Alteration of the developmental program did not occur in the course of association of GF mice with indigenous microbes derived from rat or human feces (6). SFB derived from mice and rats did not cross-colonize in rats and mice, respectively (7). There appears to be a strict limit to the interaction between the host animal and the intestinal bacteria, in accord with the concept of "autochtonous bacteria" proposed by Dubos et al. more than 30 years ago (8). Does the association of GF mice with B. thetaiotaomicron induce class II expression, expansion of IEL and IgA-producing cells, and so on after the expression of an 1,2 fucosyltransferse? What is the original host of this bacterium, mouse or human? A GDP-fucose:asialo GM1 1,2 fucosyltransferase was induced in GF mice on injury to the small intestine (9). In our study, 1,2 fucosyltransferase induction was the first event. We have no evidence, however, to suggest that this fucosylation initiates the developmental program of the intestinal mucosa, including the components in the lamina propria.

Yoshinori Umesaki
Yasushi Okada
Akemi Imaoka
Hiromi Setoyama
Satoshi Matsumoto
Yakult Central Institute for
Microbiological Research,
Yaho 1796, Kunitachi,
Tokyo 186, Japan
E-mail: hfg00420{at}niftyserve.or.jp

REFERENCES AND NOTES

1. L. Bry, P. G. Falk, T. Midtvedt, J. I. Gordon, Science 273, 1380 (1996)[Abstract].
2. Y. Umesaki, T. Sakata, T. Yajima, Biochem. Biophys. Res. Commun. 105, 439 (1982)[Medline].
3. Y. Umesaki, Y. Okada, S. Matsumoto, A. Imaoka, H. Setoyama, Microbiol. Immunol. 39, 555 (1995)[Medline].
4. J. Snel et al., Int. Syst. Bacteriol. 45, 780 (1995).
5. H. Setoyama et al., unpublished data.
6. Y. Okada, S. Matsumoto, A. Imaoka, H. Setoyama, Y. Umesaki, Infect. Immun. 62, 5442 (1994)[Abstract].
7. G. W. Tannock, J. R. Miller, D. C. Savage, Appl. Environ. Microbiol. 47, 441 (1984)[Medline].
8. C. P. Ddubos, R. W. Schaelder, R. Costelo, P. Hoet, J. Exp. Med. 10, 948 (1965).
9. Y. Umesaki and M. Ohara, Biochim. Biophys. Acta 1001, 163 (1989)[Medline].
10. We thank T. Osawa of the Yakult Central Institute for Microbiological Research for his reading of the manuscript.

Response: Development of the mouse small intestine is often viewed in terms of the cytodifferentiation of its endoderm that occurs in late fetal life, or the formation of its crypt-villus units, which is completed during the first three postnatal weeks. Umesaki et al. emphasize the importance of having a broader vision of gut development. We agree. A "trialogue" between the intestinal microbiota, the self-renewing intestinal epithelium, and the diffuse gut-associated lymphoid tissue (GALT) is probably critical in forming and maintaining this dynamic ecosystem. Studies by Umesaki et al. provide strong evidence that a component of the normal microbiota can influence the composition of the diffuse GALT. Colonization with the B. thetaiotaomicron-type strain, VPI-5482, is associated with similar composition changes. For example, an influx of IgA⁺ B cells occurs after exposure to this organism.

One should consider the diffuse GALT's composition, but also its spatial complexity, which has been hard to characterize because markers are difficult to detect with conventional immunohistochemical methods. More sensitive techniques (1) have allowed us to examine these features in mice that contain a normal (conventional) microbiota. For example, T cells populate the intraepithelial and lamina propria compartments in crypts and villi, while T cells are limited to the villus epithelium. T_H1 and T_H2 cells appear predominantly in the lamina propria of the villus (2).

Like the diffuse GALT, components of the microbiota are arranged asymmetrically along crypt-villus units: SFB attach to epithelial cells located in the upper two-thirds of the villus. Histochemical stains of unperfused small intestines obtained from specified pathogen-free conventional mice suggest that crypts are not colonized by this or other bacterial species. The asymmetric distribution of the microbiota may serve to organize components of the diffuse GALT. Conversely, the diffuse GALT may influence the spatial organization of the microbiota.

The diffuse GALT also communicates with the intestinal epithelium: Mice that lack T cells have fewer crypt epithelial cells and slower epithelial cell migration up the villus (3). Contaminating adult GF mice with SFB or VPI-5482 reveal another component of this trialogue: communication between the microbiota and the gut epithelium. VPI-5482, which was originally recovered from a human, signals the epithelium to induce and sustain 1,2 fucosyltransferase gene transcription and production of fucosylated glycoproteins and glycolipids. This is not a nonspecific response of the epithelium to bacterial colonization. Monocontamination of GF NMRI mice with two other anaerobes that normally colonize the mouse and human intestine, Peptostreptococcus micros and Bifidobacterium infantis, produces no detectable effect on fucosylated glycoconjugate production (4).

Unlike SFB, signaling occurs without direct bacterial attachment to enterocytes (5). Signaling depends on the ability of the organism to use fucose as a carbon source (5). We recently found that the B. thetaiotaomicron genome contains a locus analogous to the Escherichia coli fucose utilization regulon (6). A Tn4351 insertion renders the Fu-4 strain of B. thetaiotaomicron unable to use fucose and unable to signal enterocytes to produce fucosylated glycoconjugates. The site of insertion is the open reading frame of one of the genes within this locus (7). Monocontamination of GF mice with isogenic strains of B. thetaiotaomicron that contain engineered disruptions of each gene in the regulon should provide clues about the nature of the signal that emanates from this metabolic pathway.

To induce and sustain fucosylated glycoconjugate production in enterocytes, VPI-5482 must reach a critical population density (5). This requirement may reflect secretion of a soluble bacterial factor that produces a concentration-dependent response in the epithelium. Or there may be a density-dependent change in the metabolic properties of the bacteria that affects production of a signaling molecule--a process known as "quorum sensing" (8). In the mammalian gut, where there is a highly complex society of microorganisms, secreted signaling molecules may allow communication between (and within) bacterial species. Multiple species may cooperate to generate a concerted signal that establishes a mutually beneficial niche. Such density-dependent signaling systems may also interfere with one another if a similar set of molecules is used by different species to modulate distinct metabolic pathways. This type of interference could allow the microbiota to prevent the encroachment of pathogens. If such encroachment occurs, the response of the host may depend on the relative locations of the pathogen, components of the diffuse GALT, and members of various intestinal epithelial lineages--factors that likely are influenced by the trialogue.

J. I. Gordon
L. V. Hooper
L. Bry
Washington University School of Medicine,
St. Louis, MO 63110, USA
E-mail: jgordon{at}pharmdec.wustl.edu
T. Midtvedt
P. G. Falk
Karolinska Institute,
S-17177 Stockholm, Sweden

REFERENCES

1. M. N. Bobrow, T. D. Harris, K. J. Shaughnessy, G. J. Litt, J. Immunol. Methods 125, 279 (1989)[Medline]; K. S. Shindler and K. A. Roth, J. Histochem. Cytochem. 44, 1331 (1996)[Abstract].
2. M. S. McNevin, E. Garabedian, J. I. Gordon, unpublished observations.
3. H. Komano, et al., Proc. Natl. Acad. Sci. U.S.A. 92, 6147 (1995)[Medline].
4. L. Bry, P. G. Falk, T. Midtvedt, J. I. Gordon, unpublished observations.
5. ___, Science 273, 1380 (1996)[Abstract].
6. Z. Lu and E. C. C. Lin, Nucleic Acids Res. 17, 4883 (1989)[Medline]; Y.-M. Chen, Z. Lu, E. C. C. Lin, J. Bact. 171, 6097 (1989).
7. L. Hooper and J. I. Gordon, unpublished observations.
8. C. Fuqua, S. C. Winans, E. P. Greenberg, Ann. Rev. Microbiol. 50, 727 (1996); D. Kaiser, Science 272, 1598 (1996)[Medline].