Epithelial cells are characterised by distinct apical and basolateral membrane domains that are separated by tight junctions. Establishment and maintenance of this polarity depend on specific gene expression and protein targeting to their correct location. Our former studies, performed with renal epithelial MDCK cells, revealed a new function for galectin-3, a member of a conserved family of lectins. There, galectin-3 is required for intracellular sorting and correct targeting of non-raft-associated glycoproteins to the apical plasma membrane. In the present study, we found transport defects of the intestinal brush border hydrolases lactase-phlorizin hydrolase (LPH) and dipeptidylpeptidase IV (DPPIV) in galectin-3-null mutant mice. We could show that, in enterocytes of wild-type mice, both glycoproteins directly interact with galectin-3 and transit through non-raft-dependent apical transport platforms. Therefore, this genetic analysis provides definitive evidence for the involvement of galectin-3 in protein intracellular trafficking in vivo. Further investigations revealed that gal3-null enterocytes also exhibit striking cytoarchitecture defects, with the presence of numerous and regular protrusions located along basolateral membranes. Moreover, β-actin and villin, two characteristic markers of brush borders, become abnormally distributed along these atypical basolateral membranes in gal3–/– mice. Taken together, our results demonstrate that, in addition to a pivotal role in apical trafficking, galectin-3 also participates in epithelial morphogenesis in mouse enterocytes.
Epithelial cells display membrane polarisation that is characterised by the division of an apical plasma membrane facing the organ lumen from a basolateral domain where cell-cell and cell-extracellular-matrix contacts occur. The establishment and maintenance of this cytoarchitecture depend on the expression of specific gene products that ensure the delivery of proteins and lipids to the appropriate membrane compartment (Rodriguez-Boulan et al., 2005). Whereas basolateral targeting signals and machineries are well described, apical trafficking mechanisms remain unclear and controversial. For many years, protein segregation into lipid raft transport platforms was considered as a prerequisite for apical protein delivery (Simons and Ikonen, 1997). However, it now appears that raft- and non-raft-dependent pathways co-exist in epithelial cells, thus increasing the difficulty in drawing a clear scenario of the trafficking pathways en route to the apical plasma membrane (Delacour and Jacob, 2006).
Galectins constitute a conserved family of lectins that were initially described for their affinity for β-galactoside derivatives (Barondes et al., 1994). Consistent with their unusual ability to localise in the extra- and intracellular space, galectins are so-called `multifunctional' proteins: they participate in cell-cell, cell-matrix interactions, apoptosis and the cell cycle, modulating various normal and pathological biological processes (Hsu et al., 2006; Leffler et al., 2004).
In previous studies, we described new functions for galectins in intracellular trafficking, regulating crucial steps of protein delivery in cultured epithelial monolayers (Delacour and Jacob, 2006). First, in the HT-29 intestinal epithelial cell line, galectin-4 constitutes a major component of lipid rafts (Delacour et al., 2005). Sulphatides with long-chain hydroxylated fatty acids, also enriched in lipid rafts, were identified as high-affinity ligands for galectin-4. Moreover, inhibition of galectin-4 expression blocks protein transport to the apical plasma membrane. Therefore, the interaction between galectin-4 and sulphatides seems to play a functional role in the clustering of lipid rafts for apical delivery in HT-29 cells. Second, in renal epithelial MDCK cells, raft-independent apical carrier vesicles contain galectin-3, which interacts directly with apical cargo in a glycan-dependent manner (Delacour et al., 2006). These glycoproteins are mistargeted to the basolateral membrane in galectin-3-depleted cells, pointing to a central role for this lectin as an apical sorting mediator in the raft-independent pathway. We further demonstrated that high molecular weight clusters of apical glycoproteins are exclusively formed in the presence of galectin-3 (Delacour et al., 2007). Their stability is sensitive to elevated carbohydrate concentrations and cluster formation, as well as apical sorting, is perturbed in glycosylation-deficient cells. Therefore, glycoprotein crosslinking by galectin-3 is required for apical sorting of non-raft-associated cargo in MDCK cells.
The initial aim of this study was to examine the consequences of the absence of gal3 (also known as Lgals3) in mutant mouse enterocytes. Here we show that, despite the fact that gal3-null mutant mice are viable and do not display a major obvious phenotype (Colnot et al., 1996), this lectin is indeed directly involved in apical delivery of intestinal hydrolases in vivo. Moreover, we discovered that mutant basolateral membranes exhibit striking morphological features reminiscent of apical plasma membranes.
Apical hydrolases are mislocalised in gal3-null mutant mice
First, we used immunohistochemistry and confocal microscopy to study the distribution and localisation of galectin-3 in adult mouse small intestine. Fig. 1 depicts galectin-3 immunostaining, together with villin, a protein that interacts with actin microfilaments specifically at the brush border of enterocytes (Revenu et al., 2004). Like villin, galectin-3 was barely detected in the crypt area and was absent in goblet cells (Fig. 1A,C). Galectin-3 expression was restricted to enterocytic cells at the top of the intestinal villi. There, the lectin was distributed throughout the cytosol, mainly directly beneath the brush border and also near the basolateral membranes (Fig. 1B).
To assess the role of galectin-3 in protein trafficking in vivo, we compared the intracellular localisation of intestinal hydrolases in adult wild-type (wt) and galectin-3-null mutant (gal3–/–) mice (Fig. 2). Transverse sections of the small intestine showed that sucrase-isomaltase (SI), aminopeptidase N (APN), lactase-phlorizin hydrolase (LPH) and dipeptidylpeptidase IV (DPPIV) were almost exclusively addressed to the brush border membrane of wt mouse enterocytes (Fig. 2). In the gal3–/– mutant mouse, the apical localisation of SI and APN was not affected. By contrast, LPH and DPPIV were mislocalised in enterocytes from gal3–/– mice. Their quantity in the apical brush border membrane is significantly reduced compared to the wt mouse. The majority of LPH accumulated intracellularly in a punctate pattern, whereas about 40% of DPPIV was mis-sorted to the basolateral membrane as assessed by densitometric scanning of the fluorescence images (Fig. 2). Based on these observations, we were able to distinguish two subgroups of brush border hydrolases: one in which apical targeting is insensitive to galectin-3 (SI and APN), and another one, which depends on galectin-3 expression (LPH and DPPIV).
Non-raft-associated DPPIV and LPH bind to galectin-3 in vivo
To determine whether this dependency correlated with a direct binding to galectin-3, we searched for the presence of the lectin in immunoprecipitates of each hydrolase from mouse intestinal extracts. Fig. 3A shows that galectin-3 could be readily detected on immunoblot after precipitation of LPH and DPPIV. This interaction was inhibited by addition of galactose or lactose to the precipitation buffer indicating that galectin-3 interacts with LPH and DPPIV in a carbohydrate-dependent manner. No coimmunoprecipitation could be observed between galectin-3 and SI or APN.
Next, we assessed the association of DPPIV, LPH, SI and APN with detergent-resistant membrane microdomains (DRMs) by floating analysis. Fig. 3B demonstrates that more than 60% of SI or APN was present in the floating flotillin-1-positive fractions of the gradient. By contrast, LPH did not float and remained in the detergent-soluble fractions at the bottom of the gradient. These ratios are similar to previously published data for hydrolases of the pig small intestine (Danielsen, 1995). Regarding DPPIV, two pools could be distinguished: although 30% was found in DRM fractions, about 70% was not raft-associated, this distribution is similar to that in Caco-2 cells (Alfalah et al., 2002) or HT-29 cells (Delacour et al., 2003). The dual partitioning of DPPIV in DRMs and in soluble fractions suggests that this hydrolase could follow two distinct transport pathways – one raft dependent and one non-raft dependent, as previously reported (Slimane et al., 2001). This may account for the differences between the LPH and DPPIV distributions in the mutant intestine. Despite a shared dependency on galectin-3, LPH and DPP-IV respond differently in the absence of galectin-3 in vivo. The intracellular accumulation of LPH could be explained by a failure of galectin-3 dependent sorting, thus resulting in a complete blockage in the cytoplasm. In the case of DPPIV, the absence of galectin-3 might exclusively affect the non-raft-associated fraction, and not have any effect on correct targeting of the raft-associated DPPIV pool to the apical membrane as suggested by Fig. 2. A basolateral localisation of this enzyme in gal3–/– mice can be explained by a loss of galectin-3 dependent sorting of non-raft-associated DPPIV, which later results in basolateral trafficking.
Why does non-raft-associated LPH accumulate intracellularly whereas the majority of DPPIV enters the plasma membrane? First, it has to be considered that in contrast to transfections of cells in culture (Delacour et al., 2006; Delacour et al., 2007), this is the first in vivo analysis of a sorting defect following constitutive galectin-3 inactivation. Here, perturbations in cellular transport might be bypassed by adaptative mechanisms with varying preferences for the proteins affected. Previously, we have observed a basolateral mis-sorting of LPH in galectin-3 depleted MDCK cells (Delacour et al., 2006). Here, galectin-3 is the major if not the sole galectin, suggesting that compensatory cellular mechanisms for galectin-3 are lacking in this cell line. However, in the murine small intestine, galectin-2 and galectin-4/6 are expressed in addition to galectin-3 (Nio et al., 2005) and it has been demonstrated that galectin-4 is involved in apical trafficking in intestinal cells (Danielsen and van Deurs, 1997; Delacour et al., 2005). In contrast to galectin-3, galectin-4 plays a role in the stabilisation of lipid rafts and thus facilitates apical targeting of raft-associated cargo. Hence, based on the sorting patterns of raft-associated hydrolases and on the intestinal expression of other members of the galectin family, we conclude that alternative apical transport pathways may partially compensate for the absence of galectin-3 in mouse enterocytes. In addition, DPPIV exhibits high capacities to follow different transport routes. More than a decade ago, Matter et al. revealed that unlike other hydrolases, apical delivery of DPPIV in Caco-2 cells occurred in two waves and that DPPIV reaches the apical membrane by transcytosis (Matter et al., 1990). Moreover, the competence of DPPIV to be conducted via the basolateral membrane to the apical cell surface or to take a direct route, changes during development of the polarised monolayer. At early stages in the development of Fisher rat thyroid cells, a significant quantity of DPPIV is transcytosed from basolateral to apical followed by the establishment of a direct pathway to the apical membrane after polarisation (Zurzolo et al., 1992). Hence, depending on the developmental state of the epithelial monolayer the targeting profile of DPPIV varies between direct and indirect routes. Therefore, it is also likely that the basolateral localisation of DPPIV in galectin-3-depleted enterocytes is based on an immature differentiation of the epithelial monolayer.
Accumulation of Rab8-positive endosomes close to basolateral membranes of gal3–/– enterocytes
In light of the observations that intracellular trafficking pathways of LPH and DPPIV are affected in enterocytes from gal3–/– mice, we studied the morphology of secretory organelles by immunohistochemistry and confocal analysis. No differences between gal3–/– and wt enterocytes were detected by Golgi-staining with anti-giantin, late endosomal staining with anti-Lamp1 or recycling endosomal staining with anti-Rab11 antibodies (Fig. 4). However, a significant difference could be observed when enterocytes were stained with anti-Rab8. In contrast to a diffuse cytosolic and apical labelling of Rab8 in wt enterocytes, in the absence of galectin-3 Rab8-positive endosomes accumulated in close proximity to the basolateral cell poles (Fig. 4). This suggests that the equilibrium of Rab8-positive endosomes is disturbed in galectin-3 depleted enterocytes.
Enterocytes from gal3–/– mice show abnormal morphologies
We next compared the enterocyte structure of wt and gal3–/– mice by electron microscopy (Fig. 5). It first appeared that gal3–/– enterocytes were filled with numerous intracellular vesicles and vacuoles of variable size. This could be explained by the partial block in protein trafficking that occurs in the absence of galectin-3 and corroborates the accumulation of endosomal organelles observed by immunohistochemistry. As assessed by immuno-EM, these organelles were positive for LPH. Fig. 6 depicts immunogold staining of the apical brush border of wt and gal3–/– enterocytes, whereas significant quantities of intracellular LPH were exclusively detected in gal3–/– enterocytes. Hence, our observations suggest that in the absence of galectin-3 a significant proportion of LPH remains trapped intracellularly, most likely in Rab8-positive endosomes. This is of particular interest because the involvement of a Rab8-positive compartment in post-Golgi apical protein trafficking has recently been described (Sato et al., 2007).
Finally, Fig. 5 reveals intriguing membrane abnormalities in gal3–/– enterocytes: although no obvious difference in the organisation and morphology of apical microvilli or tight junctions was detected (Fig. 5A-D), striking alterations at the lateral and basal membranes of enterocytes from gal3–/– mice were observed (Fig. 5E-H). Instead of the rare and limited membrane interdigitations typical of small intestinal enterocytes in wt mice, we found extensive infoldings along the basolateral membranes in gal3–/– enterocytes. Frequently, protrusions into the lateral space between two adjacent cells were observed, which could be interpreted as `microvilli-like' structures. To gain more information, the integrity of the epithelial monolayer was studied by immunofluorescence using villin, β-actin or ezrin, which are essential structural components of apical microvilli (Fievet et al., 2007), as well as the basolateral marker molecule E-cadherin. The distribution of the apical brush border marker villin was also perturbed and shifted towards the lateral membranes of enterocytes in gal3–/– mice (Fig. 7). Similar effects could be observed for actin microfilaments, which are normally concentrated at the terminal web, but accumulated adjacent to the lateral membranes in the absence of galectin-3 expression (Fig. 7). Moreover, these two markers colocalised along the abnormal basolateral membranes in gal3–/– enterocytes (Fig. 8, supplementary material Movies 1 and 2). Ezrin, an additional component of microvilli, is retained in the cytosol of gal3–/– mice (Fig. 7), thus demonstrating that in addition to the mistargeting of non-raft apical markers, the distribution of key structural elements of the microvillar cytoskeleton is deeply altered in mutant enterocytes. All these alterations were observed throughout the small intestine of several mutant animals. By contrast, Fig. 7 shows that the accumulation of E-cadherin in the lateral membranes of enterocytes was not altered in the mutants, which suggests that basolateral trafficking is not impaired in the absence of galectin-3.
Our in vivo data provide evidence for two additional cellular functions of galectin-3, which may act independently or may even be linked with each other. First, corroborating our in vitro data, galectin-3 is essential for intracellular trafficking of non-raft-associated intestinal hydrolases. After entering the secretory pathway, galectin-3 binds to non-raft-associated glycoproteins such as LPH and DPPIV and this association is necessary for their targeting to the proper membrane compartment. Second, an unexpected and stunning consequence of galectin-3 mutation in vivo is the formation of basolateral infoldings or `microvillus-like' protrusions and a basolateral accumulation of actin and villin.
Interestingly, relocalisation of villin and actin from the apical to the basolateral pole, associated with apparently similar morphological changes has been reported in small intestines of rat treated with colchicine (Hasegawa et al., 1987; Pavelka et al., 1983) and also in the proximal renal tubule of mice during reperfusion after ischaemia (Brown et al., 1997; Molitoris et al., 1988). In parallel, these studies revealed a redistribution of apical proteins to the two membrane domains of epithelial cells. Since galectin-3 is upregulated in ischaemia-induced renal failure (Nishiyama et al., 2000), it might antagonise basolateral targeting of apical proteins and thereby induce perturbations in the cytoarchitecture of epithelial cells.
Could the morphological abnormalities observed here be directly linked to the trafficking defects? In intestinal and renal epithelia, the assembly of apical microvilli indeed depends on the presence of villin at the apical domain (Dudouet et al., 1987; Shibayama et al., 1987). It has been additionally shown that exogenously added villin recruits actin, modulates the state of actin microfilaments beneath the plasma membrane and induces the growth of microvilli at the cell surface of fibroblasts (Friederich et al., 1989). Similarly, ezrin has been demonstrated to promote microvilli morphogenesis in retinal pigment epithelium (Bonilha et al., 1999). Thus, basolateral microvillus-like structures could be consequently generated after the redistribution of brush border elements.
The question arises: can the absence of galectin-3 interfere with the apical positioning of these brush border elements? Here, ezrin would constitute an appropriate candidate: it is an O-glycosylated, non-raft-associated protein (Louvet-Vallee et al., 2001; Stickney et al., 2004). Nevertheless, in vivo, ezrin, villin and β-actin were not co-precipitated with galectin-3 (data not shown). This suggests that either these cytoskeletal elements interact with the lectin by very weak binding forces, or, more likely, that their mislocalisation could be an indirect, possibly more general effect on epithelial organisation.
During this study, we observed that galectin-3 is located in the upper half of intestinal villi, in fully mature enterocytes. A similar distribution has also been recently described for gal3 mRNA, by in situ hybridisation (Nio et al., 2005). Such a restricted distribution of galectin-3 suggests that the lectin may be involved in the terminal differentiation process of mouse enterocytes, probably by maintaining their polarisation state.
Despite the striking abnormalities discovered in this study, gal3-null mutant mice are viable and exhibit no obvious phenotype when maintained in highly protected and controlled animal house conditions. The defects so far discovered in gal3–/– mice are relatively mild (Bichara et al., 2006; Colnot et al., 1998; Colnot et al., 2001; Iacobini et al., 2005) and no intestinal deficits have ever been recorded (F.P., unpublished data). It will now be interesting to change the environment of these animals, for example by supplying different food diets. The lack of galectin-3 may turn out to be deleterious once placed in these challenging conditions.
In summary, we have shown that galectin-3 is involved in apical protein sorting in vivo and also in the epithelial organisation per se.
Materials and Methods
Galectin-3 was detected with rabbit polyclonal antibodies essentially as described before (Delacour et al., 2007). Monoclonal antibodies directed against villin and polyclonal antibodies against ezrin were generously provided by S. Robine and M. Arpin (Curie Institute, Paris, France). E. M. Danielsen (Panum Institute, Copenhagen, Denmark) and B. Nichols (Baylor College of Medicine, Houston, TX) generously supplied us with antibodies directed against DPPIV/APN and LPH/SI. H. P. Hauri (Biozentrum, Basel, Switzerland) and A. Wandinger-Ness (New Mexico Health Sciences Center, Albuquerque, NM) generously provided us with antibodies directed against lamp1 and Rab11. The anti E-cadherin, anti-Rab8 and anti-flotillin-1 antibodies were obtained from BD Biosciences (Mountain View, USA). Anti-β-actin antibodies are from Cell Signaling (Boston, MA). Polyclonal anti giantin antibodies were purchased from Covance (Berkeley, CA).
Wild-type and gal3-null mutant (Colnot et al., 1998) mice used in this study were of 129Sv background. The animals were maintained in a specific pathogen-free animal house facility and handled respecting the French regulation for animal care.
Preparation of tissue samples, immunostaining and confocal fluorescence microscopy
Mice were killed by cervical dislocation, the small intestine was dissected, rinsed with phosphate buffered saline (PBS, pH 7.4) and cut into 1 cm pieces. Tissue samples were fixed for 2 hours in either 4% formaldehyde or in 60% ethanol:30% chloroform:10% acetic acid, then embedded in paraffin and processed for immunohistochemistry. Confocal images of fixed cells were acquired on a Leica TCS SP2 microscope using a 63× water planapochromat lens (Leica Microsystems, Wetzlar) essentially as described before (Jacob and Naim, 2001). For 3D reconstruction, fluorescent images of 35 layers were recorded over a range of 8 μm on a Leica DMI 6000 fluorescence microscope. The images were processed by iterative deconvolution and 3D rendering provided by the velocity software package (Improvision, Coventry, England).
Co-immunoprecipitation, membrane preparation and floating analysis
Following tissue sample preparation, lysates of intestinal pieces were processed for coimmunoprecipitation essentially as described before (Delacour et al., 2007). The preparation of membranes was carried out on ice with ice-cold buffers supplemented with protease inhibitors. After rinsing the small intestine with PBS, 100 mg pieces were cut and homogenised in 1 ml lysis buffer (25 mM HEPES, 150 mM NaCl, pH 6.5). Cell debris was removed by centrifugation for 10 minutes at 1000 g. The supernatant was pelleted by centrifugation for 30 minutes at 100,000 g in a SW41 rotor. The pellet was resuspended in 1 ml lysis buffer and extracted for 30 minutes with 1% Triton X-100 on ice before loading on a 30%, 26%, 2.5% Nycodenz step gradient. After centrifugation for 20 hours at 178,000 g in a SW41 rotor, 1 ml fractions were collected from the top. Each hydrolase was isolated by immunoprecipitation from each fraction, followed by SDS-PAGE analysis and immunoblot. Flotillin-1 was detected by immunoblot from TCA-precipitated fractions.
For electron microscopy investigations, tissue samples were cut into small pieces of about 1 mm3 and were immersion-fixed in 2.5% paraformaldehyde, 2.5% glutaraldehyde and 0.05% picric acid in 0.067 M cacodylate buffer of pH 7.4 for 2 hours at 4°C (Elsasser et al., 1993). Standard procedures for dehydration and embedding in Epon were used. For immunostaining, samples were fixed in 0.1 M cacodylate buffer, pH 7.35 containing 1% paraformaldehyde. The samples were dehydrated in a graded series of alcohol, embedded in Lowicryl K4M (Polysciences Ltd.) and polymerised at –20°C under UV illumination (360 nm) for 48 hours. Thin sections were incubated with monoclonal anti-LPH in a dilution of 1:50 and visualised using a 10 nm immunogold-conjugated goat anti-mouse antibody solution (British Bio Cell International) at a dilution of 1:50, both in 1% BSA in PBS. In both cases, thin sections were stained with uranyl acetate and lead citrate before examination using an EM 109 electron microscope (Zeiss, Oberkochen, Germany).
We are grateful to B. Agricola, M. Dienst and T. Straube for technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft (DFG), Bonn, Germany: grants JA 1033 (to R.J.) and Sonderforschungsbereich 593 (to R.J.). Financial support from GEFLUC, La Ligue Contre le Cancer and l'Association pour la Recherche contre le Cancer (ARC) was allocated to F.P.
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/121/4/458/DC1
↵* These authors have contributed equally to this work
- Accepted November 14, 2007.
- © The Company of Biologists Limited 2008