Rabbit M cells and dome enterocytes are distinct cell lineages

Epithelial cells are frequently exposed to pathogens that can proliferate in the mucosal environment. Secreted mucus is believed to be the main way of trapping and eliminating most of these pathogens, but mucosal immunity is necessary to prevent their action on epithelial surfaces completely. This is achieved by specific lymphoid tissues that underlie the different mucosae. In gut-associated lymphoid tissues (GALTs), sampling of luminal antigens is achieved by specialized epithelial cells, i.e. M cells, that bind and rapidly transport pathogens from the apical plasma membrane to the basolateral invagination, which contains enough lymphoid cells to trigger an immune response (for reviews, see Neutra et al., 1996; Niedergang and Kraehenbuhl, 2000).

In the small intestine, pluripotent epithelial stem cells reside deep within the crypts and, through commitment and proliferation of lineage precursors, give rise to progressively differentiating absorptive enterocytes and mucous cells, which migrate up towards the crypt mouth and along the villi to be shed from the tips a few days later (for reviews, see Bjerknes and Cheng, 1999; Cheng and Leblond, 1974; Potten and Loeffler, 1990). In GALT, M cell origin and differentiation steps are still a matter of controversy. It has been postulated that, like epithelium cells from the small intestine, they arise and segregate directly from stem cells of crypts. Bye et al. (Bye et al., 1984) have found that transitional forms of M cells at the neck of crypts and upward share some morphological and functional properties (i.e. binding and transport of inert or living particles) with fully differentiated M cells. These findings are supported in mouse and rabbit studies by other authors using different markers or morphological parameters (Clark et al., 1993; Gebert et al., 1999; Gebert et al., 1992; Gebert and Posselt, 1997; Giannasca et al., 1994; Pappo, 1989; Sicinski et al., 1986). Conversely, several authors have challenged the lineage hypothesis. They propose that mature enterocytes could switch to the M cell phenotype under the influence of lymphoid cells from the underlying follicles (Kerneis et al., 1997; Savidge, 1996; Smith and Peacock, 1980) or of pathogens from the lumen (Borghesi et al., 1996; Meynell et al., 1999; Savidge et al., 1991), followed by de-differentiation to the enterocyte phenotype at the apex of domes (Sierro et al., 2000). This has been strengthened by the nature of the role of follicular lymphoid cells both in vivo, in the induction of new Peyer’s patches in mouse duodenal mucosa, and in vitro in the acquisition of the M cell phenotype by polarized Caco-2 cells (Kerneis et al., 1997).

We have decided to revisit the M cell differentiation pathway and focus on two important questions that remain unsolved. First, do pre-M cells, previously detected by several authors, originate directly from proliferative cells of the crypts, or could they derive from still poorly differentiated enterocytes of the upper part of the crypts or higher in the dome? Second, are pre-M cells sufficiently numerous to represent the total pool of mature M cells of the median part of the dome or is there a need for conversion of enterocytes into M cells?

Several markers can now be considered to be reliable for selective labeling of M cells in rabbit GALT (Gebert and Hach, 1993; Gebert et al., 1992; Jepson et al., 1993a; Jepson et al., 1992). More recently, we have found new mucin-like markers that are selective for rabbit M cells or dome enterocyte apical surfaces (Lelouard et al., 1999; Lelouard et al., 2001; Maury et al., 1995), and allow simultaneous visualization of M cells and enterocytes in entire domes. We have used multi-labeling to detect M cell and enterocyte precursors in three different GALTs and have designed a rabbit GALT epithelium dissociation method with ethylene diamine tetra acetic acid (EDTA) (Lelouard et al., 1999) that allows visualization of labeled cells continuously from the flank of the dome to the bottom of crypts. We have also assessed micro-dissected domes to obtain large surface views of the follicle associated epithelium (FAE) and cryostat sections.

We identified M cells in the mid-crypt area, among rapidly dividing epithelial cells, and showed that the spatial organization of M cells and dome enterocytes was already established at the neck of crypts. These observations strongly favor the hypothesis that both cell types constitute distinct cell lineages derived from common stem cells.

New-Zealand albino rabbits weighing 2 to 3 kg were obtained from the Institut National de la Recherche Agronomique (Montpellier, France). Animals were housed and cared for according to French regulation 87-848 and to EEC-L358.

Monoclonal antibodies (MAbs) to rabbit intestinal glycoproteins, namely MAb 58 and MAb 214, have previously been described and characterized (Lelouard et al., 1999; Lelouard et al., 2001), and are detailed in Table 1. Goat anti-mouse IgG coupled to FITC or TRITC were from Biosys (Compiègne, France); mouse anti-vimentin (clone V9) conjugated to Cy3, lectins VVA-FITC from Vicia villosa and WGA-RITC from Triticum vulgaris were obtained from Sigma (St Louis, MO). The BU-1 MAb was kindly provided by the Mayo Foundation (Rochester, MN).

Epithelium sheets from rabbit appendix, distal Peyer’s and caecal patches were detached with EDTA as previously described (Lelouard et al., 1999). Briefly, tissues were incubated in dissociation buffer containing 200 mM sucrose, 40 mM EDTA and protease inhibitors. After 1 to 2 hours, epithelium sheets were released and fixed for 1 hour with 2% formaldehyde in 100 mM potassium phosphate buffer, pH 7.4. Epithelial sheets were further micro-dissected under a stereo-microscope (see Fig. 1). Sheets were permeabilized with 0.2% Triton X-100 in 50 mM Tris-HCl buffer, pH 7.4, and labeled in a droplet of MAb followed by FITC- or RITC-coupled secondary antibodies. Double-labeling with two MAbs was performed by indirect labeling with the first MAb and FITC- or RITC-coupled secondary goat anti-mouse antibody, followed by incubations with an unrelated MAb, then 4,6-dichlorotriazinyl-aminofluorescein (DTAF)-coupled MAb 58 or Cy3-coupled anti-vimentin MAb. Lectins were used at 5 μg/ml concentration in the same buffer.

Whole domes were micro-dissected from fixed tissues, permeabilized and labeled as above. Sections (8 μm) were cut from fixed tissues in a Reichert 2700 cryostat permeabilized and labeled as above. Labeled epithelial sheets, domes and sections were mounted with Mowiol containing 2.5% 1.4-diazabicyclo[2.2.2]octane (DABCO). Samples were observed under a Leica TCS 4D confocal microscope (Wetzlar, FRG) equipped with an argon-krypton laser. Epithelial sheets, pictures of whole domes and cryostat sections were maximal projections of 5 to 60 focal planes, depending on the thickness of the sample. All experiments with Peyer’s patches were performed on the two distal ones, which gave the most reproducible labeling with several markers (Lelouard et al., 2001).

20 ml of 5 mg/ml bromodeoxyuridine (BrdU) in sterile phosphate-buffered saline was administered intraperitoneally to rabbits. The animals were killed 1 hour later and the appendix, distal Peyer’s and caecal patch epithelium were detached with EDTA as described above. The first epithelial sheets were obtained approximately 1 hour after the beginning of dissociation and were immediately fixed at room temperature with 2% formaldehyde in 100 mM phosphate buffer, pH 7.4, for 1 hour. Sheets were rinsed and permeabilized for 5 minutes in Tris-saline containing 0.05% Tween 20. Labeling was then performed with MAb BU-1, which is specific to BrdU (Gonchoroff et al., 1985; Gonchoroff et al., 1986).

We recently described a method to dissociate FAE from rabbit appendix using EDTA (Lelouard et al., 1999). The released epithelial sheets frequently displayed a ring shape, which was formed by one dome with its adjacent surrounding colonic epithelium (Fig. 1A). The top of the dome was often torn off upon shaking, thus producing the ring shape. Crypts associated with the dome and adjacent colonic epithelia were generally intact. It was, thus, possible to unfold the epithelial sheets by gently pulling with tweezers (Fig. 1B), and to further dissect them and isolate the dome from the colonic epithelium by tearing them apart (Fig. 1C). Most crypts remained connected to the dome. A single dome was usually surrounded by 40 to 50 crypts (Fig. 1C). These crypts were short, ranging from 45 to 70 μm long, and often branched (see DIC pictures in Fig. 2). In caecal patch epithelium detached with EDTA, crypts were organized in several rows (Fig. 1D). They were grouped into three or four, sometimes branched (Fig. 1E), and much longer than in the appendix (170 to 230 μm). EDTA treatment of Peyer’s patches released dome epithelial sheets whose crypts were very often torn off. This was possibly due to the fact that villi surrounding every dome were separated from each other, thus weakening the epithelium that broke during dissociation, contrary to other GALTs. However, a few rare domes with crypts were sometimes observed.

To enhance the reliability of recognition and selective discrimination of M cells and dome enterocytes, especially in crypts, we used a combination of several markers on the three different GALTs described above. These markers, labeled by MAbs or lectins, are summarized in Table 1. They gave complementary labeling of M cells and dome enterocytes (see below).

The different epithelial lineages began their differentiation in the proliferative area of the villus crypts. It was therefore important to localize the first pre-M cells relative to this area in dome crypts. Proliferative cells were detected by in vivo 1 hour pulse labeling with BrdU, and M cells were identified by vimentin labeling (vim⁺ cells), which is characteristic of M cells in rabbit intestinal epithelium (Gebert et al., 1992; Jepson et al., 1992). Labeling was performed in isolated permeabilized FAE. In the appendix, pulse labeling with BrdU indicated that the proliferative areas were located in crypts, but also at the crypt neck (Fig. 2A-D). The first vim⁺ cells were also located in these areas (Fig. 2C,D), and some of them were co-labeled with BrdU (Fig. 2D, inset). Many epithelial cells weakly expressed cytoplasmic vimentin at the crypt neck. Vimentin-labeled cells rapidly increased in number to finally reflect the proportion of mature M cells present in the mid-part of the dome. Along the crypt-dome axis, vimentin expression in each cell increased and the intermediate filament structural arrangement was progressively organized, from the apico-basal perinuclear bundles in the crypt to the typical cage structure surrounding the lymphoid cell-containing pocket (Fig. 2E-G). In Peyer’s patches, the first vim⁺ cells were also found in the proliferative area of crypts, as illustrated in Fig. 2H, and co-labeling with BrdU was evident in several cells (Fig. 2H inset). In the caecal patch, the first vim⁺ cells were also found at mid-crypt in spots labeled by BrdU (Fig. 2I). From mid-crypt upwards, vim⁺ cells represented an important subpopulation of cells. At the neck of crypts, apical VVA co-labeling of vim⁺ cells confirmed that they were M cells (not shown).

As pre-M cells arise directly from mitotic cells of the crypts, it appeared important to investigate whether the proportion of pre-M cells was similar to that of mature M cells in the dome flanks. The first approach involved serial cryostat cross-sectioning of intestinal tissue. Epithelium in crypts and villi from normal ileum was devoid of vimentin labeling (Fig. 3A). On the dome side of crypts from Peyer’s patches, starting from the proliferative BrdU-labeled area (Fig. 3B), numerous regularly spaced vim⁺ cells were detected in a proportion similar to that observed for M cells on the dome (compare Fig. 3C with 3D). MAb 58, the only known apical marker of M cells in rabbit appendix and Peyer’s patch (Table 1), confirmed that these cells were M cells (vim⁺58⁺ cells) (Fig. 3C,D). A few vim⁺58⁺ cells were also detected on the villus side of crypts (Fig. 3C) and on the adjacent villus (Fig. 3D), as in the appendix and caecal patches (not shown). This indicated that even M cells recently reported on villi facing the dome (Borghesi et al., 1999) arise, already pre-differentiated, from dome crypts.

We looked at the relative distribution of M cells and enterocytes to determine whether it was the same at the base of the dome and at mid-height, where lymphoid cells penetrate the epithelium and M cells are known to be functional. We performed double-labeling on micro-dissected domes with MAb 58 and MAb 214, the only markers specific to the apex of dome enterocytes in all rabbit FAE (see Table 1). At the crypt neck of Peyer’s patches, round 214⁺ dome enterocytes and 58⁺ M cells were already labeled and never co-localized (Fig. 4A). The spatial organization of M cells and enterocytes at the base and on the lateral side of the dome was similar, three to four M cells surrounding each enterocyte (Fig. 4A,B). M cells of Peyer’s patches are known to have varying apical surfaces (Jepson et al., 1993b). We looked for correlations between the apical surface of M cells and the size of vimentin-delimited pockets. Large vimentin-outlined pockets corresponded to large apical M cell surfaces, whereas small pockets corresponded to small apical M cell surfaces (Fig. 4C). Staining of M cells by MAb 58 even revealed the existence of such small apical surfaces (Fig. 4B), which were undistinguishable by scanning electron microscopy (not shown).

In appendix, M cells and enterocytes were distributed in a very regular mosaic pattern from the base to a short distance from the top of the dome, with every M cell surrounded regularly by 5 to 7 dome enterocytes (Fig. 4D). In caecal patches, double-labeling of enterocytes with MAb 214 and M cells with WGA revealed that the spatial arrangement between M cells and enterocytes was already established at the crypt neck (not shown) as in the appendix and Peyer’s patches.

The data presented in this study indicate that, in rabbit FAE, M cells and dome enterocytes represent distinct pre-programmed lineages already separated before the neck of crypts. The two lineages would be derived from common stem cells, as are epithelial cells in small intestine (Cheng and Leblond, 1974). This is substantiated by the following findings:

(1) a vimentin filamentous cytoplasmic network already appeared in the zone of crypts where rapidly dividing cells incorporate BrdU, in the three GALTs studied;

(2) the vimentin expression pattern was approximately one out of two cells in GALT crypts, in continuity with what is observed in the dome flanks, whereas it was absent in crypts of the small intestine;

(3) the mosaic-like relative distribution of M cells and dome enterocytes at the neck of crypts was similar to that observed on the dome side.

These findings were achieved using a newly designed method for labeling whole FAE epithelium dissociated with EDTA (Lelouard et al., 1999) combined with labeling of entire micro-dissected domes and cryostat sections. This allowed us to visualize markers deep in the crypts, down to the zone of rapidly dividing cells pulse labeled with BrdU, and extensively revealed the relative distribution of M cells and enterocytes from the crypts in three different GALTs.

Detection results depend upon the selectivity of the markers used. Vimentin has been widely used as an excellent marker for rabbit M cells on dome flanks (Gebert et al., 1992; Jepson et al., 1992). In the crypts, vim⁺ cells were not enterocyte precursors, as vimentin labeling was absent in the crypts of small intestine villi. The other M cell markers used included the glycosidic epitopes recognized by MAb 58 (Lelouard et al., 1999) and lectin receptors (Gebert and Hach, 1993; Jepson et al., 1993a), depending on the GALT under study. In the FAE, they clearly labeled M cells and displayed a pattern complementary to MAb 214-labeled enterocytes (Lelouard et al., 1999; this study). They always co-labeled vim⁺ cells as soon as their expression began in the crypt.

Crypts are known to be the region where stem cells divide and segregation of cell lineages takes place in the small intestine (Cheng and Leblond, 1974; Gordon, 1989; Potten and Loeffler, 1990). Previous studies have favored a separate cell lineage for M cells, different from that of enterocytes, originating from crypts and followed by a progressive transition from immature to mature M cells (Bye et al., 1984; Gebert et al., 1999; Gebert et al., 1992; Gebert and Posselt, 1997; Giannasca et al., 1994; Sicinski et al., 1986). The presence of M cell markers that outline radial strips at the base of domes starting from the mouth of crypts suggests a clonal origin for M cells (Clark et al., 1993; Gebert et al., 1999; Gebert and Posselt, 1997). Stronger evidence of early committed M cells is the presence of immature M cells in the upper part of crypts, as recognized by their morphology in electron microscopy and their binding to UEA1 lectin (Gebert et al., 1999). In this paper, we monitored M cell precursors deeper in crypts, down to the zone of rapidly dividing cells, where vimentin-labeled pre-M cells already represent about half the population, as in the flanks of dome where mature M cells lie. The presence of pre-M cells in the dividing zone is crucial and favors the hypothesis of an early commitment of M cell lineage, before terminal differentiation occurs in the upper crypt.

It has been proposed that M cells are derived from differentiated dome enterocytes via an induction step, owing to the influence of lymphocyte penetration into the epithelial monolayer (Kerneis et al., 1997; Savidge, 1996) and/or by an interaction with bacteria (Meynell et al., 1999), and that M cells would revert back to enterocyte phenotype at the top of domes (Sierro et al., 2000). Our results are not in line with such a conversion. The model of conversion of Caco2 cells into M cells (Kerneis et al., 1997) represents the only in vitro model currently available for studying the biology of M cells and interactions between epithelial cells, bacteria (Schulte et al., 2000) and lymphoid cells. The fact that the Caco2 cell line can acquire an M cell phenotype by interaction with lymphoid cells from mouse Peyer’s patches does not necessarily mean that fully differentiated enterocytes do the same in vivo. It is known that Caco-2 cells retain crypt cell properties (Grasset et al., 1985; Grasset et al., 1984), and display multipotent phenotypes (Engle et al., 1998), which could indicate that this cell line might rather behave as intestinal crypt cells that could still differentiate, depending on the local environment. The fact that Kerneis et al. (Kerneis et al., 1997) succeeded in inducing Peyer’s patch formation in vivo, by injecting lymphoid follicular cells in mouse duodenum, reveals the crucial role of the local lymphoid environment in the early commitment of progenitor epithelial cells and GALT formation.

Our data highlight a striking and unchanged spatial organization of M cells and dome enterocytes along the crypt-dome axis, with a constant ratio between the two populations in three different GALTs. This again is not compatible with a conversion of enterocytes into M cells along the dome flanks, which should lead to a concomitant increase in M cells and a decrease in enterocyte population. Meynell et al. (Meynell et al., 1999) have reported a rapid increase in M cells that are able to transport microspheres across the FAE of rabbit Peyer’s patches after exposure to Streptococcus pneumonia. They hypothesize that this increase corresponds to enterocyte conversion. However, when sampling microspheres after bacterial stimulation, they observed less than 25 M cells/mm², which is very few compared with the total pool of M cells. We alternatively propose that in rabbit Peyer’s patches, many M cells may be inaccessible to bacterial-sized antigen or microspheres, owing to their small surface area, which is almost completely covered by adjacent enterocyte microvilli (see Fig. 4B). Consistent with this idea is the fact that M cells that bind microspheres display apical surfaces of more than 100 μm² (Meynell et al., 1999). After antigenic stimulation, the surface area of M cells should increase after recruitment of lymphoid cells in their pocket. The recruitment of lymphoid cells has actually been shown by injection of non-invasive Shigella flexneri (Sansonetti et al., 1996) or Streptococcus pneumonia (Borghesi et al., 1996) in rabbit Peyer’s patches, and we observed that large apical M cell surfaces seemed to be associated with large pockets. Enlarged apical surfaces should allow sampling of bacterial-sized antigens or microspheres (Meynell et al., 1999). Our data favor this interpretation rather than the alternative proposed expansion of M cells by conversion from fully differentiated enterocytes upon contact with micro-organisms (Meynell et al., 1999; Savidge et al., 1991). The fact that many small M cell apical surfaces were only visible by immunostaining, shows the difficulty of evaluating M cell numbers in rabbit Peyer’s patches by scanning electron microscopic analysis. Moreover, the shape change of M cells and the increase of both their apical surface and their pocket upon lymphocyte recruitment might enhance the chance of detecting M cells on electron microscope sections. This could lead to an apparent increase in the number of M cell profiles in the presence of pathogens (Borghesi et al., 1999). Our data and those of Sansonetti et al. (Sansonetti et al., 1996) indicate that future studies should focus on the size, shape and relative amount of M cells and enterocytes after bacterial challenge.

In a recent paper, Sierro et al. (Sierro et al., 2000) observed in mouse that apoptosis was not present in the zone of M cell disappearance, whereas apoptotic enterocytes were numerous at the very top of domes. They suggested a reverse differentiation of M cells to enterocyte phenotype before entering into apoptosis. This was based on the fact that the cell migration rate of BrdU-labeled cells does not slow down in the zone of disappearance of M cells, as expected if release diminished the number of cells. However, M cells represent only 10% of the cell population in mouse (Clark et al., 1993), and extrusion or not of such a low number of M cells would probably not affect the cell migration rate. The situation is different in rabbit FAE: contrary to mouse FAE and rabbit adjacent villi, apoptosis was absent in both the area of the sharp disappearance of M cells and at the top of domes (our unpublished results). Hence, if apoptosis in rabbit could not explain the disappearance of M cells, it could also not explain that of dome enterocytes. A possible explanation is that epithelial cells could be rapidly sloughed off from the dome by anoikis in the early stage of apoptosis (or even before) in the area where the M cell phenotype disappears and also at the tip of domes. This hypothesis is supported by the presence of vimentin-labeled luminal debris and M cell desquamation in this region (Owen and Piazza, 1998).

In conclusion, the presence of M cell precursors has been documented for several years (Bye et al., 1984; Gebert et al., 1999; Giannasca et al., 1994). In addition, we show that M cell precursors can already be detected in the zone of rapidly dividing cells, and that the spatial distribution of M cells and dome enterocytes in crypts reflects that observed in the domes flank. The mosaic organization of M cells and enterocytes, already present in the zone of rapidly dividing cells, suggests that interactions between neighboring cells might involve lateral specification events as in Notch signaling (Artavanis-Tsakonas et al., 1999), which may be important to investigate in the future. All this provides strong evidence in favor of an early commitment of M cell and enterocyte distinct lineages. This could very well be induced by the influence of lymphoid cells (Kerneis et al., 1997) during the first stem cell division steps.

We thank Dr P. Mangeat, Dr M. Neutra, Dr S. Baghdiguian and Dr A. M. Martinez for helpful discussions. We gratefully acknowledge F. Mérezègue for excellent technical help. This work was supported by the Centre National de la Recherche Scientifique (UMR 5539, and Cell Biology project 96033), the Association pour la Recherche sur le Cancer (contract 9829), and the Ligue Nationale contre le Cancer. H.L. was a fellow of the Délégation à la Recherche et au Développement and the Association pour la Recherche sur le Cancer.