OUP user menu

A longitudinal study of infant faecal microbiota during weaning

Fabien Magne, Wahiba Hachelaf, Antonia Suau, Ghazalia Boudraa, Irène Mangin, Mahmoud Touhami, Karim Bouziane-Nedjadi, Philippe Pochart
DOI: http://dx.doi.org/10.1111/j.1574-6941.2006.00182.x 563-571 First published online: 1 December 2006

Abstract

For infants, the introduction of food other than breast milk is a high risk period due to diarrheal diseases, and may be corroborated with a shift in the faecal microbiota. This longitudinal study was the first undertaken to understand the effect of the supplementation on the infant's faecal microbiota and particularly the bifidobacteria. Eleven infants were enrolled. Their faecal microbiota were analysed using temporal temperature gradient gel electrophoresis (TTGE) with bacterial and bifidobacterial primers. In parallel, bifidobacterial counts were followed using competitive PCR. Three periods were distinguished: exclusive breastfeeding (Bf period), weaning (i.e. formula-milk addition, W period) and postweaning (i.e. breastfeeding cessation, Pw period). The bifidobacterial counts were not modified, reaching 10.5 (Log10 cells g−1 wet weight). In the TTGE profiles, the main identified bands corresponded to Escherichia coli, Ruminococcus sp. and Bifidobacterium sp., more precisely Bifidobacterium longum, Bifidobacterium infantis and Bifidobacterium breve. For both TTGE profiles, the analysis of the distance suggested a maturation of the faecal microbiota but no correlation could be established with the diet. Despite a high interindividual variability, composition of the faecal microbiota appeared more homogenous after weaning and this point may be correlated with the cessation of breastfeeding.

Keywords
  • weaning
  • formula milk
  • bifidobacteria
  • intestinal microbiota
  • 16S rRNA gene
  • temporal temperature gradient gel electrophoresis (TTGE)

Introduction

It is well documented that in economically developing countries diarrheal and respiratory illnesses are more common among infants introduced early to solid foods than among infants who remain exclusively breast fed (Foote & Marriott, 2003). Furthermore, several studies showed a protective effect of exclusive breastfeeding on acute diarrhea (Mitra & Rabbani, 1995). In 2001, the World Health Organization issued a revised global recommendation that mothers should breast feed exclusively for 6 months (Gupta et al., 2002).

The intestinal microbiota exerts a barrier against the development of pathogenic bacteria in the digestive tract and is mandatory to the establishment of oral tolerance to food antigens (Gaboriau-Routhiau & Moreau, 1996). The intestinal microbiota of infants can be influenced by several environmental factors; among them, the type of feeding is the most important (Mackie et al., 1999). Exclusive breastfeeding has a specific effect on the composition of microbiota when compared with formula-fed infants. Human milk may favor the growth and dominance of bifidobacteria and may repress the development of other obligate and facultative anaerobes, particularly Bacteroides spp., Clostridium spp. and Enterobacteria. In comparison, the microbiota of formula-fed infants is colonized more frequently by these anaerobes, in addition to bifidobacteria (Mackie et al., 1999).

However, the addition of food supplements to the diet of nursing infants, which probably causes major shifts in the faecal microbiota, has scarcely been studied. Nevertheless, a study using bacteriological culture has observed changes in the infant microbiota, when breast milk was supplemented with solid food. In this study, the counts of Enterococci and Bacteroides increased in the group of infants (6 months old) who received solid food as well as breast milk, whereas enterobacteria and bifidobacteria remained constant (George et al., 1996). However, the type of feeding was not controlled and could vary between infants, so it was difficult to estimate the impact of the food on the infant microbiota. The difficulty in understanding the effects of supplementation lies partly in the fact that, in human infants, supplementary foods are usually added gradually, and the total daily intake of breast milk declines progressively until weaning is completed.

Our study is the first longitudinal study to analyse the faecal microbiota before, during and after weaning and focused particularly at this transition period and the nature of the dietary supplements to understand the effect of the supplementation on the infant's faecal microbiota and particularly the bifidobacteria. All healthy infants were supplemented with the same artificial milk and three periods were observed: exclusive breast milk, breast milk supplemented with artificial milk (weaning) and artificial milk alone (postweaning). The impact of diet on the faecal microbiota at weaning was investigated using bacterial and bifidobacterial PCR-temporal temperature gradient gel electrophoresis (PCR-TTGE). Moreover, the bifidobacterial counts were followed using competitive PCR.

Materials and methods

Subjects

Infants exclusively breast-fed from birth were recruited in the Amilcar Cabral clinic of Oran (Algeria). Parents who were interested in introducing a formula milk were informed of the study. Their written consent was obtained after reviewing the study protocol. This research protocol observed the tenets of the Declaration of Helsinki. Eleven infants were enrolled in this study. Formula milk was given to the infants when their mothers decided to start weaning. Two formula milks (starter and follow-up) were used according to infant ages and their compositions are shown in Table 1. Breastfeeding was supplemented with the formula milk between 6 and 18 weeks of age (age±SD, 12.9±4.9 weeks). In this study, three periods were distinguished: (1) the breastfeeding period (Bf period) corresponding to exclusive breastfeeding, (2) the weaning period (W period) defined as addition of formula-milk and (3) postweaning period (Pw period), which started after cessation of breastfeeding. The ages of infants are presented in Table 2.

View this table:
1

Composition of the formula milks (per 100 mL)

Formula milks
StarterFollow-up
Energy (kJ)289.5268.00
Protein (g)1.511.62
Carbohydrate (g)7.558.3
Lactose (g)6.184.18
Maltodextrin (g)1.364.12
Fat (g)3.672.70
Minerals (g)0.360.47
Sodium (mg)16.2029.7
Potassium (mg)62.1087.75
Chloride (mg)44.5568.85
Calcium (mg)55.3560.75
Phosphorus (mg)27.6848.60
View this table:
2

Age of infants over the experimental periods

Enrollment11 (9 boys and 2 girls)
Effective weight (g)4144 ± 733 [2950–5200]
Bf period
Age (weeks)5.3 ± 2.8 [2.0–11.0]
W period
Age (weeks)12.9 ± 4.9 [6.0–21.0]
Pw period
Age (weeks)20.8 ± 10.5 [7.0–42.0]
  • Values are means ± SD [ranges]. Bf period, breastfeeding period; W period, weaning period; Pw period, postweaning period.

Faecal samples were collected during the three periods and stored at −20°C. The numbers of analysed samples per infant (means±SD) in the three periods were 3.3±1.2 for the Bf period, 2.6±2.2 for the W period, 4.3±2.3 for the Pw period.

Extraction and purification of total DNA

DNA was extracted from faecal samples (125 mg) using a bead-beating method adapted from Godon et al. (1997). The protocol was described in detail in a previous study (Mangin et al., 2006). The amount and integrity of DNA were estimated using 2% (w/v) agarose gel electrophoresis containing ethidium bromide (0.1 ng mL−1), in 1 × Tris-Borate-EDTA (TBE).

Bacterial TTGE

The primers S-D-Bact-339-a-S-20 (5′-CTC CTA CGG GAG GCA GCA GT-3′) and S-D-Bact-788-a-A-19 (5′-GGA CTA CCA GGG TAT CTA A-3′) were used to amplify the variable regions 3 and 4 of the bacterial 16S rRNA genes. A GC-rich sequence (5′-CCC CCC CCC CCC CGC CCC CCG CCC CCC GCC CCC GCC GCC C-3′) was added to the 5′ end of the primer S-D-Bact-788-a-A-19. The protocol has been previously described (Magne et al., 2006). The Dcode universal mutation detection system (Bio-Rad, Hercules, CA) was used for sequence-specific separation of amplicons (Magne et al., 2006). Additionally, known bacterial strains were used to standardize band migration and gel curvature among different gels. This ladder consisted of the following organisms listed in their migration order: Bacteroides sp., Enterococcus faecium, Staphylococcus epidermidis, Escherichia coli and Bifidobacterium longum.

Bifidobacterial TTGE

Specific primers for bifidobacterial TTGE have been validated (Satokari et al., 2001): forward primer S-G-Bif-164-a-S-18 (GGG TGG TAA TGC CGG ATG), and reverse primer S-G-Bif-662-a-A-18 (CCA CCG TTA CAC CGG GAA). A GC-rich sequence (5′-CGC CCG CCG CGC GCG GCG GGC GGG GCG GGG GCA CGG GGG G-3′) was added to the 5′ end of the primer S-G-Bif-662-a-A-18. PCR was performed using the Taq DNA polymerase (AmpliTaq Gold; Perkin-Elmer Corporation, Foster City, CA) as previously described (Mangin et al., 2006). The Dcode universal mutation detection system was used for sequence-specific separation of amplicons (Mangin et al., 2006). Additionally, references representing known bifidobacteria strains were loaded to allow standardization of band migration and gel curvature among different gels. This ladder consisted of the following organisms, frequently isolated in the human faecal microbiota: Bifidobacterium adolescentis, Bifidobacterium bifidum, Bifidobacterium angulatum, Bifidobacterium dentium, Bifidobacterium infantis, Bifidobacterium gallicum, Bifidobacterium pseudocatenulatum and Bifidobacterium animalis ssp. lactis.

Analysis of TTGE profiles

TTGE analysis was based on the method of Deplancke et al. (2002) who have developed an original strategy. Gel patterns were analysed using Diversity Database 2.1, part of the Discovery Series (Bio-Rad).

Bands were detected automatically. Each band was defined through its relative intensity and relative front (Rf). The relative intensity was the intensity of a particular band in a lane expressed as a percentage of the total intensity data in the lane. Rf was the distance from the top of a defined lane to the band. In order to compare several gels, a normalized Rf derived from Rf was used to classify bands. This normalized Rf was based on the migration of a marker. A band set stored classified bands defined as unique band types according to their normalized Rf.

Comparisons of TTGE profiles were performed using Dice's similarity coefficient (Dsc) analysis based entirely on the results of band classification. Dsc values were compared, based on the presence or absence of bands. Dice's coefficient is defined as follows:

D sc=[2j/(a+b)] where j is the number of common bands between samples A and B; a and b are the total number of bands in samples A and B, respectively. This coefficient ranges from 0 (no common bands) to 1 (identical bands patterns). Consequently, the distance between two TTGE profiles was: Dist=1–Dsc.

Firstly, infant faecal microbiota were analysed individually in order to show the maturation of the ecosystem. So the distance between TTGE profiles was calculated according to the first sample collected and also to the preceding sample. Secondly, for each infant, a distance was calculated between the different periods (Bf and W, Bf and Pw, and W and Pw). Mean values of these distances were analysed to obtain a general picture of the faecal microbiota maturation in each period. In this case two analyses were carried out. First, for each infant, TTGE profiles of W and Pw periods were compared to all profiles of Bf period. The average distances were calculated for Bf/W and Bf/Pw transitions, representing the evolution of the faecal microbiota over time. Second, TTGE profiles of W and Pw periods were compared with the preceding periods (Bf and W respectively). The average distances were calculated for Bf/W and W/Pw transitions, representing the speed of the faecal microbiota evolution.

Identification of bands frequently retrieved

For identification, bands were excised from TTGE gels, cloned and sequenced. The recovery of DNA includes sterile excision of bands, which were kept overnight at 4°C in 50 μL of sterile water.

PCR was performed with Taq DNA polymerase using primers S-D-Bact-339-a-S-20 and S-D-Bact-788-a-A-19 or S-G-Bif-164-a-S-18 and S-G-Bif-662-a-A-18 without a GC-rich sequence. PCR amplicons were purified and concentrated using QIAquick spin PCR Purification Kit (QIAGEN, S.A. Courtaboeuf, France). Their concentrations and sizes were estimated on a 1% agarose gel electrophoresis containing ethidium bromide (0.1 ng mL−1). The purified products were cloned into pGEM®-T (Promega Corporation, Madison, WI) as specified by the manufacturer.

Sequencing reactions were performed by Sequencia (Evry, France). The insert was sequenced using two plasmid-targeted primers, T7 (5′-TAA TAC GAC TCA CTA TAG GGC GA-3′) and SP6 (5′-ATT TAG GTG ACA CTA TAG AAT AC-3′).

Newly determined sequences were compared to those available in public databases [Ribosomal Database Project (RDP) and GenBank®] in order to ascertain their closest relatives. A molecular species comprised all sequences with at least 98% sequence similarity (reference strains, isolates, clones) (Suau et al., 1999).

Competitive PCR

Bifidobacterial quantification has been described previously (Mangin et al., 2006). The competitor was serially 10-fold diluted and coamplified using PCR with a constant amount of target DNA from the infant faecal sample. PCR amplifications were carried out with the standard PCR mix described above using primers S-G-Bif-164-a-S-18 and S-G-Bif-662-a-A-18. The relative intensities from individual bands were measured using Quantity One software (Bio-Rad) and a band intensity ratio between the target and the competitor was calculated. According to the competitor copy number, the number of 16S rRNA genes was deduced and divided by a mean copy value of 4 to obtain a cell equivalent value (cells g–1 wet weight) (Mangin et al., 2006).

Statistical analyses

The data for the three periods were given as means±SD or±SEM for the 11 infants. Differences between the means were checked for significance using the paired t-test.

Results

Infants

In this study, 11 cases of diarrheas were diagnosed in seven infants. The occurrences of diarrhea were as follows: one infant developed diarrhea in all three periods, two infants in the W and Pw periods and four infants in the W period.

Bacterial diversity

The faecal microbiota was analysed using PCR-TTGE with bacterial primers. An example of band patterns is presented in Fig. 1a. The band set contained a total of 63 bands with different combinations among the TTGE profiles. The occurrence of bands within the three periods was as follows: 35 bands in the breastfeeding period (Bf), 31 bands in the weaning period (W), and 42 bands in the postweaning period (PW). In the Bf period, 49% of the classified bands occurred in just one infant, 31% in two infants, and 20% in more than two infants. In the W period, 52% of the classified bands were retrieved in just one infant, 23% in two infants, and 26% in strictly more than two infants. In the Pw period, 36% of the classified bands occurred in just one infant, 36% in two infants, and 29% in strictly more than two infants.

1

(a) TTGE profiles representing the bacterial diversity of an infant from all samples collected over time (weeks). M, ladder; ♦, breastfeeding period; ▪,weaning period; ◻, postweaning period. (b) Representation of the distance. The distance between TTGE profiles calculated according to the first sample collected is illustrated for each infant overtime () and to the preceding sample (Embedded Image). (c) Mean values of band intensities observed in bacterial TTGE within the three periods. Three bands were analyzed: bands A (Ruminococcus sp; Embedded Image), B (E. coli; Embedded Image) and C (Bifidobacterium sp.; Embedded Image). Bf period: breastfeeding period; W period: weaning period; Pw period: postweaning period. * The relative intensities of band A and B were statistically different with the Bf period (P<0.05). § The relative intensity of band A was statistically different with the W period (P<0.05). None of the differences are statistically significant for the band C (P<0.05).

The number of bands per lane varied from one to 11. In the Bf period, one to 10 bands were detected (mean±SEM 3.6±0.6). The number of bands in the W period was between one and six (mean±SEM 3.4±0.3). The TTGE profiles of the Pw period were composed of one to 11 bands (mean±SEM 4.4±0.5). When samples were gathered according to periods, Bf, W and Pw, the mean band number did not differ.

The example presented in Fig. 1a emphasizes the dramatic differences between periods, not only using visual comparison, but also using calculation of the distance over time between the patterns with Dice's coefficient (Fig. 1b).

To obtain a general picture of the faecal microbiota maturation, the means of the distances were calculated according to the different periods and reported in Table 3. The average distances for Bf/W and Bf/Pw transitions increased; moreover, the distances between Bf/W and Bf/Pw were significantly different (P<0.05). This distance alteration was a consequence of a change in the band composition in the TTGE profiles; consequently this revealed a shift in the faecal microbiota of infants. Conversely, no difference existed between the distances of the periods Bf/W and W/Pw. The transition from Bf to W period or from W to Pw period slightly altered the faecal microbiota evolution.

View this table:
3

Distances of bacterial and bifidobacterial TTGE profiles

Distances
TTGE profilesBf and W periodsBf and Pw periodsW and Pw periods
Bacteria0.59*± 0.070.76*± 0.040.61 ± 0.04
Bifidobacteria0.39*± 0.060.61*± 0.050.58 ± 0.05
  • Values are means ± SEM. Bf period, breastfeeding period; W period, weaning period; Pw period, postweaning period. The differences between the distances of Bf/W periods and W/Pw periods were not statistically significant (P<0.05).

  • * The distances between Bf/W periods and Bf/Pw periods were significantly different (P<0.05).

Bacterial band identification

Band identification is summarized in Table 4, including their sequence similarities with closest relatives in GenBank®. Nevertheless these data should be interpreted with caution, as only a partial 16S rRNA gene sequence was analysed.

View this table:
4

Identification of the most frequent bands and numbers of infants colonized

NamesTTGE typeRf (0–1)Closest relativeBf periodW periodPw period
ABacteria0.50Ruminococcus sp.1/113/116/11
BBacteria0.59E. coli9/119/119/11
CBacteria0.90–0.97Bifidobacterium sp.11/1111/1111/11
DBifidobacteria0.52B. infantis/longum3/117/116/11
EBifidobacteria0.54B. breve4/114/117/11
  • Bf period, breastfeeding period; W period, weaning period; Pw period, postweaning period; Rf, relative front.

Three bands were frequently retrieved and their relative intensity was compared through the different periods (Fig. 1c). The detection of band A (corresponding to Ruminococcus sp.) was altered when feeding changed: it was detected in one infant in the Bf period, three infants in the W period and six infants in the Pw period. Among infants, one infant had this band in his TTGE profile during the three periods, two infants in W and Pw periods, and three infants only in the Pw period. This band represented a small part of the bacterial population (c. 4% of all bands within profiles) during Bf and W periods and increased at the Pw period, reaching 20%. The band B (corresponding to E. coli) was frequently detected in nine infants during the three periods. Among them, six infants had this band in their TTGE profiles during the three periods, two during the Bf and W periods, one in the Bf and Pw periods, one in the W and Pw periods and one in the Pw period. Its intensity was not different at the Bf and W periods, reaching 15% of the band population, but it significantly increased during the Pw period, reaching about 30%. A cluster of bands (‘band C’), which migrated at the bottom of the gels, was observed in all infants studied during the three periods. Preceding studies showed that multiple bands assigned at the bottom of the gel corresponded to Bifidobacterium species. Bifidobacterium strains identified in the human faecal microbiota could not be separated using bacterial TTGE. Bands corresponding to Bifidobacterium sp. were dominant in the infant faecal microbiota and their intensity reached 70% of the band population during the Bf period. In contrast with bands A and B, the Bifidobacterium band intensity did not vary significantly over time. Nevertheless, during the W and Pw periods, the intensity of ‘band C’ slightly decreased and reached 63% and 58% of the band population.

Bifidobacterial quantification

To evaluate the population of bifidobacteria in infant faecal samples, 108 samples were analysed using competitive PCR with primers specific for bifidobacteria. This genus was not detected in seven samples of four infants: one sample for two infants, three samples of another one and four samples for the last one. For these samples, TTGE profiles obtained using bacterial primers did not harbour ‘band C’. For each period, infant faecal samples were analysed using competitive PCR. Mean (±SEM) counts (Log10 cells g−1 wet weight) were 10.7±0.2 for the Bf period, 10.5±0.2 for the W period and 10.5±0.2 for the Pw period. Mean counts of bifidobacteria were not significantly different.

Bifidobacterial diversity

A similar strategy has been used to analyse profiles obtained from the bifidobacterial population. An example of PCR-TTGE patterns for bifidobacterial amplicons is presented in Fig. 2a. This gel represented all faecal samples collected from an infant.

2

(a) TTGE profiles representing the bifidobacterial diversity of an infant from all samples collected over times (weeks). ♦, breastfeeding period; ▪,weaning period; ◻, postweaning period. (b) Representation of the distance. The distance between TTGE profiles calculated according to the first sample collected is illustrated for each infant overtime (—) and to the preceding sample (Embedded Image). (c) The mean values of specific band intensity observed in bifidobacterial TTGE within the three periods. Two bands were studied: band D (B. infantis/longum; Embedded Image) and Band E (B. breve;Embedded Image. Bf period: breastfeeding period; W period: weaning period; Pw period: postweaning period.None of the differences was statistically significant (P<0.05).

The band set contained a total of 32 bands, which were found in different combinations among the TTGE profiles. Twenty-six bands were detected within the Bf period, 29 in the W period and 28 in the Pw period. In the Bf period, 42% of the classified bands occurred in just one infant studied, 35% in two infants, and 23% in strictly more than two infants. In the W period, 55% of the classified bands were retrieved in just one infant studied, 28% in two infants, and 17% in strictly more than two infants. In the Pw period, 28% of the classified band occurred in just one infant studied, 36% in two infants, and 36% in strictly more than two infants.

The TTGE profiles obtained were composed of one to 10 bands. One to seven bands were detected in samples collected during the Bf period (mean±SEM 3.3±0.5), one to nine bands in the W period (mean±SEM 3.8±0.5) and one to 10 bands in the Pw period (mean±SEM 3.6±0.4). The mean band numbers from these three periods were not significantly different.

In order to analyse bifidobacterial TTGE profiles, the distance using the Dice coefficient was calculated between samples (Fig. 2b) and clustered according to the periods (Table 3). The distances increased when the W and Pw periods were compared to Bf. The distance between Bf and W and between Bf and Pw were significantly different. The increasing of the distance corresponded to a modification of bands over time in the TTGE profiles. This difference may be corroborated with a shift within the bifidobacterial population. No difference was detected between the Bf/W and W/Pw transitions. The transition from Bf to W did not produce more changes than from W to Pw.

Bifidobacterial band identification

Patterns obtained from bifidobacterial TTGE were compared in order to study the prevalence of bands. Two different bands were frequently retrieved through the different periods (bands D and E). Their identification is reported in Table 4. The detection of band D (corresponding to B. infantis/longum) and band E (corresponding to B. breve) increased over time when feeding changed (Table 4). The band D was detected in the three periods for three infants, in the W and Pw periods for two infants, in the W period for two infants and in the Pw period for one infant. The band F was detected in three periods for three infants, in the Bf and W periods for one infant and in the Pw period for four infants. In addition, relative intensities were compared between the different periods (Fig. 2c). When feeding changed, for both bands (bands D and E) no variation was observed. The band E represented 39–61% of the band population.

Discussion

This work is the first longitudinal study of changes in microbiota through the weaning period. The analysis of the faecal microbiota of 11 infants using bacterial and bifidobacterial TTGE revealed a high interindividual variability in the dominant microbiota, which slightly decreased after cessation of breastfeeding. Bifidobacterial counts, estimated using a competitive PCR, were not altered during the study.

One to 11 bands composed bacterial profiles and the mean value was between three and four during the Bf period. The bifidobacterial profiles of these infants were composed of one to seven bands and the mean band number was also between three and four during the Bf period. These results were in agreement with preceding studies using denaturing gradient gel electrophoresis (DGGE), which analysed changes in the composition of faecal infant microbiota from 5 days to 5 months after birth. In these studies, the number of bands in bacterial DGGE varied from one to six after 1 month (Favier et al., 2002), and one to four bands were observed in bifidobacterial DGGE for each sample (Favier et al., 2003). In our study the number of bands observed in TTGE profiles, either bacterial or bifidobacterial, was not altered when feeding changed.

Analysis of profiles obtained from bacterial TTGE allowed the construction of a bacterial band set of 63 bands and a bifidobacterial band set of 32 bands. At the same time, only 12 bifidobacterial species have previously been associated with humans: Bifidobacterium adolescentis, Bifidobacterium infantis, Bifidobacterium longum, Bifidobacterium bifidum, Bifidobacterium breve, Bifidobacterium catenulatum, Bifidobacterium pseudocatenulatum, Bifidobacterium angulatum, Bifidobacterium gallicum, Bifidobacterium inopinatum, Bifidobacterium dentium, and Bifidobacterium denticolens, the last three being found primarily in the oral cavity (Lauer, 1990; Biavati & Mattarelli, 1991; Crociani et al., 1996; McCartney et al., 1996; Matsuki et al., 1999). The high number of bifidobacterial bands observed in our study could be due to a TTGE artefact as previously described (Satokari et al., 2001). Actually, bifidobacterial species possess from two to five operons of 16S rRNA genes, which could be different (Mangin et al., 1996). It could lead to several bands for one species. Consequently, the number of bands may overestimate the number of species in the infant faecal microbiota. Nevertheless, the number of bacterial or bifidobacterial bands detected was important during every period. Moreover, during the Bf and W periods, a majority of bands detected in TTGE profiles (for bacteria and bifidobacteria) was detected in just one infant. These results show a high interindividual diversity in the dominant faecal microbiota. However the number of bands detected only in one infant decreased after weaning, showing an increase of homogeneity in infant faecal microbiota. It could be related to the cessation of breastfeeding. Compositions of the breast milk had been demonstrated to be different between women: the individual characteristics of breast milk may influence the infant intestinal microbiota. Moreover breastfeeding favoured contact between infants and their mothers. Consequently, the skin microbiota may also contribute to the maturation of the infant's faecal microbiota. For example, in a study, 96% of infants, whose parents were Staphylococcus aureus skin carriers, had S. aureus in their faeces and 91% had the same strain as at least one of the parents (Lindberg et al., 2004). These results suggested that S. aureus from parental skin could establish readily in the infantile gut, and perhaps the same process could be applied for other bacteria. To summarize, parameters such as breast milk composition and physical contact between mothers and their infants may lead to a specific intestinal microbiota which characterize each infant.

The evolution of the faecal microbiota was estimated using a distance method. When the distance was calculated according to the first period (Bf), a difference was observed between the W and Pw periods, suggesting a maturation of the faecal microbiota. Nevertheless, when the distance was calculated according to the preceding period, no difference was observed. The faecal microbiota evolved in these infants but no acceleration was observed. There was no difference between distances at alimentation transition, suggesting no correlation with the diet. In this study, the ages and the durations of the periods were different for each infant. So it was difficult to attribute the maturation of the faecal microbiota to the diet (Bf, W or Pw) or to the time.

Analysis of bacterial and bifidobacterial TTGE profiles revealed frequent bands. In the bacterial profiles, two bands (Bands A and B) were excised, reamplified and sequenced. The detection of band A, which corresponded to Ruminococcus sp., increased with time and was detected in six infants during the Pw period. In a preceding study using the DGGE method, Ruminococcus sp. belonged to dominant species identified in the microbiota of two babies, until the age of about 3.5 months (Favier et al., 2002). Moreover, counts of Ruminococcus strains using a culture-dependant method are usually present at low (≤5 log10 CFU g−1) or moderate (6–8 log10 CFU g−1) levels in faeces of infants (before and after 1 month, respectively) (in Mackie et al., 1999).

The band B, which corresponded to E. coli, was frequently isolated within the three periods. The majority of infants had this band at the Bf period and conserved it during the W and Pw periods.

In the bottom of the gels, a cluster of bands was visualized in all infants. These bands corresponded to Bifidobacterium sp. and could not be separated in bacterial TTGE (data not shown). It can be assumed that band intensity is proportional to the relative abundance of a particular species in the population (Godon et al., 1997; Calvo-Bado et al., 2003): bands within the profile represent the dominant bacteria within the ecosystem. Bruggemann et al. (2000) supported that profiles of bacterial communities generated using PCR-based methods are reasonable estimations of an in situ dominant community. Based on band intensity, bifidobacteria were predominant, compared to other molecular species present in the faecal microbiota of infants, and were not altered by the alimentation transitions. In a previous study, the percentage of cultivable bifidobacteria within the total population was about 30% in the faeces of a 1-month-old baby (Favier et al., 2003). Preceding studies using culture methods had also identified Bifidobacterium as a dominant bacterial group in the majority of infant faecal microbiota, reaching 8.5 (log10 cells g−1) at 6 months (83% of infants) (George et al., 1996), 7.3 at 12 months (72% of infants) (Sepp et al., 1997) and 9.7 months at 10–18 months (86% of infants) (Guerin-Danan et al., 1997). Our study confirmed that bifidobacteria represented a dominant fraction of the faecal bacterial population in these infants, reaching 10.5 log10 cells g−1 and was not altered by the diet modification. Nevertheless, during the Pw period, the intensities of bands A and B increased significantly, whereas bifidobacterial bands (‘band C’) tended to diminish. This decrease may be correlated with the cessation of breastfeeding and lead to the development of other bacteria.

Strains of Ruminococcus, as well as some bifidobacteria, have been defined as the major producers of extracellular glycosidases that degrade oligosaccharide chains of mucin (Hoskins & Boulding, 1981; Falk et al., 1998). The resulting mono- and disaccharides can be used by other faecal populations that cannot degrade these chains (Falk et al., 1998). Consequently, the enzymatic potential of these bacteria suggests that these organisms have an essential or at least substantial role in the intestinal ecosystem or host. Moreover, the high levels of bifidobacteria may prevent intestinal colonization by other bacteria. Among bifidobacteria, two species were frequently detected in the bifidobacterial PCR-TTGE profiles: B. infantis/longum and B. breve. Preceding work has shown that certain bifidobacterial species, such as B. infantis, B. longum and B. breve, were often detected in the intestinal microbiota of breast-fed and formula-fed infants (Kleessen et al., 1995). Using 16S rRNA gene primers specific for bifidobacterial species, Matsuki et al. (1999) have found that B. breve was present in 70% of the faecal samples from infants followed by B. infantis and B. longum.

To the best of our knowledge, this study constitutes the first longitudinal investigation of the faecal microbiota at weaning. Our results confirmed the dominance of bifidobacterial species and their high level during breastfeeding without alteration during weaning. Ruminococcus sp., E. coli and Bifidobacterium species (such as B. longum/infantis and B. breve) were the most frequently retrieved species in the dominant microbiota using PCR-TTGE.

In conclusion, no specific pattern was observed during weaning when the bacterial and the bifidobacterial microbiota were investigated. In contrast, a high interindividual variability has been observed. Nevertheless, the composition of the faecal microbiota appeared more homogenous after weaning and this may be correlated to the cessation of breastfeeding. In order to evaluate the consequence of the cessation of breastfeeding, it would be interesting to study the evolution of bacterial groups composing the human faecal flora of infants.

Acknowledgements

We acknowledge Balamurugan Ramadass for the reading and improvement of this manuscript.

References

View Abstract