Abstract
Galectin-4 (G4) is a member of a family of soluble galactoside-binding lectins found in various mammalian tissues. To determine the function of this protein in colorectal tissue, we separately produced the N- and C-terminal carbohydrate binding domains (CBD) of rat G4 as a recombinant glutathione S-transferase (GST) fusion protein (G4-N and G4-C) and examined the tissue binding site(s) of each CBD by light and electron microscopy (LM and EM). At the LM level, both fusion proteins stained the intercellular borders of the surface-lining epithelial cells of colorectal mucosa. At the EM level, two proteins recognized spatially close but distinct subcellular structures. G4-N stained electron-lucent flocculent substances freely located in the intercellular spaces, whereas G4-C bound to the lateral cell membranes demarcating the intercellular spaces. These findings suggest that colorectal G4 may be involved in crosslinking the lateral cell membranes of the surface-lining epithelial cells, thereby reinforcing epithelial integrity against mechanical stress exerted by the bowel lumen.
Keywords
A number of soluble galactoside binding lectins have been isolated from various mammalian tissues. All these proteins show well-conserved amino acid sequence in their carbohydrate binding domains (CBDs) and therefore are believed to constitute a distinct protein family recently designated galectins (Barondes et al. 1994; Kasai and Hirabayashi 1996). To date, 10 different types of galectins have been cloned and characterized. They include galectin-1 (Clerch et al. 1988), −2 (Gitt et al. 1992), −3 (Cherayil et al. 1989), −4 (Oda et al. 1993), −5 (Gitt et al. 1995), −6 (Gitt et al. 1998), −7 (Madsen et al. 1995), −8 (Hadari et al. 1995), −9 (Wada and Kanwar 1997), and −10 (Leonidas et al. 1995). These are expressed as 14−36−kD cytosolic proteins without a secretion signal peptide (Barondes et al. 1994), which have one (galectin−1, −2, −3, −5, −7, and −10) or two (galectin−4, −6, −8, and −9) CBDs within a single molecule. Of galectin−4, −6 and a splicing isoform of galectin−9, which are exclusively localized in gastrointestinal (GI) tract epithelia, the most well-studied is galectin-4 (G-4). This protein has been immunocytochemically detected as an adherens junction protein in oral epithelium (Chiu et al. 1992, 1994), as a structural protein forming cytoplasmic membrane thickening in esophageal epithelium (Wasano and Hirakawa 1995), and as a protein forming detergent-insoluble complexes with brush-border enzymes in small intestinal epithelium (Danielsen and van Deurs 1997). Although G-4 is known (Gitt et al. 1998) to be expressed at equal levels in small intestine and colon, much less information is available concerning immunoelectron microscopic localization of this protein in colorectal tissue. This is probably due to the high water solubility of colorectal G-4 (Huflejt et al. 1997). In this study, we separately cloned cDNA encoding N- and C-terminal CBDs (the protein domains will be hereafter referred to as N-CBD and C-CBD) of rat G-4 and surveyed the tissue binding site(s) of each CBD on tissue sections of rat colorectal tissue. The very characteristic staining pattern of the surface-lining epithelial cells by each CBD suggests a previously undescribed functional role of G-4 in this tissue.
Materials and Methods
Reagents
The reagents for SDS-PAGE and agarose gel electrophoresis were obtained from Bio-Rad (New York, NY). Restriction enzymes and DNA standards were from Fermentas (Vilinius, Lithuania). RT-PCR kits were purchased from Cetus (Emeryville, CA). The PCR cloning vector pGEM-T and T4 ligase were from Promega (Madison, WI). The expression vector pGEX-2T, goat anti-GST (Schistosoma japonicum origin) antibody, and GST purification kits equipped with glutathione-Sepharose were purchased from Pharmacia (Uppsala, Sweden). XL-1 blue and BL21(DE3)pLysS competent cells were obtained from Stratagene (La Jolla, CA) and Novagen (Madison, WI), respectively. Unless stated otherwise, all other reagents of analytical grade were from Sigma (St Louis, MO).
DNA Cloning
All the manipulations of nucleic acids such as restriction, ligation, transformation, selection, cell culture, gel electrophoresis, and elution were done using standard protocols (Sambrook et al. 1989). To obtain cDNA encoding N-CBD and C-CBD of rat G-4, we synthesized two sets of primers; two anti-sense downstream primers, Primer 1 (5′-GTT
Expression and Isolation of recombinant N-CBD and C-CBD Proteins
The cDNA encoding N-CBD or C-CBD was released with EcoRI and BamHI restriction enzymes, gel-purified, and subcloned in between the same sites in the pGEX-2T expression vector containing a GST (Schistosoma japonicum origin) gene and a thrombin cleavage sequence upstream from the site. The host strain BL21(DE3)pLysS was transformed with the plasmid and inoculated on ampicillin-containing agar plates. The resulting colony was picked up in distilled water, boiled, and then colony PCR was done using the above primer sets. The PCR-positive bacteria were cultured in 200 ml SOC medium until absorbance at 600 nm reached 0.6. Protein production was induced by adding 0.4 mM IPTG at 30C for 2 hr. To isolate recombinant protein, the bacteria were cooled on ice, centrifuged into a pellet, and resuspended with 5 ml protein extraction solution (PES) (58 mM Na2HPO4, 18 mM KH2PO4, 75 mM NaCl, 2 mM EDTA, 2 mM benzamidine, 2 mM aminohexanoic acid, and 1 mM PMSF). For isolation of C-CBD protein, which contains two cysteine residues at amino acid positions 246 and 281, 0.4 mM iodoacetamide was added to the PES solution. The bacteria were lysed with an ultrasonicator (Tomy; Tokyo, Japan) at a maximal output twice for 10 sec. After centrifugation to remove cell debris, the clear supernatant was loaded into 1 × 3-cm lactosyl-Sepharose. After washing with 100 ml PES, bound protein was eluted with PES containing 100 mM lactose. The eluted protein was concentrated with an Amicon ultrafiltration membrane Y-10, dialyzed against PES to remove lactose and stored at −80C. Crude extract from noninduced or IPTG-induced bacteria, and purified GST-fused N-CBD and C-CBD proteins (hereafter referred to as GST-N-CBD and GST-C-CBD) were analyzed by SDS-PAGE. Both fusion proteins were incubated with thrombin and the digestion products were analyzed by SDS-PAGE. Nonfused GST protein was expressed in bacteria containing pGEX-2T without any specific DNA insert, purified through glutathione-Sepharose resin, and used for a cytochemical control experiment.
Histocytochemistry Using GST-N-CBD and GST-C-CBD
Ten Wistar rats (6 weeks old) were sacrificed with CO2 gas according to the guidelines of our institutional animal welfare committee. The ascending colon (1 cm distal from the ileocecal junction), the descending colon (1 cm distal from the colon passing between the distal end of the stomach and the cranial mesenteric root) (Hebel and Stromberg 1986), and the rectum were obtained and opened. Fecal balls, if present, were removed by forceps and viscous contents were washed away with a jet stream of acetate-buffered saline (ABS; 150 mM NaCl, 10 mM sodium acetate, pH 7.4). Each tissue specimen was cut into tissue blocks with a razor blade, fixed with ice-chilled 4% paraformaldehyde (PFA) in acetate buffer (100 mM sodium acetate adjusted to pH 7.4 with acetic acid), and kept in this solution until postfixation with 1.2% glutaraldehyde (GA) in the same buffer for 30 min just before cryosectioning. For GST-C-CBD staining at the EM level, several tissue blocks were chopped into 0.1-mm-thick slices with a Tissue Chopper (Sorvall; Newtown, CT), with the muscular layer left intact. These partially sliced tissue blocks were then washed in ABS with extensive shaking for 10 min before postfixation with GA as above. This chop and wash step was carried out because we noted that the intercellular borders of the epithelium located near the cut edges of the tissue blocks (hand-cut with a razor blade during initial PFA fixation) were stained with GST-C-CBD more intensely than those located in the middle part of tissue blocks.
For LM histochemistry, the tissue specimens were soaked overnight in 0.6 M sucrose and embedded into Tissue-Tek OCT compound (Miles; Elkhart, IN). Cryosections 5 μm thick were cut in a cryotome (American Optical; Buffalo, NY) and picked up onto poly-L-lysine-coated glass slides. The sections were covered by blocking solution (BS) containing 3% bovine serum albumin in ABS for 30 min to mask nonspecific protein binding sites. The sections were incubated with GST-N-CBD or GST-C-CBD (0.1 μg/ml BS) for 30 min, washed with three changes of ABS for 5 min each, and then overlain with goat anti-GST antibody diluted 1000 times with BS for 30 min. After washing in ABS three times for 5 min each, the sections were stained with rabbit anti-goat IgG labeled with FITC diluted 500 times with BS for 30 min, washed as above, and examined under an epifluorescence microscope equipped with an automatic camera (Olympus; Tokyo, Japan).
For EM cytochemistry, tissue specimens were infused with 2.3 M sucrose in 0.1 M acetate buffer overnight, attached to aluminum stubs, and frozen in liquid nitrogen. Ultrathin frozen sections about 100 nm thick were cut with a dry type diamond knife in an Ultracut S equipped with an FCS cryoattachment (Leica; Vienna, Austria), and picked up onto collodion-coated nickel grids. All the staining steps were done by transferring the grids from a droplet to the next droplet with a platinum wire loop. The grids were inverted on a droplet of ABS to remove sucrose and transferred onto a BS droplet. The sections were stained and washed in the same way as for LM staining steps, except that rabbit anti-goat IgG antibody labeled with 15-nm colloidal gold (Bio-Cell; Cardiff, UK) was used instead of that labeled with FITC. The sections were finally stained with an aqueous 1% uranium salt solution for 10 sec and briefly coated with an aqueous 0.5% methylcellulose solution (Tokuyasu 1973). The grids were inverted onto a filter paper, kept at 60C overnight, and examined under a 1200 EX electron microscope (JEOL Datum; Akishima, Japan).
In control experiments, GST-N-CBD or GST-C-CBD was omitted or replaced with nonfused GST. GST-N-CBD or GST-C-CBD was preincubated with 50 mM thiodigalactoside (TDG) for 30 min, which is the most potent inhibitor of galactoside binding of galectins. Anti-GST or secondary antibody was omitted.
Results
Preparation of Recombinant N-CBD and C-CBD as a GST Fusion Protein
After RT reaction using intestinal poly(A)RNA followed by PCR using two primer sets, we obtained about 480-BP and 460-BP products (Figure 1), which corresponded to the expected molecular size of cDNA encoding N-CBD and C-CBD, respectively. The DNA product was ligated into the expression vector containing a GST gene and a thrombin cleavage sequence upstream from the cloning site, and recombinant GST-N-CBD or GST-C-CBD induction was triggered with IPTG. SDS-PAGE analysis of a crude extract from noninduced and induced bacteria revealed IPTG-dependent induction of GST-N-CBD and GST-C-CBD protein at molecular weights of about 43 and 42.5 kD, respectively (Figure 2). The size of each fusion protein matched the sum of GST (26 kD) plus N-CBD (17 kD) or C-CBD (16.5 kD). Successful production of fusion protein was further confirmed by incubation with thrombin (Figure 2), which released a 17-kD band from GST-N-CBD and a 16.5-kD band from GST-C-CBD. Both fusion proteins were water-soluble, retained sugar-binding capability, and could be purified by binding to lactosyl-Sepharose. For isolating GST-C-CBD that contains two cysteine residues, iodoacetamide was added to the extraction solution. In the absence of an alkylating agent, the sugar-binding capability of this domain is inactivated within several minutes after sonication, even when maintained at 4C, probably due to oxidation of the cysteine residue(s), as is the case with other types of galectins.
Agarose gel electrophoresis of two PCR products encoding N-CBD (Lane B) and C-CBD (Lane C). The size of the products corresponds well to the expected molecular weight of cDNA (481 BP) encoding N-CBD and cDNA (459 BP) encoding C-CBD. No product is seen in a sample from RT-PCR without addition of any primer (Lane D). Arrows indicate the positions of DNA molecular weight standards (Lanes A and E) from lower to upper; 400 and 500 BP.
SDS-PAGE of protein samples from various steps of the recombinant protein preparation. Arrowhead indicates the position of GST-N-CBD (Lane C) or GST-C-CBD (Lane D) protein band in a crude extract from IPTG-induced bacteria. Neither band can be seen in a crude extract of noninduced bacteria (Lane B). The recombinant GST-N-CBD and GST-C-CBD protein purified through lactosyl-Sepharose as about 43 kD (Lane E) and 42.5 kD (Lane F) protein bands, respectively. Thrombin cleavage of GST-N-CBD (Lane G) and GST-C-CBD (Lane H) yields, in addition to 26-kD GST (small arrowhead), 17-kD N-CBD (large arrowhead), and 16.5-kD C-CBD (double large arrowheads). Lane I, loaded with thrombin only. Six arrows (Lanes A and J) indicate the positions of protein molecular weight markers from lower to upper: 14.3, 18.4, 24.0, 34.7, 45.0 and 66.0 kD.
GST Portion of the Fusion Protein as a Useful Cytochemical Marker
Two CBDs of G-4 were expressed as a GST fusion protein to examine their tissue binding site(s) by indirectly detecting the GST sequence as a cytochemical marker. GST is a ubiquitous enzyme that is involved in the transfer of sulfur from its donor (glutathione) to various kinds of acceptors and is therefore widely distributed within almost all kinds of tissues or cells. Therefore, before commencing our histochemical study, it was important to exclude the following possibilities: (a) whether the GST portion of the fusion protein reacts or binds to intrinsic glutathione remaining in cryosections, and (b) whether the anti-GST antibody, which was raised against GST of Schistosoma japonicum origin, crossreacts with intrinsic GST of rat origin. To check on these, several control experiments were performed. (a) GST-N-CBD or GST-C-CBD was omitted from the staining steps or replaced with plain nonfused GST. (b) The fusion protein was preincubated with TDG, the most potential inhibitor of the sugar-binding capability of galectins. (c) Anti-GST antibody or secondary antibody was omitted. All staining at both LM and EM levels vanished in all these control experiments (not shown), indicating that the staining was not an artifact caused by the interaction between intrinsic glutathione and the GST portion of the fusion protein, or between intrinsic rat GST and anti-Schistosoma japonicum GST antibody. This was also supported by the finding that GST-N-CBD and GST-C-CBD apparently recognized different subcellular structures as mentioned below.
Histocytochemistry Using GST-N-CBD and GST-C-CBD
For simplicity, the terms “GST-N-CBD and GST-C-CBD” are hereafter replaced by “G4-N and G4-C,” because it is apparent from the above control experiments that the protein sequence responsible for the tissue binding of the recombinant protein is not its GST but its N-CBD or C-CBD portion.
Figure 3 shows fluorescence microscopic views of G4-N and G4-C binding sites in rat colorectal tissue. G4-N recognizes the intercellular borders of the surface-lining epithelial cells. On the other hand, G4-C binds not only to the same site as does G4-N but also to the apical balloon-like cytoplasm of goblet cells scattered throughout the colon mucosa, including the surface-lining and crypt epithelium. The staining pattern of the intercellular borders was different between G4-N and G4-C. The former stained granular structures mainly located in the lower half of the intercellular borders, whereas the latter labeled smooth linear structures extending throughout the intercellular borders from the apical to basal surface of the epithelium. No apparent labeling was seen on the apical or basal surface of the surface-lining epithelial cells with either G4-N or G4-C. However, in some areas of the sections, G4-C also decorated the luminal surface of some crypt cells (Figure 3D), but it is uncertain at the LM level whether this represents apical cell structure or secretions merely attached to the site. All the fluorescent structures completely vanished after the control stainings, as mentioned above.
Figures 4 and 5 show G4-N and G4-C binding sites at the EM level using a secondary antibody labeled with colloidal gold. Both G4-N and G4-C again recognize the intercellular spaces of the surface-lining epithelial cells. However, the labeling site was apparently distinct. G4-N recognized electron-lucent, flocculent substances found in the intercellular spaces (Figure 4). In contrast, G4-C intensely stained the lateral cell membranes, especially the membranes of long cell processes forming well-developed interdigitations (Figure 5). When the specimens were chopped and washed before cryosectioning, the staining density of G4-C was enhanced, in contrast to a significant decrease in G4-N staining. Although the reason for this enhancement of G4-C staining after the treatment is unknown, it appears probable that the procedure enabled G4-C to access its ligand(s) by removing the G4-N-positive substances filling the intercellular spaces. After staining with either fusion protein, no apparent labeling was observed on any cell-cell junction, basal cell membrane, or basal lamina (Figures 4 and 5). The secretory granule contents of goblet cells were almost entirely lost during the preparation for ultracryosectioning, and therefore it was impossible to determine what structure is responsible for the positive G4-C staining seen in the luminal surface or apical cytoplasm of the cells. In the control experiments, all labeling completely vanished (not shown).
Discussion
We have previously demonstrated the immunocytochemical localization of G-4 in esophageal epithelium (Wasano and Hirakawa 1995). Despite efforts using various fixation protocols and the same high-titer antibody as employed for the demonstration of G-4 in the esophageal epithelium, we failed in our subsequent light and electron microscopic immunocytochemical study aimed at the subcellular localization of G-4 in colorectal tissue, which has been shown to be one of the richest sources of the protein (Leffler et al. 1989; Oda et al. 1993; Gitt et al. 1998). Chiu et al. (1992) have found that G-4 in the oral and upper esophageal epithelium is very water-insoluble and was detectable by immunoelectron microscopy without any loss or translocation of the protein after an initial fixation with acetone. Similarly, Danielsen and van Deurs (1997) have found that G-4 in small intestine forms detergent-insoluble (when isolated at low temperature) complexes with brush-border enzymes and was detectable on ultracryosections after weak fixation with 0.1% GA and 2% PFA. In contrast, Huflejt et al. (1997) have reported using a colon epithelial cell line T84 in which, although the cells express high concentrations of G-4, the protein is highly water-soluble and is readily lost or translocated unless special fixation and permeabilization protocols were performed. This unexpected solubility of colorectal G-4 may be one of the factors that make it difficult to demonstrate its immunocytochemical localization in this tissue.
Fluorescence micrographs of colon mucosa stained with G4-N (
Electron micrographs of the surface-lining epithelial cells stained with G4-N. Gold particles can be seen on flocculent substances located in the intercellular spaces (
To circumvent this technical problem, we separately cloned cDNA encoding N-CBD and C-CBD of G-4 and expressed each of them as a recombinant GST fusion protein. The resulting fusion proteins were used to clarify the endogenous ligand(s) on both LM and EM sections of rat colorectal tissue. In all parts of the colon from the ascending colon to the rectum, the intercellular borders of the surface-lining epithelial cells were intensely stained with both G4-N and G4-C. Because even at the LM level the binding pattern of G4-N appeared to be different from that of G4-C, we investigated the binding site(s) in more detail with EM using colloidal gold as a cytochemical marker. The binding sites of G4-N and G4-C were strikingly different. The former recognized flocculent substances located in the intercellular spaces, whereas the latter closely attached to the lateral cell membranes of the cells. This finding indicates the following possibilities. (a) Two CBDs within a G-4 molecule recognize distinct ligands, i.e., ligand located in the intercellular spaces and ligand associated with (or anchored to) the lateral cell membranes. (b) The opposite lateral cell membranes can be cross-bridged with two G-4 molecules via these two ligands. (c) By crosslinking lateral cell membranes in this way, G-4 may be involved in cell-cell adhesion of the colorectal surface-lining epithelial cells. This idea is further supported by the following findings. First, the architecture of the G-4 molecule, consisting of two structurally distinct domains with different sugar specificities (Oda et al. 1993), enables crosslinkage of ligands possessing distinct sugar moieties. Second, it has been shown (Huflejt et al. 1997) that G-4 is externalized at the basolateral surface of polarized T84 epithelial cells and is associated with other cellular components at this location. It has also been shown (Huflejt et al. 1997) that surface coating with recombinant G-4 enhances the attachment of T84 cells in vitro. Finally, a differential cDNA screening study (Rechreche et al. 1997) using mRNA transcript probes from normal and cancerous human colorectal tissue has demonstrated that the G-4 mRNA expression is significantly downregulated in the tumor tissue. Therefore, what is the physiological meaning of the lateral crosslinkage of the surface-lining epithelial cells by G-4? One likely answer to this is that the surface-lining epithelium of this level of the GI tract is exposed to mechanical stress due to the formation of viscous feces (or fecal balls in some animal species, including rat). It is possible that the specialized intercellular crosslinkages formed by G-4 may play a role in reinforcing physical cell-cell contact and thereby stabilizing the structural integrity of the epithelial cells against the stress coming from the bowel lumen. However, one should take account of the possibility that G-4 is not the only galectin responsible for the crosslinkage of the ligands at this tissue site. This is based on the facts that no direct evidence for immunolocalization of G-4 at this tissue site has been provided and that several other galectins, including galectin-3, −6, and a splicing isoform of galectin-9, are shown (Leffler et al. 1989; Wada et al. 1997; Gitt et al. 1998) to be co-expressed in the intestinal epithelium. In particular, the latter two and G-4 are structurally related galectins with two CBDs within a single protein molecule, and are exclusively localized in GI tract epithelia.
Electron micrographs of the surface-lining epithelial cells stained with G4-C. Gold particles can be seen on the lateral cell membranes and long cell processes interdigitating (ID) with those of adjacent cells (
We were unable to detect any significant labeling on the basal cell membrane or basal lamina with either G4-N or G4-C. This result is inconsistent with the observation (Huflejt et al. 1997) that laminin, a ubiquitous glycoprotein forming basal lamina, may be one of the ligands for G-4. This finding is also in contrast to the observation by the same authors that a large amount of G-4 is accumulated near or at the basal cell membrane in polarized T84 epithelial cells. The explanation for this discrepancy remains unknown. We also found that, when G4-C containing two cysteine residues was isolated under nonreducing conditions, the protein rapidly lost its sugar-binding capability, probably due to oxidation of the residues as with other galectins possessing the same residues (Whitney et al. 1986; Hadari et al. 1995). Because the inactivated G4-C no longer binds to lactosyl-Sepharose, it is presently unknown how this domain of G-4 behaves in vivo after externalization into the ECM, where rich oxygen molecules are present, and whether this domain still plays some functional role after loss of its sugar-binding capability.