Journal of Microbiological Methods 94 (2013) 3746

Contents lists available at SciVerse ScienceDirect

Journal of Microbiological Methods

journal homepage: www.elsevier.com/locate/jmicmeth

Comparison between terminal-restriction fragment length polymorphism

(T-RFLP) and quantitative culture for analysis of infants' gut microbiota
Fei Sjberg a,, Forough Nowrouzian a, Ignacio Rangel b, Charles Hannoun a, Edward Moore a,
Ingegerd Adlerberth a, Agnes E. Wold a
a
b

Department of Infectious Diseases; Institute of Biomedicine, University of Gothenburg, Guldhedsgatan 10A, S-413 46 Gothenburg, Sweden
Department of Clinical Medicine; School of Health and Medical Sciences, rebro University, 70182 rebro, Sweden

a r t i c l e

i n f o

Article history:
Received 20 December 2012
Received in revised form 30 March 2013
Accepted 4 April 2013
Available online 11 April 2013
Keywords:
T-RFLP
Quantitative culture
Clone
Infantile gut
Microbiota diversity
16S RNA

a b s t r a c t
The infantile intestinal microbiota is a major stimulus for immune maturation. Both culture and DNA-based
methods can be used for microbiota characterization, but few studies have systematically compared their performance for analysis of the gut microbiota. Here, we examined fecal samples obtained on six occasions between
one week and 12 months of age from six vaginally delivered infants. After quantitative aerobic and anaerobic
culture of the samples on selective and non-selective media, DNA was extracted from the fecal samples and
analyzed regarding 16S rRNA gene polymorphism by terminal-restriction fragment length polymorphism
(T-RFLP). A database was constructed for direct identication of T-RFLP peaks by analysis of pure-culture bacteria
and analysis of a limited number of samples by 16S rRNA cloning and sequencing. Bacterial genera present at
>106 CFU/g feces, as determined by quantitative culture, were generally readily detected by T-RFLP, while
culture on selective media was more sensitive in detecting facultative anaerobes with lower population counts.
In contrast, T-RFLP more readily than culture detected several anaerobic species, also taxa that could not be
identied using the database. T-RFLP readily identied bacteria to the genus level and also provided some
sub-genus discrimination. Both T-RFLP and culture identied Bidobacterium, Clostridium and Bacteroides
spp. among the most common colonizers of the infantile microbiota throughout the rst year of life.
T-RFLP analysis showed that microbiota complexity was high in the rst weeks of life, declined to a minimum at 12 months of age, and thereafter increased again. Principal component analysis revealed that
early samples (1 week6 months) chiey differed between individual infants, while 12-month samples
were similar between children, but different from the early samples. Our results indicate that T-RFLP has
high sensitivity and adequate taxonomic discrimination capacity for analysis of gut microbiota composition,
but that both culture and molecular based analysis have limitations and both approaches may be needed to obtain a full picture of the complex gut microbiota.
2013 Elsevier B.V. All rights reserved.

1. Introduction
The human gastrointestinal tract harbors a complex microbial ecosystem, comprised mainly of strictly anaerobic species (Savage, 1977).
The intestinal microbiota is a major stimulus for the immune system (Crabbe et al., 1968) and we and others have shown that a low

Abbreviation: T-RFLP, terminal-restriction fragment length polymorphism; T-RF,

terminal-restriction fragment; CCFA, cycloserine cefoxitin fructose egg yolk agar; CFU,
colony forming units; PCA, principal component analysis; O2PLS-DA, orthogonal 2 partial least squares-discriminant analysis.
There is no conict of interest. Founding sources had no involvement in study design;
in the collection, analysis and interpretation of data; in the writing of the report; and in the
decision to submit the article for publication.
Corresponding author at: Department of Infectious Diseases, Clinical Bacteriology,
Guldhedsgatan 10A, S-413 46 Gothenburg, Sweden. Tel.: +46 31 3424623; fax: +46
31 3424975.
E-mail address: fei.sjoberg@microbio.gu.se (F. Sjberg).
0167-7012/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.mimet.2013.04.002

complexity of the intestinal microbiota in young infants is associated

with increased risk of later allergy development (Abrahamsson et al.,
2012; Bisgaard et al., 2011; Wang et al., 2008).
The microbiota is established gradually as infants acquire bacteria
from the birth canal, from skin contacts with hospital staff and
family, from foods and ambient air (Adlerberth and Wold, 2009).
Due to the initially high oxygen content, facultative bacteria, such
as Streptococcus, Enterococcus, Enterobacteriaceae (Ellis-Pegler et al.,
1975) and Staphylococcus species (Lindberg et al., 2011) are the rst
to establish in the baby's gut. They are followed by anaerobic bacteria,
starting with relatively oxygen tolerant groups, such as Bidobacterium
and Bacteroides spp., and followed by more and more oxygen-sensitive
species until a complex adult type microora has built up after some
years (Adlerberth and Wold, 2009). With decreasing oxygen tension,
facultative bacteria decline in numbers. Hence, the facultative:anaerobe
ratio decreases from roughly 1:1.5 in newborn infants to 1:200 in adults
(Ellis-Pegler et al., 1975).

38

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

Most basic knowledge regarding the infant gut microbiota derives

from culture-based studies. In recent years, molecular techniques
have emerged permitting identication of bacteria that cannot be
cultured, mainly very strict anaerobes. Heterogeneity of the ubiquitous bacterial 16S rRNA gene, 1500 bp in length, is most commonly
used for identication, e.g. by cloning and sequencing. An alternative
is to amplify parts of the 16S rRNA gene by PCR followed by multiparallel sequencing, i.e. next generation sequencing, in which very
large numbers of sequences can be read. Molecular techniques permit
identication of non-culturable anaerobes present in high population
numbers. However, subdominant bacterial populations may escape
detection, including facultative anaerobes that constitute b1% of the
gut microbiota. For example, to detect a species present at 106 CFU/g
among 1011 bacteria/g requires that 100,000 sequences be analyzed.
An alternative approach is to digest PCR amplied 16S rRNA genes by
restriction enzymes and separate the fragments according to size. In
terminal-restriction fragment length polymorphism (T-RFLP), the
terminal fragment is uorescence-labeled during the PCR reaction
which permits its detection. Each bacterial taxon (genus or species)
produces a T-RF (terminal-restriction fragment) of a certain molecular
weight (Liu et al., 1997) and a complex bacterial community generates
a series of DNA fragments differing in size. A disadvantage is that the
identity of the T-RF is not immediately apparent; databases are needed
to translate fragment size to genus/species.
Few studies have directly compared the sensitivity of culture and
DNA based methods regarding characterization of the gut microbiota.
Hayashi et al. (2002) compared strictly anaerobic culture and 16S
rRNA clone libraries. Culture was able to identify approximately 30% of
the bacterial cells identied in the microscope count and 75% of the sequenced clones were not detected by culture. In contrast, Wilson and
Blitchington (1996) found that 16S rRNA cloning only identied slightly
more bacterial groups than did culture. T-RFLP and culture were compared regarding identication of meat spoiling bacterial communities;
lactic acid bacteria present at 107 CFU/g on average were readily
detected, while enterobacteria present at 103 CFU/g escaped detection
(Nieminen et al., 2011). A comparison between culture and T-RFLP in
characterization of bacterial communities on rice leaves showed that
certain genera were only detected by culture-independent, DNA-based
analyses, while other bacteria were only found using a culture-based approach. The latter mainly included Gammaproteobacteria such as Pseudomonas, but there were no reports on the respective population levels
of the bacteria that were identied, or missed, by T-RFLP (Ferrando et al.,
2012). Despite the renewed interest in the commensal microbiota, and
notwithstanding extensive searches of the literature, we have not
found any systematic comparison of the advantages and limitation between T-RFLP and culture when assessing complex bacterial communities, such as the gut microbiota.
The aim of this study was to compare T-RFLP with quantitative culture for analysis of the intestinal microbiota in infants. Parallel analysis
of fecal samples by T-RFLP and quantitative aerobic and anaerobic culture using a range of selective media permitted us to calculate the sensitivity of T-RFLP in relation to culture for different groups of gut bacteria
and investigate the utility of T-RFLP for characterization of the infantile
microbiota. A limited number of samples were also analyzed by cloning
and sequencing. Further, the work included development of a database
enabling rapid identication of terminal-restriction fragments (T-RFs)
derived from infantile gut bacteria to the genus or species level.

development (Adlerberth et al., 2006, 2007). None of the studied infants

had received any antibiotics during the study period. Five were exclusively breastfed until 46 months of age while one (infant B) received
formula from birth. Solid foods were introduced at 46 months of age.
Informed consent was obtained and the Medical Ethics' Committee of
University of Gothenburg approved the study.
2.2. Fecal sample collection and culture
Stools were collected by the parents on six occasions during the rst
year of life (at 1, 2 and 4 weeks and at 2, 6 and 12 months of age; called
samples 16). The samples were transported in an anaerobic milieu in a
gas-tight sachet and cultured aerobically and anaerobically on selective
and non-selective media within 24 h after voiding as previously
described (Adlerberth et al., 2006, 2007). The detection limit was 300
(102.5) colony forming units (CFU)/g of feces. In brief, a calibrated
spoonful of feces was serially diluted and cultivated on non-selective
media (Columbia blood agar for aerobic culture and Brucella blood
agar for anaerobic culture) and 9 selective media (Drigalski for
Enterobacteriaceae, Staphylococcus agar for Staphylococcus, Bacteroides
bile esculin agar for Bacteroides, Beerens agar for Bidobacterium,
Enterococcosel agar for Enterococcus, Sabouraud agar for yeasts, Rogosa
agar for Lactobacillus and Cycloserine Cefoxitin Fructose egg yolk agar
(CCFA)for Clostridium difcile. Anaerobic spore formers (clostridia)
were quantied by anaerobic culture of alcohol-treated samples on
Brucella blood agar. Bacterial colonies of different morphologies were
enumerated on appropriate media, subcultured for purity and speciated
using a combination of biochemical tests and genus or species-specic
PCRs as previously described (Adlerberth et al., 2007). Pure-culture
isolates were stored frozen at 80 C as were remaining feces.
2.3. Pure-culture and type strains
For construction of a T-RF database, pure-culture strains deriving
from culture of infants' microbiota and culture collection strains
were used. In total, 42 gut bacterial strains were obtained from the
Culture Collection of the University of Gothenburg (CCUG), 36 of which
were type strains of the respective species (Supplement Table B). Infant
pure-culture isolates were cultured and subcultured for purity and their
16S rRNA gene was sequenced (Section 2.6.2) to conrm their species
identity.
2.4. Extraction of DNA and PCR of 16S rRNA genes
2.4.1. DNA extraction from feces
Bacterial DNA was extracted from 160 mg (wet weight) of frozen
infant feces using the QIAamp DNA Stool Mini Kit (Qiagen, Hildens
Germany). DNA yield was increased by adding an extra step, in
which 1012 glass beads (2 mm in diameter) were added to the fecal
ASL buffer suspension and shook at 1200 rpm (IKA vibrax Shaker,
Labortechnik, Germany) for 30 min at 5 C. DNA was quantied spectrophotometrically (v3.3 NanoDrop Technologies, Wilmington, USA).

2.1. Subjects

2.4.2. DNA extraction from pure-culture and type strains

DNA from Enterobacteriaceae, Bidobacterium spp. and Lactobacillus
spp. was extracted by incubation in 50 l lysis buffer (10 mM Tris HCl,
1 mM EDTA and 10 mM saline) at 95 C for 10 min, followed by
centrifugation for 3 min at 12,000 g. Other bacteria were lysed
following the InstaGene Matrix protocol (Bio-Rad Laboratories,
Richmond, USA). The DNA concentrations were measured with the
NanoDrop spectrophotometer.

Six vaginally delivered infants (called A-F) born in 19982002 and

followed to one year of age were studied. They were selected from the
Swedish ALLERGYFLORA birth-cohort recruited to examine the relation between early intestinal colonization pattern and later allergy

2.4.3. PCR amplication of bacterial 16S rRNA genes

Bacterial 16S rRNA genes were amplied using the universal
primers ENV1 (5-6-FAM-AGA GTT TGA TII TGG CTC AG -3, Escherichia
coli nr. 8-27) and ENV 2 (5-CGG ITA CCT TGT TAC GAC TT-3, E. coli nr.

2. Materials and methods

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

1511-1492) (Gutell, 1994). For T-RFLP, the forward primer was uorescently labeled with 6-FAM at 5. The PCR mixture contained 50 ng DNA
(25 and 75 ng were also tested; 25 ng DNA generated smaller T-RF
peaks), 25 l Hot Start Taq Master Mix (Qiagen), 0.3 M primer,
1 mM MgCl2 and water in a nal volume of 50 l (fecal samples) or
25 l (pure-culture strains and the reagents were cut in halves). The
PCR reaction was performed in an Eppendorf Mastercycler gradient
(Eppendorf, Hamburg, Germany) using the following program: an
initial activation of Taq polymerase at 95 C for 15 min, 25 cycles (28
and 32 were also tested) of 94 C for 1 min, 50 C for 45 s and 72 C
for 2 min, with a nal extension at 72 C for 7 min. Negative controls
containing all DNA extraction reagents, but without bacteria, were
included throughout the whole analysis to detect DNA contamination
from reagents or instruments.
2.5. T-RFLP analysis
2.5.1. Restriction enzyme digestion
Puried PCR products from fecal samples (100 ng) were digested
with 16 U (4, 10, and 20 U were also tested) of the MspI endonuclease
restriction enzyme (New England Biolab, Ipswich, USA) at 37 C for
5 h in a nal volume of 5 l. For pure-culture bacteria, 30 ng PCR
products were digested with 16 U MspI in a nal volume of 10 l. The
reaction was stopped by heating at 65 C for 20 min.
2.5.2. Fragment analysis
Fluorescently labeled T-RFs were detected using an ABI PRISM 310
genetic analyzer (Applied Biosystems, California, USA) with pop 4 gel.
Resolution (number and quality of peaks) was compared using different
volumes of digested PCR amplicons (0.5, 0.7, 1, 1.5, 2.5, 3.8 and 4.8 l),
run voltages (9.9, 12.2 kV and 15 kV), injection times (5 s and 10 s)
and size standards (ROX 500, ROX1000, combined ROX 500 and
ROX1000, LIZ 600 and LIZ 1200 Applied Biosystems). Reproducibility
was investigated by running samples in duplicate.
The fragment lengths were analyzed by GeneMapper, version 4
(Applied Biosystems) using Local Southern Method. True peaks
were separated from noise within a fragment length range between
28 and 1000 bp according to Abdo et al. (2006). Briey, the data were
standardized by dividing the area of each peak by the total peak area
of the particular sample. The standard deviation (SD) of the dataset
was then computed assuming that the true mean was zero. Peaks
with an area of 3 SD above the mean were considered as true signals.
They were collected and removed; whereafter the process was iterated
until no more true peaks were identied. T-RFs that differed by no
more than 1 bp in different analyses were considered as identical.
All fecal samples were analyzed in two independent PCRs and subsequent T-RFLP analyses. Only fragments represented in both runs were
considered and their peak sizes and areas were averaged.
2.5.3. In silico analysis
To obtain the expected T-RF size of different bacterial species,
in silico digestion was performed (RDP website: http://rdp8.cme.msu.
edu/html/TAP-trp.html). Cutting sites for the MspI enzyme in their
16S rRNA genes were detected and the size of the terminal fragments
were calculated and compared with observed T-RF size obtained from
T-RFLP analysis of the same type strains.
2.6. Cloning and sequencing

39

solution (2 l) was transformed into JM109 high efciency competent

cells (Promega) or Bioblue chemically competent cells (BIOLINE, UK).
Transformed cells were cultured on triplicate Luria-Bertani agar plates
containing ampicillin, IPTG and X-Gal (Promega) incubated overnight at 37 C. A total of 927 white colonies were analyzed (51292
colonies/sample).
2.6.2. Sequencing of the cloned 16S rRNA gene
Clone inserts were amplied with M13 vector primers: M13 forward
5-CGC CAG GGT TTT CCC AGT CAC GAC and M13 reverse 5-TCA CAC
AGG AAA CAG CTA TGA C (TAG Copenhagen A/S, Denmark) using the
following PCR program: 95 C for 15 min, 30 cycles of 94 C for 30s,
55 C for 30 s and 72 C for 90 s, with a nal extension step at 72 C
for 10 min. PCR products (4 l) were veried by agarose gel electrophoresis, inserts with approximately 1750 nt were passed to a second PCR
using 16S universal primers as described for analysis of fecal isolates.
PCR products (4 l) were again veried by agarose gel electrophoresis.
Clones of around 1500 nt size were digested using 0.2 l of the 16S
PCR amplicon, 8 units of MspI and their fragment size were analyzed.
Within one species, there are several enzyme restriction sites along
the 16S gene. Complete digestion yields a single T-RF (the digestion
site nearest the forward primer). When using inadequate enzyme concentration other digestion sites may be detected. Such partial digestion
was used when analyzing clones in order to obtain a more complex pattern. Clones with the same partial digestion T-RF prole were considered to be identical and at least two representative clones of each
unique T-RF prole were selected for sequencing.
Sequencing PCR was performed using the Bigdye terminator
v3.1 cycle sequencing kits (Applied Biosystems) and puried M13 PCR
product was used as template. The 20 l reaction mix contained 2.5 l
of the M13 PCR product, 2 l Bigdye terminator, 2 l buffer and 20 M
of primer M13 forward, M13 reverse or universal primer 907r (5-CCG
TCA ATT CCT TTA AGT TT-3 E. coli nr. 907926). Amplication was
done with the following program: 25 cycles of 96 C for 30 s, 55 C
for 15 s and 60 C for 4 min. Amplicons were precipitated in 50 l
95% ethanol and 2 l sodium acetate buffer (3 M) for 20 min at
room temperature. After centrifugation (14,000 g, 20 min) the
pellet was washed with 250 l 70% ethanol and centrifuged again
(14,000 g, 5 min). The supernatant was removed and the remaining
uid evaporated (room temperature, 1 h). The pellet was dissolved in
15 l formamide. DNA was sequenced on an ABI PRISM 3100 Avant genetic analyzer (Applied Biosystems) and the result was analyzed using
the DNA sequencing analysis program, v5.2 (Applied Biosystems) and
aligned using FASTA at the European Bioinformation Institute's public
database http://www.ebi.ac.uk/Tools/sss/fasta/.
2.7. Statistical analysis
Principal component analysis (PCA) and orthogonal 2 partial least
squares-discriminant analysis (O2PLS-DA) were used to obtain an overall pattern of the infantile gut microbiota (SIMCA P + version 12,
Umetrics AB, Ume, Sweden). PCA depicts the inherent structure of
multivariate data and major correlations between observations and variables (Ramette, 2007). O2PLS-DA is a regression variant of PCA by
which correlation between a multivariate dataset and a Y variable can
be tested (Westerhuis et al., 2010). Wilcoxon matched pairs T-test
was used to compare the diversity of the T-RFs obtained at different
ages (Graphpad Prism 5, San Diego, USA).

A clone library was constructed from nine fecal samples obtained

from four infants. These samples were selected based on their high
abundance of T-RFs that could not be identied from cultured bacteria.

3. Results

2.6.1. Ligation and transformation

PCR products (70 ng DNA) were ligated into pGEM-T Easy Vector Systems (Promega, USA) according to the manufacturer's protocol. Ligation

3.1.1. T-RFLP optimization

In T-RFLP, identication of a bacterial taxon is based on the size of
the terminal fragment obtained after digestion of the 16S rRNA gene.

3.1. T-RFLP methodology

40

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

It is crucially important that a single labeled fragment (T-RF) is produced after PCR and digestion of the 16S rRNA gene from a single
taxon. If more than one fragment is generated, each of these can be
interpreted as a unique bacterial species, i.e. a pseudo-T-RF. First
the PCR reaction was optimized using pure-culture strains. Up to
25 cycles, only 1500 nt amplicons, representing the full length 16S
rRNA gene, were produced, while after >25 cycles, fragments of other
sizes appeared that could be interpreted as T-RFs. Further, using b16 U
of the restriction enzyme MspI per 100 ng PCR amplicon led to appearance of pseudo-T-RFs as a result of incomplete digestion.
The optimal T-RFLP run conditions were found to be: a mixture of
1 l of digested 16S amplicon, 9 l of formamide and 0.5 l of LIZ 1200
being denatured at 95 C for 3 min and then placed immediately on
ice, a 5 s injection time and separation for 80 min at 15 kV. A run voltage of 15 kV gave more accurate fragment size and shorter run time
than 12 kV.
After optimization, T-RFLP reproducibility was investigated by
repeated analysis of the same samples. Between analyses performed
on the same day, T-RFs differed by 0.2 nucleotides, this increased to
1 nucleotides when the same sample was run on different days. We
decided to analyze all fecal samples in two independent PCRs and subsequent T-RFLP analyses and to include only T-RFs appearing in both
analyses. T-RFs differing no more than 1 nucleotides were considered
as a single taxon; their peak sizes and areas from the duplicates were
averaged.
3.1.2. Identication and exclusion of pseudo T-RFs
A troubling observation was that we repeatedly observed quite large
T-RF peaks representing nucleotide length 210(1) and 540(1)
(found 24 and 29 times, respectively, in the 36 fecal samples analyzed).
These fragments were neither generated when analyzing pure-culture
isolates, nor cloned DNA from fecal samples, suggesting that they
were pseudo-T-RFs. We found that these two peaks appeared only in
samples in which the Bacteroides-specic T-RF (size ~90) represented
>25% of the total peak area, i.e. samples in which Bacteroides spp.
were quantitatively dominant. In silico digestion analysis of the 16S
rRNA gene of Bacteroides type strains revealed three MspI enzyme
restriction sites (CCGG) situated at around 90, 210 and 540 nucleotides
in B. vulgatus and B. fragilis. B. ovatus and B. thetaiotaomicron only have
two sites (90 and 540) in. Complete digestion by MspI should generate
a single terminally labeled 90 nt fragment. If digestion at this site would
not be complete, digestion at one of the other MspI sites could generate
fragments of 210 or 540 nt, respectively. In samples with large quantities of Bacteroides DNA, such incomplete digests could be numerous
enough to pass the threshold and generate visible T-RFs. This hypothesis
was supported by a lack of the 210 nt fragment in the three fecal samples
that, according to culture, contained B. ovatus or B. thetaiotaomicron, but
no other Bacteroides species. Another source of variation that may explain
the occurrence of more than one peak per genome is that the 16S rrn operons may appear in several copies in the bacterial genome and that these
copies may differ slightly in sequence. If such nucleotide polymorphisms
affect restriction enzyme sites, T-RFs of different length may appear
(Crosby and Criddle, 2003). As the T-RFs of 210 and 540 nt were deemed
to represent pseudo-T-RFs, fragments of these sizes were excluded from
the analyses.
3.1.3. Discrepancies between true and observed T-RF size
We compared observed T-RF size from T-RFLP analysis of type strain
bacteria with the theoretical T-RF size obtained by in silico digestion
of the sequence of the same strain. Signicant discrepancies were
observed, i.e. the lengths of long fragments were underestimated as migration time increased non-linearly with size (Supplement Table S1).
Analysis of MW size standard markers conrmed that an increment in
size gave an increment in migration time that gradually diminished
with increasing MW (Supplement Fig. S1).

A further complication in translating migration time to size was

that the PCR digests were labeled with the uorophore FAM that has a
MW of 376 while the standard was labeled with LIZ whose MW is not
revealed by the manufacturer. However, since the true T-RF sizes
were underestimated already at short T-RFs, the LIZ uorophore should
have much higher MW than FAM. The LIZ MW standard was chosen
based on its large number of markers and their even distribution across
a broad range.

3.1.4. Construction of a T-RF database

Since observed migration time of T-RFs could not be translated
into a particular bacterial taxon using in silico digestion, we decided to
develop our own database for T-RFs identication. T-RFs of 96 bacterial
species could be identied through T-RFLP analysis of pure-culture
bacterial isolates from infant's stools and culture collection type
strains. Cloning and sequencing of bacterial DNA from nine infant
fecal samples added another 34 species, yielding a database consisting
of the observed T-RF sizes of 130 bacterial species. All 16S rRNA sequences of isolates and clones in the database were submitted to the
European Nucleotide Archive and the assigned accession numbers were
HE974918HE974988.

3.1.5. Taxonomic discrimination capacity of T-RFLP

A concern in microbiological methodology is the level of taxonomic
discrimination that can be achieved. Culture enables typing to the species level and below, while16S rRNA gene based technique could not,
which is inadequate for many purposes. Table 1 shows, for some genera,
the number of unique species, or unique T-RFs, that were identied
using the database. For example, our database contained ten different
Bacteroides species found in the human intestinal microbiota. These
ten species collectively generated seven unique T-RFs, suggesting that
T-RFLP was able to taxonomically differentiate groups of related or
even individual Bacteroides species. As seen in Table 1, most genera
were represented by several T-RFs, indicating that the level of taxonomic discrimination for T-RFLP in our hands was at, or below, the genus
level. A notable exception was Enterobacteriaceae, which yielded a single T-RF in almost all samples, and a second T-RF appeared in some
samples from which Enterobacter was cultured. Staphylococci also
generated a single T-RF, making it impossible to distinguish between
S. aureus and coagulase-negative staphylococci (Table 1). This can be
explained by the close taxonomic proximity of many species within
clinically relevant genera, which have been separated based on clinical
importance and not taxonomic relatedness. Interestingly, T-RFLP was
able to differentiate several groups within the streptococcus genus, a

Table 1
Level of taxonomic resolution of culture and T-RFLP.
No of unique taxa
Genus
Facultative bacteria
Enterobacteriaceae
Enterococcus
Staphylococcus
Streptococcus
Obligate anaerobes
Bacteroides
Bidobacterium
Clostridium
Lactobacillus

Database
(species)
7
3
2a
9
10
9
13
9

T-RFLP
(T-RFs)
2
2
1
6
7
3
11
8

Based on enzyme digestion using MspI.

a
Staphylococci were only identied as either S. aureus, or coagulase-negative staphylococci, based on the coagulase test.

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

task that is very difcult even by sequencing the entire 16S rRNA gene
(Brugger et al., 2012).

3.2. Analysis of infant intestinal microbiota development by T-RFLP and

culture
3.2.1. Analysis of the microbiota of infant F by T-RFLP
The microbiota composition was examined by T-RFLP analysis of
fecal DNA from six infants (AF) sampled on six occasions (1, 2 and
4 weeks of age, and 2, 6 and 12 months of age). Fig. 1 shows the
T-RFLP pattern of the stool microbiota of infant F. Each peak (T-RF)
represents a bacterial taxon (genus or species). As evident from the
gure, certain T-RFs (e.g. 30, 31, and 149) were present in the early
samples but then disappeared. Using the database, these were identied
as Lactobacillus mucosae (30 nt), L. vaginalis (31) and Staphylococcus
(149). As pointed out above, S. aureus and coagulase-negative staphylococci generate the same T-RF and cannot be distinguished by T-RFLP
(Table 1). Other T-RFs, e. g. 129 and 130, representing Bidobacterium
spp., were present in infant F's gut microbiota throughout the
12 months. A large 486 nt T-RF was present from 1 week to 6 months;
it was identied as Enterobacteriaceae. This peak was no longer detectable by T-RFLP at 12 months of age. Different Enterobacteriaceae species
could not be resolved by T-RFLP (Table 1).
Culture revealed that infant F was colonized by E. coli from 1 week to
12 months of age (data not shown). Over this period the population
size decreased from 109.4 to 10 7.9 CFU/g feces, which reects the
increased competition from anaerobes. Thus the disappearance of
the Enterobacteriaceae peak at 12 months of age was due to its relatively
lower abundance in this sample. Similarly, the Staphylococcus peak
(149) also disappeared after 2 months of age, although infant F was
colonized by Staphylococcus during the entire period, but in decreasing
population levels.
Interestingly, several T-RFs were present in the one and two-week
samples, absent at one month of age, and then reappeared in the
2, 6 and 12-month samples. These included T-RFs identied as
Parabacteroides (86), Bidobacterium (126), Coprobacillus catenaformis
(532) and several unidentied T-RFs (53, 69 and 422).
As expected, the 6 and 12-month samples were dominated
by obligately anaerobic bacteria such as; Bacteroides spp. (90, 93),
Clostridium indolis (213), C. ramosum (287), Akkermansia muciniphila
(265), Lactobacillus casei/paracasei (562), L. fermentum (571) and
unidentied species represented by T-RFs 136, 282, 516 and 535.

Fluorescence intensity (0-500 unit)

30,31

41

3.2.2. Multivariate analysis of microbiota colonization pattern analyzed

by T-RFLP
An overall picture of the microbiota composition as analyzed by
T-RFLP was obtained by principal component analysis (Fig. 2). Here,
each observation (sample) is positioned based on the T-RFs found in
that sample. In general, samples obtained between 1 week and 6 months
of age (samples 15), clustered by infant, suggesting that each infant had
a partly unique microbiota. In contrast, all 12-month samples (A6F6)
and the 6-month samples from two children (A5 and F5) appeared in vicinity of one another, but far from the early samples. Thus, the late samples were more similar to one another than to the early samples from the
respective infants.
To identify T-RFs that produced the difference between early
and late samples, we employed O2PLS-DA, a regression variety
of principal component analysis. The late samples (A6, B6, C6, E6,
F6, A5, and F5) were compared against the rest of the samples
(outlier D6, whose microbiota was clearly different from all other samples, was excluded). O2PLS-DA showed that early samples were distinguished from late ones by the presence of the following T-RFs: 486
(Enterobacteriaceae), 161 (Propionbacterium acnes), 148 (Staphylococcus),
181 (Lactobacillus gasseri), 130 (Bidobacterium infantis/breve), 547
(Streptococcus sanguinis/salivarius), 129 (Bidobacterium spp) and
30 (Lactobacillus mucosae), as well as unidentied T-RFs 158 and
702. The late samples were distinguished from the early ones by
the presence of T-RFs such as 218 (Eubacterium rectale/eligens),
215 (Ruminococcus gnavus), 77 (Bacteroides spp), 188 (Clostridium
bartlettii/lituseburense/bifermentans), 514 (Clostridium perfringens),
216 (Ruminococcus obeum), all representing strict anaerobes, as
well as the unidentied T-RFs 136, 201, 215, 277 and 516.
The 12-month sample from child D (D6) did not group with any of
the other samples (Fig. 2). O2PLS-DA analyzing D6 in relation to the
other 35 samples revealed that D6 contained several T-RFs that were
unique to this sample: i.e. 220, 273, 279, 447 and 491. In addition, this
sample contained T-RFs that were only found in two samples deriving
from other infants i.e. 176 (L. acidophilus) and the unidentied T-RFs
206, 297, 311 and 529.

3.3. Comparison between methodologies

3.3.1. Detection limits of T-RFLP and culture
As all samples had been analyzed by quantitative culture on selective and non-selective media, we could investigate the ability that a

129,130 149

T-RFno.(id)a

486

1w

25 (14)

2w

30 (19)

1mo

19 (12)

53 86
69 126

422

2mo

532

6mo

21 (11)
33 (19)

12mo

38 (21)
90,93

136

213

265, 282, 287

516, 535, 562,571

Nucleotide length (0-600)

Fig. 1. T-RFLP analysis of fecal microbiota composition in an infant followed during the rst year of life. Bacterial DNA was extracted from fecal samples obtained at 1and 2 weeks of
age and at 1, 2, 6 and 12 months of age and analyzed by T-RFLP. Each peak represents a terminal restriction fragment of a specic length that corresponds to a bacterial phylotype,
usually a genus. A database was constructed from T-RFLP analysis of known bacterial species or cloned and sequenced bacterial 16 s rRNA genes, and this database was used to
identify individual T-RFs in fecal samples. T-RF no. denotes the number of T-RFs identied in the sample. a the numbers of T-RF that could be identied using the database. The
infant depicted corresponds to infant F in Figs. 2 and 3.

42

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

F6/E6
R2X[1] = 0.120029

R2X[2] = 0.0880433

Ellipse:Hotelling T2(0.95)

Fig. 2. Microbiota development over the rst year in six infants. Fecal samples from six infants (AF) obtained on six occasions over the rst year of life were analyzed by T-RFLP and
the T-RF pattern was analyzed by principal component analysis. (1 = 1 week, 2 = 2 weeks, 3 = 1 month, 4 = 2 months, 5 = 6 months and 6 = 12 months samples). The position of the sample is generated by the totality of T-RFs presented in the sample.

particular bacterial group was detected by culture or T-RFLP, respectively. As shown in Fig. 3, certain facultative bacteria were more
often detected by culture than that by T-RFLP. This was true of staphylococci and enterococci (Fig. 3). Enterobacteriaceae, were, however,
almost as often detected by T-RFLP as by culture. Aerobic streptococci
were readily detected by T-RFLP, but seldom by culture, which could
be explained by the fact that we did not include any selective culture

Enterococcus
Staphylococcus
Streptococcus
Enterobacteriaceae

medium for streptococci; these bacteria were only detected on the

non-selective media when present in very high counts.
T-RFLP identication of several obligate anaerobic bacteria that
were very often missed by culture, including Veillonella, Coprobacillus,
and Ruminoccous, as well as a number of T-RFs which could not be
resolved.
Fig. 4 shows the fecal population levels, determined by culture, in
samples that were positive or negative for a bacterial group by
T-RFLP. Regarding enterococci, only a sample with very high population
counts (10 9 CFU/g) was positive, while staphylococci could be detected
in samples were they were present at 106.9 1.2 CFU/g. Bacteroides spp.
were detected by T-RFLP in all samples which were positive by culture.
This might be explained by the high population levels of this species
(>108 CFU/g in all samples, Fig. 4). However, with the exception of
enterococci, T-RFLP was quite sensitivity; 81% of the bacteria that

Bacteroides
Bifidobacterium
12

C. difficile
Collinsella
Coprobacillus
Lactobacillus
Prevotella
Propionibacterium

T-RFLP
Culture

Ruminococcus

Log CFU/g faeces (culture)

Clostridium

10
8
6
4

Detected
by T-RFLP

= No
= Yes

Sutterella

Veillonella

ro
nte

s
m
ile ides
us
ae
cu
riu
fic
ce
cill
oc
ro
dif
cte
ria
ba
a
e
o
loc
cte
t
m
t
b
y
c
c
u
Ba
ph
idi
ido
ba
La
f
r
a
i
o
t
t
r
s
B
S
te
Clo
En

12

24

36

us

cc

co

Number of samples
Fig. 3. Frequency of samples yielding a bacterial taxon by culture or T-RFLP. Altogether
36 fecal samples (6 samples each from 6 infants) were cultured quantitatively and
analyzed by T-RFLP. The bars indicate the number of samples in which a certain taxon
was detected by T-RFLP (lled bars) or culture (hatched bars), respectively. Facultative
bacterial taxa are indicated in bold.

Fig. 4. Fecal population levels of culturable bacteria detected and not detected by T-RFLP.
Population levels of different bacterial genera in 36 fecal samples were determined by
quantitative culture. The population levels of a particular genus were compared between
samples in which this genus was also detected by T-RFLP (lled symbols) and samples
which were negative for the genus in question by T-RFLP (open symbols). The bars indicate medians.

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

were present at >10 6 CFU/g in fecal samples (as judged by culture),

were detected by T-RFLP.

43

curve (Fig. 5A). On average, the number of T-RFs detected by T-RFLP

were 22 (1 week), 21 (2 weeks), 17 (1 month), 19 (2 months), 26
(6 months) and 36 (1 year). Thus, the number of T-RFs tended to
decline between 1 week and 1 month of age (p = 0.09), while a clear
increase in the number of peaks was seen between 1 and 12 months
of age (p = 0.03). Roughly half of the peaks that were present in
early samples (1 and 2 weeks) had disappeared at around 1 month
and were not seen again in the same infant, while the other half
appeared again one or several times in the following samples. The identity of the T-RFs that disappeared after the early weeks differed between
infants, and no general pattern could be discerned.
In total, we identied 845 T-RFs in the 36 samples examined by
T-RFLP, representing 146 unique T-RFs (bacterial phylotypes). Using
the T-RF database (Supplement data) comprising T-RFs from 130
different species, we were able to identify 553/845 (65%) of the
total T-RFs or 72/146 (49%) of the unique T-RFs. On average, culture
identied around half as many unique bacterial taxa as did T-RFLP
(average 41%, range: 2563% in individual infants, Fig. 5B). There
was no trend toward increased complexity in the late samples using
culture (Fig. 5B), indicating that the anaerobic species acquired after
6 months of age were, to a large extent, missed. Interestingly, the
same was true of the very early samples, which did not appear more
complex than the 12-month sample using culture. Thus, a number of
bacteria present in the early samples were also missed by culture.

3.3.2. Most prevalent bacterial taxa according to culture and T-RFLP

analysis
We compared which bacterial genera/species were most prevalent
in the examined infants, depending on whether the samples were analyzed by culture or by T-RFLP. Table 2 shows the genera/species identied in at least four infants out of six on different occasions. Both T-RFLP
and culture identied Bidobacterium, Clostridium and Bacteroides spp
as common colonizers of the infantile microbiota throughout the rst
year of life. Using culture, E. coli and staphylococci were found in all
infants and enterococci in 5/6. Using T-RFLP, Enterobacteriaceae was
identied in all infants up to 6 months of age while staphylococci
were only prevalent in the 1 week sample and enterococci were not
identied as a major colonizer. In contrast, Streptococcus spp, Veillonella
and several anaerobes, which were among the most prevalent colonizers according to T-RFLP, were not found to be equally prevalent,
using culture for detection. Interestingly, the early samples (1
2 weeks) yielded several unidentied T-RFLP fragments that clearly
differed from the unidentied T-RFs that were prevalent in the late
samples (6 & 12 months). Unidentied T-RFs were least common
in the 1 and 2 months samples.
3.3.3. Microbiota complexity as analyzed by T-RFLP and culture
T-RFLP is ideally suited to determine microbiota complexity; as
each taxon is represented by a peak; the number of peaks can be calculated and used as a measure of complexity. As seen in the analysis
of infant F (Figs. 1, 5A), there were more T-RFs in the 1 and 2-week
samples (25 and 30, respectively) than in the 1 and 2-month samples
(19 and 21, respectively).
We examined microbiota complexity in all six infants as a function
of time, i.e. the number of taxa (species or genera) identied at each
sampling occasion. Fig. 5A shows the total number of T-RFs distinguished by T-RFLP in each fecal sample collected from the six studied
infants (AF). Interestingly, with the exception of infant B, a higher
number of bacterial taxa were detected in the 12 week samples
than in the 12 month samples, yielding a V-shaped complexity

3.3.4. Analysis by cloning and sequencing of the 16S rRNA gene

In order to expand our T-RF database with non-cultivable bacteria,
we cloned and sequenced PCR amplied 16S rRNA genes from nine
fecal samples that contained many T-RFs that could not be identied
from analysis of cultured isolates. A total of 927 clones were obtained
and after agarose gel screening of the products, 834 clones were
successfully analyzed using T-RFLP (Table 3). Clones with the same
partial digestion T-RFLP prole were grouped together and at least
two representatives of each clone were selected for sequencing. The
average length of the sequences was 500 nt and for sequences showing
b98% similarity with previously published sequences, near full length
sequencing of the 16S rRNA gene was performed. Taxonomic classication revealed that 64% of the clones belonged to the Bacteroidetes

Table 2
Most prevalent bacterial groups in infantile fecal microbiota as determined by culture and T-RFLP.
Number of infants (n) yielding a specied bacterial group in feces
1 week
Culture
E. coli
Bidobacterium
Co-neg Staph
Bacteroides
Enterococcus
Clostridium
S. aureus
T-RFLP
Enterobacteriaceae
Bacteroides
Bidobacterium
Clostridium
Staphylococcus
Veillonella
Streptococcus
595#
702#

2 weeks

1 month

2 months

6 months

12 months

6
6
6
5
4
4
4

E. coli
Co-neg Staph
Bidobacterium
S. aureus
Bacteroides
Enterococcus
Clostridium

6
6
5
5
4
4
4

E. coli
Bidobacterium
S. aureus
Lactobacillus
Co-neg Staph
Bacteroides
Enterococcus

6
6
6
6
5
4
4

E. coli
Bidobacterium
Co-neg Staph

E. coli
Bidobacterium
Co-neg Staph

S. aureus
Bacteroides
Enterococcus

6
6
6
5
4
4

Bacteroides
Enterococcus
Clostridium
S. aureus

6
6
6
5
5
4
4

E. coli
Co-neg Staph
Bidobacterium
Bacteroides
Enterococcus
Clostridium

6
6
5
5
5
5

6
5b
5
5
4
4
4
4
4

Enterobacteriaceae
Bacteroides
Streptococcus
Bidobacterium
Barnesiella
Clostridium
595#
876#

6
5b
5
5
4
4
4
4

Bidobacterium
Enterobacteriaceae
Bacteroides
Veillonella
Lactobacillus
Streptococcus

5
5
4b
4
4
4

Enterobacteriaceae
Bidobacterium
Bacteroides
Bacteroides
Propionbacterium
Clostridium
Streptococcus

5
5
5b
4a
4
4
4

Enterobacteriaceae
Bacteroides
Bacteroides
Veillonella
Bidobacterium
Clostridium
Coprobacillus
Lactobacillus
495#
876#

6
6b
5a
5
5
5
4
4
4
4

Veillonella
Clostridium
Bacteroides
277#
Bidobacterium
Ruminococcus
Lactobacillus
Coprobacillus
Enterobacter
Selenomonas
136#
495#
516#

6
6
6ab
6
5
5
5
5
4
4
4
4
4

Fecal samples from six infants at six time points were analyzed by culture and T-RFLP. Bacterial taxa identied in at least 4/6 infants by culture or T-RFLP, respectively, are indicated
and n refers to the number of infants in whom the respective taxon was found. Italic = anaerobic bacteria, *Coagulase-negative Staphylococcus. #Unidentied T-RFs are presented
by their respective fragment sizes. a b = different Bacteroides with T-RF size around 80 and 90, respectively.

44

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

50

T-RFLP

Culture

40

No. of species

40

No.of T-RFs

50

30
20
10

30
20
10

A
143
69

B
91
73

C
136
63

D
144
63

E
169
76

F
162 Total
59 Id.(%)

A
68

B
58

D
68

C
41

E
55

F Infant
40 Total

Fig. 5. Microbiota complexity as a function of age in six infants. Fecal samples from infants AF were collected on six occasions over the rst year of life (1and 2 weeks of age and at
1, 2, 6 and 12 months of age), analyzed by T-RFLP and identied (when possible) by using a database. (A) The number of T-RFs detected in each sample is depicted by a black bar,
hence each infant is represented by six bars, one for each culture occasion. Below, the total number of T-RFs identied in that infant is indicated, as well as the proportion of T-RFs
that could be identied by using the database (Id %). Most infants showed a high complexity in the early samples, followed by a decrease in complexity around 12 months of age
(3rd4th bar), and again increased complexity in the last samples (6 and 12 months of age). (B) Analysis of the same samples by quantitative aerobic and anaerobic culture
revealed fewer bacterial taxa and did not reveal the biphasic complexity pattern seen with T-RFLP.

phylum, mainly represented by the genera Bacteroides, Parabacteroides,

Barnesiella and Prevotella. The second most prevalent phylum was the
Firmicutes (25%) which was represented by three classes, among which
class Clostridia was most prevalent with genera such as Ruminococcus,
Clostridium, Eubacterium, Faecalibacterium and Lachnospira. Negativicutes
were represented by Veillonella and Dialister and the Bacilli class by
Streptococcus and Lactobacillus. Proteobacteria (mainly Enterobacteriaceae)
were readily detected, while the genus Bidobacterium was only detected
in a single sample (Table 3).

bacterial isolates of known species to construct a database to enable

identication of the T-RFs in a sample. Cloning and sequencing, followed
by T-RFLP, provided additional data regarding the identity of T-RFs
representing non-cultivable bacteria. Using this approach, knowledge
on the T-RFs corresponding to more than a hundred bacterial species
common in the infantile gut microbiota were obtained. This database
allowed us to directly identify approximately 60% of the peaks found
in fecal samples from six infants followed from birth to 12 months of
age. Thus, despite extensive efforts, a substantial proportion of the infantile gut microbiota still consisted of unidentied bacteria, which is a
clear limitation of the T-RFLP method when employed on gut microbiota
samples.
T-RFLP in most cases permitted identication at the genus level and,
in many cases; several distinct T-RFs represented a certain genus. In
accordance, Nieminen et al. (2011) concluded that the taxonomic resolution level of T-RFLP method was in between genus and species. Low
taxonomic resolution is a signicant problem when massive parallel
sequencing, Next Generation sequencing, is used, as this is based on
analysis of only parts of the 16S rRNA gene. For example, sequencing
based on 400-bases of the 16S rRNA gene only gives a chance of 89%

4. Discussion
In the present study, we optimized T-RFLP for analysis of the
microbiota of infants and examined its performance using fecal samples
from six infants who were followed from 1 week to 1 year of age.
In theory, the T-RF size may be calculated from knowledge of the
specic enzyme cleavage sites (in silico digestion). In practice, this
was impossible as migration time was not lineally related to T-RF size
across the size span, in accordance with earlier reports (Hahn et al.,
2001; Pandey et al., 2007). Therefore, we used T-RFLP analysis of

Table 3
Dominant bacterial groups as detected by cloning and sequencing.
Phylum

Dominant generaa

No. of clones
Infant C

Class
Bacteroidetes
Bacteroidia
Firmicutes
Clostridia
Negativicutes
Bacilli
Proteobacteria
Gamma
Alpha
Beta
Actinobacteria
Actinobacteria
Werrucomicrobia
Werrucomicrobiae
Total no. of clones

Infant D

Infant E

Infant F

6 months 12 months 2 weeks 12 months 2 weeks 6 months 12 months 6 months 12 months

Bacteroides, Parabacteroides

53

43

Rumin, Clost, Euba & Faecaib

Veillonella
Streptococcus

61

30

101

17
2

14
14
6

9
6

Enterobacteriaceae
Koporiimonas
Sutterella

Bidobacterium

34

60

123

30

535

(64)

1
20

13
10
2

59
9

13
5
1

117
74
15

(14)
(9)
(2)

8
5

53

81
5
2

(10)
(0.006)
(0.002)

(0.002)

(0.004)
834

Akkermansia
58

69

61

69

124

Total % of
clones

57

3
101

244

51

Cloning was performed on nine fecal samples from four infant (C, D, E and F). Taxonomic classications of clones were achieved after sequencing analysis a Genera representing 15% of
the sequences b Ruminococcus, Clostridium, Eubacterium & Faecalibacterium.

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

to perform identication at the genus level, for 50-bases which was

commonly used in many early pyrosequencing studies, this drops to
52% (Wang et al., 2007). A signicant problem was, however, posed
by Enterobacteriaceae, which was present in all samples and readily
detected, but which could not be resolved further by T-RFLP, except
for a distinct T-RF produced by some species within the Enterobacter
genus. Also this is in accordance with previous reports (Nieminen
et al., 2011). The 16S rRNA gene heterogeneity varies among bacterial
taxa and many species in the Enterobacteriaceae family are 96% identical
in the 16S rRNA gene, which makes it practically impossible to differentiate, e.g., E. coli, Klebsiella, and Proteus, by 16 S rRNA-based methods. In
contrast different Bacteroides spp. share only 47%96% sequence identity, explaining why Bacteroides could be typed almost to the species
level; ten different Bacteroides species yielded seven unique T-RFs. It is
possible that taxonomic resolution could be increased by using more
than one enzyme for cleavage of the amplied 16S rRNA gene. We initially tested the restriction enzymes MspI and AluI but subsequently
used only MspI as database construction would be very labor intensive
using two enzymes. MspI has been proven to have the highest resolving
delity among 18 enzymes (Engebretson and Moyer, 2003), and yields
the highest number of unique T-RF patterns when analyzing human
colonic ora (Matsumoto et al., 2005; Wang et al., 2004). Our PCA analysis of T-RFLP data showed that MspI could separate and cluster the
different fecal samples in accordance with other published studies
(Palmer et al., 2007; Yatsunenko et al., 2012).
Comparison between T-RFLP and anaerobic and aerobic culture
using selective and non-selective media, revealed that T-RFLP identied
2.5 times more different taxa than did culture-based analysis. This was
primarily due to the identication of several obligate anaerobic bacteria
by T-RFLP that were very often missed by culture, including Veillonella,
Coprobacillus, and Ruminoccous, as well as a number of T-RFs which
could not be resolved. Ruminococcus is very sensitive to oxygen exposure and thereby difcult to culture unless all handling of samples are
performed under strict anaerobic conditions. Bacteroides was more
frequently detected by T-RFLP than by culture, which is in agreement
with previous observations (Arumugam et al., 2011). The reverse was
true for bidobacteria, while lactobacilli and Clostridium difcile were
equally readily detected by both methods. Thus, several of the anaerobic
bacteria colonizing the newborn infant may be detected by culture,
provided that selective media are available and the species is not
exquisitively oxygen sensitive.
Cloning and sequencing mainly indentied Bacteroides, some clostridia, Veillonella and Enterobacteriaceae in the analyzed fecal samples.
Since only approximately 100 clones were examined from each sample,
only bacteria present in population levels well above 109 CFU/g feces
could be expected to be detected by this method. However, this does
not explain the inability to identify Bidobacterium spp, which were,
according to culture, present in equal or higher population levels as
Bacteroides. For 16S rRNA gene based analyses, several factors can inuence the detection limit: 1) The degree of t of the universal PCR
primers. Bidobacteria belong to the Actinobacteria phylum characterized by high GC content and several universal primers may not be
ideally suited to detect these bacteria (Farris and Olson, 2007). 2) The
number of 16S operon copies which varies from 1 to 15 between bacterial species (Acinas et al., 2004). For example, Bidobacterium has 3.6
copies on average, as compared to 6 in Bacteroides (according to the
ribosomal rna operon copy number database http://rrndb.mmg.msu.
edu/search.php). 3) The efcacy by which bacterial DNA can be
extracted from the sample. DNA might more easily be released from
Gram-negative bacteria such as Bacteroides, and Enterobacteriaceae,
while some Gram-positive bacteria have a very thick and sturdy cell
wall. We believe that this may be the major reason for the very poor
detection of enterococci by T-RFLP, despite the fact that most samples
had >106 CFU/g of enterococci. We used the Qiagen extraction kit identied by McOrist et al. (2002) as the most effective among four commercial fecal DNA extraction kits and complemented with glass bead

45

beating of the sample to further enhance the possibility of breaking

the cell wall. This might still be inadequate for extraction of DNA from
Gram-positive bacteria.
In contrast to cloning and sequencing, T-RFLP was, in most cases,
able to detect bacteria that were present in a fecal sample above a
population of 10 6 CFU/g. This makes the method better suited than
sequencing for analyzing complex microbiota with great variations
in population densities between different taxa theoretically at
least 10,000 sequences must be analyzed to obtain one read of a
bacterium present at a level of 10 6 CFU/g within a total population
of 10 11 CFU/g. The advantage of T-RFLP is that the fragments are
separated by size before being detected. Thus, the presence of 100
to 1000-fold excess of Bacteroides DNA does not necessarily hamper
the detection of a subdominant species, provided that it has a T-RF
which clearly differs in size from these dominant bacteria.
Culture-based analysis showed superior sensitivity for facultative
bacteria, mainly because they were often present in comparatively
low population levels. Although obligate anaerobes are quantitatively
dominant in the fecal microbiota, especially in adults, facultative
bacteria may have special relevance in terms of immune stimulation
and shaping of the infant's immune system, for example by being
able to translocate, i.e. pass intact and viable across the mucosa, a
property not shared by most strict anaerobes (Berg, 1999; Steffen
et al., 1988).
Analysis of the T-RFLP patterns of infantile fecal samples yielded
an interesting picture that, to the best of our knowledge, has not
been reported before. It appeared that in most infants, a quite high
bacterial diversity characterized the one- and two-week fecal samples,
but that microbiota complexity decreased thereafter to reach a nadir
at around 1 month of age. Thereafter, complexity increased markedly
again, being highest in the last sample collected at 12 months of age.
It is well known that microbiota complexity increases over the rst
year of life as more and more anaerobic bacterial species settle in the
gut (Adlerberth and Wold, 2009). However, the high bacterial complexity of the samples obtained at 1 and 2 weeks of age was an unexpected
nding. Several bacterial species that were found in early fecal samples
disappeared from the microbiota after the rst weeks and did not
reappear. Some of these may derive from the birth canal, as all infants
examined here were vaginally delivered, or from the environment.
Such non-professional gut colonizers may have the chance to expand
in the rst weeks but will not be able to withstand the competition as
more professional gut bacteria are acquired. Furthermore, they may
be more sensitive to antibacterial peptides produced in the mucosa in
response to colonization. Previous reports have shown that bacteria
originating in the mother's vaginal ora can be cultured from the baby's
gastric contents immediately after birth (Dominguez-Bello et al., 2010),
but that many of these bacteria cannot colonize the infant's gut.
Using principal component analysis, we found that the early
samples had a relatively unique microbiota in each infant, but that
the microbiota of older infants was more similar to one another.
This convergence agrees with previous results (Kurokawa et al.,
2007). Towards the end of the rst year of life, a more complex
microbiota develops, probably as a result of exposure to a larger variety
of bacteria after weaning, and perhaps also due to changes in the gut
milieu after termination of breast-feeding. Breast milk creates a milieu
in the infantile gut that suppresses growth of many bacterial species
(Adlerberth and Wold, 2009). Our results suggest that T-RFLP is fairly
sensitive and suitable for rapid characterization of the infant gut
microbiota. However, facultative bacteria in moderate population levels
will only be detected using culture on selective media. Furthermore,
several T-RFs remained unidentied, there is no simple method to
know from which bacteria they derive, especially if they are not present
in very high populations and thereby may escape detection by cloning and sequencing or next generation sequencing. Thus, a full characterization of the infantile gut microbiota requires a combination of
methods.

46

F. Sjberg et al. / Journal of Microbiological Methods 94 (2013) 3746

Acknowledgments
This study is supported by the Swedish Medical Research Council
(grants K2009-57X-14455-08-3 and K2002-745X-14455-01A), the
European Commission (IMMUNOBIOTA, QLRT-2000-00538), the Medical faculty of Gothenburg University (LUA-ALFGBG-11006), the Torsten
and Ragnar Sderbergs Foundation, the Cancer and Allergy Foundation
and the Wilhelm and Martina Lundgrens Foundation.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://
dx.doi.org/10.1016/j.mimet.2013.04.002.
References
Abdo, Z., Schuette, U.M., Bent, S.J., Williams, C.J., Forney, L.J., Joyce, P., 2006. Statistical
methods for characterizing diversity of microbial communities by analysis of terminal
restriction fragment length polymorphisms of 16S rRNA genes. Environ. Microbiol. 8,
929938.
Abrahamsson, T.R., Jakobsson, H.E., Andersson, A.F., Bjorksten, B., Engstrand, L.,
Jenmalm, M.C., 2012. Low diversity of the gut microbiota in infants with atopic
eczema. J. Allergy Clin. Immunol. 129 (434440), e432.
Acinas, S.G., Marcelino, L.A., Klepac-Ceraj, V., Polz, M.F., 2004. Divergence and redundancy of 16S rRNA sequences in genomes with multiple rrn operons. J. Bacteriol.
186, 26292635.
Adlerberth, I., Wold, A.E., 2009. Establishment of the gut microbiota in Western infants.
Acta Paediatr. 98, 229238.
Adlerberth, I., Lindberg, E., Aberg, N., Hesselmar, B., Saalman, R., Strannegard, I.L., Wold,
A.E., 2006. Reduced enterobacterial and increased staphylococcal colonization of
the infantile bowel: an effect of hygienic lifestyle? Pediatr. Res. 59, 96101.
Adlerberth, I., Strachan, D.P., Matricardi, P.M., Ahrne, S., Orfei, L., Aberg, N., Perkin, M.R.,
Tripodi, S., Hesselmar, B., Saalman, R., Coates, A.R., Bonanno, C.L., Panetta, V., Wold,
A.E., 2007. Gut microbiota and development of atopic eczema in 3 European birth
cohorts. J. Allergy Clin. Immunol. 120, 343350.
Arumugam, M., Raes, J., Pelletier, E., Le Paslier, D., Yamada, T., Mende, D.R., Fernandes,
G.R., Tap, J., Bruls, T., Batto, J.M., Bertalan, M., Borruel, N., Casellas, F., Fernandez, L.,
Gautier, L., Hansen, T., Hattori, M., Hayashi, T., Kleerebezem, M., Kurokawa, K.,
Leclerc, M., Levenez, F., Manichanh, C., Nielsen, H.B., Nielsen, T., Pons, N., Poulain,
J., Qin, J., Sicheritz-Ponten, T., Tims, S., Torrents, D., Ugarte, E., Zoetendal, E.G.,
Wang, J., Guarner, F., Pedersen, O., de Vos, W.M., Brunak, S., Dore, J., Antolin, M.,
Artiguenave, F., Blottiere, H.M., Almeida, M., Brechot, C., Cara, C., Chervaux, C.,
Cultrone, A., Delorme, C., Denariaz, G., Dervyn, R., Foerstner, K.U., Friss, C., van de
Guchte, M., Guedon, E., Haimet, F., Huber, W., van Hylckama-Vlieg, J., Jamet, A.,
Juste, C., Kaci, G., Knol, J., Lakhdari, O., Layec, S., Le Roux, K., Maguin, E., Merieux, A.,
Melo Minardi, R., M'Rini, C., Muller, J., Oozeer, R., Parkhill, J., Renault, P., Rescigno,
M., Sanchez, N., Sunagawa, S., Torrejon, A., Turner, K., Vandemeulebrouck, G., Varela,
E., Winogradsky, Y., Zeller, G., Weissenbach, J., Ehrlich, S.D., Bork, P., 2011. Enterotypes
of the human gut microbiome. Nature 473, 174180.
Berg, R.D., 1999. Bacterial translocation from the gastrointestinal tract. Adv. Exp. Med.
Biol. 473, 1130.
Bisgaard, H., Li, N., Bonnelykke, K., Chawes, B.L., Skov, T., Paludan-Muller, G., Stokholm,
J., Smith, B., Krogfelt, K.A., 2011. Reduced diversity of the intestinal microbiota during
infancy is associated with increased risk of allergic disease at school age. J. Allergy Clin.
Immunol. 128 (646652), e641e645.
Brugger, S.D., Frei, L., Frey, P.M., Aebi, S., Muhlemann, K., Hilty, M., 2012. 16S rRNA
terminal restriction fragment length polymorphism for the characterization of
the nasopharyngeal microbiota. PLoS One 7, e52241.
Crabbe, P.A., Bazin, H., Eyssen, H., Heremans, J.F., 1968. The normal microbial ora as a
major stimulus for proliferation of plasma cells synthesizing IgA in the gut. The
germ-free intestinal tract. Int. Arch. Allergy Appl. Immunol. 34, 362375.
Crosby, L.D., Criddle, C.S., 2003. Understanding bias in microbial community analysis
techniques due to rrn operon copy number heterogeneity. Biotechniques 34, 790794
(796, 798 passim).
Dominguez-Bello, M.G., Costello, E.K., Contreras, M., Magris, M., Hidalgo, G., Fierer, N.,
Knight, R., 2010. Delivery mode shapes the acquisition and structure of the initial
microbiota across multiple body habitats in newborns. Proc. Natl. Acad. Sci. U. S. A.
107, 1197111975.

Ellis-Pegler, R.B., Crabtree, C., Lambert, H.P., 1975. The faecal ora of children in the
United Kingdom. J. Hyg. (Lond.) 75, 135142.
Engebretson, J.J., Moyer, C.L., 2003. Fidelity of select restriction endonucleases in determining microbial diversity by terminal-restriction fragment length polymorphism.
Appl. Environ. Microbiol. 69, 48234829.
Farris, M.H., Olson, J.B., 2007. Detection of Actinobacteria cultivated from environmental
samples reveals bias in universal primers. Lett. Appl. Microbiol. 45, 376381.
Ferrando, L., Fernandez Manay, J., Fernandez Scavino, A., 2012. Molecular and culturedependent analyses revealed similarities in the endophytic bacterial community
composition of leaves from three rice (Oryza sativa) varieties. FEMS Microbiol. Ecol.
80, 696708.
Gutell, R.R., 1994. Collection of small subunit (16S- and 16S-like) ribosomal RNA
structures: 1994. Nucleic Acids Res. 22, 35023507.
Hahn, M., Wilhelm, J., Pingoud, A., 2001. Inuence of uorophor dye labels on the
migration behavior of polymerase chain reactionamplied short tandem repeats
during denaturing capillary electrophoresis. Electrophoresis 22, 26912700.
Hayashi, H., Sakamoto, M., Benno, Y., 2002. Phylogenetic analysis of the human gut
microbiota using 16S rDNA clone libraries and strictly anaerobic culture-based
methods. Microbiol. Immunol. 46, 535548.
Kurokawa, K., Itoh, T., Kuwahara, T., Oshima, K., Toh, H., Toyoda, A., Takami, H., Morita,
H., Sharma, V.K., Srivastava, T.P., Taylor, T.D., Noguchi, H., Mori, H., Ogura, Y.,
Ehrlich, D.S., Itoh, K., Takagi, T., Sakaki, Y., Hayashi, T., Hattori, M., 2007. Comparative
metagenomics revealed commonly enriched gene sets in human gut microbiomes.
DNA Res. 14, 169181.
Lindberg, E., Adlerberth, I., Matricardi, P., Bonanno, C., Tripodi, S., Panetta, V., Hesselmar, B.,
Saalman, R., Aberg, N., Wold, A.E., 2011. Effect of lifestyle factors on Staphylococcus
aureus gut colonization in Swedish and Italian infants. Clin. Microbiol. Infect. 17,
12091215.
Liu, W.T., Marsh, T.L., Cheng, H., Forney, L.J., 1997. Characterization of microbial diversity by determining terminal restriction fragment length polymorphisms of genes
encoding 16S rRNA. Appl. Environ. Microbiol. 63, 45164522.
Matsumoto, M., Sakamoto, M., Hayashi, H., Benno, Y., 2005. Novel phylogenetic assignment database for terminal-restriction fragment length polymorphism analysis of
human colonic microbiota. J. Microbiol. Methods 61, 305319.
McOrist, A.L., Jackson, M., Bird, A.R., 2002. A comparison of ve methods for extraction
of bacterial DNA from human faecal samples. J. Microbiol. Methods 50, 131139.
Nieminen, T.T., Vihavainen, E., Paloranta, A., Lehto, J., Paulin, L., Auvinen, P., Solismaa,
M., Bjorkroth, K.J., 2011. Characterization of psychrotrophic bacterial communities
in modied atmosphere-packed meat with terminal restriction fragment length
polymorphism. Int. J. Food Microbiol. 144, 360366.
Palmer, C., Bik, E.M., DiGiulio, D.B., Relman, D.A., Brown, P.O., 2007. Development of the
human infant intestinal microbiota. PLoS Biol. 5, e177.
Pandey, J., Ganesan, K., Jain, R.K., 2007. Variations in T-RFLP proles with differing
chemistries of uorescent dyes used for labeling the PCR primers. J. Microbiol.
Methods 68, 633638.
Ramette, A., 2007. Multivariate analyses in microbial ecology. FEMS Microbiol. Ecol. 62,
142160.
Savage, D.C., 1977. Microbial ecology of the gastrointestinal tract. Annu. Rev. Microbiol.
31, 107133.
Steffen, E.K., Berg, R.D., Deitch, E.A., 1988. Comparison of translocation rates of various
indigenous bacteria from the gastrointestinal tract to the mesenteric lymph node.
J. Infect. Dis. 157, 10321038.
Wang, M., Ahrne, S., Antonsson, M., Molin, G., 2004. T-RFLP combined with principal
component analysis and 16S rRNA gene sequencing: an effective strategy for
comparison of fecal microbiota in infants of different ages. J. Microbiol. Methods
59, 5369.
Wang, Q., Garrity, G.M., Tiedje, J.M., Cole, J.R., 2007. Naive Bayesian classier for rapid
assignment of rRNA sequences into the new bacterial taxonomy. Appl. Environ.
Microbiol. 73, 52615267.
Wang, M., Karlsson, C., Olsson, C., Adlerberth, I., Wold, A.E., Strachan, D.P., Martricardi,
P.M., Aberg, N., Perkin, M.R., Tripodi, S., Coates, A.R., Hesselmar, B., Saalman, R.,
Molin, G., Ahrne, S., 2008. Reduced diversity in the early fecal microbiota of infants
with atopic eczema. J. Allergy Clin. Immunol. 121, 129134.
Westerhuis, J.A., van Velzen, E.J., Hoefsloot, H.C., Smilde, A.K., 2010. Multivariate paired
data analysis: multilevel PLSDA versus OPLSDA. Metabolomics 6, 119128.
Wilson, K.H., Blitchington, R.B., 1996. Human colonic biota studied by ribosomal DNA
sequence analysis. Appl. Environ. Microbiol. 62, 22732278.
Yatsunenko, T., Rey, F.E., Manary, M.J., Trehan, I., Dominguez-Bello, M.G., Contreras, M.,
Magris, M., Hidalgo, G., Baldassano, R.N., Anokhin, A.P., Heath, A.C., Warner, B.,
Reeder, J., Kuczynski, J., Caporaso, J.G., Lozupone, C.A., Lauber, C., Clemente, J.C.,
Knights, D., Knight, R., Gordon, J.I., 2012. Human gut microbiome viewed across
age and geography. Nature 486, 222227.