Identification of blottin: A novel gastric trefoil factor family‐2 binding protein

The trefoil factor family (TFF) peptides are important in gastro‐intestinal mucosal protection and repair. Their mechanism of action remains unclear and receptors are sought. We aimed to identify and characterise proteins binding to TFF2. A fusion protein of mouse TFF2 with alkaline phosphatase was generated and used to probe 2‐D protein blots of mouse stomach. The resulting spots were analysed by MS. The protein identified was characterised by bioinformatics, rapid amplification of cDNA ends, in situ hybridisation (ISH) and immunohistochemistry (IHC). Functional assays were performed in gastrointestinal cell lines. A single major murine protein was identified and named blottin. It was previously unknown as a translated product. Blottin is also present in rat and human; the latter gene is also known as GDDR. The predicted full‐length proteins are 184 amino acids long (20 kDa), reducing to 164 amino acids (18 kDa) after signal peptide cleavage. ISH of gastrointestinal tissues shows abundant blottin mRNA in gastric surface and foveolar epithelium. IHC shows cytoplasmic staining for blottin protein, and by immunoelectron microscopy in mucus granules and Golgi stacks. Previous work showed that blottin is down‐regulated in gastric cancers. Blottin contains a BRICHOS domain, and has 56% similarity with gastrokine‐1. Cultured HT‐29 cells express blottin and show increased DNA synthesis with antiblottin antibody; however, this effect is reversed by the immunising peptide. We have identified and characterised a TFF2‐binding protein produced by gastric epithelium. Blottin may play a role in gastrointestinal mucosal protection and modulate gut epithelial cell proliferation.

The TFF peptides are expressed in simple epithelia including the gut, salivary and lacrimal glands, and breast [1,2], where they are cosecreted with mucins [1]. There is tissue specificity along the GI tract: TFF1 and 2 are normally gastric, whereas TFF3 is intestinal [18], and they are up-regulated in wounded gut, particularly in the ulcer-asociated cell lineage (UACL) [19]. They are potent motogens which assist epithelial repair [2,4,9,20,21]. In tumours of the GI tract, their staining patterns may help to predict clinical outcome; TFF3 expression in gastric carcinomas is prognostic for a poor outcome [22].
In the stomach, the TFF1 knockout mouse suffers severe loss of mucus-secreting cells, that results in gastritis, followed by the development of adenomas and eventually carcinomas [23]. The TFF2-KO mouse displays slightly elevated gastric acid secretion and increased susceptibility to indomethacin-induced damage [24] and shows immunological alterations [25]. The TFF3-KO mouse suffers high sensitivity to colonic ulceration which may be fatal [26].
It is clear that the TFF peptides play an important role in gut function but their mechanism of action has yet to be elucidated. It is logical to postulate the existence of receptors, and the evidence to support the presence of TFFbinding proteins was recently reviewed [27]. Whole rat gut and electrically tight epithelial cell monolayers secrete chloride when exposed to basolateral TFF1 or TFF3 [28]. Ex vivo studies on stomach wounds reveal that serosal TFF2 protects against indomethacin-induced ulcerative damage better than luminal TFF2, and the peptide stimulates HT-29 cell motility [4]. Direct binding of b-gal-TFF3 to gutfrozen sections yields signals on the basolateral membranes of crypts of Lieberkuhn [29]. Iodinated TFFs have been tracked in pulse-chase infusions in rats and found in TFF2-secreting gastrointestinal epithelial cells, implying a basolateral uptake [30,31]. In yeast two-hybrid studies, mouse TFF1 binds to Muc2 and Muc5AC through their von Willebrand factor -C cysteine-rich domains [32]. These observations imply the existence of TFF receptor(s) which assist in activating the appropriate output behaviour, and which could become clinically relevant as pharmacological targets.
Herein we further explore TFF2 biology in the gut through the use of a proteomic approach to identify a binding protein, and a genomic approach to its characterisation.

Generation of TFF2-AP fusion protein
Full-length mouse TFF2 cDNA [25] from stomach tissue was inserted in frame at the 5' end into the secretory embryonic alkaline phosphatase (AP) fusion vector APTag2 [33] as follows. Primers were designed to add a Hind III cloning site at either end of the mTFF2 sequence and to remove the stop codon. The primers designed were as follows: forward: 5' ggg aag ctt atg ggc cct cga ggt gcg 3'; reverse: 5' c cca agc ttt gta gtg aca atc ttc cac ag 3'. Plasmid clones were sequenced to confirm fidelity.
Plasmids with or without mTFF2 were transfected into subconfluent human HEK293 cells using Superfect (Qiagen), and culture supernatants containing the fusion protein were collected after 6 days, centrifuged to remove cellular debris (4006g, 10 min), and filter sterilised.

Tissue binding of TFF2-AP
Organs from young adult C57 Bl/6J mice were either fixed in formalin and embedded in paraffin, or snap-frozen for cryosectioning. Gastric epithelium was stripped from the stroma by incubation for 15 min at 377C in 30 mM EDTA in calcium and magnesium-free DMEM, and either fixed in 4% paraformaldehyde (PFA) at room temperature for 30 min for wholemount in situ hybridisation (ISH) or immunohistochemistry (IHC), or homogenised as discussed below. Endogenous AP activity was quenched with 20% glacial acetic acid in methanol for 30 s, followed by washing in PBS. Sections or wholemount preparations were incubated in mTFF2-AP or control supernatants as described [33]. AP activity was revealed by coprecipitation of nitroblue tetrazolium (NBT) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP; both from Roche). Specificity of the binding was confirmed by the abrogation of staining when 2 mM unglycosylated, hTFF2 (a generous gift of Dr. L. Thim, Novo Nordisk, Denmark) was included. Sections were counterstained with nuclear fast red, and mounted in Glycergel (Dako). Wholemount ISH was performed using digoxigenin (DIG)-labelled probes on embryos (E12.5 to E18.5) and tissues fixed in 4% PFA, and processed as described previously [34].

Protein electrophoresis, binding of TFF2-AP and peptide identification
Standard protocols for 1-D [35] or 2-D [36] SDS-PAGE were used. The mouse tissues were homogenised on ice in a glass-Teflon homogeniser either directly in Laemmli buffer, or in 62.5 mM Tris (pH 6.8) containing 2% SDS with EDTA-free protease inhibitors ('Complete', Roche). Proteins were electroblotted to PVDF (BioRad) membranes at 800 mA for 60 min. Blots were not subjected to any renaturation steps.
Membranes were probed with mTFF2-AP; corresponding spots on silver-stained gels run in parallel were identified, then excised for identification of tryptic fragments by nano-LC-QTOF-MS (LC quadrupole TOF MS) [37]. The peptide ion spectra were crossmatched with the National Center for Biotechnology (NCBI, USA) database using the MS-Tag programme.

RNA isolation and analysis
The study was performed in accordance with appropriate local ethical guidelines. Total RNA was isolated from gastrointestinal tissue using TRIzol (Invitrogen, Paisley, UK), as previously described [38]. Rapid amplification of cDNA ends (RACE) was performed with the SMART-RACE kit (BD Biosciences Clontech UK, Oxford, UK), according to the manufacturer's instructions (primer sequences available on request). We generated a cDNA probe for human blottin from a sequence-verified I.M.A.G.E. Consortium cDNA clone (I.M.A.G.E. ID 1693411) [39] using the Rediprime ™ II DNA Labeling System and Redivue ™ [a-32 P]dCTP (both from Amersham Biosciences, UK). Our study of normal gastric antrum and gastric carcinomas of distal intestinal type using SAGE has been reported previously [38].

Antiserum production
Following identification of the TFF2-AP-binding protein, two peptides were synthesised with an additional cysteine at the N-terminus: one from the N-terminal (residues [29][30][31][32][33][34][35][36][37][38][39][40][41][42][43][44][45][46][47][48] which was predicted to be highly antigenic and hydrophilic, and one from the more hydrophobic C-terminal (residues 145-164). Peptides were conjugated to keyhole limpet haemocyanin (Imject, Pierce Chemical) and used as the immunogen for generation of antisera in New Zealand white rabbits. Only the N-terminal peptides produced immunoreactive antisera. Antibodies were affinity-purified using iodoacetyl gel, as described by the manufacturer (Ultralink, Pierce Chemical, Rockford, IL, USA). Briefly, 7 mg of a peptide were covalently linked to 2 mL packed volume of gel in coupling buffer, excess binding sites were quenched with 50 mM cysteine HCl, and the gel washed in PBS. Rabbit antiserum (1.5 mL)was applied to the column for 60 min at room temperature, then washed with six volumes of PBS and eluted into 861 mL fractions using 100 mM glycine (pH 2.5). Collecting tubes contained 50 mL of IM Tris base (pH 9.0) to neutralise the protein solutions. Fractions 3-5 contained most of the eluted antibodies and were pooled, dialysed against PBS in four sequential 600 mL steps over 18 h. Antibody concentration was 545 mg/mL, to which was added 0.1% BSA as carrier protein.

Tissue studies
Paraffin blocks were cut at 5 mm for 35 S riboprobe ISH [40] or IHC using the above antibody at 1:100, then a detecting goat antirabbit antibody conjugated to Alexa-546 was used at 1:200 (Molecular Probes, USA). Mouse mAbs to the human TFF1 (IgG1, affinity purified) and TFF2 (IgM, culture supernatant, also reactive against mouse TFF2) were used at 1:400 and 1:2, respectively, and detected using subclass-specific goat antimouse second layers conjugated to Alexa-488 and Alexa-633, respectively, again at 1:200. Sections were mounted in VectaShield Hard Set containing DAPI. Photomicrographs were obtained using SmartCapture X version 2.5.9 software (www.digitalscientific.co.uk) on an Olympus BX61 fluorescence microscope, and collages were made in Photoshop 7.0. The only manipulations undertaken were adjusting brightness and contrast.

Wounding assays
Cell lines were grown to confluence in 24-well plates and wounded with a sterile tip. Cells were washed twice in serum-free DMEM, then exposed to 10 mM hTFF2 or 3.4 nM mouse or human EGF in quadruplicate. Controls received equivalent volumes of PBS. Each group was exposed to either preimmune rabbit serum or rabbit antihuman blottin serum at 1:100, diluted in DMEM plus preimmune rabbit serum, producing a constant 2% rabbit serum. Wound widths (800 6 10 mm) were measured with an eyepiece graticule at five positions in each well for up to 30 h. Identical positions were scored by using an indexed XY plate holder. Rates of cell migration were computed from the regression lines of wound widths versus time.

Growth assays
Five-thousand cells per well were incubated overnight in 96well plates in 100 mL DMEM with 2% preimmune rabbit serum, rinsed with serum-free DMEM before adding antihuman blottin at final dilutions of 1:100, 1:200, 1:400 and 1:800, with preimmune serum added to maintain a constant 2% total. Half of the wells received 10 mg/mL immunising peptide or PBS alone. Plates were removed at intervals up to 9 days, washed twice with PBS, drained and frozen at 2207C until DNA analysis. Cell DNA was assayed using Hoechst 33258 binding [44]. All assays were performed in quadruplicate.

Statistical analyses
Wounding and growth assays were designed to test for differences in either cell migration or growth rates as measured by serial observations in the same wells at the same position (migration), or DNA content of a series of separate 96-well plates (growth). Data were subjected to linear regression, ANOVA and multiparameter analyses (Student-Newman-Keuls Multiple Comparisons Test) using Instat 3 software (GraphPad, San Diego, USA). The Null hypothesis was rejected if differences were likely at a frequency of 1:20 or more (equivalent to a p value of ,0.05).

Generation of mouse TFF2-AP fusion protein and tissue binding
Following the generation of the TFF2-AP fusion protein, mouse tissues from the skin, lung, liver, kidney and regions of the GI tract were analysed for binding activity. The only mouse tissue which gave a strong signal was the stomach, in which strong staining was seen at the foveolar surface ( Fig. 1B), compared to the control AP-alone incubation (Fig. 1A). This binding could be partially inhibited by the presence of hTFF2 (2 mM) in the binding buffer ( Fig. 1C).
Binding was also seen further down the gastric glands, and some of this staining was not displaced by hTFF2. A similar foveolar staining pattern was seen for human stomach probed with the same fusion protein ( Fig. 1D and E), which was partially blocked by hTFF2 (Fig. 1F).

Protein electrophoresis, binding of TFF2-AP and peptide identification
Mouse tissue extracts were subjected to SDS-PAGE electrophoresis, transferred to PVDF membranes and probed with mTFF2-AP (Fig. 1G). A strong 1-D signal at around 20 kDa was seen in stomach epithelium alone, and which gave a single 2-D signal at a pI of 6.9 (Fig. 1H, arrow). The corresponding silver-stained protein was identified by nano-LC-QTOF-MS as the protein product of a cDNA (RIKEN cDNA 1810036H07, see below), which was at that time uncharacterised in the mouse genetic databases. Due to the number of membrane blots used to identify the protein, we called it blottin. A separate small doublet on the 2-D gastric epithelial blots (Fig. 1H, arrowhead) at an apparent MW of about 66 kDa, and approximate pI of 5 could not be identified due to lack of material.

Bioinformatics and structure of blottin genes, mRNAs and proteins
Comparison of the protein sequence from the MS with the public gene expression databases using NCBI's BLAST search tools yielded matching cDNAs from the RIKEN Consortium in Japan, which were highly expressed in murine stomach (RIKEN cDNA 1810036H07) [45] ( Fig. 2A).
For each species, primers were designed from the existing sequences for 5' and 3' RACE, which provide full-length mRNA sequences. The human, mouse and rat blottin mRNA sequences (GenBank accession numbers AY943908, AY943907 and AY 943909) are presented and compared in Supplementary Fig. 1, along with a table of the current NCBI nomenclature, mRNA accession numbers and UniGene cluster numbers for blottin in each species. The three blottin transcripts average around 750 bp in length and show strong evolutionary conservation. The corresponding genes were identified by searching the public genomic databases with the cDNAs. The gene structure and full sequence for blottin in each species are also presented and compared in Supplementary Fig. 1. All three genes contain six exons spanning 7-8 kb with a similar exon-intron structure. The blottin gene maps to mouse chromosome 6D2, human chromosome 2p13.3 and rat chromosome 4q34; these regions exhibit conservation of synteny.
The mouse, rat and human blottin amino acid sequences are presented in Fig. 2A. The predicted full-length translation products contain 184 amino acids containing a transmembrane helical signal peptide, with post-translational cleavage 20 amino acids from the N-terminus, indicating that the mature protein is either expressed on the extracellular surface or secreted. The predicted and observed mature blottin protein is around 18 kDa, with an observed (mature) pI of 6.9. Predicted myristylation sites in the N (one) or C (three) terminal regions were not confirmed by our MS analyses. The mouse blottin protein sequence has been submitted to the Swiss-Prot/TrEMBL database where its accession number is Q9CQS6.
The blottin protein contains a so-called BRICHOS domain (residues 54-151; Fig. 2A). These domains show evolutionary conservation from C. elegans to humans, but are of unknown function. They are present in a range of proteins, including lung surfactants, proteins associated with dementia and another gastric protein [46] known as gastrokine-1 (GKN1), previously described as CA-11, AMP-18 and foveolin [47,48]. GKN1 shows strong homology with blottin. At the protein level, mouse blottin and mouse GKN1 are 24% identical and 56% similar, as demonstrated by their comparison in Fig. 2B. At the genomic level, blottin and GKN1 share the same intron-exon structure and are located side-by-side on the same chromosomes in human, mouse and rat. were precoloured molecular weight standards (10 mg/ lane). Blots exposed to AP supernatant alone gave no signals (not shown). Lanes: 1, kidney; 2, lung; 3, standards; 4, distal colon; 5, proximal colon; 6, small intestine; 7, duodenum; 8, stomach; 9, standards; 10, liver. Note lane 9 contains a small overspill from lane 8. 2-D electrophoresis of gastric epithelium (H) used 250 mg of protein, and was treated as for the 1-D blot. Arrow: strong single spot (blottin) at approximate pI 6.9, MW 20 kDa. Arrowhead: a small mTFF2-AP-positive doublet which could not be analaysed further, at approximately 66 kDa, pI 5.

Blottin mRNA and protein expression
In a previous study, we performed large-scale gene expression profiling using SAGE to compare normal gastric antral mucosa with gastric adenocarcinomas of intestinal type [38]. We showed that gastrin, MUC5, TFF1 and TFF2 were highly expressed in normal stomach but were down-regulated in the tumours [38]. The same expression pattern was seen with GKN1 by SAGE; this was subsequently validated by ISH and IHC [38,47]. The SAGE data (not shown) for blottin were similar: blottin was relatively highly expressed in normal antral mucosa (with tag TTTAGGATGA, at 15 transcripts per 10 000 total, thus 0.15% of mRNA) but was absent from the gastric tumours studied (zero transcripts identified) [38].
For tissue localisation of mouse blottin mRNA, we generated a digoxigenin-labelled antisense oligonucleotide probe for ISH). This was used on wholemount mouse gut tissues from a series of embryos from day E12.5. Blottin transcripts were detected only in the stomach, and only after E17.5, where there was strong staining at the mucosal surface in the body and pyloric regions, as well as Brunner's glands of the proximal duodenum ( Supplementary Fig. 2). . This is identical to the translation product predicted from the mRNA ORF (GenBank mRNA accession number AY943907). Such predictions are the source of the human and rat protein sequences (AY943908 and AY943909). Comparison is performed with CLUSTALW, in which red indicates small and hydrophobic amino acids (AVFPMILWY); blue indicates acidic amino acids (DE); magenta indicates basic amino acids (RHK); and green indicates amino acids with hydroxyl or amine side chains (STYHCNGQ). In the 'consensus' line, the degree of conservation is denoted as follows: '*' indicates identical residues; ':' indicates conserved substitutions, according to the colours above; and '.' indicates semiconserved substitutions. Primary translation products are predicted to be 184 amino acids long, with a signal peptide that is posttranslationally cleaved 20 amino acids from the N-terminus (arrow). The proteins contain a BRICHOS domain (residues 54-151). Blottin contains five cysteine residues which are indicated by black boxes in the consensus line and are conserved between species. The proteins are clearly highly conserved: for example, the human and mouse proteins (including signal peptide) show 70% identity and 80% similarity. The 20-mer peptide at residues 29-48 was used for antibody production. (B) Mouse blottin and GKN1 proteins. Sequence similarity searches show that blottin is similar to GKN1 at the mRNA and protein levels. Here, mouse blottin protein has been aligned with mouse GKN1 [47]. The proteins show 24% identity and 56% similarity, such that the structural description above applies equally to GKN1. The only exception is that the first cysteine residue in blottin, indicated by a black box containing 'C' in the consensus line, is absent from GKN1.
The same oligonucleotide probe to mouse blottin was 35 Slabelled and was used on formalin-fixed paraffin embedded mouse tissue sections, whereas a 35 S-labelled antisense riboprobe to human blottin was used for human tissue sections. As with the wholemount samples, we found strong signals for blottin mRNA at the mucosal surface and adjacent foveolar cells of the stomach (Fig. 3A and B). This contrasts with the protein localisation pattern seen with direct binding of the fusion protein mTFF2-AP, with which staining was also seen in lower portions of the gastric glands in both human and mouse ( Fig. 1B and E). The specificity of the signal is indicated by its reduction in the presence of human TFF2 (Fig. 1F) and this staining pattern may reflect uptake of blottin by cells deeper in the glands.
By immunofluorescent microscopy, mouse blottin (red immunofluorescence) was seen in the stomach glands of both frozen and formalin-fixed paraffin embedded tissue, and was clearly separated from TFF2 signals (electronically coloured white immunofluorescence) lower in the glands, but also seen in the surface mucus layer (Fig. 3D), in contrast to the image for the control section without primary antibodies (Fig. 3C). The same was true with the antiserum against human blottin on human material, where the location of TFF1 (green immunofluorescence, Fig. 3G) was similar to blottin (red immunofluorescence, Fig. 3H), and again clearly separate from TFF2 (white immunofluorescence, Fig. 3G and H). We further showed that direct binding of mTFF2-AP to mouse stomach frozen sections was inhibited only by the presence of hTFF2, and not by either the immunising peptide or the antibody against blottin (data not shown).
We checked that our antibodies were specific by using the human gastric carcinoma line AGS, since these cells do not express GKN1 or blottin. We then transfected these cells with an expression vector containing GKN1. Thereafter, protein extracts of these cells could be demonstrated by Western blotting to contain GKN1 (unlike the parental cell line). However, these cells lacked any staining for blottin, unlike the gastric extract run simultaneously (Supplementary Fig. 3). Thus our antibodies to blottin and GKN1 do not crossreact.

EM of mouse stomach
We then proceeded to study the ultrastructural localisation of blottin by immuno-electron microscopy using the specific blottin antiserum. Again, mouse gastric epithelium showed abundant blottin in the surface and foveolar (pit) cells. Specificity of the signal was confirmed by its abrogation by coincubation with the immunising peptide (data not shown). Binding was restricted to the mucus granules, with evidence of secretion into the mucus layers at the luminal surface (Fig. 3I), and the Golgi stacks (Fig. 3J). No cell membrane, junctional complex, ER, nuclear or mitochondrial staining was seen.

Blottin expresssion in vitro
In vitro assays of gene function are enabled by the use of cell lines. We tested three human carcinoma lines for blottin immunoreactivity: one gastric (HGT-101), one colorectal (HT-29) and one breast (MCF-7). The two gastrointestinal cell lines expressed blottin. HGT-101 cells were stained by double immunofluorescence for TFF1 or TFF2 and blottin; some cells express each protein separately, and in areas of confluence some cells coexpressed TFF1 and blottin (Fig. 4). HT-29 cells did not stain for TFF2 (not shown). MCF-7 cells, although they express TFF1 and TFF3, did not express blottin and were not studied further.

Wound and growth assays of blottin function in cell lines
Since only the gastrointestinal cell lines expressed the blottin protein, we performed wound assays, similar to those reported by Marchbank et al. (1998) for TFF peptide-like responses [9], to test whether the antibody to blottin was able to interfere with short-term epithelial restitution in vitro. Neither HGT-101 nor HT-29 cells were affected to any significant extent by the presence of up to a 1:100 dilution of the antibody to human blottin (not shown), compared to preimmune serum at the same total concentration (2%).

Growth assays
Using a strategy similar to the wounding assays, we studied the effects of antiblottin antibody on cell growth. Assays were performed in a total of 2% rabbit serum. There was little effect on HGT-101 cells (not shown). The growth rate of HT-29 cells was increased by the antiblottin antiserum. This effect was partly blocked by the immunising peptide, which on its own caused reduced growth (Fig. 4E).

Discussion
We report here, in confirmation of our earlier work [49], the discovery and characterisation of a murine gastric foveolar protein which binds to mouse TFF2, a peptide which itself plays a central role in gastrointestinal wound healing [2,3]. This was achieved through proteomic methods by using a direct binding approach on mouse tissue sections, isolating the gastric epithelium involved, confirming the signals on 1-D electrophoretic blots, and then identifying a major signal on 2-D gels and blots. The identified and isolated protein was sequenced and compared to the online protein and genetic databases. The results show a well-conserved protein on syntenic chromosomes in the mouse, human and rat, confirming and extending both our previous report [49] and the recent results of Du et al. [50,51] and those of Westley et al. [52]. Our RACE results for the human mRNA sequence (GenBank accession number AY943908) differ slightly from the two mRNA sequences so far published (AF494509, from the GDDR paper [50]; and AY358664), as explained in detail in Supplementary Fig. 1. However, our mRNA sequence agrees with both published ESTs and with the genomic sequence, and thus we believe it to be correct. Interestingly, one of the differences (in AY358664) is a 24 bp deletion which starts at the existing translation start site and creates a new one, in-frame: presumably this could therefore represent a splice variant, although the sequences involved are not those of standard GT/AG intron-exon boundaries.
The predicted amino acid sequence has a 20-mer helical signal peptide, a more hydrophilic N-terminal region, and a more hydrophobic C-terminus. This correlates well with the positions of predicted four myristylation sites, one towards the N-terminus, and three near the C-terminus. We did not, however, find these adducts on peptide fragmentation analysis, so this prediction cannot yet be substantiated. Indeed at the level of immuno-EM, mouse gastric foveolar cells displayed no gold particles in either the lipid bilayer or junctional complexes of the cell membrane. This is in marked contrast to the abundant particles found specifically inside the mucus granules, Golgi apparatus, and ER and in the secreted mucus layer. We conclude that either the potential myristylation sites may be rapidly modified, or any lipid membrane insertion is a minor event. Histological evidence at the light microscopic level of close proximity of blottin to foveolar apical surface membranes may thus reflect the positions of secretory granules ready to undergo exocytosis, as well as extracellular material recently ejected.
We were careful to analyse our mass spectrometric data to enquire about further possible adducts. We were unable to find evidence of either N-or O-glycosylation products, and we conclude that these modifications do not occur at any significant levels in the tissues studied.
We observed that there was some binding of the mTFF2-AP ligand to portions of the gastric glands which did not express the blottin mRNA (Fig. 1). Possible explanations include: (1) the blottin synthesised in mucous neck and more superficial cells is secreted and taken up by cells deeper in the gastric gland; (2) cells migrating from the stem cell zone express blottin mRNA transitorily, but retain some of the blottin they synthesised. Either way, there may be unbound blottin which could adsorb the ligand in free solution as applied to the sections, and reveal an apparently paradoxical signal. Previously we reported a similar discordant distribution of TFF2 mRNA and protein in human gastric foveolar epithelium [53]. It is thus possible that the trefoil-blottinbinding phenomena in vivo could be more complex than we expect. These interactions clearly need further study.
We have studied the expression of mouse and human gastric blottin using species-specific rabbit antisera and 35 S-riboprobe ISH. Both proteins and their mRNAs are highly expressed in normal gastric surface and foveolar (pit) epithelial cells. The antibody study has been confirmed in murine stomach at the electron microscopic level, where blottin appears to be cosecreted with mucin granules, in an analogous manner to TFF1, which is expressed in the same cells [54]. The protein staining patterns were abrogated by coincubation with the immunising peptide, confirming their specificity. We have observed that the location of blottin expression is distinct from its putative binding partner TFF2, which is found deeper in the gastric glands. The reason for this is unexplained, but may reflect the need for glandular secretions to be able to flow upwards into the mucus gel layer, in which interactions between TFF and mucin are held to be important in maintaining the integrity of the mucus barrier [10]; the secretion of blottin may modulate this process.
In our cultured HGT-101 gastric cancer cells, we noted that TFF1 and blottin could be coexpressed, but not blottin with TFF2. In contrast, TFF2 and blottin were seen in adjacent but not the same cells. These differences may reflect some residual differentiation capacity in the cell line. The colonic HT-29 cells expressed blottin protein at similar levels, but the levels of TFF1 were low by immunofluorescence, and TFF2 was absent (data not shown).
We are confident that our antiserum to blottin is specific by comparison with GKN1, since Western blots of the AGS gastric line transfected with a GKN1 protein expression vector stain only for GKN1, but lack any blottin signal (Supplementary Fig. 3). We have also found that the binding of blottin antibodies to Western blots of murine gastric epithelium can be displaced by the immunising peptide, but that antibodies to GKN1 are not (data not shown), in a similar manner to the binding of mTFF2-AP fusion protein, which can be displaced by hTFF2. This indicates that our reagents are specific.
A similar predicted protein in humans was recently reported [50] (GenBank AF494509). These authors used a subtractive library to screen for down-regulated genes in gastric cancers. In addition, a homologous rat EST was identified. These authors later showed that a gastric cell line 7901 displayed reduced growth when transfected with a GDDR vector [51]. However, there were no data suggesting any link to TFFs or other gastric proteins.
A further recent report confirms that a human homologue of the mouse blottin protein can be extracted from gastric cytosol and cell lines. That protein, which was named TFIZ1 by the originating group, binds to TFF1 by a covalent linkage on its free cysteine 58 [52]. These authors did not report any binding of TFF2 with TFIZ1, which our data would predict. It is possible that when TFF1 is covalently bonded to TFIZ1 there is steric hindrance that disallows further TFFs (or other molecules) to bind. The fact that the TFF2-blottin, but not TFF1-blottin, binding was detected ex vivo in the mouse suggests that there may be species differences in the native binding preferences between these proteins. These aspects will be of great interest to study further.
We have attempted to explore blottin's functional role by using two blottin-expressing gastrointestinal cell lines in in vitro assays of growth and short-term wounding (modelling mucosal restitution in vivo). We reasoned that blottin secreted by the cells may act in to an autocrine manner to modulate their behaviour, and that any effect may be disrupted by the presence of antibodies to the protein. We did not find any significant effects in cell wounding assays in either gastric (HGT-101) or colonic (HT-29) cell lines, both of which express the blottin protein. However, blottin antiserum altered the growth of HT-29 cells, as measured by the rate of DNA synthesis. The antiserum appeared to accelerate cell growth, whereas the immunising peptide reduced growth, an effect reversed by adding blottin antiserum. The gastric cell line HGT-101, paradoxically, did not show this effect. These results are corroborated by those of Du and colleagues [51], who recently reported that the human protein homologue to mouse blottin, GDDR, when transfected into SGC 7901 human gastric cancer cells in vitro, decreased their growth, as measured by an MTT assay. The mechanism of these effects is as yet unknown. We were interested to find that the blottin protein could be expressed ectopically in the reparative epithelial monolayer in the colon of Crohn's disease, which suggests that the apparent paradoxical expression of blottin by HT-29 cells has some in vivo equivalent. We cannot yet ascribe a function for the protein in this context, but note that there is a similar ectopic expression of TFF1 and TFF2, normally confined to the stomach, in the surface monolayer epithelium of Crohn's disease and other inflammatory conditions of the gastrointestinal tract, in particular the ulcer-associated cell lineage (UACL) [55,56], generally considered to be attempts to repair those epithelia [2,19,57].
Receptors are commonly expressed in a polarised manner in monolayer epithelia. For example, the EGF-receptor is found in a basolateral position in gut epithelia [58], while TFFs may affect chloride ion pumping when applied to basolateral aspects of cell monolayers or whole gut preparations in the Ussing chamber, but not using an apical route [11,28]. We do not find the expression of blottin to be polarised in this manner, and conclude that it may be premature to suggest that blottin is a true TFF-receptor, in the classical pharmacological sense. However, like the TFFs, blottin's mode of action may be enhanced in a wounding situation, where exposure of a denuded epithelium would allow api-cally secreted blottin to access the basolateral domain and play a role in repair, perhaps by influencing mucus viscosity.
In conclusion, we describe a new murine protein, blottin, which binds to mouse TFF2 and is found in the gastric surface and foveolar epithelial cells. The protein is well conserved in mammals, has a human homologue GDDR/ TFIZ1, and has similarities with another gastric foveolar protein, GKN1 [47,48]. The genes for both proteins are well conserved on syntenic chromosomes across species. Blottin may inhibit cell proliferation, in contrast to GKN1 which is a mitogen [48]. Blottin is seen in the reparative surface epithelium of the colon in the UACL of Crohn's disease, in a location similar to that we have previously reported for TFF1 [55,56]. We suggest that blottin may also play a role in gastrointestinal mucosal maintenance and repair through interactions with the TFF peptides.