Mutualistic/commensal breakdown activities of bidobacteria on carbohydrates. In gut
ecosystems, bacteria can exploit mutualistic as well as commensal and competitive activities during
metabolism of various carbon sources available in such an environment. us, we were interested to
evaluate possible trophic relationships towards carbohydrates between bidobacterial taxa found in
various ecological niches. Specic bi-association of bidobacterial taxa were selected based on their
common ecological origin (e.g., human-, porcine-, or rabbit-gut). eir growth performances on car-
bon sources that are commonly available in their particular ecological niche were assayed during
co-cultivation (bi-association) and compared to those achieved when the strains were grown separately
(mono-association). Such trials suggested that when cultivated together these bidobacterial taxa were
taking an evident benet as supported by the enhanced cell densities in the bi-associations compared to
the mono-associations (Fig.3). ese growth-benets were in a few cases evident for both strains such
as in the case of the B. magnum and B. cuniculi combination when these strains were cultivated on xylan
or on starch (Fig.3). Such ndings were further supported by the evaluation of transcriptomic changes,
employing an RNAseq approach, observed for these bidobacterial strains when cultivated together
as compared to the situation where these strains were assayed separately. Notably, in bi-associations
where a benet was noticed in terms of cell numbers for both strains, such as for B. magnum and B.
cuniculi cultivated on starch, complementarity in terms of the transcription of the genetic repertoire
for the metabolism of this complex sugar was observed for these strains. In fact, under such conditions
the alpha-glucoside phosphorylase-specifying genes, i.e. BMAGN_0016 and BCUN_1467, as well as
amylase-encoding genes, i.e. BMAGN_0612 and BCUN_0313, of both B. magnum and B. cuniculi were
shown to be transcribed (Fig.3). Interestingly, while these latter enzymes are predicted to be intracellular,
two genes encoding putative extracellular pullulanases (BCUN_0354 and BCUN_0356), were shown to
be transcribed. Even though the detected transcription level for these genes was low, it may be that such
enzymes allowed a partial de-branching of starch so as to allow internalization of the released degrada-
tion products into the bacterial cell for complete degradation. e resulting starch-degradation products
such as maltose and alpha-glucosides are presumed to be internalized by means of ABC transporters
(BMAGN_0006-BMAGN_0009 and BCUN_2019-BCUN_2022) and/or PTS systems (BMAGN_0338
and BCUN_1552), of which the corresponding genes were shown to exhibit increased transcription
(2.37 fold and 10.42 fold with p < 0.001, respectively) relative to mono-association conditions (Fig.3).
us, on this substrate both strains appear to co-operate in order to achieve starch degradation thanks
to the collective action of their extracellular amylases, perhaps resulting in the production of a larger
amount of starch-derivatives compared to that achieved when the strains are cultivated separately on this
substrate. In contrast, when these strains were co-cultivated on xylan, we only noticed an up-regulation
of the gene encoding a beta-xylosidase (BCUN_1638) in B. cuniculi. Similarly, B. cuniculi exhibited a
modest transcriptional upregulation of the genes specifying the putative uptake and degradation machin-
ery for xylose (BCUN_0705-BCUN_0707 and BCUN_1645), whereas B. magnum showed no modula-
tion in the beta-xylosidase-encoding gene, yet down-regulation of genes involved in xylose transport.
e latter ndings suggest that B. magnum modulates gene expression in order to support growth of B.
cuniculi and allows this latter strain to participate in xylan degradation and harvesting of the deriving
xylose. Since none of these enzymes are predicted to be extracellular, an alternative explanation for this
behavior is partial degradation of xylan to xylo-oligosaccharides during media preparation. ese deg-
radation products can be the target for specic uptake transporters, allowing their complete breakdown
inside the bacterial cells. Another scenario was noticed for the co-cultivation of B. stellenboschense and
B. biavatii grown in the presence of glycogen, where only the latter strain seems to take advantage of
www.nature.com/scientificreports/
8
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
the presence of the other strain (Fig. 3). In fact, under these circumstances B. biavatii substantially
enhanced transcription of various genes encoding enzymes predicted to be involved in the hydrolysis
of glycogen such as predicted glycosyl hydrolases of the GH13 and GH77 families, as well as a glycogen
phosphorylase (Fig. 3). In contrast, B. stellenboschense did not reveal any transcriptional modulation
of genes with predicted functions in glycogen utilization when the strains were co-cultivated (Fig. 3).
Similarly, data concerning co-cultivation of B. longum subsp. suis and B. thermacidophilum subsp. por-
cinum on starch is also consistent with a cross-feeding scenario. In fact, under these growth conditions,
qPCR assays revealed that cell numbers of B. longum subsp. suis are enhanced (≥ 2 fold) with respect to
those noticed when this strain was cultivated on its own on this substrate. Furthermore, transcriptome
analyses revealed a signicant induction (≥ 8 fold, p < 0.001) of genes encoding a complete ABC system
involved in the up-take of maltose, which is a starch-derived glycan, of B. longum subsp. suis when grown
with B. thermacidophilum subsp. porcinum in the presence of starch as the sole carbon source (Fig.3).
In contrast, even though the B. thermacidophilum subsp. porcinum genome is predicted to encode a
Figure 3. Evaluation of possible bidobacterial cross-feeding by a transcriptomics approach. (Panel a)
reports the abundance, observed through quantitative qRT-PCR, of eight bidobacterial species cultivated in
MRS supplemented with four dierent carbohydrates. ese species were either grown on their own (mono-
association) or in the presence of another bidobacterial strain (bi-associations) sharing the same ecological
niche. e ve case studies analysed are named progressively with letters from A to D, corresponding to: B.
cuniculi and B. magnum grown on starch (A), B. cuniculi and B. magnum grown on xylan (B), B. biavatii
and B. stellenboschense grown on glycogen (C) and B. thermacidophilum subsp. porcinum and B. longum
subsp. suis grown on starch (D). (Panel b) shows the transcriptional fold change of genes encoding enzymes
in the breakdown of glycans observed in the ve case studies, named progressively with letters from A to
D. Functional annotation of enzymes are indicated in orange while the functional annotation of transporter
encoding genes and the predicted glycan specicity is highlighted in green.
www.nature.com/scientificreports/
9
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
secreted GH13 family enzyme (BPORC_0608), the transcription of genes predicted to be involved in
starch metabolism of B. thermacidophilum subsp. porcinum did not appear to be aected by the presence
of B. longum subsp. suis (Fig.3).
Overall, our ndings suggest that bidobacteria access carbohydrates that are commonly present in
their ecologic niches employing trophic interactions that may vary from commensalism to mutualistic.
Furthermore, these concerted breakdown activities may similarly exert positive eects on other members
of the gut microbiota and thus promote an expansion of the gut glycobiome.
Bidobacteria and mammalian gut adaptation. Bidobacteria have predominantly been iden-
tied in the GIT of mammals
1
. However, their functional contribution to the microbiota residing in
this body compartment has not been thoroughly investigated. We therefore looked for the presence of
GH-encoding bidobacterial DNA sequences among 44 out of 136 available gut metagenome data sets
Figure 4. Data mining for bidobacterial GH genes and bidobacterial pathways for carbohydrate
degradation in adult and infant fecal metagenome data sets and an infant fecal metatranscriptome
data set. Bar plots above the heatmaps show the relative abundance of bidobacteria in the analysed
samples. Heatmaps in the upper part depict the coverage obtained by alignment of adult and infant fecal
metagenomic data sets, or infant metatranscriptome data sets to predicted bidobacterial GH-encoding
genes. In order to compare results for datasets with dierent sizes, all coverage values were normalized
as obtained from a 10 million read dataset. Heatmaps in the lower part of the image represent the
coverage obtained by alignment of the same datasets to genes constituting the bidobacterial pathways
for carbohydrate degradation. Relevant GH genes and pathways involved in the metabolism of glycans are
highlighted in red. Pathways designations are identical to those indicated in Fig.1.
www.nature.com/scientificreports/
10
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
from healthy adults, enrolled in the Human Microbiome Project
19
that had shown consistent presence
of bidobacteria based on MetaPhlAn proling
34
, as well as in nine metagenomes and related meta-
transcriptomes of healthy infant gut microbiota, sequenced by the Broad Institute (NCBI bioproject ID
63661). e coverage of each gene included in the bidobacterial pangenome was computed based on
the metagenomic reads with >98% full-length identity. As displayed in Fig. 4, the average abundance
of Bidobacterium genes ranged from 0.2% to 35.9%, with a distinctly higher prevalence of bidobac-
teria in the faecal metagenomes from infants. is is consistent with the catabolic potential for HMOs.
Interestingly, among the most frequently represented bidobacterial genes in the metagenomic data sets
from adult, the presence of an extensive repertoire of GH-encoding genes, such as those specifying
GH3, GH13, GH43, GH51 and GH77 (involved in the breakdown of complex plant carbohydrates), is
noteworthy (Fig.4). Such ndings reinforce the notion that despite the relative paucity of bidobacte-
ria detected in the adult human gut, their functional contribution to the human gut microbiome may
be important in terms of expanding the overall glycobiome of the large intestine, thereby aecting the
overall gut physiology. Furthermore, in the metagenome datasets from infants the relatively high bido-
bacterial abundance is also reected by the prevalence of bidobacterial GH-encoding genes, such as
those specifying members of the GH2, GH20, GH29, GH33, GH36, GH42, GH84, GH95, GH101 and
GH112 families, involved in the degradation of milk-related carbohydrates such as lactose and HMOs,
as well as in mucin degradation (Fig. 4). Notably, GH-specifying genes encoding β -hexosaminidases,
lacto-N-biosidases, galacto-N-biose/lacto-N-biose phosphorylases, α -L-fucosidases, sialidases, neurami-
nidases, N-acetyl β -glucosaminidases, hyaluronidases and endo-α -N-acetylgalactosaminidases are not
widespread in adult metagenomes (Fig.4). Availability of metatranscriptomes corresponding to the ana-
lysed infants’ metagenomes revealed pronounced transcription of the bidobacterial GH gene predicted
to be involved in milk, HMO and mucin degradation (Fig.4), clearly supporting the key functional roles
exploited by these GHs in the infant gut microbiota.
To further elucidate the functional contribution of bidobacteria to the human microbiota we screened
the metagenomic and metatranscriptomic datasets for the presence of genes encompassing carbohydrate
degradation pathways that had been predicted to be present in the bidobacterial pan-genome (Figs1
and 4). Both the adult and infant metagenomes showed a high abundance of genes for the Bidobacterium
phosphoketolase-dependent, or so-called bif shunt, pathway, as expected, as well as genes involved in
galactose degradation I, melibiose degradation, lactose degradation III, glycogen degradation I, glycogen
degradation II and sucrose degradation IV pathways (Fig.4), accompanied by presence of genes involved
in starch degradation V, arabinose degradation I, xylan degradation and xylose degradation I pathways.
e presence of such pathways illustrates their presumed importance for adaptation of bidobacteria to
the human gut environment. Interestingly, the infant’s metatranscriptomic datasets correspond to the
presence of the bif shunt (as expected) and pathways for galactose, lactose, glycogen and sucrose deg-
radation, while no or low transcriptional activity was detected that corresponded to catalytic pathways
for the plant polysaccharides arabinose, xylan and xylose (Fig.4), reecting the paucity of these carbo-
hydrates in an infant’s diet.
Conclusions
Bidobacteria may be considered as key representatives of the mammalian gut microbiota, especially
during the rst phase of their host’s life. However, very little is known about their genetic strategies
to colonize and persist within the gut, and to get access to the nutrients available in this environment.
e current study highlights the very extensive saccharolytic features displayed by members of the
Bidobacterium genus, revealing how these bacteria metabolize specic carbohydrates available in their
particular ecological niche. Comparison of the glycobiome identied in the genus Bidobacterium with
those identied in other members of the human gut microbiota revealed their unique and important
contribution in terms of GHs involved in the breakdown of complex plant carbohydrates such as arab-
inoxylan, galactan and starch. e impact of bidobacteria in the breakdown of dietary carbohydrates
is also crucial for the establishment and reinforcement of trophic relationships between members of the
gut microbiota. In fact, both mutualistic as well as commensal interactions in the mammalian gut can
be carbohydrate-driven
6,7
. Here, we have shown how a simple bidobacterial community may co-operate
between themselves as well as with other members of the gut microbiota in the utilization of specic
glycans, commonly available in the mammalian gut, by means of cross-feeding activities so as to provide
growth benets to one or both members of such a community as well as with the other members of
the gut microbiota. Such ndings support the concept of the existence of a social intelligence of bido-
bacterial communities in the harvesting and metabolism of glycans available in their ecological niches,
which regulate the dynamics of the gut microbiota relationships. However, it is reasonable to expect
that in much more complex microbial communities such as those identied in the mammalian gut,
cross-feeding activities as exemplied by bidobacteria also involve other members of the gut microbi-
ota, perhaps generating an even larger benecial eect. A survey of human gut metagenomic datasets,
representing both adult and infant samples, revealed that, notwithstanding their relatively low abundance
in adults, the functional contribution of bidobacteria to the enzymatic arsenal directed at degradation of
complex carbohydrates is relevant. In this context, the expansion of the GH repertoire dedicated to the
metabolism of infant dietary sugars, such as HMOs, as well as mucin is noteworthy. Another clear sign
of advanced bidobacterial adaptation to its mammalian host is represented by the identication in a
www.nature.com/scientificreports/
11
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
small number of bidobacterial taxa which include B. bidum, B. longum subsp. infantis
20,33
, as well as B.
biavatii, B. crudilactis, B. kashiwanohense, B. stellenboschense and B. mongoliense, of metabolic repertoires
involved in the breakdown of host-derived glycans such as HMOs and mucin.
Materials and Methods
Bacterial strains, growth conditions. All Bidobacterium strains were cultivated in an anaer-
obic atmosphere (2.99% H
2
, 17.01% CO
2
and 80% N
2
) in a chamber (Concept 400, Ruskin) on De
Man-Rogosa-Sharp (MRS) broth (Scharlau Chemie, Barcelona, Spain) supplemented with 0.05% (w/v)
L-cysteine hydrochloride and incubated at 37 °C. Cell growth on semi synthetic MRS medium sup-
plemented with 1% (wt/vol) of a particular sugar was monitored by optical density at 600 nm using a
plate reader (Biotek, Vermont, USA). e plate reader was run in discontinuous mode, with absorbance
readings performed in 60 min intervals, and preceded by 30 sec shaking at medium speed. Cultures were
grown in biologically independent triplicates and the resulting growth data were expressed as the mean
of these replicates. Carbohydrates were purchased from Sigma (Milan, Italy) or Carbosynth (Berkshire,
UK). All the carbohydrates were dissolved in water and then sterilized by ltration using 0.2 micron l-
ter size and then added to autoclaved MRS with the exception of xylan (Poly(β -D-xylopyranose[1→ 4])
(Sigma, Aldrich) which was autoclaved with MRS.
In the case of HMO experiments, human milk samples were kindly provided by the Western rust
Human Milk Bank (Irvinestown, Co. Fermanagh, Ireland). e isolation of oligosaccharides from pooled
samples (n ≥ 3) was performed as described previously
35
. Briey, samples were defatted by centrifugation
at 4 °C (3850 × g, 20 min, Sorvall RC6 plus
®
). Caseins were precipitated at pH 4.6. Aer neutralization,
large peptides were removed by ultraltration (5 kDa molecular-weight cut-o, Millipore
®
Helicon S10
Spiral Cartridge). e permeates were freeze-dried and stored at − 80 °C until further processing. To
remove lactose and residual peptides, the extracts were re-solubilized in MilliQ
®
water and applied to a
Sephadex G-25 column (Pharmacia, Uppsala, Sweden; 92 × 2.6 cm). Elution was performed with deion-
ized water (5 mL/min). Fractions were monitored for peptides according to Bradford
36
and the lactose
content was determined by HPAEC
35
. Fractions low in peptide- and lactose content were pooled and
used for further characterization.
Bidobacterial and hosts phylogenetic reconstruction. e phylogeny of the 47 was recon-
structed as described previously
9,10
. Phylogeny of the eukaryotic hosts was reconstructed through the
SUPERFAMILY web tool
37
.
Bidobacterial transcriptomics identication through RNAseq assays. B. biavatii, B. boum,
B. stellenboschense, B. coryneforme, B. cuniculi, B. gallinarum and B. pseudocatenulatum were selected
as representatives for each of the seven phylogenetic groups. ese strains were grown in MRS sup-
plemented with glucose, glycogen, lactose, xylose, rhamnose, mannose, trehalose or HMO as a carbon
source. In order to evaluate cross-feeding activities between two bidobacterial strains, bacteria were
co-cultivated on MRS supplemented with a particular glycan [RS2-resistant starch (Sigma, Aldrich),
xylan (Poly(β -D-xylopyranose[1→ 4]) (Sigma, Aldrich) or glycogen (Sigma, Aldrich)], and growth was
monitored by the evaluation of optical density at 600 nm followed by qRT-PCR using strain-specic
primers (see below).
When growth was observed during logarithmic phase, cell pellets were resuspended in 1 ml of
QUIAZOL (Quiagen, UK) and placed in a tube containing 0.8 g of glass beads (diameter, 106 μ m; Sigma).
e cells were lysed by shaking the mix on a BioSpec homogenizer at 4 °C for 2 min (maximum set-
ting). e mixture was then centrifuged at 12,000 rpm for 15 min, and the upper phase containing the
RNA-containing sample was recovered. e RNA sample was further puried by phenol extraction and
ethanol precipitation according to an established method
38
. Quality and integrity of the RNA was checked
by Agilent 2200 Tape Station Nucleic Acid System (Agilent Technologies, Palo Alto, Calif.). One hundred
ng of total RNA was used as the starting input for RNA-Seq library preparation. Briey, 100 ng of total
RNA was treated with Ribo-Zero rRNA removal kit for Gram-positive bacteria (Epicentre, Madison,
WI, U.S.A.) to remove rRNA according to the supplier’s instructions. e yield of rRNA depletion was
checked by Agilent 2200 Tape Station Nucleic Acid System (Agilent Technologies, Palo Alto, Calif.).
en, rRNA-depleted RNA samples were fragmented using RNaseIII (Life Technologies, USA) followed
by size evaluation using Experion (BioRad, UK). Whole transcriptome library was constructed using the
Ion Total-RNA Seq Kit v2 (Life Technologies, USA). Barcoded libraries were quantied by qRT-PCR
and each library template was amplied on Ion Sphere Particles using Ion One Touch 200 Template
Kit v2 (Life Technologies, USA). Samples were loaded into 316 Chips and sequenced on the PGM (Life
Technologies, USA). Sequencing reads were depleted of adapters, quality ltered (with overall quality,
quality window and length lters) with FastqMcf (https://code.google.com/p/ea-utils/) and aligned to the
respective bidobacterial reference genome through BWA
39
with high stringency cut-os (99% nucleotide
identity) in order to accurately map reads of co-cultivation datasets on the correct genome. Counts of
reads overlapping ORFs were performed using HTSeq (http://www-huber.embl.de/users/anders/HTSeq/
doc/overview.html) and analysis of the count data was performed using the R package DESeq2
40
. When
the RNAseq analyses of mono-associations, DESeq2 output consists of fold induction values determined
www.nature.com/scientificreports/
12
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
as the normalized number of transcripts identied for a given gene for bacterial cells cultivated on MRS
containing a specic carbohydrate, relative to the number of identied transcripts for that same gene
when the strain was grown on MRS containing glucose (reference condition). In case of RNAseq anal-
ysis of bi-associations, DESeq2 output consists of fold-induction values determined as the normalized
number of transcripts identied for a given gene for bacterial cells cultivated in mono-association relative
to the normalized number of identied transcripts for that same gene when the strain was cultivated in
bi-association, using MRS supplemented with the same carbohydrate for both conditions. DESeq2 data
normalization and dierential gene expression analysis are based on the negative binomial distribution,
which takes into consideration the overall transcript abundance identied for each analysed sample
40
.
Evaluation of the cell density by qPCR. e co-cultivation with other bacteria was monitored by
quantitative PCR (qPCR). e copy-number of a gene for a given strain used in the co-cultivation exper-
iments was evaluated and compared to the growth rate of each individually cultivated microorganism.
qPCR was performed using the CFX96 system (BioRad, CA, USA). Primers used in this study are listed
in Table S2. Each PCR reaction mix contained the following: 7.5 μ l 2× SYBR SuperMix Green (BioRad,
CA, USA), 5 μ l of DNA dilution, each of the forward and reverse primers at 0.5 μ M and nuclease-free
water was added to obtain a nal volume of 15 μ l. PCR products were detected with SYBR Green uo-
rescent dye and amplied according to the following protocol: one cycle of 95 °C for 3 minutes, followed
by 39 cycles of 95 °C for 5 s and 60 °C for 20 s. Melting curve: 65 °C to 95 °C with increments of 0.5 °C/s.
In each run, negative controls (no DNA) for each primer set were included. Standard curve was built
using the CFX96 soware (BioRad).
Bidobacterial gene survey of human gut metagenomic and metatranscriptomic datasets.
We surveyed the presence of bidobacterial genes into the microbial diversity of the healthy gut of
44 adults and nine infants. To this end we implemented a mapping-based pipeline to detect the pres-
ence and quantify the coverage of these gene categories annotated from the newly sequenced strains
into the Illumina deep shotgun-sequenced metagenomic data sets derived from stool samples of the
Human Microbiome Project (HMP)
19
and into Illumina deep shotgun-sequenced metagenomic and
metatranscriptomic data sets obtained from stool samples of nine healthy infants. e mapping was
performed using BowTie2
41
using multiple-hit mapping and “very-sensitive” policy. e mapping was
post-processed with a custom script to retain those matches with at least 98% full-length identity with
respect to at least one reference gene by threshold-ing the BowTie2 score at −12 for the 100 nt-long
adults and infant datasets (−6 is the penalty for a high-quality mismatch). Genes covered by matching
reads for less than 90% of their full length were discarded and the nal gene-wise average coverage was
computed using SAMtools
42
and BEDtools
43
. In the nal matrices, GHs genes and genes constituting
carbohydrate-degrading pathways were collapsed into GH families and MetaCyc pathways. e estima-
tion of the relative abundance of bidobacteria (at the genus level) was performed with MetaPhlAn
34
.
Data Deposition. e RNAseq data were deposited in SRA database under the following study acces-
sion numbers: PRJNA239567 and PRJNA277297.
References
1. Ventura, M., Turroni, F., Motherway, M. O., MacSharry, J. & van Sinderen, D. Host-microbe interactions that facilitate gut
colonization by commensal bidobacteria. Trends in microbiology 20, 467–476, doi: 10.1016/j.tim.2012.07.002 (2012).
2. Bergey, D. H., Goodfellow, M., Whitman, W. B. & Parte, A. C. 1 online resource (2 v.) (Springer, New Yor, 2012).
3. Turroni, F. et al. Diversity of Bidobacteria within the Infant Gut Microbiota. PLoS One 7, e36957, doi: 10.1371/journal.
pone.0036957 (2012).
4. Milani, C. et al. Assessing the fecal microbiota: an optimized ion torrent 16S rNA gene-based analysis protocol. PLoS One 8,
e68739, doi: 10.1371/journal.pone.0068739 (2013).
5. Turroni, F. et al. Exploring the diversity of the bidobacterial population in the human intestinal tract. Applied and environmental
microbiology 75, 1534–1545, doi: 10.1128/AEM.02216-08 (2009).
6. El aoutari, A., Armougom, F., Gordon, J. I., aoult, D. & Henrissat, B. e abundance and variety of carbohydrate-active
enzymes in the human gut microbiota. Nat ev Microbiol 11, 497–504, doi: 10.1038/nrmicro3050 (2013).
7. oropatin, N. M., Cameron, E. A. & Martens, E. C. How glycan metabolism shapes the human gut microbiota. Nature reviews.
Microbiology 10, 323–335, doi: 10.1038/nrmicro2746 (2012).
8. Ventura, M. et al. Genome-scale analyses of health-promoting bacteria: probiogenomics. Nature reviews. Microbiology 7, 61–71,
doi: 10.1038/nrmicro2047 (2009).
9. Milani, C. et al. Genomic encyclopedia of type strains of the genus Bidobacterium. Applied and environmental microbiology 80,
6290–6302, doi: 10.1128/AEM.02308-14 (2014).
10. Lugli, G. A. et al. Investigation of the evolutionary development of the genus Bidobacterium by comparative genomics. Applied
and environmental microbiology 80, 6383–6394, doi: 10.1128/AEM.02004-14 (2014).
11. Powell, S. et al. eggNOG v4.0: nested orthology inference across 3686 organisms. Nucleic acids research 42, D231–239, doi:
10.1093/nar/gt1253 (2014).
12. Ventura, M., Canchaya, C., Fitzgerald, G. F., Gupta, . S. & van Sinderen, D. Genomics as a means to understand bacterial
phylogeny and ecological adaptation: the case of bidobacteria. Antonie Van Leeuwenhoe 91, 351–372, doi: 10.1007/s10482-006-
9122-6 (2007).
13. Lombard, V., Golaconda amulu, H., Drula, E., Coutinho, P. M. & Henrissat, B. e carbohydrate-active enzymes database
(CAZy) in 2013. Nucleic acids research 42, D490–495, doi: 10.1093/nar/gt1178 (2014).
14. Pousaeva, ., Fitzgerald, G. F. & van Sinderen, D. Carbohydrate metabolism in Bidobacteria. Genes & nutrition 6, 285–306,
doi: 10.1007/s12263-010-0206-6 (2011).
www.nature.com/scientificreports/
13
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
15. ennie, E. A. & Scheller, H. V. Xylan biosynthesis. Current opinion in biotechnology 26C, 100–107, doi: 10.1016/j.copbio.2013.11.013
(2014).
16. Foschia, M., D. P., Sensidoni, A. & Brennan, C. S. e eects of dietary bre addition on the quality of common cereal products.
Journal of Cereal Science 58, 216–227 (2013).
17. Lee, F. J., usch, D. B., Stewart, F. J., Mattila, H. . & Newton, I. L. Saccharide breadown and fermentation by the honey bee
gut microbiome. Environmental microbiology 17, 796–815, doi: 10.1111/1462-2920.12526 (2015).
18. Gill, S. . et al. Metagenomic analysis of the human distal gut microbiome. Science 312, 1355–1359, doi: 10.1126/science.1124234
(2006).
19. Human Microbiome Project, C. Structure, function and diversity of the healthy human microbiome. Nature 486, 207–214, doi:
10.1038/nature11234 (2012).
20. Turroni, F. et al. Genome analysis of Bidobacterium bidum PL2010 reveals metabolic pathways for host-derived glycan
foraging. Proc Natl Acad Sci USA 107, 19514–19519, doi: 10.1073/pnas.1011100107 (2010).
21. iyohara, M. et al. An exo-alpha-sialidase from bidobacteria involved in the degradation of sialyloligosaccharides in human
mil and intestinal glycoconjugates. Glycobiology 21, 437–447, doi: 10.1093/glycob/cwq175 (2011).
22. arp, P. D. et al. Pathway Tools version 13.0: integrated soware for pathway/genome informatics and systems biology. Briengs
in bioinformatics 11, 40–79, doi: 10.1093/bib/bbp043 (2010).
23. Wang, L. & Wise, M. J. Glycogen with short average chain length enhances bacterial durability. Die Naturwissenschaen 98,
719–729, doi: 10.1007/s00114-011-0832-x (2011).
24. Goh, Y. J. & laenhammer, T. . Insights into glycogen metabolism in Lactobacillus acidophilus: impact on carbohydrate
metabolism, stress tolerance and gut retention. Microbial cell factories 13, 94, doi: 10.1186/s12934-014-0094-3 (2014).
25. Jones, S. A. et al. Glycogen and maltose utilization by Escherichia coli O157:H7 in the mouse intestine. Infection and immunity
76, 2531–2540, doi: 10.1128/IAI.00096-08 (2008).
26. Henrissat, B., Deleury, E. & Coutinho, P. M. Glycogen metabolism loss: a common marer of parasitic behaviour in bacteria?
Trends in genetics: TIG 18, 437–440 (2002).
27. Misof, B. et al. Phylogenomics resolves the timing and pattern of insect evolution. Science 346, 763–767, doi: 10.1126/
science.1257570 (2014).
28. Wyatt, G. . e Biochemistry of Sugars and Polysaccharides in Insects. Advances in Insect Physiology 4, 287–360 (1967).
29. Egan, M. et al. Cross-feeding by Bidobacterium breve UCC2003 during co-cultivation with Bidobacterium bidum PL2010
in a mucin-based medium. BMC microbiology 14, 282, doi: 10.1186/s12866-014-0282-7 (2014).
30. Sonnenburg, J. L., Chen, C. T. & Gordon, J. I. Genomic and metabolic studies of the impact of probiotics on a model gut
symbiont and host. PLoS Biol 4, e413, doi: 10.1371/journal.pbio.0040413 (2006).
31. Barile, D. & astall, . A. Human mil and related oligosaccharides as prebiotics. Current opinion in biotechnology 24, 214–219,
doi: 10.1016/j.copbio.2013.01.008 (2013).
32. itaoa, M. Bidobacterial enzymes involved in the metabolism of human mil oligosaccharides. Advances in nutrition 3,
422S–429S, doi: 10.3945/an.111.001420 (2012).
33. Sela, D. A. et al. e genome sequence of Bidobacterium longum subsp. infantis reveals adaptations for mil utilization within
the infant microbiome. Proceedings of the National Academy of Sciences of the United States of America 105, 18964–18969, doi:
10.1073/pnas.0809584105 (2008).
34. Segata, N. et al. Metagenomic microbial community proling using unique clade-specic marer genes. Nature methods 9,
811–814, doi: 10.1038/nmeth.2066 (2012).
35. Marino, . et al. Method for mil oligosaccharide proling by 2-aminobenzamide labeling and hydrophilic interaction
chromatography. Glycobiology 21, 1317–1330, doi: 10.1093/glycob/cwr067 (2011).
36. Bradford, M. M. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of
protein-dye binding. Analytical biochemistry 72, 248–254 (1976).
37. Wilson, D. et al. SUPEFAMILY–sophisticated comparative genomics, data mining, visualization and phylogeny. Nucleic acids
research 37, D380–386, doi: 10.1093/nar/gn762 (2009).
38. Sambroo, J., E. F. F. & Maniatis, T. Molecular cloning: a laboratory manual. Cold Spring Harbor, NY: Cold Spring Harbor
Laboratory, 545 (1989).
39. Li, H. & Durbin, . Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760, doi:
10.1093/bioinformatics/btp324 (2009).
40. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for NA-seq data with DESeq2. Genome
biology 15, 550, doi: 10.1186/s13059-014-0550-8 (2014).
41. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nature methods 9, 357–U354, doi: Doi 10.1038/
Nmeth.1923 (2012).
42. Li, H. et al. e Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079, doi: 10.1093/bioinformatics/
btp352 (2009).
43. Quinlan, A. . & Hall, I. M. BEDTools: a exible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842,
doi: 10.1093/bioinformatics/btq033 (2010).
Acknowledgements
We thank GenProbio srl for nancial support of the Laboratory of Probiogenomics. is work was
nancially supported by a FEMS Jensen Award to FT, and by a Ph.D. fellowship (Spinner 2013, Regione
Emilia Romagna) to S.D. DvS and FT are members of e APC Microbiome Institute, while DvS is also
a member of the Alimentary Glycoscience Research Cluster, both funded by Science Foundation Ireland
(SFI), through the Irish Government’s National Development Plan (Grant numbers SFI/12/RC/2273
and 08/SRC/B1393, respectively). is work was also partially supported by Fondazione Caritro, by
the EU FP7 (PCIG13- GA-2013-618833), and by MIUR “Futuro in Ricerca” E68C13000500001 to NS.
Furthermore, this project has been funded in part with funds from the National Institute of Allergy and
Infectious Diseases, National Institutes of Health, Department of Health and Human Services, under
Contract No. HHSN272200900018C.
Author Contributions
C.M. designed all of the experiments, performed bionformatics analyses as well as pathways predictions
and wrote the manuscript. G.A.L. and L.M. performed genomics and comparative genomics bioinformatic
analyses. N.S., D.V.W. and M.S. were involved in screening of metagenomics datasets. S.D., F.T., C.F.,
www.nature.com/scientificreports/
14
SCIENTIFIC RepoRts | 5:15782 | DOI: 10.1038/srep15782
M.M. and A.V. performed all the experiments involving RnaSeq sequencing while A.H., S.A., B.S., J.L.,
R.H. and D.M. were involved in growth of bidobacterial strains on dierent substrates. M.V., A.M. and
D.V.S. conceived the study, revised and approved the manuscript. All authors reviewed the manuscript.
Additional Information
Supplementary information accompanies this paper at http://www.nature.com/srep
Competing nancial interests: e authors declare no competing nancial interests.
How to cite this article: Milani, C. et al. Bidobacteria exhibit social behavior through carbohydrate
resource sharing in the gut. Sci. Rep. 5, 15782; doi: 10.1038/srep15782 (2015).
is work is licensed under a Creative Commons Attribution 4.0 International License. e
images or other third party material in this article are included in the article’s Creative Com-
mons license, unless indicated otherwise in the credit line; if the material is not included under the
Creative Commons license, users will need to obtain permission from the license holder to reproduce
the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/
