Kaiso is a BTB/POZ zinc finger protein originally described as an interaction partner of p120ctn. In cultured cell lines, Kaiso is found almost exclusively in the nucleus, where it generally acts as a transcriptional repressor. Here, we describe the first in situ immunolocalization studies of Kaiso expression in normal and cancerous tissues. Surprisingly, we found striking differences between its behavior in monolayers of different cell lines, three-dimensional cell culture systems, and in vivo. Although nuclear localization was sometimes observed in tissues, Kaiso was more often found in the cytoplasm, and in some cell types it was absent. In general, Kaiso and p120ctn did not colocalize in the nucleus. To examine this phenomenon more carefully, tumor cells exhibiting strong nuclear Kaiso staining in vitro were injected into nude mice and grown as xenografts. The latter showed a progressive translocation of Kaiso towards the cytoplasm over time, and even complete loss of expression, especially in the center of the tumor nodules. When xenografted tumors were returned to cell culture, Kaiso was re-expressed and was once again found in the nucleus. Translocation of Kaiso to the cytoplasm and down-regulation of its levels were also observed under particular experimental conditions in vitro, such as formation of spheroids and acini. These data strongly imply an unexpected influence of the microenvironment on Kaiso expression and localization. As transcriptional repression is a nuclear event, this phenomenon is likely a crucial factor in the regulation of Kaiso function.
- human tumor
The p120ctn binding partner Kaiso was originally discovered by a yeast two-hybrid approach using p120ctn as bait (1). p120ctn is an Armadillo protein first identified as a prominent tyrosine kinase substrate implicated in cell transformation by Src (2), and in ligand-induced receptor signaling through various tyrosine kinase receptors (3, 4) . p120ctn binds to the juxtamembrane domain of classical cadherins (reviewed in refs. 5, 6 ), where it modulates cell-cell adhesion by regulating cadherin turnover and stability at the cell surface (7–9) . In E-cadherin-negative cancer cells, p120ctn localizes aberrantly to both the cytoplasm and the nucleus (10, 11) . Cytoplasmic p120ctn may promote metastasis through interactions with Rho GTPases (12), but aside from the Kaiso connection, roles for p120ctn in the nucleus have not been described.
There is mounting evidence that Kaiso is a nuclear protein that plays a role in transcription repression. Kaiso contains an amino-terminal BTB/POZ domain (Broad Complex, Tramtrak, Bric à brac/Pox virus and Zinc finger) and a carboxyl-terminal region with three zinc finger motifs of the C2H2 type (1). Other known members of the BTB/POZ family have roles in development or cancer, and are also transcriptional repressors. For example, the human BCL-6 and promyelocytic leukemia zinc finger proteins are causally involved in non-Hodgkin's lymphoma and acute promyelocytic leukemia, respectively (13, 14) . The BTB/POZ proteins APM-1 and mGCL inhibit growth of cervical carcinoma and osteosarcoma cells, respectively (15, 16) , whereas the BTB/POZ protein HIC1 and its relative HRG22 are candidate tumor suppressors in a variety of human cancers (17). Kaiso homodimerizes via its POZ domain (1), which may contribute to the assembly of large multiprotein complexes. As in the case of other POZ family members, Kaiso interacts with histone deacetylases through the corepressor N-CoR (18, 19) , thereby controlling gene expression. Kaiso's description as a transcriptional repressor is also based on the specific recognition by its zinc fingers of an MGMG motif (where M is a 5-methylcytosine; refs. 19, 20 ). Thus, despite the absence of a canonical methyl-binding domain, Kaiso may belong to a group of complexes or proteins, including MeCP2, MeCP1, and methyl-binding domain-1, which are involved in repressing DNA transcription through their attraction to methylated DNA (21, 22) . CAST analysis, however, revealed that Kaiso also binds to the DNA sequence CTGCNA in a methylation-independent manner mediated by zinc finger-2 and zinc finger-3 (23). The relative importance of the methylation-dependent and -independent binding motifs is unknown. Like HIC1 (17), Kaiso may repress transcription by a dual mechanism. Several cancer-related gene fragments, such as regulatory sequences of the genes Rb, Xist, S100A4 (mts-1), CDH1 (E-cadherin), matrilysin (MMP7), MTA2 (metastasis-associated gene 2) and Wnt11, have been proposed as candidates for Kaiso binding (19, 20, 23, 24) . On the other hand, the core rapsyn promoter was recently shown to be up-regulated by a complex of Kaiso with δ-catenin in subsynaptic nuclei of neuromuscular junctions (25).
DNA binding and transcription repression by Kaiso may be modulated by interaction with p120ctn. Interaction between p120ctn and Kaiso occurs via the Arm repeats 1 to 7 of p120ctn and a noncontiguous Kaiso domain flanking the carboxyl-terminal DNA-binding zinc finger domain (1). This close physical juxtaposition of the p120ctn-binding site of Kaiso with its DNA-binding motif may explain the reported finding that DNA binding of Kaiso is inhibited by nuclear p120ctn (23, 26) . Association of p120ctn with E-cadherin occurs via its Arm repeats 1 to 5 and 7 (7). Thus, Kaiso and E-cadherin likely bind the same p120ctn domain in a mutually exclusive manner (1, 27) .
Previous reports describing almost exclusive nuclear localization of Kaiso have been based entirely on in vitro observations in cell cultures (1, 28) . Here, we describe the first in situ studies of Kaiso localization in normal and cancerous human tissues. Surprisingly, Kaiso was frequently found in the cytoplasm, and was also often absent. Moreover, Kaiso localized differently in the same cell lines depending on whether the cells were grown in vitro, as two-dimensional monolayers with increasing cell densities on various substrates, as three-dimensional cultures, or as xenografts in nude mice. These data reveal a striking microenvironmental effect that is crucial in regulating Kaiso localization, expression, and activity.
Materials and Methods
Cell Culture. HT29 and SW48 cell lines were purchased from the American Cell Type Culture collection (ATCC, Rockville, MD). HT29 human colorectal adenocarcinoma cells were maintained in DMEM (Invitrogen, Carlsbad, CA) with 15% FCS, nonessential amino acids (Invitrogen), 0.4 μM sodium pyruvate, 0.2 units/ml penicillin, 0.2 μg/ml streptomycin and 2 mmol l-glutamate. SW48 human colon adenocarcinoma cells were maintained in LB15 (Invitrogen) supplemented with 10% FCS, 0.2 units/ml penicillin, 0.2 μg/ml streptomycin, and 2 mmol l-glutamate. MCF-10A, a spontaneously immortalized, nontransformed human mammary epithelial cell line (29), was kindly provided by J. Brugge (Harvard Medical School, Boston, MA). These cells were grown in a special supplemented medium as described previously (30).
Primary Antibodies. Kaiso was detected using two mouse monoclonal antibodies (6F and 12H, each at a concentration of 4 μg/mL), and a rabbit polyclonal antibody (designated pAb R and used at dilution 1:250; refs. 1, 28 ). We have also generated a new pAb (S1337) against peptide C-NVTDGSTEFEFIIPESY, which corresponds to a COOH-terminal sequence of human and mouse Kaiso. The procedure used has been described previously (31). For p120ctn recognition, monoclonal antibody (mAb) pp120 (Transduction Laboratories, Lexington, KY) was used at a 1:200 dilution. Staining for E-cadherin was done with mAb human epithelial cadherin-1 (Takara, Kyoto, Japan) at a concentration of 1.3 μg/mL. For staining of carbonic anhydrase IX, the mouse monoclonal antibody M75 (32) was used at a dilution of 1:50. Antibody against HIF-1α (Transduction Laboratories, Erembodegem, Belgium) was used at a 1:80 dilution. Double staining for Kaiso and p120ctn was done either with mAb 6F in combination with FITC-conjugated pp120 mAb (Transduction Laboratories, Lexington, KY) or with pAb S1337 together with mAb pp120.
Preparation of Cells and Tissues for Immunodetection. HT29 and SW48 cells were grown on membrane filters, fixed in 4% formalin for 1 hour and embedded in 5% agar in PBS before embedding in paraffin. Frozen or paraffin-embedded sections of normal and cancerous tissues from different patients were obtained from the Department of Pathology at the University of Antwerp, as well as a tissue microarray containing biopsies of inflammatory breast cancers (IBC) from 32 patients prior to initiation of chemotherapy (33). Archival formalin-fixed paraffin-embedded tissue samples from human prostate cancers (PCas) were obtained from the Department of Pathology, University Hospital of Liège, Belgium. These samples were surgically obtained from 45 patients who had undergone a radical retropubic prostatectomy for localized prostate cancer. Patients who had received prior hormonal therapy, chemotherapy, or radiation therapy were not included in the investigation. All patients had a clinically confined tumor, classified as stage T1 (n = 19) or T2 (n = 26) according to the tumor-node-metastasis classification (34). After histopathologic examination, 19 patients were classified as having tumors of pathologic stage pT2. In 16 patients, extracapsular extension of the tumor (stage pT3A) was observed. The remaining 10 patients were categorized as having PCas of stage pT3B because their tumors showed evidence of invading the seminal vesicle. All PCas were evaluated according to the Gleason scoring system (35). Tumors were classified as high grade (n = 24) when the Gleason score was 7 or above, and as low grade (n = 21) when it was 6 or below. One tissue block per patient containing the most representative tumor-bearing areas was selected considering the capsular status (pathologic stage) and the Gleason score stated in the pathological report. Serial sections were made and stained as described below. Decreased expression of p120ctn and E-cadherin was noted in several PCas, and down-regulation of each of these proteins correlated with high tumor grade (to be reported in detail elsewhere).
Xenografts in Nude Mice. Athymic nu/nu mice 8 weeks of age were purchased from Iffa Credo, France. Following a 7-day adaptation period, each mouse was inoculated s.c. in each flank with 5 × 106 HT29 or SW48 tumor cells in 100 ml sterile PBS. This experiment was done twice. In the first series, mice with HT29-derived tumors were sacrificed (one per time point) on days 6, 11, 14, and 20 after inoculation, whereas mice with SW48-derived tumors were sacrificed on days 6, 11, 20, and 32 after inoculation. All tumors grew progressively, although HT29 xenografts grew faster than SW48 xenografts. In a second series, mice with HT29 xenografts were sacrificed (one per time point) on days 2, 3, 4, 6, 9, 12, and 31 after inoculation, whereas mice with SW48 xenografts were sacrificed on days 4, 6, 8, 9, 13, 23, and 38 after inoculation. Half of each tumor was fixed in 4% formalin before embedding in paraffin, and the other half was frozen in liquid nitrogen and stored at −80°C. When tumors were very small, the left tumor was prepared for paraffin embedding and the right tumor was prepared for frozen sections. Cryosections and paraffin sections were made for the different growth stages of the tumors.
Ex vivo Culture of Xenografted Cells. On the day of excision of xenografted cells, 1/4 of the tumor was explanted. Cells were disaggregated in PBS containing 1 mg/mL collagenase A and 0.05% trypsin, and incubated at 37°C on a rotator for 3 hours. After washing in medium, half of the cells were prepared for cytospin analysis. The remaining cells were grown on cover slips until they were subconfluent (1 week after explantation for HT29 cells, and 2 weeks after explantation for SW48 cells), and then used for immunofluorescence staining. The human origin of these cultures was verified by appropriate species-specific antibodies.
Immunocytochemistry and Immunohistochemistry. Paraffin and cryostat sections of 5 μM/L were used. Immunodetection in cryosections has been described (31), as has the protocol for paraffin sections (cf. E-cadherin in ref. 36). Sections were incubated with primary antibodies for 1 hour at room temperature. Staining of paraffin sections was completed by a biotinylated secondary antibody, streptavidin-peroxidase, and 3,3′-diaminobenzidine. In the case of double staining, a standard Kaiso staining was followed by application of anti–carbonic anhydrase IX antibody M75. The latter was revealed using a DAKO EnVision alkaline phosphatase kit in combination with fuchsin + substrate-chromogen (DakoCytomation, Heverlee, Belgium).
Samples were examined with an Olympus BX51 microscope, and images were recorded with a Coolsnap camera (Photometrics, Tucson, AZ) using RSImage software (Roper Scientific, Trenton, NJ).
Cell Culture Under Hypoxic and Normoxic Conditions. HT29 and SW48 cells were seeded on glass cover slips in six-well plates and incubated under normal conditions overnight. They were exposed to hypoxia by incubation in a humidified chamber that was continuously purged with a mixture of 1% O2, 5% CO2, and 94% N2, as described by Fordel et al. (37). Medium was changed every 2 days with appropriately gas-equilibrated media. After 1, 3, 4, 5, and 7 days, cell cultures were harvested and fixed with methanol or paraformaldehyde under hypoxic conditions. Control cells were also cultured simultaneously under standard normoxic conditions.
Two-dimensional and Three-dimensional Growth Conditions. Monolayers of MCF-10A cells were grown on clean glass cover slips, either uncoated or coated with GFR-Matrigel (BD Biosciences, Erembodegem, Belgium). Cells were plated as sparse (15-20% of confluency), subconfluent (70-80%), confluent (90-95%), and very confluent (100%) cultures. Monolayers of HT29 were grown on uncoated glass cover slips. Spheroids of HT29 cells were made in static cultures according to the method of Yuhas et al. (38). In order to obtain three-dimensional cultures of MCF-10A, single cells were seeded on a Lab Tek II eight-well glass chamber slide (Nunc, Roskilde, Denmark) with a solidified layer of GFR-Matrigel measuring 1 to 2 mm in thickness. Cells were grown under the same conditions as those described earlier (30). Formation of acini-like structures was monitored for 21 days. Immunofluorescence staining was done on days 0.5, 1.5, 3, 7, 14, and 21.
Immunofluorescence. Cells were rinsed briefly with PBS and fixed either with ice-cold 100% methanol for 15 minutes at −20°C, or with 4% paraformaldehyde in PBS for 25 minutes at room temperature. Immunostaining was done as previously described (30, 39) . Antibodies were incubated for 1 hour or overnight at 4°C. Secondary antibodies were Alexa 488/594-coupled anti-mouse immunoglobulin or Alexa 488/594-coupled anti-rabbit immunoglobulin antibodies (Molecular Probes, Eugene, OR). Counterstaining was done with Vectashield mounting medium containing 4′,6-diamidino-2-phenylindole or propidium iodide (Vectorlabs, Burlingame, CA). Monolayer cultures were examined with a Zeiss Axiophot photomicroscope and images were recorded with a MicroMax camera (Princeton Instruments, Trenton, NJ) and Metamorph software (Image Universal Corporation, New York, NY). Differential interference contrast pictures of very confluent monolayers were taken using a Leica DM IRE2 microscope equipped with an HCX PL APO 63×/1.30 glycerin-corrected 37°C objective and a Coolsnap HQ camera. Wide-field fluorescence images were monitored with the same microscopic setup using a monochromator system with a 50 W Xenon lamp for excitation. The fluorescence dyes Alexa 488 and 594 were detected using a standard B/G/R filter cube (Leica). Blind deconvolution was carried out afterwards using the Leica Deblur software. Confocal images of three-dimensional cultures were made with an inverted Zeiss LSM 410 or LSM 510 confocal laser scanning microscope.
Predominantly Cytoplasmic Localization of Kaiso In vivo. Previous analyses of Kaiso had been carried out exclusively in cultured cells. Here, we examined localization of Kaiso by immunocytochemistry in a broad range of both normal and cancerous tissues, including ovary (two normal samples, one papillary serous adenocarcinoma, and a liver metastasis of the latter), oviduct (two normal samples), breast (one fibroadenoma and 10 adenocarcinoma samples), testis tumors (three embryonal carcinomas, two seminomas, and one paratesticular sarcoma), squamous cell carcinomas of lung (two samples) and esophagus (two primary tumors and one lymph node metastasis), adenocarcinomas of the stomach (two samples) and colon (seven primary tumors and four liver metastases), normal colon of neonate (one sample) and adult (one sample), and endocrine tumor of the pancreas (one sample). Besides this broad screening, we also focused on larger series of PCas (45 samples) and IBCs (32 samples).
In contrast to in vitro observations, we detected only a few human cell types that were positive for Kaiso at the nuclear level. These included oocytes ( Fig. 1A ), eosinophilic granulocytes in the muscularis propria ( Fig. 1B), submucosa (not shown) and lamina propria of the colon ( Fig. 1J and K), and cells from certain tumors, e.g., a seminoma ( Fig. 1C), an undifferentiated paratesticular sarcoma ( Fig. 1D) and 3 out of 10 breast adenocarcinomas ( Fig. 1E). Nuclear Kaiso staining was often associated with cytoplasmic immunopositivity. In general, however, most cells were completely negative for nuclear Kaiso, including tumor cells of several breast adenocarcinomas ( Fig. 1F), stromal and granulosa cells of ovary ( Fig. 1A), normal glandular epithelium of the prostate ( Fig. 1G), hepatocytes ( Fig. 1H), pancreatic endocrine tumor cells ( Fig. 1I), and normal epithelial cells of the colon ( Fig. 1J and K). In normal prostate epithelium, some cytoplasmic staining was detectable in basal cells, although there was no or only weak staining in the cytoplasm of secretory epithelial cells and periacinar myofibroblasts ( Fig. 1G). In each of 45 PCas analyzed, Kaiso was weakly expressed in the cytoplasm and apparently completely absent from the nucleus (exemplified in Fig. 2A for a high grade tumor; Gleason score 7). In serial sections, the same tumor cell field showed retention of E-cadherin and p120ctn at cell-cell contacts ( Fig. 2B and C). Heterogeneous loss of these two proteins was seen in high-grade tumors, but there was no evidence for nuclear p120ctn (to be described in detail elsewhere). Kaiso staining that is weak in the cytoplasm and absent from the nucleus was also seen in 8 out of the 32 IBC samples analyzed ( Fig. 1L). Conversely, the other 24 IBC samples were clearly positive for Kaiso in both nucleus and cytoplasm ( Fig. 1M-O). However, staining intensity showed intratumoral heterogeneity, and in some cases immunopositive dots in the perinuclear cytoplasm were observed (arrows, Fig. 1M-O).
Staining of consecutive sections of a primary papillary serous carcinoma of the ovary ( Fig. 2D-F) showed that Kaiso (D) and p120ctn (E) were colocalized in the cytoplasm, whereas E-cadherin was expressed with varying intensities at cell-cell contacts (F). This pattern was also apparent in stainings of a liver metastasis of this same ovarian cancer ( Fig. 2G-I), where high cytoplasmic Kaiso and p120ctn levels were obvious, whereas E-cadherin was focally absent ( Fig. 2I).
Another interesting staining pattern was seen in a squamous cell carcinoma of the esophagus ( Fig. 2J-L), its metastasis in the lymph nodes ( Fig. 2M), and some adenocarcinomas of the breast ( Fig. 2N and O). These tumors displayed a gradient of Kaiso expression levels and a diversity of localizations, with prominent nuclear staining in cancer cells at the border between tumor nodule and surrounding tissue or in infiltrating parts of the tumor, and with weaker nuclear or mainly cytoplasmic staining in the central parts of the tumor. Convincing nuclear p120ctn was not seen in any of these tumors. At the tumor border, the nuclear Kaiso staining was accompanied by p120ctn staining in the cytoplasm and by decreased E-cadherin and p120ctn at cell-cell contacts (e.g., Fig. 2K and L). Cytoplasmic Kaiso and p120ctn were also observed in the peritumoral stroma ( Fig. 2J and K). In general, there was little evidence of p120ctn and Kaiso colocalization in the nucleus, but the proteins were frequently observed together in the cytoplasm.
Staining with a newly made pAb (S1337) raised against a carboxyl-terminal epitope of Kaiso not recognized by mAb 6F, mAb 12H, or pAb R gave similar results. An example is shown of IBC with highly positive cells ( Fig. 1O, to be compared with Fig. 1N). Examples of negative staining in the nuclei are given for epithelial cells of normal colon crypts, whereas eosinophilic granulocytes in the lamina propria showed positive nuclei ( Fig. 1K, to be compared with Fig. 1J). This confirms the specificity of the Kaiso staining in tissue sections.
Cells with high cytoplasmic Kaiso levels and absent or weak expression in the nucleus were also common in several normal and benign tissues, e.g. ciliated epithelial cells of the oviduct ( Fig. 3A ) or bronchus ( Fig. 3B), biliary duct cells ( Fig. 3C), smooth muscle cells of the muscularis propria of the colon of a neonate ( Fig. 3D), neural ganglia of the stomach ( Fig. 3E), and breast fibroadenoma cells ( Fig. 3F). In general, cytoplasmic staining in these tissues was manifested in a punctuate pattern, often in a polarized fashion ( Fig. 3A-C), although smooth muscle cells showed dots only close to the nucleus ( Fig. 3D), which was quite comparable to the staining results in some cases of IBC ( Fig. 1M-O).
Subcellular Translocation and Loss of Kaiso Expression in Human Xenografts. In order to investigate the prominent differences between in vitro and in vivo Kaiso expression, we did xenograft experiments using human HT29 and SW48 tumor cell lines. HT29 cells normally express p120ctn at cell-cell contacts, whereas SW48 cells express only a minor amount of p120ctn (7).
In a first xenograft experiment, cryosections and paraffin sections were made from tumors at different time points after s.c. injection in athymic mice. Stainings in both types of sections were comparable. Results of paraffin sections of HT29 xenografts are presented in Fig. 4 , and data from SW48 experiments were similar. In both cases, we observed a striking change in the subcellular localization and the amount of Kaiso as the injected cells grew into tumors in vivo. All injected cells showed Kaiso-positive nuclei in vitro, as tested by immunofluorescence and immunohistochemistry of cell cultures embedded in paraffin. In contrast, Kaiso localization in the growing tumors evolved from a diversity of patterns ( Fig. 4A and B) to a nearly complete loss of expression ( Fig. 4E-H). Cells with nuclear Kaiso were observed only occasionally, but always at the border of younger tumor nodules, where they had close contact with the surrounding stroma. At an intermediate time point ( Fig. 4C and D), a possible translocation of Kaiso from the nucleus towards the cytoplasm was detected, especially in the form of granules located around the nucleus. Here again, use of pAb S1337 yielded quite similar results, with nuclear Kaiso staining in young xenografts ( Fig. 4I) and loss of nuclear staining at later stages ( Fig. 4J).
Because diverse patterns of Kaiso expression were already present in tumors excised 6 days after inoculation ( Fig. 4A), the data were extended by a second xenograft experiment in which mice were sacrificed at earlier stages (see Materials and Methods). Again, in vivo findings clearly differed from the in vitro stainings. In general, the results were similar to those of the first xenograft experiment, with diverse Kaiso localization patterns apparent 2 days after inoculation (data not shown). At 6 days after inoculation, xenografts were stained for the endogenous hypoxia-regulated marker carbonic anhydrase IX (32). Interestingly, the center of the tumor nodules stained positively for carbonic anhydrase IX but lacked nuclear Kaiso, whereas the tumor edges showed the inverse staining pattern ( Fig. 4K). Tumor cells nearby blood vessels showed lower carbonic anhydrase IX levels but prominent nuclear Kaiso ( Fig. 4L).
HT29 xenografts at 31 days and SW48 xenografts at 38 days after inoculation were essentially Kaiso-negative, also at tumor-host interfaces. Remarkably, when these tumor cells were put into ex vivo culture, they remained Kaiso-negative for at least 2 days ( Fig. 5A for HT29), but later on, they recovered nuclear staining ( Fig. 5E, at 6 days after explantation). In the case of HT29, p120ctn staining was consistently positive at cell-cell contacts ( Fig. 5B and F), and colocalization of p120ctn with Kaiso was not detected. SW48 cells showed no p120ctn staining, as expected (7).
Influence of the Microenvironment on Kaiso Expression. In order to determine the possible effect of hypoxia on Kaiso expression, we did a comparative study on cultures of HT29 and SW48 cells grown under normoxic versus hypoxic conditions. Cultures at low density invariably showed nuclear Kaiso (illustrated for HT29 cells in Fig. 6A-D ). When cell cultures became very confluent, Kaiso disappeared from the nucleus. This was again observed under both hypoxic and normoxic conditions ( Fig. 6E-H). Hypoxia was monitored by staining for carbonic anhydrase IX and HIF-1 (data not shown). Loss of Kaiso was seen a few days earlier under normoxic conditions, which can be explained by the more rapid formation of a superconfluent layer. Upon double staining for p120ctn, no influence of either oxygen pressure or culture density on p120ctn expression was observed ( Fig. 6I and J).
We then analyzed the subcellular localization of Kaiso in two-dimensional versus three-dimensional cultures. We first aimed at confirming a relationship between cell density and Kaiso protein expression in HT29, SW48, and MCF-10A cells. Only cultures that achieved the highest density at the time of fixation showed regions negative for Kaiso in the nucleus. In these regions, cells had piled up, forming semi-three-dimensional cultures.
Next, we examined Kaiso location in full three-dimensional structures. HT29 spheroids, formed during 5 days on a bacterial Petri dish or on an agar base, were collected and immunostained for Kaiso, p120ctn, and E-cadherin. In comparison with subconfluent HT29 monolayers, Kaiso expression was strongly decreased in these three-dimensional structures ( Fig. 6K and L), and the inner cells of the spheroids in particular completely lost nuclear Kaiso. In contrast, p120ctn and E-cadherin were still expressed at cell-cell contacts.
The formation of acini-like structures by MCF-10A cells was analyzed in three-dimensional Matrigel cultures. These cells were also grown on glass cover slips, either uncoated or coated with Matrigel. In the latter controls, strong nuclear and weak cytoplasmic Kaiso staining was observed as long as the cultures were subconfluent ( Fig. 7A-D ). Morphogenesis in Matrigel of hollow structures of 1 to 2 mm thickness was compared with published data (30). Sizes of the forming glandular structures perfectly matched these data in time, and (almost) completely hollow acini were formed by days 10 to 14. At early time points (day 0.5, 1.5, and 3), when cells were still single or had formed only very small aggregates, Kaiso expression had already decreased significantly in the nucleus, whereas cytoplasmic levels slightly increased ( Fig. 7E and F). At days 7 and 14, nuclear Kaiso could no longer be detected, and only cells at the outer rim (reported as a polarized layer; ref. 30) showed some cytoplasmic Kaiso staining ( Fig. 7G and H). Strikingly, apparently single cells sticking to the acini or shed by them contained nuclear Kaiso (arrows, Fig. 7G and H). Finally at day 21, when typical acini had formed, Kaiso immunopositivity completely disappeared ( Fig. 7I and J). Stainings with E-cadherin and p120ctn antibodies showed satisfactory antibody penetration into these relatively large three-dimensional structures (exemplified in Fig. 7K and L).
We have shown that the intracellular localization of Kaiso can deviate dramatically from what has been reported for various cultured cells. Kaiso is almost always nuclear in cells grown as two-dimensional monolayer cultures, but usually cytoplasmic or even absent in vivo. The data we obtained using several highly specific mAbs and pAbs recognizing at least two different Kaiso epitopes were essentially identical. By directly switching the cells between in vitro and in vivo growth conditions (xenografts and xenograft explants), we have shown that the nuclear exclusion of Kaiso is completely reversible. Comparable results were obtained from three-dimensional cell culture models: Kaiso translocated from the nucleus towards the cytoplasm, and eventually disappeared nearly completely in HT29 spheroids cultured in liquid medium and in MCF10A acini-like spheroids grown in reconstituted basement membrane (Matrigel). Our findings reveal a novel and striking cell density–dependent translocation and/or down-regulation of Kaiso in a wide variety of cell types and tissues, both normal and tumoral ones.
In vivo, nuclear Kaiso was sometimes observed at tumor borders, in infiltrating parts of particular human tumors, and in metastases, whereas the majority of normal human tissues and primary tumors had Kaiso-negative nuclei. The reproducibility of these findings was shown by analysis of two larger tumor series: a set of 45 PCas of either low or high grade, and a set of 32 IBCs. The PCas invariably showed Kaiso-negative nuclei, whereas the IBCs showed intratumor and intertumor variability ranging from negative to quite strong Kaiso staining in the tumor cell nuclei.
Kaiso is generally described as a transcription repressor (19, 20, 22–24) . Interestingly, the majority of candidate Kaiso target genes identified thus far, i.e., CDH1 (E-cadherin), S100A4 (mts-1), matrilysin (MMP7), MTA2, and Wnt11 have been linked with development and/or cancer. The absence of nuclear Kaiso in many of the human tissues examined may reflect a requirement for the expression of certain of these Kaiso-repressible genes. E-cadherin is the prototypic cell-cell adhesion molecule in epithelial cells and is a renowned tumor and invasion suppressor (40). Both MTA1 and MTA2 are components of NuRD ATP-dependent chromatin remodeling and histone deacetylase complexes. Whereas the transcription factor complex that contains MTA-1 is highly expressed in metastatic cells, a housekeeping role is suggested for the complex that contains the homologous MTA2 (41). Both S100A4 and matrilysin are known for their importance in malignant invasion (42, 43) . Down-regulation of the Wnt11 proximal promoter by Kaiso was recently shown in the Xenopus model system (24). The Wnt11 protein contributes via a noncanonical, β-catenin-independent Wnt signaling pathway to planar cell polarity and morphogenetic movements such as convergent extension. More particularly, inhibition of Xenopus Kaiso by antisense morpholino oligonucleotides or by nuclear p120ctn was shown to interfere with normal gastrulation movements, axial elongation, and neural fold closure (24). Expression of WNT11 is up-regulated in various human tumor cell lines and tumors (44), including high-grade prostate tumors and androgen-independent prostate cancer cell lines and xenografts (45). Treatment of IEC6 rat intestinal epithelial cells with Wnt11 down-regulates E-cadherin-mediated cell-cell contacts and stimulates cell migration and contact-independent cell growth (46). This implicates nuclear Kaiso in control of particular cell migration events. The diversity of Kaiso localization in many of the tumors described herein may be related to these observations, although the molecular details and functional consequences remain poorly understood.
Our data suggest that the microenvironment has a crucial causative role in regulating Kaiso expression and localization patterns. Modulating the microenvironment of HT29, SW48, and MCF-10A cells changed Kaiso levels and switched its localization between the cytoplasm and the nucleus. The fact that these phenomena were readily reversible suggests that they are not related to changes in DNA methylation. Differences between autonomous growth of cells in two and three dimensions may be important determinants. Other factors, such as the presence or absence of stroma and/or oxygen levels were suggested to be instrumental with the finding that nuclear Kaiso was frequently maintained along the periphery of tumor xenografts, and was also associated with the more malignant parts in several tumors. In keeping with this, we observed an inverse relationship between expression of carbonic anhydrase IX, a frequently used marker for hypoxia, and nuclear Kaiso in young HT29 xenografts.
In order to define the microenvironmental factors that modulate Kaiso expression patterns more precisely, we tested several relevant parameters directly. In vitro cultivation of HT29 or SW48 cells under hypoxic conditions did not by itself abolish expression of nuclear Kaiso. Loss of nuclear Kaiso was observed in dense cultures, but this happened under both normoxic and hypoxic conditions. This suggests that progressive Kaiso loss in tumor cell nuclei in xenografts is also mainly related to high cell density rather than to hypoxia. One possibility is that threshold levels of paracrine down-regulating factors are more easily reached under these conditions. Indeed, three-dimensional spheroids of HT29 cells grown without supplementation with extracellular matrix components also lost nuclear Kaiso. Likewise, three-dimensional cultures of MCF-10A cells in Matrigel showed translocation of nuclear Kaiso to the cytoplasm, but this could occur early after seeding, even at the single-cell stage. Later on, when a polarized differentiated phenotype appeared in acini-like structures, Kaiso expression was lost completely. In contrast, when these cells were cultured on either uncoated or Matrigel-coated glass cover slips, strong nuclear Kaiso was retained. Our data indicate that close cell contact with either the extracellular matrix in a three-dimensional matrix, or with fellow cells in dense cultures or spheroids simulating a three-dimensional tissue, could trigger Kaiso down-regulation. This agrees well with the paucity of nuclear Kaiso in normal tissues, where cells are bounded in three dimensions either by fellow cells or by extracellular matrix.
Presently, we do not know why a small fraction of normal cell types express nuclear Kaiso or why only particular tumor cells do so. Consistent expression of nuclear Kaiso in two-dimensional cultures of tumor cell lines may point to a dedifferentiation process resulting from insufficient contact with extracellular matrix or surrounding cells. For some tumor types, a similar dedifferentiation with consequent insensitivity to microenvironmental cues and acquisition of a higher migratory behavior may explain the occurrence of nuclear Kaiso. Alternatively, nuclear Kaiso in particular tumors may be directly induced by activated oncogenes or loss of tumor suppressor genes, and lead to a more invasive phenotype by virtue of the transcriptional activities of Kaiso. The expression of nuclear Kaiso in cell types with diffusely infiltrating growth patterns, such as eosinophilic granulocytes and paratesticular sarcoma cells, fits this hypothesis. Nuclear Kaiso in infiltrating parts of otherwise coherent tumor nodules and in metastases may reflect the dedifferentiation and higher migratory activities at these locations. However, the nature of most genes known to be candidate targets for inhibition by Kaiso, predicts an inhibition rather than a stimulation of invasion by nuclear Kaiso. It may therefore be worthwhile in the future to analyze three-dimensional cultures of MCF-10A cells transformed with various oncogenes (47, 48) for subcellular expression patterns of Kaiso in relation to loss of cell polarity and acquisition of invasive properties. It is also important to keep in mind that tumors are influenced by the state of activation of the surrounding stroma (49). From the above, it is evident that not all findings on Kaiso subcellular localization can be readily explained. For instance, the observation that clinically localized PCas express no nuclear Kaiso regardless of grade, shows that aggressive tumors do not invariably have nuclear Kaiso.
Kaiso's interaction partner p120ctn has been reported to be partly nuclear in E-cadherin-deficient tumor cell lines (10) and in a minor fraction of lobular breast cancers with complete loss of E-cadherin (50). Hence, it is surprising that we were unable to detect colocalization of Kaiso with p120ctn in the nucleus. In contrast, the presence of both proteins in the cytoplasm was readily seen in various tumors, including the xenografts. For pancreatic carcinomas, p120ctn was reported to occur in both the cytoplasm and nucleus, particularly in the case of high-grade undifferentiated tumors (51). Recent studies on breast cancer revealed frequent cytoplasmic p120ctn in lobular carcinomas and atypical lobular hyperplasias, but rarely in ductal tumors (50, 52) . Cytoplasmic p120ctn is known to modulate the activity of small GTPases, to contribute to growth factor-induced cell migration, and to regulate assembly and stability of E-cadherin-dependent cell junctions (6). It is conceivable that Kaiso is retained in the cytoplasm or excluded from the nucleus by virtue of its interaction with cytoplasmic p120ctn. Although it has been shown that nuclear p120ctn relieves Kaiso-mediated transcriptional repression in cultured cells, whereas cytoplasmic p120ctn does not (24, 26) , the common cytoplasmic location of Kaiso in vivo is consistent with a mechanism whereby p120ctn inhibits Kaiso transcriptional repression by sequestering it in the cytoplasm rather than by directly blocking its binding to DNA. Physical dissociation between cytoplasmic p120ctn and nuclear Kaiso, as we observed in particular tumors or tumor fields, may thus contribute to tumor progression. On the other hand, we cannot presently exclude that other p120ctn-independent mechanisms exist for cytoplasmic sequestration and down-regulation of Kaiso. Further studies are needed to elucidate the exact role of possible Kaiso-p120ctn interactions in the cytoplasm, as well as a possible correlation with activity of small GTPases.
In summary, we have shown that the microenvironment contains crucial information that modulates Kaiso expression and subcellular localization. Our observations point to an important role for nuclear Kaiso in particular cases of dynamic cell behavior and dedifferentiation. Further research may determine the exact underlying rules that are likely to differ on a case-by-case basis. Given that high levels of nuclear Kaiso observed in monolayers in vitro are not representative of nonpathologic situations in vivo (or in vitro three-dimensional), experimental conditions for studying the physiologic role(s) of Kaiso in either normal tissues or malignant tumors need to be thoughtfully reconsidered.
Grant support: FWO, the Geconcerteerde Onderzoeksacties of Ghent University, Fortis Verzekeringen (Belgium), and Interuniversity Attraction Poles Programme, Belgian Science Policy. Stichting Emmanuel van der Schueren and by a BOF grant from Gent University (A. Soubry). J. van Hengel is a postdoctoral fellow with the FWO (Fund for Scientific Research, Flanders).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Drs. J. Brugge and J. Debnath for MCF-10A cells, protocols, and for their three-dimensional culture; Dr. P. Brouckaert for his help with the xenografts; Dr. G. Berx for constructive discussions; Hilde Hellemans, Petra D'Hooge, and Barbara Gilbert for technical assistance; and Dr. Amin Bredan for editorial assistance.
- Received June 8, 2004.
- Revision received December 3, 2004.
- Accepted January 6, 2005.
- ©2005 American Association for Cancer Research.