We assessed the fecal microflora of 10 healthy infants and 27 infants with atopic eczema during breast-feeding and after weaning. The atopic infants had less frequently Gram-positive species among the most predominant aerobes and smaller total cell counts. Further differences were associated with more extensive manifestations, seen as higher bacteroides and lower bifidobacteria counts. Weaning resulted in decreased bacteroides counts in atopic and total cell counts in healthy infants and diminished predominance by bifidobacteria in both. In conclusion, the most prominent question raised by these data is whether Gram-positive bacteria may have distinctive importance in protection against atopic sensitization. Further studies aiming to answer this question are warranted.
The prevalence of atopic diseases has rapidly increased in Western societies during recent decades . Indirect epidemiological evidence has led to the suggestion that this increase could be related to reduced exposure to certain infections and environmental microbes due to high hygiene circumstances [2–5]. Postnatal allergen challenge of the immature immune system may predispose the infant toward development of potentially harmful type 2 T-helper cell (Th2)-polarized memory . The ‘hygiene hypothesis’ is based on an assumption that such sensitization is less likely to occur in the simultaneous presence of adequate microbial stimulus directing the immune responses toward the Th1 cytokine phenotype . This suggested association between microbial stimulus and allergy is most directly supported by the recent finding that perinatal administration of specific exogenous bacteria (i.e. probiotics originally isolated from a healthy gut) halved the subsequent occurrence of atopic eczema in at-risk infants .
The gut microbiota is the major source of microbial stimulus in infancy. Recent experimental data indicate that the predilection to induce Th1 or Th2 phenotype cytokine responses may vary between different bacterial groups . Thus if indeed microbes have a role in the etiology of allergy these data raise the question as to whether the composition of gut microbiota in allergic infants could be associated with the development of allergic symptoms and could consequently constitute a new treatment target. Studies indicating that the composition of gut microbiota is different between allergic and healthy infants support this view [10,11]. Components of the potentially pathogenic microflora (and therefore their antagonists such as bifidobacteria and lactobacilli) have been suggested to have a particular importance in this respect [7,11]. This is based on the idea that tolerance is not formed towards pathogenic microbes and consequently they may have greater potential to stimulate the Th1 responses than commensal bacteria against which tolerance exists [7,12]. Conversely, such microorganisms may cause a vicious circle via enhanced gut inflammation that compromises the gut barrier, which leads to aberrant antigen uptake and thus increased risk for atopic sensitization . On the basis of experimental data microflora could also be associated with food allergy due to enhanced elimination of food antigens via effects on immunological defense [13–17], degradation, permeation and presentation [18–22].
There is rapidly increasing evidence that the intestinal microflora could constitute a potential treatment target in allergic disease [8,10,11,23]. The dynamic nature of the gut microbiota, influenced by factors such as diet, aging and gastrointestinal disease is a challenge for such targeting. Therefore in the present study our aim was to characterize the composition of gut microflora longitudinally with respect to allergic symptoms and dietary change.
2.1Subjects and design
The study population comprised of 27 atopic infants and 10 healthy infants during and after exclusive breast-feeding. None of the selected infants had been treated with antibiotics. All the atopic infants fulfilled the Hanifin criteria for atopic eczema in children, i.e. atopic eczema was confirmed if the following three major features were detected: pruritus, typical morphology and distribution, and chronic relapsing course (duration of 1 month or longer) . It was also suspected that they were sensitized to several basic foods as supported by positive radioallergosorbent and/or skin prick/patch tests [25,26]. The symptoms of atopic eczema had begun at a mean age of 2.1 months (range 0.1–6 months). During breast-feeding 11 of the atopic infants additionally had gastrointestinal symptoms (such as constipation, loose stools and vomiting) suggestive of milk hypersensitivity. After weaning, cow milk allergy was diagnosed in 17 atopic infants (of whom eight exhibited gastrointestinal symptoms during breast-feeding) by a double-blind placebo-controlled challenge as previously described . Healthy infants were age-compatible and had no gastrointestinal or other diseases or symptoms.
After breast-feeding, the atopic infants were weaned to an amino acid derived formula (SHS International Ltd., Liverpool, UK), while the healthy infants received an adapted infant formula (Valio Ltd., Helsinki, Finland). Breast milk and dietary intakes were monitored as previously described .
Fecal samples were collected from all infants while they were exclusively breast-fed at a mean (95% confidence intervals (CI)) age of 5.8 (4.9–6.6) months and after cessation of breast-feeding at a mean age of 11.9 (10.5–13.3) months. From atopic infants a follow-up sample was also collected at a mean age of 12.8 (11.7–13.8) months, the diversification of their diet being slower than in healthy infants. The extent of atopic sensitization and the severity of atopic eczema were assessed at the respective time points as described below.
The study was approved by the Committee on Ethical Practice of Tampere University Hospital, Finland, and written informed consent was obtained from the children's parents.
2.2Evaluation of the severity of atopic disease
The extent of sensitization of atopic infants was evaluated by serum total IgE concentration (Phadebas IgE Prist, Pharmacia, Uppsala, Sweden) and the severity of atopic eczema by the SCORAD method, which was established by the European Task Force on Atopic Dermatitis to improve the objectivity of such evaluation . In brief, the extent was estimated using the rule of nines, the intensity being the sum of the scores for erythema, edema and/or papules, excoriation, lichenification and dryness. The subjective manifestations, including pruritus and sleep loss, were assessed from parents’ estimations. SCORAD was obtained with the calculation: extent/5+3.5×intensity+subjective score.
2.3Plate culture analysis of selected microbial populations
The parents collected two fecal specimens from diapers after defecation. These were immediately cooled to 6–8°C and delivered to researchers within 24 h. One sample was frozen at −75°C until fluorescence in situ hybridization (FISH) analysis and the other immediately analyzed by traditional plate culturing. All the fecal samples collected during and after breast-feeding were analyzed fresh by two qualitative cultivation procedures. Firstly, the samples were anaerobically cultured on cycloserine-cefoxitin-fructose agar (Oxoid, Basingstoke, UK) for the detection of Clostridium difficile and aerobically on 110 agar (Oxoid) and Dixon agar, respectively for Staphylococcus aureus and yeasts. Secondly, in order to study the most dominant aerobic flora in the feces the samples were cultured on MacConkey agar (Difco, Detroit, MI, USA), blood agar (Lab M, Bury, UK) and cystine-lactose electrolyte deficient agar (Lab M). Different colonies were counted and identified. From these data, 1–3 of the most predominant colony types were characterized according to their growth on selective media, colonies, color, and cell morphology.
2.4FISH analysis of selected gut genera
The fecal samples with adequate amount of fecal material for quantitative analysis (n=15) were processed, bacterial cells fixed and FISH assay performed as previously described [28,29]. In brief, cells were fixed overnight in 4% (v/v) paraformaldehyde at 4°C, washed twice in PBS, and stored at −20°C in a PBS:ethanol (1:1) solution. Subsamples of the fixed cells were hybridized overnight in hybridization buffer with 5 ng μl−1 Cy3 indocarbocyanin-labeled oligonucleotide probe. Bifidobacteria were enumerated with probe BIF164 (5′-CATCCGGCATTACCACCC) , bacteroides with BAC303 (5′-CCAATGTGGGGGACCTT) , lactobacilli/enterococci with LAB158 (5′-GGTATTAGCA(T/C)CTGTTTCCA) , and clostridia belonging to the Clostridium histolyticum group (from here onwards referred to as clostridia) with HIS150 (5′-TTATGCGGTATTAATCT(C/T)CCTTT) . Total cell numbers were counted using 4′,6-diamidino-2-phenylindole (DAPI) as nucleic acid stain.
Fisher's exact test was used to compare atopic and healthy infants with respect to the presence of plate culture-identified bacterial groups. Fisher's least significant difference (LSD) method was used for comparisons between more than two groups. The differences between the atopic and healthy infants in the occurrence of Gram-positive and Gram-negative species among 1–3 of the most predominant aerobes were assessed by exact unconditional test . Bacterial numbers enumerated by FISH were transformed into log10 values and are presented as group means with the 95% CI in parentheses. Age-dependent bacterial figures were analyzed using analysis of covariance with the age of the infant included as a covariant. Inter-group changes in bacterial numbers were analyzed by the Wilcoxon signed rank test. The Spearman rank correlation (expressed as ρ describing linear relationship between two variables) was calculated to study the relationships between the number of different FISH-enumerated genera, serum total IgE and SCORAD scores. All analyses, except for the unconditional exact test, were performed using computer software SPSS for Windows™ release 9.0.1 (SPSS, Inc., Chicago, IL, USA).
3.1The gut microflora in atopic and healthy infants during exclusive breast-feeding
Gram-positive species (enterococci, staphylococci or streptococci) were among the most predominant aerobes in 30% (8/27) of the samples from atopic infants, whereas the respective value for healthy infants was 70% (7/10; P=0.04; Fig. 1). On the species level Klebsiella spp. were found among the predominant aerobes only in the samples from atopic infants (P=0.08), whereas predominance by Streptococcus viridans was unique to healthy infants (P=0.02; Fig. 1). In the microscopic assessment of DAPI dyed cells, the atopic infants were found to have smaller total numbers of microbes in feces than the healthy controls (P=0.06; Table 1). No statistically significant differences were observed in the bacterial concentrations within the genera enumerated by the oligonucleotide probes (Table 1).
Table 1. The presence of selected predominant gut genera in the fecal flora of atopic and healthy infants as assessed by FISH during exclusive breast-feeding and after cessation of breast-feeding
aAge adjusted where relevant.
bAll atopic infants whose fecal samples contained enough biomass for FISH analysis.
cPrior to this comparison the variances in the number of bifidobacteria between atopic infants with and without GI symptoms and healthy babies had been equalized for statistical purposes by removing an outlier from the data (an atopic infant with gastrointestinal symptoms, log mean 5.15).
dDifference to the respective numbers of bacteria compared (in bold) reaches significance level P<0.08.
eDifference to the respective value compared (in bold) reaches significance level P<0.05.
fDiffers significantly (P<0.05) from the respective value determined during breast-feeding.
Time of detection
Number of cells per gram of feces (the log meana of each group with 95% CI in parentheses)
Healthy infants (n=10)
GI symptoms during BF (n=5)
No GI symptoms during BF (n=10)
Positive CM challenge (n=6)
Negative CM challenge (n=9)
Total cell counts
Total IgE in serum (kU l−1)
The values for atopic infants during breast-feeding additionally show the group divided based on the presence of gastrointestinal symptoms during breast-feeding as well as on the outcome of cow milk challenge during weaning. The mean serum total IgE concentration and mean SCORAD scores with interquartile ranges are provided to illustrate the extent of atopic sensitization and severity of atopic eczema of the atopic infants at enrolment. BF, breast-feeding; GI, gastrointestinal; ND, not determined; Positive/negative CM challenge, exhibited/did not exhibit an allergic response to cow milk.
3.2The gut microflora and the extent of atopic sensitization
There was a tendency toward lower bifidobacteria concentration in atopic infants with gastrointestinal symptoms in comparison to atopic infants without these symptoms (P=0.07) and healthy babies (P=0.08; Table 1). The feces of infants who were later diagnosed cow milk allergic contained higher numbers of bacteroides during breast-feeding than atopic infants with negative cow milk challenge (P=0.14) and healthy infants (P=0.06; Table 1). The cow milk allergic infants had more extensive atopic sensitization and severe clinical symptoms than the other atopic infants as indicated by higher total concentration of IgE in serum and SCORAD scores, respectively (Table 1).
During weaning a negative correlation was observed between the fecal concentration of clostridia and SCORAD scores; the more extensive and severe the eczema (i.e. higher SCORAD scores), the smaller the number of clostridia enumerated in the follow-up samples collected after cessation of breast-feeding (ρ=−0.71; n=17), P=0.01. A similar negative correlation was observed between the SCORAD scores and the total cell counts in the follow-up samples (ρ=−0.49), P=0.04. Conversely, a positive correlation was observed between the concentration of total IgE in serum and the number of lactobacilli/enterococci in feces (ρ=+0.66), P=0.03.
3.3The gut microflora and weaning
The proportion of bifidobacteria from the total cell counts decreased after cessation of breast-feeding in both groups alike; from 50% (95% CI 28–73) to 20% (9–32), P=0.04 in allergic infants and from 49% (27–71) to 26% (10–42), P=0.07 in healthy infants (Table 1). The proportion of the total cell counts detected by the four probes used in FISH analyses decreased along with the decreasing concentration of bifidobacteria; prior to weaning these probes covered 56% (95% CI 41–71%) of the total microbes detected by DAPI, whilst only 28% (19–37%) after cessation of breast-feeding. In healthy infants, total cell counts were also seen to decrease significantly after cessation of breast-feeding (Table 1), whilst in atopic infants, the number of bacteroides decreased during the follow-up from 8.27 (7.60–8.94) to 7.46 (6.37–8.54), P=0.02. The numbers of other bacterial genera detected by FISH were not influenced by further diversification of the diet during the follow-up period in atopic infants (data not shown). No significant changes following weaning were observed in the plate culture assays (Fig. 1).
In this study we demonstrated that Gram-positive species were less frequently among the most predominant culturable aerobic bacteria in the fecal flora of atopic than healthy infants. Interestingly, the recent in vitro study by Hessle and co-workers  indicated that Gram-positive bacteria (enterococcal, staphylococcal and streptococcal species among them) are more potent inducers of Th1 cytokines interleukin (IL)-12 and interferon-γ than Gram-negative bacteria, which correspondingly are more potent inducers of Th2 cytokine IL-10. These studies taken together raise the question as to whether a sufficient Gram-positive microflora could have a greater importance than Gram-negative flora in providing counter-regulation for the immune system in early infancy when it is universally Th2-skewed. Further studies are needed to answer this question; at the present juncture there is indication that exposure to Gram-positive bacteria can promote the formation of Th2-mediated oral tolerance  and prevent atopic sensitization , but it has not been assessed whether similar results could also be obtained by Gram-negative species. After the first days of life aerobic bacteria account for only a fraction of the total bacteria present in the gut , therefore further studies should also assess whether the balance between Gram-positive/Gram-negative anaerobic biota of atopic infants differs from that of healthy infants. It should be noted, however, that the first bacteria inhabiting the gut must be capable of oxidative metabolism and recent studies draw emphasis towards the importance of the initial bacterial exposure in the development of gut barrier and tolerogenic responses [36,37]. It has also been shown that the initial colonizers persist longer as part of the microflora than those adopted at a later stage and thus the observations made in this study may reflect the initial colonization events [38,39].
The plate culture analyses carried out in this study covered the most common microbial species that are known to cause gastrointestinal symptoms in children. In addition, many of the bacteroides species enumerated by FISH are putrefactive by reason of their proteolytic activities and clostridia species potentially pathogenic by toxin formation . Bifidobacteria and lactobacilli, on the other hand, are considered beneficial to the host and can restrain the harmful microflora [40,41]. Our data indicate that these microorganisms are not aberrantly present in atopic infants as hypothesized . This suggests that as individual species or genera, potentially pathogenic microbes or their antagonists do not have a distinctively important role in preventing or promoting the initial development of atopy. However, it is also possible that aberrance has existed during the first weeks of life but is no longer detectable at this age as suggested by our previous study on infants at 3 weeks and 3 months of age . Also the relatively small study population, although very well defined, does not allow an indisputable conclusion.
Some deviation from the microbiota of healthy infants was associated with specific manifestations, i.e. gastrointestinal symptoms and risk of developing hypersensitivity to cow milk, suggesting that the composition of the microflora may be associated with the extent of allergic sensitization. The existence of such association is supported by the correlations found among the atopic infants between the bacterial groups enumerated by FISH and serum total IgE or SCORAD scores after weaning. Whether the presence of these bacterial groups is affected by the manifestations of atopy or whether they actually have atopic sensitization preventing or promoting properties can only be speculated upon. The correlations that were observed after weaning could also be artifacts resulting from association between diversification of the diet and severity of atopic sensitization, in particular as the correlations involving clostridia and lactobacilli/enterococci are not in agreement with previous studies [11,42,43].
The response of the digestive microflora to weaning was analyzed separately in atopic and healthy infants due to different formulae used for the intervention. Our study is in agreement with previous findings demonstrating the bifidogenic nature of breast milk in comparison to solid foods  and extends the finding to allergic infants weaned to amino acid derived formula among which the decrease in the number of bifidobacteria was particularly clear. As there is strong evidence supporting the beneficial effects of bifidobacterial flora, it may be worthwhile to consider bifidobacterial supplementation alongside weaning to such formulae [40–42]. In contrast with healthy infants, the total cell counts did not decrease in response to weaning among atopic infants, which may be explained by the lower initial counts at the beginning of weaning. On the other hand, the decrease in the number of bacteroides which was unique to atopic infants could be either a response to the intervention by the amino acid derived formula or alleviated manifestations of atopy with aging (data not shown).
In conclusion, the results warrant further studies to assess the possible significance of the balance between colonization by Gram-positive and Gram-negative species in the prophylaxis of atopy. The results also tentatively indicate that the presence of any single species or genera does not have a unique importance in the initial development of atopic disease but may affect or be affected by the extent of atopic sensitization. Further studies are needed to uncover the correlation between microflora and atopic sensitization.
We gratefully acknowledge Tuija Poussa, MSc, for able statistical consultation. The project was supported by the Academy of Finland, the Development Center of Finland (Tekes) and the Medical Research Fund of Tampere University Hospital. E.A. was supported by a scholarship from the Center for International Mobility (CIMO).