Squamous cell carcinomas (SCCs) of the oral cavity, pharynx and larynx remain a significant public health problem. They represent 2–3% of all malignancies, and their incidence, particularly that of oral SCCs (OSCCs), is increasing in Western countries.1 In spite of improved therapeutic procedures, the prognosis of OSCC patients remains poor and considerably lower than that of other neoplasias.2 This fact can be attributed to several factors: failure to respond to available chemotherapy, late presentation of the lesions and lack of suitable markers for early detection and prognosis.3, 4 Hence, the finding of novel tumour markers, particularly those associated with tumour cell invasion and spreading, can help provide a more accurate evaluation of prognosis and a more efficient management of the disease.
PA2.26 antigen was identified in our laboratory as a cell-surface protein induced in murine epidermal keratinocytes and dermal fibroblast-like cells during wound healing and chemical carcinogenesis.5 Sequence analysis of the isolated cDNA and biochemical characterization of the protein revealed that murine PA2.26 is a small mucin-like transmembrane glycoprotein of about 45 kDa,6 highly homologous to the rat alveolar type I cell marker T1α and the podocyte-associated glycoprotein podoplanin.7, 8 Murine PA2.26 nucleotide sequence is almost identical to that of OTS-8 and gp38, markers of the osteoblastic cell lineage and stromal cells in peripheral lymphoid tissues, respectively,9, 10 and completely matches the nucleotide sequence of RANDAM-2, a recently discovered membrane glycoprotein expressed in neuronal cells during mouse neurogenesis.11 Expression of PA2.26/T1α is developmentally regulated and has also been detected in the choroid plexuses, ependyma, mesothelia, endothelia of lymphatic vessels and ciliary epithelia, located at the apical surface and microvillous projections.6, 12 Other homologues include the rat glycoproteins RT140 and E11,13, 14 the canine receptor for the influenza C virus gp40 and its human homologue gp36,15, 16 and the platelet aggregation-inducing sialoglycoprotein Aggrus.17 The generation of knock-out mice containing a targeted mutation in the T1α /PA2.26 gene locus has provided important insights concerning the involvement of this glycoprotein in normal tissue development. T1α null mice die at birth due to respiratory failure caused by anomalies in the development of alveoli.18 These mice also exhibited defects in lymphatic but not blood vessel pattern formation, associated with lymphedema, dilation of lymphatic vessels and diminished lymphatic transport.19 Although the precise function of T1α /podoplanin/PA2.26 in normal tissues remains unknown, T1α null mouse defects have been attributed to disruption of epithelial-mesenchymal signalling18 and to impaired cell to substratum adhesion and migration.19
Our previous studies suggested that induction of PA2.26 in mouse epidermal cells and tumours was related to cell migration and malignant progression.6, 20 For this reason, we focused on characterizing the human homologue of mouse PA2.26 and on studying its involvement in human cancer. In our study, we report the molecular characterization of human PA2.26. Both mouse and human PA2.26 exhibit similar structural and biochemical characteristics and, likely, share the same function. We show that PA2.26 is induced in a subset of human early OSCCs, associated with downregulation of membrane E-cadherin protein expression. Furthermore, PA2.26 promotes the formation of cell-surface extensions and diminishes cell-cell adhesiveness when expressed in cultured human keratinocytes. These results suggest that PA2.26 might play an important role in tumour progression in human oral cancer.
Mouse PA2.26 sequence was tested for similarity to known sequences deposited in the GenBank database using BLAST software.21 Four overlapping IMAGE cDNA clones, 41487, 154633, 881369 and 1031768, containing a cDNA sequence highly homologous to that of mouse PA2.26 were purchased from the Medical Research Council Center (Cambridge, UK, Human Genome Mapping Project) and sequenced with an ABI PRISM 377 machine (PE Applied Biosystems, Foster City, CA). A CpG-rich region extending from exon 1 to intron 1 was predicted by the GRAIL/cpg program at the Human Genome Mapping Project (http://menu.hgmp.mrc.ac.uk/menu-bin/Nix/Nix.pl). The chromosomal localization of the gene was performed by using the BLASTN and TBLASTN programs against the human genome (National Cancer for Biotechnology Information, NCBI). PA2.26 sequence was assigned to the NT_004577 contig localized at 1p36.13.
A partial human PA2.26 cDNA containing the full coding region was amplified from a human placenta cDNA library (Clontech, Palo Alto, CA) using the following oligonucleotides: 5′-CGGGATCGATGTGGAAGGTGTCAGC-3′ and 5′-CCGCTCGAGGGCCACAGAAGTCAGAAACG-3′ (nucleotides 196–763 of the sequence reported in Fig. 1a). PCR conditions were 30 cycles: 95°C for 1 min, 59°C for 1 min and 72°C for 1 min. The resulting 568 bp cDNA fragment was cloned into the pGEM-T Easy vector (Promega, Madison, WI), sequenced in both directions, and subcloned into the pcDNA3 expression vector (Invitrogen, San Diego, CA).
To express human PA2.26 fused to the N-terminal end of the enhanced green fluorescent protein (PA2.26-GFP), the PA2.26 cDNA fragment containing the full coding region was amplified by PCR using oligonucleotides 5′-GTGCTGGAATTCCCCGATGTGG-3′ and 5′-TCAGGTACCCTGGGCGAGTACCTTCC-3′, that introduce an EcoRI site (GAATTC) at the 5′ end before codon 1 and a KpnI site (GGTACC) at the 3′ end, after codon 162, eliminating the stop codon. The product was subcloned into the pEGFP-N1 expression vector (Clontech). Sequencing confirmed the correct ligation of the PA2.26 cDNA fragment.
Northern blot analysis
Commercial Northern blots containing poly (A)+ RNA from adult human tissues (Clontech) were probed with PA2.26 and β-actin cDNAs according to the manufacturer's instructions. The probe for PA2.26 was a fragment of 568 bp cloned from placenta (see above).
Production of antibody
A polyclonal Ab against human PA2.26 was raised in a rabbit by immunization with the synthetic peptide Leu-Glu-Gly-Gly-Val-Ala-Met-Pro-Gly-Ala-Glu-Asp-Asp-Val-Val, comprising amino acids 37–51 of the protein ectodomain (P37-51), coupled to keyhole limpet hemocyanin (Isogen Bioscience BV, Maarssen, The Netherlands). A subcutaneous injection of 500 μg of P37-51 (333 μg of the protein carrier) in complete Freund's adjuvant was given at multiple sites in the back of the animal, followed by 2 booster injections in incomplete adjuvant at 2-week and 2-month intervals. The antiserum was concentrated by ammonium sulfate precipitation (50% saturation) and dialyzed against PBS/sodium azide. IgGs were purified by affinity chromatography through protein A-Sepharose 4B (Sigma-Aldrich, Madrid, Spain).
In vitro transcription/translation reaction
The PA2.26 cDNA subcloned into the pcDNA3 vector was in vitro transcribed and translated in the TNT-coupled wheat germ or rabbit reticulocyte extract systems (Promega) according to the manufacturer's instructions. The products of the reaction were fractioned on 12% SDS-PAGE gels and PA2.26 detected by Western blotting.
cDNA transfection and cell culture conditions
HeLa and HaCaT cell lines were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (Gibco Invitrogen, Barcelona, Spain) and antibiotics (25 μg/ml amphotericin B, 100 μg/ml ampicillin and 32 μg/ml gentamicin; Sigma-Aldrich). Cultures were maintained on plastic plates at 37°C in a 5% CO2 humidified atmosphere. HaCaT keratinocytes (passage 33) were kindly provided by Dr. Norbert E. Fusenig (German Cancer Research Center, Heidelberg, Germany). HeLa cells were obtained from the American Type Culture Collection (Rockville, MD).
For transient transfections, cells were seeded on glass coverslips 24 hr before DNA transfer and then transfected with the PA2.26 cDNA subcloned into the pcDNA3 vector, using Effectene Transfection Reagent (Quiagen, Valencia, CA). Immunofluorescence analysis was done 24 hr after transfection.
For stable transfections, cells cultured on plates were transfected with the PA2.26-EGFP construct or the empty pEGFP-N1 vector, as above. Transfected cells were selected by growing in DMEM containing 10% fetal calf serum and 0.5 mg/ml of G418 for 3 weeks. Individual clones were isolated with cloning rings and characterized for PA2.26 expression by flow cytometry analysis and Western blotting.
Western blot analysis and glycosidase digestion
For detection of PA2.26 in Western blots, cells or tissues (obtained from the “Hospital Universitario de la Princesa”, Madrid, Spain) were lysed in buffer RIPA (0.1% SDS, 0.5% sodium deoxycholate, 1% Nonidet P-40, 150 mM NaCl, 50 mM Tris-HCl, pH 8.0) and a cocktail of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 μg/ml aprotinin and 2 μg/ml leupeptin). Samples containing the same amount of protein (30 μg) were run on 10% or 7% SDS-PAGE and transferred to Immobilon P membranes (Millipore, Bedford, MA). Filters were immunoblotted with the PA2.26 Ab (at a dilution of 1:400 in PBS containing 1 mg/ml BSA) or preimmune serum at an equivalent dilution. As an additional control, the Ab was preincubated with an excess of either P37-51 (3 μg/ml) or an unrelated peptide (Gly-Ala-Ser-Lys-Cys-Asp-Gly-Phe-Arg-Ser, Pcontrol) for 30 min. As secondary Ab, anti-rabbit IgG coupled to horseradish peroxidase (Nordic Immunological Laboratories, Tilburg, The Netherlands) was used. For detection of E-cadherin and α-tubulin in Western blots, the mAbs ECCD-2 and DM1A (Sigma-Aldrich) were used at 1:100 and 1:10,000 dilution, respectively. Peroxidase activity was detected using an enhanced chemiluminescence kit (ECL, Amersham, Arlington Heights, FL).
PA2.26 digestion with glycosidases was performed as previously described6 in whole cell lysates (in buffer RIPA) obtained from HeLa cell transfectants, followed by Western blotting using the PA2.26 polyclonal Ab.
Paraffin-embedded sections of OSCCs and premalignant lesions were obtained from the archive of the Department of Clinical Pathology at the “Hospital Universitario de la Princesa” (Madrid, Spain). OSCC patients did not receive any therapeutic treatment prior to local resection surgery. All cases were confirmed to have negative lymphatic nodes by histology and were therefore classified as intraoral-confined disease. Clinical follow-up data were retrieved from the oral carcinoma database of the Department of Oral and Maxillofacial Surgery, Hospital Universitario de la Princesa. The minimum follow-up period was 3 years. The mean follow-up was 5 years.
PA2.26 immunostaining was performed directly in deparaffinized sections by either the Envision plus peroxidase method or the Labelled streptavidin-biotin (LSAB) 2 system, alkaline phosphatase (Dako Cytomation, Glostrup, Denmark) or the avidin-biotin alkaline phosphatase method. The Ab was used at 1:400 dilution before and after preincubation with P37-51 and Pcontrol peptides. Sections were dehydrated in graded ethanols, cleared in xylene and mounted in Permount after counterstaining with hematoxylin and eosin. In all cases recorded as positive for PA2.26 immunostaining, labeling was proven to be blocked by preincubation of the Ab with P37-51. E-cadherin immunostaining in tumour sections was performed by heat-induced antigen retrieval in a press cooker with EDTA solution, pH 8.0, for 2 min, using mAb 4A2C7 at 1:500 (Zymed Laboratories, San Francisco, CA).
Confocal immunofluorescence microscopy analysis
Double immunofluorescence detection of PA2.26 and ezrin in HaCaT and HeLa cells transiently transfected with PA2.26 cDNA was performed on cells fixed with 3.7% formaldehyde in PBS, permeabilized with 0.05% Triton X-100. PA2.26 polyclonal Ab (before and after preincubation with P37-51 and Pcontrol peptides, see above) and anti-ezrin mAb 3C12 (Sigma-Aldrich) were used at 1:100 dilution. FITC-labeled anti-rabbit or TRITC-labeled anti-mouse IgGs (Jackson, West Grove, PA) were used as secondary antibodies, respectively. For F-actin staining, phalloidin coupled to Rhodamin was used. Staining of nuclei was performed in a solution of 4′, 6-diamino-2-phenilindole (DAPI, 1 μg/ml).
Immunofluorescence detection of E-cadherin and β-catenin in PA2.26-EGFP stable transfectants was carried out on confluent cells grown on glass coverslips, fixed in cold methanol,22 using the mAbs 4A2C7 (Zymed Laboratories) and C19220 (Transduction Laboratories, Lexington, KY) at 1:100 and 1:250 dilutions, respectively, and appropriate TRITC-labeled secondary Abs.
Double-label immunofluorescence detection of PA2.26 and ezrin or E-cadherin in tumour sections was performed after heat-induced antigen retrieval and proteinase K digestion for 10 min at 37°C. Sections were blocked in PBS containing 1% BSA and incubated with PA2.26 Ab and anti-ezrin 3C12 or anti-E-cadherin 4A2C7 mAbs. Alexa Fluor 588 goat anti-rabbit IgG or Alexa Fluor 594 goat anti-mouse IgG (Molecular Probes Europe BV, Leiden, The Netherlands) were used as secondary Abs. Tissue sections were then mounted in Mowiol and examined with a confocal microscope.
Confocal laser scanning microscopy was performed with a Leica TCS-SP2 (Leica Lassertechnik GmbH, Heidelberg, Germany) adapted to an inverted Leitz DMIRB microscope. Images were taken using a 63× (NA 1.4 oil) Leitz Plan-Apochromatic objective. Ten to 25 optical sections of 0.3 μm were made for 3-dimensional reconstruction through the whole depth of the cells, and projections of 5–6 sections were used for reconstruction of particular cellular domains. For tissue samples, 12 optical sections of 0.5 μm were made. Images were assembled using Leica confocal software 2.0.
Cell aggregation assay
Ca2+-dependent cell aggregation assays were performed according to Navarro et al.22 Briefly, cell cultures were dissociated into single-cell suspensions under E-cadherin-saving conditions. Cells were suspended in 10 mM Hepes, pH 7.4, 150 mM NaCl containing 0.5% BSA and 10 mM CaCl2 or 1 mM EGTA and incubated under gyratory shaking for 60 min. The degree of cell aggregation was expressed by the aggregation index 1−(N60/N0), where N0 and N60 indicate the initial number of particles and the number of particles after 60 min of aggregation, respectively. All N0 and N60 measurements were done in duplicate, and the experiments were repeated at least twice.
cDNA cloning, exon/intron organization and chromosomal mapping
Several ESTs with significant homology to the mouse PA2.26 mRNA6 were identified by database comparison. Four overlapping IMAGE clones containing the full-length cDNA were acquired from the Medical Research Council Center (Cambridge, UK, Human Genome Mapping Project) and their sequences determined. The 2737 nt cDNA obtained (Fig. 1a) contains a single polyadenylation consensus sequence AATAAA and an open reading frame (ORF) encoding a 162-amino acid protein highly homologous to the 172-amino acid mouse PA2.26 protein. The human PA2.26 protein exhibits the typical structure of a type-I membrane mucin-like protein with an ectodomain containing a high proportion of potential O-glycosylation serine and threonine residues and a presumptive spanning membrane domain followed by a short cytoplasmic tail of 9 amino acids (Fig. 1a,b). The cytoplasmic domain of the human protein conserves a cluster of 3 basic amino acids (Fig. 1b) shared by transmembrane proteins of microvillal location23 that appears to be responsible for PA2.26 binding to proteins of the ezrin, radixin, moesin (ERM) family.6
Database comparison of the complete 2737 nt cDNA sequence presented in Figure 1a revealed the presence of several ESTs belonging to the same transcriptional unit, further extending the 5′ UTR end by 8 nt (access to GenBank BE563074/AA301065), by 37 nt (AU133042) and by 114 nt (BE887331), indicating the possible existence of other transcription initiation sites. Thus, a larger cDNA sequence registered in the GenBank (under accession no. NP-006465) contains an alternative ATG start codon 5′ upstream of the sequence presented in Figure 1a, predicting a new ORF, which includes the coding sequence reported here plus additional 76 N-terminal amino acids. We believe that this larger, alternative polypeptide is unlikely to be expressed in human tissues because it lacks a signal peptide in the amino terminus for protein delivery to the plasma membrane, which is conserved in the mouse, rat and canine homologues.6, 7, 15 In addition, we show in this work that the size of the protein expressed by our cloned cDNA after transfection into mammalian cells coincides with that of the endogenous PA2.26 protein detected in human tissues by Western blotting (see below). The ORF encoding human PA2.26 is identical to that described for gp36 and T1α-2, the human homologues of canine gp40 and rat T1α, respectively.16, 24 However, the reported gp36 and T1α-2 cDNA sequences are incomplete (768 and 865 bp, respectively) and have many changes in UTRs with respect to the PA2.26 sequence reported here. In addition, a 1013 bp cDNA sequence for human podoplanin has been delivered to GenBank (under accession no. AF390106). This sequence is identical to that reported here for PA2.26 and corresponds to positions 21–1033 of the cDNA sequence shown in Figure 1a. Alignment of the PA2.26 cDNA sequence with human genomic sequences deposited in GenBank predicts that the human PA2.26 gene contains 6 exons and 5 introns. The lengths of introns and exons are shown in Figure 1c. All the intron-exon junctions follow the canonical AG-GT rules. The first exon contains the initiation codon ATG and a 5′ UTR of 203 nt. The last exon contains the termination codon followed by a 1769 nt UTR, which includes a 200 nt AluSc sequence in the reverse orientation at the 3′ end (Fig. 1a). The largest intron (>22746 nt) lies between exons 1 and 2. A 432 nt CpG island was predicted that comprises the last 182 nt of the exon 1 and the first 250 nt of intron 1. We could assign the PA2.26 sequence to the NT_004577 contig localized to the short arm of chromosome 1 at 1p36.13.
Expression studies in normal tissues
Northern blot analysis with RNA isolated from various adult human tissues using a PA2.26 probe containing the full coding sequence amplified from a placenta cDNA library (see Material and Methods) revealed 2 mRNA species of about 2.7 kb and 0.9 kb (Fig. 2). The larger transcript corresponds to the estimated size of the full-length human PA2.26 cDNA presented in Figure 1a. The origin of the 0.9 kb message is unknown at the present time. Clear expression of PA2.26 transcripts was observed in skeletal muscle, placenta and heart and, at a lower level, in lung. The 2.7 kb transcript was the most abundant of the 2 transcripts in the majority of these tissues. No specific signal could be detected in several other tissues, such as brain, liver, kidney and pancreas, even after long exposure of the autoradiography.
Production of a specific polyclonal antibody
Since our previous mAb generated against the mouse PA2.26 antigen6 was found not to recognize the human protein, we synthesized a peptide comprising amino acids 37–51 (P37-51) of the extracellular domain of human PA2.26, a region that does not contain glycosylation residues. This peptide was used to immunize rabbits (see Material and Methods). To test whether the obtained polyclonal Ab recognized PA2.26, we synthesized the protein encoded by the PA2.26 cDNA by a coupled transcription/translation reaction using a cell-free wheat germ extract system. A polypeptide of about 21 kDa corresponding to the core protein of PA2.26 was synthesized, as detected in a Western blot (Fig. 3a). Preincubation of the Ab with P37-51 but not with an unrelated peptide (Pcontrol) prevented recognition of the 21 kDa polypeptide, indicating that the Ab was specific for the P37-51 amino acid sequence of PA2.26. The same result was obtained using a rabbit reticulocyte extract system to transcribe/translate the gene (data not shown). To test whether the Ab recognized the PA2.26 protein expressed in mammalian cells, PA2.26 cDNA was transfected into HeLa cells, which do not express endogenous PA2.26. As shown in Figure 3b, the Ab specifically recognized a protein of about 38–40 kDa, which corresponds to the glycosylated mature form of PA2.26, as demonstrated by treatment with glycosidases (not shown). PA2.26 was neither expressed in parental HeLa cells nor in cells transfected with the empty vector. The Ab recognized a polypeptide of the expected size, 61 kDa, corresponding to PA2.26 expressed as a C-terminal fusion protein with EGFP (PA2.26-EGFP, Fig. 3d). To confirm the O-glycosylation of human PA2.26 in the fusion protein, cell lysates were digested with glycosidases, followed by Western blotting with the PA2.26 Ab. Sequential digestion with neuraminidase and O-glycosidase yielded a molecular mass reduction of about 11 kDa, while treatment with O-glycosidase alone did not affect the mobility of the protein (Fig. 3d), indicating that human PA2.26 protein is heavily O-glycosylated and contains sialic acid, just as the mouse PA2.26 protein.6 We also studied whether the Ab recognized the native PA2.26 protein expressed in human tissues. As shown in Figure 3c, protein forms of the same size as that of the exogenous PA2.26 protein expressed in HeLa cell transfectants were detected in placenta, lung and testis lysates by Western blotting, indicating that the protein encoded by the cloned cDNA has the same molecular mass than the endogenous PA2.26 protein. The level of PA2.26 protein expression in lung was lower compared with placenta and testis, which is in agreement with the Northern blot data (Fig. 2). Testis (which was not screened for PA2.26 expression by Northern blot hybridization) was used in this assay because it proved to be positive for PA2.26 expression by immunohistochemistry (not shown).
PA2.26 is expressed in oral squamous cell carcinomas
Since a preliminary RT-PCR analysis showed PA2.26 mRNA expression in some human OSCC cell lines (data not shown), we studied the expression of PA2.26 in primary oral tumours by immunohistochemistry. We found that PA2.26 is induced in about 25% (15 of 61 cases) of early OSCCs located at floor of mouth and tongue (Table I). A small number (n = 6) of gingival SCCs were also screened and found not to express PA2.26 (data not shown). No significant correlation was observed between PA2.26 expression and tumour differentiation grade (Table I). No association was found, either, between PA2.26 expression and clinical parameters, such as the age of patients, sex, survival and locoregional recurrence. For instance, the average age of patients included in the PA2.26-positive group was 52.00 ± 10.39 (12 males and 3 females) compared to 56.23 ± 11.49 (38 males and 8 females) in the PA2.26-negative group. During the follow-up period (3 years minimum), the recurrence rate in the PA2.26-positive group was of 20% (3 of 15), whereas the mortality rate due to tumour-related causes was of 7% (1 of 15) compared to 20% (9 of 46) of patients who showed local or regional recurrences and 9% (4 of 46) of patients dead in the PA2.26-negative group.
Table I. Expression of PA2.26 in Human Oral Squamous Cell Carcinomas
No. of cases
No. of cases exhibiting positive reactivity to PA2.26 antibody
Stage I: T1, tumor 2 cm or less in greatest dimension; NO, no regional lymph node metastasis; M0, no distant metastasis. Stage II: T2, tumor size from 2 to 4 cm; N0; M0.
PA2.26 expression in OSCCs was heterogeneous, often restricted to the growth front of the tumours (Fig. 4e) and was predominantly located at the plasma membrane (Fig. 4g). PA2.26 was neither detected in the normal oral epithelium (Fig. 4a), nor in premalignant lesions (n = 10) exhibiting moderate to intense dysplasia (Fig. 4b). However, a positive staining was seen frequently in the reactive mucosa adjacent to OSCCs, including those lesions stained weakly or not stained at all with the PA2.26 Ab (Fig. 4c,d). PA2.26 expression in the hyperplastic mucosa was restricted to basal-like cells and no expression was seen in differentiating keratinocytes of suprabasal layers (Fig. 4d). Since mouse PA2.26 was found to destabilize E-cadherin-mediated cell to cell adhesions when expressed in keratinocytes,20 sections from PA2.26-positive tumours were also stained with an anti-E-cadherin mAb. We found that E-cadherin expression in the tumours was also very heterogenous, with some areas showing an intense staining and others in which E-cadherin staining was fragmented or virtually absent (Fig. 4f,h). Interestingly, some regions that expressed PA2.26 seemed to exclude the presence of E-cadherin (Fig. 4, compare e and f). In other cases, PA2.26 tended to be expressed in the layer of keratinocytes in contact with the stroma, whereas E-cadherin staining was mainly suprabasal (Fig. 4, compare g and h). Because of the preferential expression of PA2.26 in the basal layer of hyperplastic mucosa and tumour nests, we analyzed whether the presence of PA2.26 was restricted to proliferative cells by immunohistochemical staining with Ki-67, but no clear correlation between expression of this proliferation nuclear marker and that of PA2.26 was found (data not shown).
The inverse correlation observed between PA2.26 and E-cadherin expression at the surface of carcinoma cells was further analyzed by double immunofluorescence and confocal microscopy studies in a representative number of PA2.26-positive SCCs of the tongue (n = 5). In all cases, E-cadherin protein was reduced or virtually absent at cell-cell contacts in tumour nests that expressed PA2.26 (Fig. 5b–d). Conversely, a normal pattern of E-cadherin staining at cell-cell junctions was observed in regions that were negative for PA2.26 staining (insets in Fig. 5b–d). In contrast, PA2.26 and ezrin focally colocalized in vivo at the plasma membrane of tumour cells (Fig. 5f–h and insets). These results suggest that expression of PA2.26 in OSCCs might be associated with destabilization of E-cadherin-mediated cell-cell adhesion.
PA2.26 induces cell-surface protrusions and reduces cell-cell adhesiveness
To analyze the effects of PA2.26 expression on the phenotype of human epithelial cells, we transiently transfected HeLa carcinoma cells and HaCaT immortalized keratinocytes with the human PA2.26 cDNA, since these cell lines were previously found not to express PA2.26 mRNA by RT-PCR (data not shown). Immunofluorescence staining and confocal microscopy analysis showed that PA2.26 protein was directed to the cell surface (Fig. 6). Vertical optical sections indicated that PA2.26 was distributed in a clustered pattern all over the plasma membrane (Fig. 6c,d), suggesting that the antigen was concentrated at cell-surface protrusions. This was confirmed with a 3-dimensional reconstruction of the upper part of transfected cells (Fig. 6a,b). We also studied by double immunofluorescence analysis whether human PA2.26 colocalized with ezrin at plasma membrane protrusions. Ezrin was located as a punctate staining distributed uniformly throughout (underneath) the plasma membrane in HeLa cells that did not express PA2.26 (Fig. 6f,g), in a pattern typical of microvilli.25 However, in cells expressing PA2.26, ezrin was relocated to the cell edges where it colocalized with PA2.26 on membrane projections (Fig. 6e–g). A similar effect, although less pronounced, was observed in HaCaT cell transfectants (Fig. 6h–j). This milder effect could be due to the lower PA2.26 expression level at the cell surface obtained in this latter cell line. In contrast to HeLa, expression of PA2.26 in HaCaT keratinocytes was unstable, as suggested by the presence of many transfectant cells with endocytic vesicles containing PA2.26 (data not shown). Both the density and length of projections protruding from the cell edges appeared to be enhanced in transfected cells with respect to the nontransfected ones (Fig. 6e–j, arrows). Many of these PA2.26-containing projections had lengths of >5 μm (insets in Fig. 6e–g) and formed phalloidin-positive F-actin bundles at the periphery (not shown), resembling filopodia.26, 27 In contrast, in cells that did not express PA2.26, most of ezrin-positive cell-surface protrusions were microvilli or microspikes (<5 μm).
Because HeLa are not cohesive cells and have lost E-cadherin expression,28, 29 we studied the effect of PA2.26 on cell-cell adhesiveness in HaCaT keratinocytes. To this aim, HaCaT cells were stably transfected with a vector encoding PA2.26 fused to EGFP (PA2.26-EGFP). This procedure resulted in a more stable expression of PA2.26 at the cell surface when compared to the above transfection experiment. PA2.26-EGFP was also directed to cell-surface projections (Fig. 7f), where it colocalized with ezrin and F-actin (not shown), while EGFP was distributed uniformly throughout the cytoplasm in control cells transfected with the empty vector (inset in Fig. 7b). Control transfectants grown at confluence showed the typical tight cell-cell contacts of keratinocytes (Fig. 7a), decorated by a continuous line of E-cadherin and β-catenin staining (Fig. 7c,d). In contrast, aberrant cell-cell borders with irregular outlines (Fig. 7e) and intercellular spaces filled with PA2.26-containing microvilli (Fig. 7f and inset) were observed in PA2.26-EGFP transfectants. These cells exhibited a rather disorganized pattern of E-cadherin and β-catenin distribution, characterized by the presence of both proteins out of cell-cell contacts (Fig. 7g,h and insets).
To ascertain whether expression of PA2.26 in HaCaT keratinocytes impaired E-cadherin function, a Ca2+-dependent cell aggregation assay was performed in selected clones expressing varying amounts of PA2.26-EGFP (Fig. 8). Ca2+-dependent cell-cell adhesiveness decreased in PA2.26-EGFP transfectants in comparison to the parental cell line or control cells (Fig. 8b,c), in direct proportion to the level of PA2.26-EGFP expression (Fig. 8a). Although a slight reduction of E-cadherin protein levels was found in PA2.26-EGFP transfectants with respect to control cells, it did not correlate with the level of PA2.26-EGFP expression (Fig. 8a), suggesting that PA2.26 reduces HaCaT cell-cell adhesiveness by a mechanism that does not necessarily involve downregulation of E-cadherin expression.
We report in this study the full cDNA sequence of the human homologue of murine PA2.26 antigen. The human PA2.26 gene is expressed as 2 mRNAs of 2.7 and 0.9 kb in the placenta, heart, skeletal muscle and lung. These 2 transcripts have also been reported for podoplanin in human lymphatic endothelial cells.30 The size of the larger mRNA detected in human tissues by Northern blot hybridization fits the length of the cDNA reported here for PA2.26. It encodes a small mucin-type transmembrane glycoprotein of 162 amino acids practically identical to T1α-2, gp36 and podoplanin,16, 24 which is detected as a 38–40 kDa protein form in transfected cells and human tissue lysates in Western blots, using a polyclonal Ab raised against a peptide of the ectodomain. Nevertheless, the presence of an alternative mRNA of 0.9 kb in human tissues might suggest the existence of other protein isoforms not detected by this Ab.
The main finding of our study is that expression of PA2.26 (T1α, podoplanin) is induced in a subset of human early OSCCs. This observation strengthens our previous work showing the induction of PA2.26 during mouse skin carcinogenesis related to cell migration and invasion/metastasis.5, 6, 20 Both oral and skin carcinogenesis are multistep processes in which multiple genetic events lead to the disruption of normal regulatory pathways controlling cell division, differentiation and cell death.31, 32 SCCs are characterized by their ability to spread locally and regionally, and tumour infiltration involves, among other factors, alterations in the expression and/or function of cell adhesion molecules, such as E-cadherin.33, 34 Reduced expression of E-cadherin and β-catenin at the plasma membrane have been found in OSCCs associated with the degree of tumour differentiation and poor prognosis.34, 35, 36, 37 Thus, expression of E-cadherin is homogeneously reduced at late stages in poorly differentiated carcinomas, apparently by a mechanism that involves hypermethylation of the gene promoter.38, 39, 40 However, well and moderately differentiated OSCCs often show a partial loss of membrane E-cadherin and β-catenin expression, circumscribed to some tumour regions or even to certain areas of tumour nests.34, 35 PA2.26 expression in early OSCCs is heterogeneous and fragmented, frequently restricted to the invasive tumour front, and correlates with reduced membrane E-cadherin expression in vivo and with impaired cell-cell adhesiveness in vitro. These results suggest that the presence of PA2.26 in OSCCs can be related to a migratory/invasive phenotype.
The effect of PA2.26 on cell-cell adhesiveness appears to be indirect, involving the reorganization of the actin cytoskeleton by recruitment of ezrin and the induction of cell-surface protrusions. The ERM proteins have been implicated in the organization of specialized membrane domains and in the control of cell shape, cell adhesion and migration by linking transmembrane proteins to the actin cytoskeleton.25, 41 Mouse PA2.26 was shown to coimmunoprecipitate with ezrin and moesin,6 and a yuxtamembrane cluster of basic amino acids that appears to be responsible for murine PA2.26 binding to ERM proteins is conserved in the human PA2.26 endodomain. In fact, human PA2.26 colocalizes with ezrin in vivo at the surface of oral carcinoma cells and at membrane projections of HeLa and HaCaT cell transfectants. It is likely that PA2.26-induced redistribution of ezrin and formation of cell-surface extensions disrupt the anchorage of E-cadherin-catenin complexes to the cortical actin cytoskeleton. Interestingly, Pujuguet et al. show in a recent study that a constitutively active mutant form of ezrin, when expressed in cultured epithelial cells, causes extensive lamellipodia formation and disturbs E-cadherin-dependent cell-cell contacts via activation of Rac1 but not of other Rho GTPases.42 These results suggest a role of ezrin in cell-cell junction assembly. Whether the effects of PA2.26 on the cell membrane and cytoskeleton involve changes in the activity of Rho GTPases is currently being investigated. Ectopic expression of mouse PA2.26 in premalignant murine keratinocytes (MCA3D cell line) elicited a complete epithelial-mesenchymal transition, with loss of E-cadherin protein (but not mRNA) expression and acquisition of a fibroblast-like morphology associated with development of undifferentiated carcinomas.6, 20 In contrast, human PA2.26 expressed in immortal HaCaT keratinocytes failed to trigger a complete epithelial-mesenchymal transition but induced cell scattering associated with increased plasma membrane motility and reduced cell-cell cohesion. To explain this apparent discrepancy, it should be noted that HaCaT keratinocytes are nontumorigenic when injected subcutaneously or grafted as surface transplants onto nude mice, and, therefore, represent a preneoplastic stage of tumour development.43 MCA3D keratinocytes, on the other hand, can develop tumours in skin graft experiments and are considered a premalignant cell line,20 a step further in tumour progression compared to HaCaT cells. Consequently, it is likely that PA2.26 needs to cooperate with other genetic or epigenetic events to confer a full malignant phenotype in keratinocytes.
Similar effects on the plasma membrane and cell-cell adhesiveness have been recently found for dysadherin, a different mucin-type glycoprotein of 178 amino acids, also associated with cancer. Transfection of dysadherin cDNA into liver cancer cells resulted in formation of numerous microvilli and impaired cell-cell adhesiveness without affecting E-cadherin mRNA expression.44 In contrast to other known cancer anti-adhesive mucins of very large molecular masses (>200 kDa), such as MUC1 (episialin) and MUC4 (sialomucin),45, 46 PA2.26 and dysadherin belong to a group of cancer membrane mucins characterized by their relatively small size. Both PA2.26 and dysadherin appear to downregulate E-cadherin function in carcinoma cells by a novel mechanism involving a reorganization of the actin cytoskeleton.
We thank Dr. N.E. Fusenig for his generous gift of HaCaT keratinocytes. We also thank Mr. D. Megías for helping us with confocal microscopy; Dr. M. García-Gallo, Ms. E.M.G. Marazuela and Ms. C. González for skillful technical assistance; Dr. D. Sarrió for helpful suggestions and Dr. J. Renart for critical reading of the manuscript. This work was supported by grants from the “Fondo de Investigaciones Sanitarias” (FIS-01/1125 to M.Q. and FIS-02/1025 to C.G.) and the “Ministerio de Ciencia y Tecnología” (SAF2001-2361). E.M-V. was the recipient of a predoctoral fellowship from the FIS. M.M.Y. was funded by a predoctoral fellowship from the “Fundación Carolina”.