The microbial colonization of the infant GI tract is a remarkable episode in the human lifecycle. Every time a human baby is born, a rich and dynamic ecosystem develops from a sterile environment. Within days, the microbial immigrants establish a thriving community whose population soon outnumbers that of the baby's own cells. The evolutionarily ancient symbiosis between the human GI tract and its resident microbiota undoubtedly involves diverse reciprocal interactions between the microbiota and the host, with important consequences for human health and physiology. These interactions can have beneficial nutritional, immunological, and developmental effects, or pathogenic effects for the host [2
This study began with the development of a DNA microarray with nearly comprehensive coverage of the bacterial taxa represented in the available database of SSU rRNA gene sequences. Our microarray design and experimental methods were based on lessons learned in the validation of a less comprehensive SSU rDNA microarray [46
]. These previous experiments enabled us to optimize our methods for computational prediction of SSU rDNA hybridization behaviors, and to develop an experimental protocol that maximized hybridization specificity. The excellent concordance in the measurements of individual taxa determined using the new microarray design in comparison with sequencing results from corresponding SSU rDNA clone libraries () suggests that these design principles hold true for this platform across a diversity of taxa and give us confidence in both the comprehensiveness and accuracy of the results obtained with our new microarray probe set. It is important to note, however, that our methods of array design and analysis are imperfect and still evolving. Several of the unexpected species predicted by the microarray to be present in one or more samples could not be corroborated by sequencing. In most of these cases, sequence analysis of the sample(s) in question revealed that low-level cross hybridization of a highly abundant species was responsible for the false-positive prediction, a result that will be taken into consideration in future rounds of array design and analysis.
We used this microarray in a detailed, systematic, and quantitative study of bacterial colonization of the newborn human GI tract. We used freshly collected stool samples as surrogates for samples taken from the lumen and mucosa of the colon. Although there are undoubtedly differences in the population profiles of stool samples and corresponding mucosa, we found in a previous study that the profiles are nonetheless remarkably consistent—sufficiently so that individual stool samples can readily be matched to colonic biopsy samples from the same individual, based on the similarity in their bacterial profiles [15
]. Thus, we believe that the results of our temporal analysis of the bacterial populations in infant stool samples provide a useful window on the resident colonic microbiota.
In view of the importance of the symbiosis between human host and gut commensals for both human host and microbial colonist, it would be easy to imagine that the program of microbial colonization of the neonatal GI tract would have evolved under strong selective pressure, acting on both the intestinal niche and its microbial colonists, to be highly deterministic and stereotyped. We might have expected that a highly restricted group of co-evolved commensals would be exceptionally well adapted to this environment and consistently dominate the colonization process in a stereotyped fashion. Indeed, the bacteria that we found in infant and adult feces, presumably reflecting the colonic microbiota, were largely restricted to only a small subset of the bacterial world—Proteobacteria, Bacteroides, Firmicutes, Actinobacteria, and Verrucomicrobia. Yet, surprisingly, we found that in the first days to months of life, the microbiota of the infant gut, and the temporal pattern in which it evolves, is remarkably variable from individual to individual. The seemingly chaotic progression of the early events in colonization, and the similarity in bacterial composition of some early infant samples to breast milk or vaginal swabs, suggests that the bacterial population that develops in the initial stages is to a significant extent determined by the specific bacteria to which a baby happens to be exposed. Notably, these maternal “signatures” did not persist indefinitely, as evidenced by our failure to find a significantly higher correlation of the overall taxonomic profiles of baby/parent pairs from the same household versus different households.
An important exception to the tale of individuality and uniqueness in the early profiles was the remarkable similarity of the temporal profiles of the fraternal twins (babies 13 and 14) ( and ). These twins shared both a common environment and approximately 50% genetic identity, making it impossible to determine from this study to what degree each of these commonalities is responsible for their similar colonization patterns. However, evidence from this and other studies suggests that the shared environment is a major factor. One argument in favor of this view is the lack of comparable similarity in the microbial communities of other pairs that also share 50% genetic identity, including mother:baby, father:baby, and sibling:baby (unpublished data), although this dissimilarity may be due in part to their differing stages in development. Another argument in favor of a strong environmental influence is the coincidental transient appearance of specific organisms in both twins—it is hard to imagine that the appearance of a particular microbe on a particular day could be genetically programmed. Our final argument rests on evidence from a previous study that the microbiota of genetically equivalent families from a cross of inbred mice was more similar among members of the same “household” (mother and offspring who share a cage) than between households [1
The definition of a “healthy” intestinal microbiota encompasses a remarkable diversity of community profiles in the first 6 mo of life. Although diverse and idiosyncratic in the early months, these microbial communities became progressively more similar to one another (A), converging toward a generic adult-like profile (B) characterized by a preponderance of Bacteroides
and Firmicutes, common occurrence of Verrucomicrobia, and very low abundance of Proteobacteria and aerobic Gram-negative bacteria in general. We hypothesize that the earliest colonization events are determined to a large extent by opportunistic colonization by bacteria to which a baby is exposed in its environment. Common environmental exposures are likely to include the maternal vaginal, fecal, or skin microbiota, as is suggested by the observed similarity of some infants' early stool microbiota to these samples, which is consistent with previous evidence of vertical transmission of microbes [33
]. The diversity and variation would thus reflect the corresponding individuality of these accidental exposures. Over time, however, the fitness advantage of the taxa that typically dominate the adult colonic microbiota apparently overcomes the initial advantage of early-colonizing opportunists that are less well adapted to the intestinal environment. In addition, progressive changes in the gut environment, due to intrinsic developmental changes in the gut mucosa, transition to an “adult” diet, and the effects of the microbiota itself [44
], may impose increasingly stringent selection for the most highly adapted bacteria. Thus, despite the unexpectedly chaotic early months, the establishment of the gut ecosystem in human infants turns out after all to follow a conserved, conventional program.
The transformation of the intestinal microbiota to an adult-like pattern implicitly involved replacement of species found in infants, but rarely in adults, with species characteristically found in the adult colon. One potential driving force for such a demographic change might be that the adult-like community members eventually dominate by virtue of their greater ability to establish themselves stably and irreversibly once they colonize a host. We looked for evidence of this differential “stickiness” by comparing the autocorrelations over time of the abundance of each “species” (see Materials and Methods
). We found no clear evidence that the species characteristic of adult microbiota were able to establish more intrinsically stable colonization than the species characteristic of infant microbiota.
We and others have found that the individual-specific characteristics of the bacterial microbiota of adults are stable, in the sense that they remain consistently more similar within an individual over time than between individuals, for periods of a year or more, and one of the striking results of this study was the identification of relatively stable, individual-specific patterns of bacterial colonization even in the first weeks and months of life. These observations raised the interesting possibility that opportunistic colonization events in early infancy might play a significant role in defining the distinct characteristics of the same individual's microbiota into adulthood. We looked for evidence of this by comparing the intraindividual and interindividual correlations of bacterial profiles at 1 or 2 mo and 1 y, and found no significant difference (unpublished data). Thus, although these results certainly do not exclude the possibility that early colonization events play an important role in determining the adult microbiota, there does not appear to be a strong, direct correlation between the two.
Our results and conclusions differ considerably from many previous reports in several respects. One notable discrepancy between our studies and many others was the relatively low frequency and abundance of Bifidobacteria in the fecal microbiota at any age from birth to adulthood. Bifidobacteria have received disproportionate attention, in part because of their reputed beneficial effects, and many studies have reported (and reviews have repeated) that the microbiota of breast-fed infants is dominated by Bifidobacteria [17
]. We were thus surprised by, and initially skeptical of, the apparent paucity of Bifidobacteria in nearly all of our samples, and took steps to verify that our results were accurate. Bifidobacteria-specific qPCR corroborated the conclusion from our microarray results that Bifidobacteria were rarely major constituents of the GI microbiota, at least in this study population, and that in most babies, they did not appear until several months after birth, and thereafter persisted as a minority population. Although it is conceivable that there are geographical or demographic differences in the prevalence of Bifidobacteria, we suspect that the emphasis on Bifidobacteria in studies and reviews of the infant GI microbiota may be out of proportion to its prevalence, abundance, and relevance to health.
The results presented here suggest numerous future avenues of research. An intriguing feature of the bacterial population dynamics was the occurrence of abrupt shifts punctuating intervals of relative stability. Except in one case (the antibiotic treatment of baby 8), we could not readily identify a strong candidate for the cause of the shifts we observed. Some possibilities include bacteriophage outbreaks that can selectively decimate a dominant taxonomic group [52
]; stochastic, opportunistic invasion of a metabolic or anatomic niche by a fitter species; and subtle developmental or diet-induced changes in the gut environment tipping the fitness balance in the population. Other important avenues for future research will be comparing the composition and evolution of microbial communities encountered in these healthy babies to those of preterm or otherwise unhealthy babies and to investigate the effects of antibiotics, diet, and mode of delivery on the development and evolution of these communities. Even though the healthy babies in this study assumed a large range of microbial community profiles, they were similar in several respects, most notably in the major contributing phyla, in the acquisition of certain key phyla over time, and in the relative stability of their profiles over time. It may be that in other states of health or disease, we will find either species or groups that are novel to this environment, or unusual combinations of this newly defined set of “usual” species.
Importantly, although we have shown that the gut microbiota becomes increasingly stereotyped over the first year, it is clearly established that stable interindividual differences persist even in adults [15
]. When and how do these stable “intrinsic” characteristics of the microbiota of each individual develop? How long do they persist? How do the differing stabilities of colonization by different bacteria relate to their microanatomic (e.g., crypt vs. villus vs. mucous layer) or metabolic niches? Identifying the environmental and genetic factors that determine the distinctive characteristics of each individual's microbiota, and determining whether and how these individual specific features affect the host's physiology and health, will be an important goal for future investigations, in which the microarray described in this study will be a useful tool.