Juxtacrine Activation of Epidermal Growth Factor (EGF) Receptor by Membrane-anchored Heparin-binding EGF-like Growth Factor Protects Epithelial Cells from Anoikis While Maintaining an Epithelial Phenotype*

Loss of cell-matrix adhesion is often associated with acute epithelial injury, suggesting that “anoikis” may be an important contributor to cell death. Resistance against anoikis is a key characteristic of transformed cells. When nontransformed epithelia are injured, activation of the epidermal growth factor (EGF) receptor (EGFR) by paracrine/autocrine release of soluble ligands can induce a prosurvival program, but there is generally evidence for concomitant dedifferentiation. The EGFR ligand, heparin-binding EGF-like growth factor (HB-EGF), is synthesized as a membrane-anchored precursor that can activate the EGFR via juxtacrine signaling or can be released and act as a soluble growth factor. In Madin-Darby canine kidney cells, expression of membrane-anchored HB-EGF increases cell-cell and cell-matrix adhesion. Therefore, these studies were designed to test the effects of juxtacrine HB-EGF signaling upon cell survival and epithelial integrity when cells are denied proper cell-matrix interactions. Cells expressing a noncleavable mutated form of membrane-anchored HB-EGF demonstrated increased survival from anoikis, formed larger cell aggregates, and maintained epithelial characteristics even following prolonged detachment from the substratum. Physical association between membrane-anchored HB-EGF and EGFR was observed. Signaling studies indicated synergistic effects of EGFR activation and phosphatidylinositol 3-kinase signaling to regulate apoptotic and survival pathways. In contrast, although administration of exogenous EGF partially suppressed anoikis in wild type cells, it also led to an increased expression of mesenchymal markers, suggesting dedifferentiation. Taken together, we propose a novel role for membrane-anchored HB-EGF in the cytoprotection of epithelial cells.

In mammalian tissues, epithelial cells grow and differentiate only when they are in the correct context and are removed by apoptosis when they are not. Epithelial cells sense their location through specific interactions with neighboring cells. In culture, confluent epithelial cells can be maintained for extended periods of time in the absence of exogenous factors; however, at low cell density with no cell contact, epithelial cells undergo programmed cell death if exogenous growth factors are not provided (1). In addition, epithelial cells also require appropriate interaction with extracellular matrix and undergo apoptosis if such interaction is prevented. Apoptosis in response to inappropriate cell/extracellular matrix interactions is termed anoikis (2). Of note, transformed cells are resistant to anoikis, and this resistance is considered a necessary step for metastasis (3). Anoikis helps to maintain the correct cell number of high turnover epithelial tissues and is involved in a wide range of tissue-homeostatic, developmental, and oncogenic processes (2,3). Therefore, both proper cell-cell and cell-matrix interactions are required for normal growth and differentiation of the epithelial cells.
Polarized epithelial cells are susceptible to acute toxic and ischemic injury (4,5), which is accompanied by alterations in both cell-cell and cell-matrix adhesions and results in loss of cell polarity (6). This loss of cell polarity is responsible for the redistribution of integrin subunits from the basolateral to the apical membrane and contributes to the shedding of cells from the supporting basement membrane (7)(8)(9). With more sustained injury, epithelial cells undergo necrotic or apoptotic cell death (10,11). Epithelial cells that do not die are thought to participate in the regeneration and restoration of epithelial function by flattening, spreading, and migrating into areas denuded by exfoliation, where they then dedifferentiate, proliferate, and finally redifferentiate into functional epithelial cells (6). Although recent studies have documented dedifferentiation as an important event following epithelial injuries (12), previous studies did not address whether dedifferentiation of epithelial cells is essential for survival during epithelial injury.
Accumulating evidence indicates that locally produced growth factors may serve as mediators of regeneration of epithelial cells following acute injury (13,14). Specifically, previous studies have suggested that EGF, 2 or members of the EGF growth factor family, may play an important role in kidney * This work was supported by funds from the Department of Veterans Affairs and National Institutes of Health Grant DK51265 (to R. C. H.) and AHA-Scientist Development Grant 0435471N (to A. B. S.). 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. 1  tubule epithelial repair (13,15). For example, after either ischemic tubular injury or folic acid nephropathy, EGF receptor density increases (16,17), and subcutaneous injection of EGF or transforming growth factor-␣ significantly accelerates [ 3 H]thymidine incorporation and recovery of renal epithelial cell function (13). It is noteworthy that although expression of EGF decreases, expression of heparin-binding EGF-like growth factor (HB-EGF) increases in the kidney in vivo in response to acute renal tubular injury (16,18,19). Furthermore, exogenous administration of HB-EGF has been shown to be protective against ischemic injury irrespective of its administration prior to or after the induction of injury. HB-EGF is synthesized as a membrane-anchored precursor (pro-HB-EGF) that is cleaved to release a soluble growth factor (sHB-EGF) (20). It has been suggested that both isoforms of HB-EGF may serve as biologically active ligands for EGF receptors and therefore may signal via either juxtacrine or paracrine/ autocrine modes (21)(22)(23). Using MDCK cells stably expressing wild type HB-EGF and either noncleavable and hence membrane-anchored or secreted HB-EGF mutants, we have reported that expression of soluble or noncleavable membraneanchored HB-EGF mutant molecules resulted in distinctly different cell phenotypes. Although expression of soluble HB-EGF induced a transformed morphology with decreased cell-cell and cell-matrix interactions, expression of the noncleavable membrane-tethered HB-EGF mutant enhanced cell-extracellular matrix and cell-cell interactions (24). The present study was undertaken to investigate the effects of juxtacrine activation of EGFR under conditions similar to epithelial injury in vivo. The present studies utilized a model of anoikis, in which proper cell-matrix interaction was disrupted but cell-cell interactions were still possible in order to examine the potential role of membrane-anchored HB-EGF in maintenance of epithelial cell integrity.

MATERIALS AND METHODS
Cell Culture and Transfections-MDCK type II and MDCK 5aa cells were grown in Dulbecco's modified Eagle's medium as described (24). The details of HB-EGF 5aa construct, transfection, generation, and characterization of the stable cell lines has been described previously (24). In brief, to generate the noncleavable and hence membrane-anchored HB-EGF (HB-EGF 5aa ) construct, the 5-amino acid putative cleavage site (PVENP) of full-length rat HB-EGF was deleted. In addition, a C-terminal FLAG tag was added to this construct for ease of detection. MDCKII cells stably expressing this construct were named MDCK 5aa cells. Immunoblotting using anti-rat HB-EGF antibody confirmed the expression of transfected construct, and immunolocalization using anti-FLAG antibody confirmed a predominantly lateral membrane localization of the expressed protein (24). Further characterization indicated that inducers of cleavage of membrane-associated full-length HB-EGF fail to induce shedding of soluble HB-EGF in the MDCK 5aa cells, indicating that these cells expressed a membrane-associated but noncleavable form of the protein (24). Two separate clones of MDCK 5aa cells with similar expression levels that had been characterized in our previous study were used for the studies unless stated otherwise. To ensure that the observed cell death was caused by lack of attachment alone and not deprivation of critical medium components, we performed all anoikis assays in complete medium containing all supplements unless stated otherwise.
Antibodies and Reagents-The rabbit anti-rat HB-EGF polyclonal antibody was a generous gift from Dr. Li Feng (Department of Immunology, The Scripps Research Institute, La Jolla, CA). The mouse or rabbit anti-FLAG (M2), anti-vimentin, anticytokeratin antibodies, polyHEMA, and CRM-197 (a mutated, nontoxic diphtheria toxin) were purchased from Sigma. The anti-EGFR, anti-phospho-EGFR (Tyr 1173 ), anti-poly(ADP-ribose) polymerase (PARP), and anti-clusterin antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). The anti-caspase-3, Bcl-xL, Bad, Bax, and Bak antibodies and EGF were purchased from Upstate Biotechnology, Inc. (Lake Placid, NY). Anti-Bim and anti-phosphorylated focal adhesion kinase antibodies were purchased from Cell Signaling (Danvers, MA). Anti-E-cadherin antibody was purchased from Transduction Laboratories (San Jose, CA). The apoptosis kit was obtained from Roche Applied Science. Anti-rabbit or mouse rhodamine-X or fluorescein isothiocyanate were purchased from Jackson Laboratories (West Grove, PA). The EGFR kinase inhibitor PD153035, PI 3-kinase inhibitor LY294002, and mitogen-activated protein kinase (ERK1/2) inhibitor U0126 were obtained from EMD Biosciences (San Diego, CA). The anti-EGFR blocking antibody (clone 528) was purified from hybridomas obtained from ATCC at the Vanderbilt University Medical Center Antibody Core Facility. BB-94 (Batimastat, a broad range metalloproteinase inhibitor) was from British Biotechnology (Oxford, UK).
Culture on PolyHEMA-Petri dishes (10 cm) or 6-well culture dishes (Falcon; BD Biosciences) were coated with a film of polyHEMA (Sigma), following the protocol reported by Folkman and Moscona (25). This methodology was utilized to prevent cell-substratum adhesion and thereby induce anoikis. Briefly, a solution of 120 mg/ml polyHEMA in 95% ethanol was mixed overnight, centrifuged at 2,500 rpm to remove undissolved particles, and diluted 1:10 with 95% ethanol. Then 0.95 ml/mm 2 of the above polyHEMA mixture was pipetted into Petri dishes of 35-or 100-mm diameter, and dishes were left to dry at room temperature in the culture hood. Before use, poly-HEMA-coated dishes were washed twice in phosphate-buffered saline (PBS). Unless stated otherwise, cells were plated at a density of 100,000 cells ml Ϫ1 in 35-mm culture dishes or 100-mm dishes.
Detachment-induced Apoptosis Assay-Cells were cultured in medium containing complete supplements, including 10% fetal bovine serum on polyHEMA-coated dishes at a density of 100,000 cells ml Ϫ1 for 24 h. Cells were then collected by gentle centrifugation, washed with PBS, and diluted with PBS. A total of 25,000 cells were used to determine apoptosis with the cell death detection enzyme-linked immunosorbent assay kit (Roche Applied Science) according to the manufacturer's instructions. Experiments were carried out at least three times.
Immunofluorescence Microscopy-Cells grown on poly-HEMA-coated culture dishes were replated on poly-L-lysinecoated glass slides for 20 min (the minimum time for attachment of the cells) at 37°C, rinsed with ice-cold 1ϫ PBS, and fixed with 3.7% formaldehyde in PBS for 30 min on ice. Staining and imaging were performed as described (24).
Cross-linking-For cross-linking, cells from suspension culture were collected by gentle centrifugation (200 ϫ g for 10 min at 4°C) and resuspended in 1ϫ PBS (with Ca 2ϩ and Mg 2ϩ ) to avoid breaking of the cell-cell contact in clusters. The resulting cell suspension was plated in untreated plastic dishes to prevent any cell attachment and treated with disuccinimidyl suberate (a noncleavable chemical cross-linker; 1 mM in HEPES buffer, 30 min, room temperature), followed by quenching with 200 mM glycine in PBS, pH 7.2, for 15 min. Thereafter, cells were centrifuged as above, washed twice with 1ϫ PBS, lysed in radioimmune precipitation buffer, and processed for preparation of total cell lysate.
Cytosol and Mitochondrial Enriched Fractions-Cytosol and mitochondrial enriched fractions were prepared using a kit from Active Motif North America (Carlsbad, CA) and following the manufacturer's protocol.
Statistics-Graphic data are presented as mean Ϯ S.E. Statistical analysis was performed, where appropriate, using Student's t test. Differences with p Ͻ 0.05 were considered statistically significant.

Expression of Membrane-tethered HB-EGF Decreased
Anoikis-Single cell populations of wild type MDCK and MDCK 5aa cells (stably expressing the noncleavable and thus membrane-anchored HB-EGF mutant construct) were cultured on polyHEMA as described under "Materials and Methods." After 24 h, cell death was determined using trypan blue exclusion. As shown in Fig. 1A, ϳ40% of the wild type cells were dead after 24 h of culture on polyHEMA. By contrast, anoikisinduced cell death in MDCK 5aa cells was reduced by 50% compared with the wild type cells. In further studies, we used cell reattachment as a measure of cell survival. Following culture of an equal number of cells on polyHEMA for 24 h, all cells was transferred to regular culture dishes and allowed to attach for 24 h. Thereafter, cells were washed to remove all dead and/or unattached cells, photographed, detached using trypsin-EDTA, and counted. As shown in Fig. 1B, the total number of viable MDCK 5aa cells was 4 times higher compared with the wild type cells. This increase in the total number of cells was higher than the initial protection against cell death observed in MDCK 5aa cells and suggested a combined role of cell survival and cell proliferation. Indeed, basal cell proliferation in MDCK 5aa cells is higher compared with the wild type cells. 3 To analyze further whether the observed cytoprotection in MDCK 5aa cells was a transient or persistent event, we cultured equal number of wild type and MDCK 5aa cells on polyHEMA. As described above, cells were plated in complete culture medium but did not receive any further replacement of culture medium. After 7 days of culture on polyHEMA, cells were then plated on regular culture dishes. Even after 7 days of suspension culture, a significant proportion of cells in MDCK 5aa cells remained viable, as demonstrated by their ability to attach and spread. In contrast, although occasional wild type cell aggregates did adhere to the culture plates, there was no cell spreading (data not shown).
Since these measures of cell viability do not differentiate between apoptotic and necrotic cell death, we utilized an enzyme-linked immunosorbent assay-based apoptosis-specific assay. Cells were subjected to growth on polyHEMA for 24 h and collected by gentle centrifugation. Cells were then further diluted, and a total of 25,000 cells were used for the apoptosis assay. As shown in Fig. 1C, MDCK 5aa cells had a 51% decrease in anoikis compared with the wild type cells. Nuclear staining with 4Ј,6-diamidino-2-phenylindole indicated extensive nuclear fragmentation in wild type cells but only rare fragmentation in the MDCK 5aa cells (Fig. 1D).
Expression of Proapoptotic Molecules Decreased in Cells Expressing Membrane-tethered HB-EGF-A balance in cell proliferative and apoptotic pathways is essential for the survival and normal functioning of epithelial cells (1). Anoikis disturbs this balance and induces activation of proapoptotic pathways, and expression/activity of different proapoptotic molecules is up-regulated upon induction of anoikis (26,27). In MDCK II cells, anoikis induces expression/activation of various proapoptotic molecules, including Bad, Bax, Bak, Bim, and caspases (2,28,29). To test whether the observed cytoprotection against anoikis in MDCK 5aa cells resulted from prevention of the induction of apoptotic pathways, we examined expression/activation of proapoptotic molecules in MDCK 5aa cells compared with the wild type cells.
Samples from cells subjected to anoikis were collected at time 0 (when cells were plated on the polyHEMA-coated dishes) and at 4, 8, and 24 h after plating. Expression of Bak, Bax, Bim, Bad, and activated caspase-3 was determined using total cell lysate and antigen-specific antibodies. In addition, we examined poly(ADP-ribose) polymerase cleavage, since PARP is cleaved differentially during apoptosis versus necrosis, and generation of an 85 kDa band is a signature pattern of apoptosis (30). Although no differences were observed in the expression of Bak at all time points compared with the 0 h in either wild type or MDCK 5aa cells, expression of Bax, Bim, Bad, and caspase-3 all increased in a time-dependent manner in the wild type cells, with the highest expression of Bax and caspase-3 at 8 h of culture on polyHEMA-coated plates and maximum expression of Bad and Bim at 24 h. PARP cleavage to generate the 85-kDa fragment was highest at 8 h and was also associated with a time-dependent increase in its protein expression. In MDCK 5aa cells, although a minor increase in the expression of Bax was observed at early time points, its expression remained lower compared with wild type cells at all time points. Interestingly, although the pattern of Bim expression was not strikingly different in MDCK 5aa cells compared with the wild type cells, there was an apparent mobility shift in MDCK 5aa cells compared with wild type cells at all time points. It is noteworthy that phosphorylation of Bim results in a shift in its mobility, which induces its degradation and inactivation (31,32). Bad expression remained suppressed in MDCK 5aa cells at all time points of the study. Although an early activation of caspase-3 was observed in MDCK 5aa cells upon plating onto polyHEMA, expression of cleaved caspase-3 remained lower in these cells at both the 8 and 24 h time points compared with wild type cells.
In addition, the levels of PARP cleavage were lower in MDCK 5aa cells at all time points compared with the wild type cells (Fig. 2).

Mitochondrial Translocation of Proapoptotic Molecules Was Largely Prevented in MDCK 5aa Cells upon Induction of
Anoikis-The induction of apoptosis requires translocation of proapoptotic molecules from cell cytosol to the mitochondrial membrane, which induces changes in the mitochondrial pore complex and thereby activates caspases (27,33). We further determined whether the decreased expression of proapoptotic molecules in MDCK 5aa cells was also associated with similar decreases in mitochondrial translocation. Cytosol and mitochondrial enriched fractions were prepared as described under "Materials and Methods," and equal amounts of protein were resolved to determine the relative expressions of Bad, Bax, Bak, and Bim in both fractions.
Predominant expression of Bak was observed in the cytosolic fraction compared with the mitochondrial fraction, and similar to the expression in total cell lysate, no major differences were observed in the expression of Bak between wild type and MDCK 5aa cells. In contrast, expression of Bax in the wild type cells was higher in the mitochondrial fractions at 8 and 24 h postanoikis. Similar to the expression in total cell lysate, expression of Bax in MDCK 5aa cells was higher at 0 h in the mitochondrial fraction but remained lower compared with wild type cells thereafter (Fig. 3). Similarly, an increase in Bad expression in the mitochondrial fraction of wild type cells was observed, FIGURE 1. Expression of membrane-anchored HB-EGF provides cytoprotection against anoikis. A, determination of cell death upon induction of anoikis. Equal numbers of MDCK control or MDCK 5aa cells were cultured on polyHEMA-coated dishes for 24 h, and cells were then collected and subjected to trypan blue exclusion. N, trypan blue-positive cells. Values are presented as mean Ϯ S.E. (p Ͻ 0.01). B, cell attachment and proliferation as a measure of survival following anoikis. Equal numbers of MDCK control or MDCK 5aa cells were subjected to anoikis for 24 h. Thereafter, cells were collected by gentle centrifugation and replated on regular cell culture plates in the same culture medium for 24 h. Phase-contrast images were taken, cells were detached using trypsin-EDTA, and cell number was determined. Values are presented as mean Ϯ S.E. (p Ͻ 0.01). C, determination of apoptosis following anoikis. Equal numbers of MDCK control or MDCK 5aa cells were subjected to anoikis for 24 h and collected by gentle centrifugation. A total of 25,000 cells were used for the preparation of the sample for the apoptosis assay. Values represent relative percentage decrease in apoptosis in the MDCK 5aa cells and are presented as mean Ϯ S.E. (p Ͻ 0.01). D, nuclear fragmentation as a marker of apoptosis. Cells were collected by gentle centrifugation after 24 h of culture on polyHEMA and replated on poly-L-lysine-coated glass slides. Nuclei were stained using 4Ј,6-diamidino-2-phenylindole, dihydrochloride. The arrows (white for wild type cells and gray for MDCK 5aa cells) indicate that nuclear fragmentation was higher in the wild type cells compared with MDCK 5aa cells (original magnification, ϫ100).
which was highest after 8 h of anoikis. In MDCK 5aa cells, low levels of Bad expression were detected at 0 h but decreased thereafter and were lowest at 24 h after anoikis. Increased mitochondrial expression of Bim was also detected at 8 and 24 h in the wild type, whereas in the MDCK 5aa cells, Bim expression showed a mobility shift and decrease in expression after induction of anoikis. The lack of apparent expression of Bax or Bim in the cytosolic fractions was due to loading a decreased amount of protein from cytosolic fractions to match the protein amount loaded for the mitochondrial fractions.
Cells Expressing Membrane-anchored HB-EGF Formed Large Cell Aggregates and Demonstrated Increased Expression of Prosurvival Molecules-Although these results indicated that expression of membrane-anchored HB-EGF inhibited the activation of proapoptotic pathways, further studies examined whether there was simultaneous enhancement of prosurvival pathways. The role of Bcl-xL in the EGFR-dependent cytoprotection against anoikis is well documented (34,35). Therefore, we determined the expression of Bcl-xL in the same total cell lysate used to detect expression of proapoptotic molecules. As shown in Fig. 4A, similar levels of expression of Bcl-xL were detected in both cell types at the zero time point. Although the expression of Bcl-xL remained largely unaltered in the wild type cells at all time points of the study, a robust increase in the expression of Bcl-xL was observed in the MDCK 5aa cells in a time-dependent manner, with the highest expression at 24 h postanoikis (Fig. 4A).
Formation of cell aggregates upon culture in suspension is a known characteristic of MDCK II cells (36). When cultured on polyHEMA, both cell types formed cell aggregates, and the size of these aggregates increased with time. However, it was noteworthy that the aggregates formed in MDCK 5aa cells were several times larger than the cell aggregates formed in the wild type cells (Fig. 4B). Cell aggregation upon denial of appropriate cellmatrix attachment is suggested to be a mechanism of self-defense against the anoikis-induced cell death (37). Clusterin, a protein important in cell survival, has been implicated in the formation of such aggregates (38 -40). Interestingly, although expression of clusterin was highest in wild type cells at 0 h and decreased thereafter, reaching to its nadir at 24 h, minimal clusterin was detected in MDCK 5aa cells at 0 h but increased thereafter, and robust expression was detected 24 h after plating (Fig. 4A).
Prosurvival Signaling Pathways Were Highly Activated in Cells Expressing Membrane-tethered HB-EGF-Activation of prosurvival signaling pathways is critical in the regulation of cell survival, and the role of EGFR activation is well documented (34,35,41). As noted above, both membrane-anchored and soluble HB-EGF are potential ligands for EGFR, but the consequences of activation could be potentially different (20,23,24). As demonstrated in Fig. 5, anoikis induced a delayed and modest EGFR phosphorylation in the wild type cells beginning between 4 and 8 h of culture. In contrast, low levels of phospho-EGFR were detected in MDCK 5aa cells at time zero of culture on polyHEMA, with progressive increases until 24 h of suspension culture. Similar to EGFR phosphorylation, ERK1/2 phosphorylation was higher in MDCK 5aa cells compared with wild type cells and was sustained throughout the   study, with the highest phosphorylation at 24 h after induction of anoikis. In contrast, phospho-Akt decreased progressively in wild type cells but was maintained constant and high in MDCK 5aa cells. Phosphorylation of focal adhesion kinase increased in the wild type cells following culture on polyHEMA but was unchanged in the MDCK 5aa cells (Fig. 5).

Inhibition of the Signaling Pathways Indicated Combined Roles for EGFR and PI 3-Kinase Signaling in the Cytoprotection in MDCK 5aa
Cells-To assess further the role of differential activation of the prosurvival signaling pathways in the MDCK 5aa cells in the observed cytoprotection against anoikis, we used specific pharmacological and immunologic inhibitors: PD153035 (1 M), a specific and stable EGFR tyrosine kinase inhibitor and an EGFR blocking antibody (clone 528; 40 g/ml) to interfere with EGFR signaling; U0126 (10 M) to inhibit phosphorylation of ERK1/2; and LY294002 (20 M) to inhibit activation of PI 3-kinase. Cells were prepared and plated for the experiments as described before, and the respective inhibitors were added to the regular culture medium at the time of the plating of the cells. To confirm the inhibition of the desired signaling pathways with the respective inhibitors, we examined the expression of phospho-EGFR, -ERK1/2, and -Akt with or without specific inhibitors (Fig. 6A). There were no further significant increases in apoptosis in the wild type cells with inhibition of activity of any of these signaling pathways (Fig. 6B). In contrast, inhibition of EGFR signaling induced significantly higher levels of apoptosis in the MDCK 5aa cells (PD153035, 24.2 Ϯ 2.2% increase; EGFR-blocking antibody 23.49 Ϯ 3.39% increase), as did inhibition of the PI 3-kinase signaling pathways (36.5 Ϯ 4.2% increase). In contrast, inhibition of ERK-mitogenactivated protein kinase signaling did not significantly augment apoptosis (Fig. 6B). Activated caspase-3, used as a separate marker of apoptosis, confirmed the increases in apoptosis when EGFR or PI 3-kinase signaling pathways were inhibited (Fig.  6A). Simultaneous inhibition of the EGFR and PI 3-kinase resulted in an additive increase in apoptosis in the MDCK 5aa cells (Fig. 6B). It was noteworthy that although the addition of exogenous EGF decreased apoptosis by 21.0 Ϯ 2.6% in wild type cells, it decreased apoptosis in MDCK 5aa cells by only 7.5 Ϯ 2.6%.

Cytoprotection in the MDCK 5aa Cells Was Not the Result of
Dedifferentiation-MDCK 5aa cells grown under normal conditions display distinct epithelial cuboidal architecture with significantly increased cell attachment and transepithelial resistance compared with wild type cells, suggesting stronger cell-cell and cell-matrix adhesion (24). However, resistance to anoikis is an established characteristic of transformed cells (42). Therefore, to examine whether culture of MDCK 5aa cells on polyHEMA induced dedifferentiation, we examined expression of E-cadherin (an epithelial marker) and vimentin, smooth muscle actin (SMA), and fibroblast-specific protein (FSP-1) (established mesenchymal markers). Total cell lysates from cells cultured on polyHEMA for 24 h were used. Compared with wild type cells, the expression of E-cadherin was higher and expression of vimentin, SMA, and FSP-1 was lower in MDCK 5aa cells (Fig. 7A, I). We further examined expression of these proteins in response to the exogenous administration of EGF. Both wild type and MDCK 5aa cells were grown on poly-HEMA in the presence or absence of EGF (100 ng/ml) for 24 h. As shown in Fig. 7A, II, EGF administration resulted in increased expression of vimentin, SMA, and FSP-1 in the wild type cells compared with the cells grown in the absence of EGF.  NOVEMBER 9, 2007 • VOLUME 282 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 32895
However, EGF administration in the MDCK 5aa cells resulted in only minor increases in SMA and FSP-1 expression, whereas vimentin expression remained largely unaffected. EGF admin-istration did not affect E-cadherin expression in either cell type (Fig. 7A, II).
We determined cellular distribution of E-cadherin, ␤1-integrin and ZO-1 in wild type and MDCK 5aa cells following culture on polyHEMA. Since loss of lateral distribution of ␤1-integrin and E-cadherin is described during dedifferentiation and is a hallmark of epithelial injury (8), we performed co-immunofluorescence of E-cadherin or ␤1-integrin with ZO-1, a tight junction protein that is normally localized at the apical end of lateral cell-cell adhesion sites in polarized epithelial cells (43). As noted above, wild type and MDCK 5aa cells both form cell aggregates upon culture on polyHEMA. Therefore, for immunofluorescence studies, cell aggregates from 24-h suspension cultures were transferred to poly-L-lysine-coated slides, allowed to attach for 15-20 min at 37°C, and processed for staining as described under "Materials and Methods." Immunolocalization studies demonstrated that the distinct lateral membrane localization of E-cadherin and ␤1-integrin was largely preserved in MDCK 5aa cells but was largely compromised in wild type cells. In MDCK 5aa cells, ZO-1 was localized in a punctuate pattern at tight junctions similar to its localization in normal attached epithelial cells. In contrast, punctuate tight junction localization of ZO-1 was largely absent in the wild type cells (Fig. 7B, I and II). (24). In the current experiments, we examined cellular distribution of membrane-anchored HB-EGF as well as EGFR following culture on polyHEMA. Immunolocalization using anti-FLAG antibody (a FLAG epitope is present into the C terminus of the 5-aa construct) demonstrated that the lateral membrane distribution of membrane-anchored HB-EGF was preserved even during culture on polyHEMA. Also, co-immunolocalization of EGFR with ZO-1 indicated largely lateral membrane localization for EGFR in MDCK 5aa cells. By contrast, in wild type cells, ZO-1 and EGFR were distributed diffusely over the cell surface (Fig. 8A, I). As shown in Fig. 8A, II, overlay of the membranetethered HB-EGF and EGFR immunolocalization demonstrated co-localization of these proteins.

EGFR Expression Was Localized to Plasma Membrane in MDCK 5aa Cells and Co-localized with Membrane-tethered HB-EGF-We have previously demonstrated a predominantly lateral membrane localization of the membrane-tethered HB-EGF in MDCK 5aa cells grown on transwell filters
Membrane-tethered HB-EGF Physically Associated with EGFR-The co-immunolocalization studies suggested possible association between membrane-anchored HB-EGF and EGFR following culture on polyHEMA. Therefore, we performed immunoprecipitation using total cell lysates from the cells cultured on polyHEMA for 24 h with an anti-EGFR antibody. The resultant immunoprecipitate was immunoblotted with anti-FLAG antibody to detect HB-EGF. The anti-FLAG antibody recognized bands of the expected size in the immunoprecipitates from MDCK 5aa cells but not in the wild type cells (Fig. 8B, I).
We also performed chemical cross-linking of the cells grown on polyHEMA for 24 h using a noncleavable cross-linker (disuccinimidyl suberate) to determine possible protein complexing with membrane-tethered HB-EGF. Cell lysates were prepared in radioimmune precipitation-cell lysis buffer following FIGURE 7. Cytoprotection against anoikis in MDCK 5aa cells was not associated with dedifferentiation. A, expression of epithelial (E-cadherin) and mesenchymal (vimentin, smooth muscle actin (SMA), and fibroblast-specific protein (FSP-1)) markers. I, MDCK control and MDCK 5aa cells were cultured on polyHEMA for 24 h. Total cell lysate was prepared, and relative expression of epithelial and mesenchymal marker proteins was determined using antigen-specific antibody. Reprobing with actin was used for normalization. II, to determine that survival against anoikis upon paracrine EGFR is indeed associated with dedifferentiation of epithelial cells, MDCK control and MDCK 5aa cells were grown on polyHEMA in the presence or absence of EGF (100 ng/ml) for 24 h. Total cellular protein was isolated, and expression of epithelial and mesenchymal markers was examined. Protein loading was normalized with actin. Although expression of vimentin, FSP-1, or SMA increased in MDCK control cells in the presence of EGF, no major changes in the expression of the same proteins was observed in MDCK 5aa cells in the presence of EGF. B, immunofluorescent localization of membrane markers in cells cultured on polyHEMA. I, co-immunolocalization of E-cadherin and ZO-1. Please note the predominant lateral membrane location of E-cadherin (arrow) and apical punctuate location of ZO-1 (arrowhead) staining in MDCK 5aa cells (original magnification, ϫ400). Note that in the merged figure, E-cadherin is green and ZO-1 is red. II, co-immunolocalization of ␤1-integrin and ZO-1. The distinct apical punctuate staining for ZO-1 and lateral staining for ␤1-integrin is largely compromised in MDCK control cells compared with MDCK 5aa cells (original magnification, ϫ400). Green, ␤1-integrin; red, ZO-1.
cross-linking. Immunoprecipitation and immunoblotting of the non-cross-linked and cross-linked samples using anti-HB-EGF antibody showed a decrease in the intensity of the ϳ28 kDa band in the cross-linked samples compared with the noncross-linked samples, suggesting possible protein complexing (data not shown). However, the possibility of differential immunoprecipitation due to decreased availability for epitope recognition by the antibody could not be ruled out. Therefore, we performed immunoblotting of the cross-linked or normal cell lysates using anti-HB-EGF antibody. Results confirmed the observations from the immunoprecipitation experiments, since the intensity of the ϳ28 kDa HB-EGF bands obtained in noncross-linked samples decreased dramatically upon crosslinking, and several bands of higher mobility were generated. No bands were recognized in the lysates of the wild type cells with or without cross-linking (Fig. 8B, II), confirming specificity. These studies indicate the possible existence of a larger protein-protein complex. The membrane-anchored HB-EGF is The lateral membrane expression of EGFR was largely preserved in MDCK 5aa cells (original magnification, ϫ400). Green, EGFR; red, ZO-1. II, co-immunolocalization of EGFR and membrane-anchored HB-EGF. The overlay shows co-localization of membrane-anchored HB-EGF and EGFR in MDCK 5aa cells (original magnification, ϫ400). (green, EGFR; red, HB-EGF). B, examination of physical association of EGFR and other proteins with membrane-anchored HB-EGF. I, co-immuunoprecipitation with EGF receptor and membraneanchored HB-EGF. Total lysates from cells cultured on polyHEMA for 24 h were subjected to immunoprecipitation using a monoclonal anti-EGFR antibody, resolved on a 15% SDS gel, and immunoblotted with a monoclonal anti-FLAG antibody (FLAG epitope is present in the C terminus of the HB-EGF mutant construct). II, chemical cross-linking. Cells were cultured on polyHEMA for 24 h. One group of cells was subjected to chemical cross-linking using disuccinimidyl suberate. Total cell lysates were prepared and resolved on a 15% SDS-PAGE. Immunoblot was performed using an anti-HB-EGF antibody. The arrowheads indicate loss or decrease in intensity of bands, whereas arrows indicate bands generated upon cross-linking. Minus and plus signs represent noncross-linked and cross-linked samples, respectively.
known to associate with several proteins, including ␤1-integrin, CD9, and Bag-1, a protein known to enhance cell survival (44 -47); however, HB-EGF-associating proteins in our model system remain to be identified.

Inhibition of Endogenous HB-EGF Increases
Anoikis-Since MDCK 5aa cells overexpress membrane-anchored HB-EGF, we further sought to validate these findings in cells expressing endogenous HB-EGF. Since we have failed to detect HB-EGF expression in wild type MDCK II cells, we used LLCPK1/Cl4 cells for these studies. LLCPK1/Cl4 cells are clonal variants of LLCPK1 cells (48) derived from pig kidney proximal tubule (49) and express HB-EGF (50). To inhibit HB-EGF in these cells, we used CRM-197. HB-EGF serves as a receptor for diphtheria toxin, and Crm-197, a mutated nontoxic diphtheria toxin, has been shown to be a potent and specific inhibitor of HB-EGF (51). We have previously shown that, similar to human HB-EGF, porcine HB-EGF is also sensitive to . Experiments were carried out in complete culture medium as described above in order to ensure that the observed effects were due to manipulation of HB-EGF and not due to deprivation of essential nutrients. Although under normal physiological conditions, the majority of HB-EGF is expressed as uncleaved, membrane-anchored pro-HB-EGF, we also used BB-94 (Batimastat), a metalloproteinase inhibitor previously shown to inhibit cleavage of pro-HB-EGF (50), to test whether the effects observed in our study are due to inhibition of membrane-anchored HB-EGF and not the cleaved soluble form. LLCPK1/Cl4 cells growing in regular culture dishes were incubated with BB-94 (10 M), CRM-197 (25 M), or BB-94 ϩ CRM-197 for 2 h in regular culture dishes. Thereafter, cells were trypsinized, and equal numbers of cells (100,000 cells ml Ϫ1 ) were subjected to anoikis for 24 h using polyHEMAcoated dishes. The same concentrations of inhibitors were present during the culture of cells on polyHEMA. Of interest, control LLCPK1/Cl4 cells formed large cell aggregates similar to MDCK 5aa cells (shown in Fig. 4A), and the size of these aggregates further increased in cells grown in the presence of BB-94. By contrast, the size of cell aggregates markedly decreased in cells cultured in the presence of CRM-197 (Fig. 9A). Cells were collected by gentle centrifugation after 24 h of culture on poly-HEMA, and a total of 25,000 cells were subjected to an apoptosis-specific enzyme-linked immunosorbent assay, as described before. As shown in Fig. 9B, equivalent increases in anoikis were observed in cells cultured in the presence of CRM-197 and BB-94 ϩ CRM-197, respectively (34 Ϯ 5.8 and 40 Ϯ 3.9%, respectively). Anoikis in cells cultured in the presence of BB-94 alone was not significantly different from the control cells (Fig.  9B). Taken together, these data supported our findings from MDCK 5aa cells regarding a role for membrane-anchored HB-EGF in the regulation of survival against anoikis.

DISCUSSION
Although resistance to apoptosis during the pathogenesis of cancer is the most studied phenomenon relating to anoikis, anoikis represents one of the principal mechanisms underlying cell death following epithelial injury, and augmentation of cellcell contact may serve as a mechanism of cytoprotection. In squamous epithelial cells or primary mouse mammary gland epithelial cells, anoikis is inhibited by preservation of cell-cell anchorage despite loss of cell-matrix anchorage (52,53). Epithelial cells that do not die during injury may participate in the regeneration and restoration of epithelial function (13,54,55). Expression of EGFR and EGFR ligands increases in response to epithelial injuries (15,17), and exogenous administration of EGFR ligands attenuates the degree of epithelial injury and promotes repair and regeneration (13,35,41,56). However, autocrine/paracrine activation of the EGFR is usually associated with cellular dedifferentiation or transformation (57,58). EGFR inhibition decreases resistance to anoikis (59) but at the same time prevents the loss of epithelial characteristics (60). Recent studies indicate that epithelial to mesenchymal transition is a common phenomenon in response to epithelial injury (12) and thus suggest that dedifferentiation may be one of the mechanisms of EGFR-dependent protection against epithelial injury.
Among EGFR ligands, membrane-tethered HB-EGF is of special interest, because it forms complexes with different proteins, serves as an active ligand for two different receptors (ErbB1 and ErbB4), and can function in a juxtacrine manner (22,61). Both our studies and those of others suggest that biological responses to juxtacrine versus autocrine/paracrine activation of EGFR are potentially different, due to both quantitative and qualitative differences in activation of downstream signaling pathways (62). Furthermore, juxtacrine signaling requires that cells expressing the membrane-anchored ligand and receptor remain in close apposition. Our previous studies indicated that when cells are normally attached to their substratum, expression of membrane-anchored HB-EGF results in a distinctly different cell phenotype from cells expressing secreted HB-EGF (24), raising the question whether the effects of juxtacrine signaling by membrane-anchored HB-EGF could also be potentially different under detached conditions. It is important to note that in previous studies, we determined that cells expressing soluble HB-EGF underwent cellular transformation and showed significantly decreased cell-cell and cellmatrix interactions, whereas expression of the membrane-tethered HB-EGF resulted in significantly increased cell-cell and cell-matrix adhesion compared with wild type cells (24).
In the present study, we found that when grown on poly-HEMA to prevent any attachment to the substratum, MDCK cells expressing membrane-anchored HB-EGF not only were largely resistant to anoikis but also retained epithelial characteristics. These included appropriate distribution of membrane-associated proteins and persistent expression of established epithelial markers E-cadherin and lower expressions of vimentin, SMA, and FSP-1, well established mesenchymal markers. MDCK 5aa cells formed larger cell aggregates compared with the wild type cells under conditions favoring anoikis, suggesting stronger cell-cell contact. In contrast, although exogenous EGF administration to wild type cells increased cytoprotection, it also led to increased expressions of these mesenchymal markers, suggesting a potential role of dedifferentiation in such protection. These data further support the hypothesis that autocrine/paracrine activation of EGFR suppresses anoikis at least in part through induction of dedifferentiation. Our hypothesis that that membrane-anchored HB-EGF provides cytoprotection through establishment of stronger cell-cell contacts was further strengthened by the fact that LLCK1/Cl4 cells, which express endogenous HB-EGF, formed large cell aggregates when subjected to anoikis. Treatment of these cells with BB-94, a metalloproteinase inhibitor potentially capable of inhibiting any cleavage of membrane-anchored HB-EGF, further enhanced the size of these aggregates. By contrast, inhibition of HB-EGF using CRM-197 resulted in not only increased anoikis but also smaller cell aggregates. MDCK 5aa cells had increased expression of the cell-cell junction protein E-cadherin compared with the wild type cells; furthermore, MDCK 5aa cells retained normal expression of E-cadherin on the cell-cell adhesion site. In addition, there was appropriate cellular localization of ZO-1 and ␤1-integrin compared with the wild type cells, suggesting that MDCK 5aa cells were capable of maintaining their epithelial phenotype even during detached conditions. The fact that the membrane expression of EGFR in the MDCK 5aa cells was largely maintained even after 24 h of culture on polyHEMA (the time point with the highest EGFR activation) further supports juxtacrine activation of the EGFR. It is noteworthy that autocrine/paracrine activation of EGFR results in rapid receptor internalization and degradation (63).
Cytoprotection against anoikis in the MDCK 5aa cells was the result of activation of both EGFR-and PI 3-kinase-dependent pathways, whereas ERK1/2 kinase activation was not required. In this regard, a recent study using colonic epithelial cells detected a similar dependence on EGFR and PI 3-kinase activation but not mitogen-activated protein ERK1/2 kinase as the mechanism underlying resistance against anoikis. It is further noteworthy that colonic epithelial cells that formed cell-cell contacts exhibited enhanced EGFR and PI 3-kinase activation and were protected from anoikis compared with the cells that did not (64).
During anoikis, members of the Bcl-2 family function as pivotal determinants of whether cells become irreversibly committed to die. Some members of this family, including Bcl-xL, promote cell survival, whereas others, such as Bax, Bak, Bad, and Bim, promote apoptosis (27). One possible mechanism for their control of apoptosis is that their relative levels of expression and/or activation regulate the balance between apoptosis and survival (65). We observed a time-dependent but differential increase in the expression of the proapoptotic proteins in the wild type cells, with increased expression of Bax, Bad, Bim, and activated caspase-3. These findings are consistent with previously published studies demonstrating increases in the expression of Bax, Bim, or caspase-3 in MDCK cells upon induction of anoikis (28,66,67). Furthermore, PARP cleavage to generate an 85 kDa band, specific to apoptotic death, also increased with time. In contrast, in MDCK 5aa cells, expression of proapoptotic proteins decreased progressively and was lower compared with the control cells at all time points. In addition, a shift in the mobility of the Bim was observed (4 h after culture on polyHEMA). Such a mobility shift has been correlated with its phosphorylation and degradation, and a role for EGFR activation has been demonstrated in such phosphorylation (68 -70).
Cell-cell contact in the absence of cell-matrix adhesions has been shown to be cytoprotective in many cell types (52,53). We also found that EGFR activation increased in a time-dependent manner. In this regard, Shen and Kramer (71) reported a role for cell-cell contact-dependent EGFR activation in protection against anoikis in squamous epithelial cells. Of note, in our study, cells were initially plated as a population of single cells, which subsequently formed cell aggregates, and the size of these aggregates increased progressively with time. EGFR phosphorylation in MDCK 5aa cells also increased in a time-dependent manner, with the highest receptor phosphorylation observed at 24 h after growth on polyHEMA. Taken together, we postulate that juxtacrine activation of the EGFR upon cellcell contact in the MDCK 5aa cells underlies the observed EGFR phosphorylation, leading to regulation of pro-and antiapoptotic pathways. This conclusion is further supported by our observation that expression of clusterin, a glycoprotein involved in cell aggregation and cell survival, increased in a time-dependent manner in MDCK 5aa cells.
Akt activation has been previously shown to play a key role in providing cytoprotection against anoikis-induced apoptosis (72), and Akt phosphorylation was increased in MDCK 5aa cells. However, inhibition of EGFR phosphorylation resulted in only modest inhibition of Akt phosphorylation, and combined inhibition of EGFR and PI 3-kinase signaling resulted in a higher degree of apoptosis, suggesting the PI 3-kinase activation was in large part independent of EGFR signaling. Since higher Akt phosphorylation was observed in the MDCK 5aa cells even at the 0 h time point of anoikis, when no substantial differences in the EGFR phosphorylation were observed between control and MDCK 5aa cells, we postulate that the increased Akt phosphorylation in the MDCK 5aa cells was due to complexing of HB-EGF with other proteins. HB-EGF is known to interact with ␤1-integrin through its interaction with CD-9, as well as with Bag-1 (44,47,73). All of these proteins have established roles in cell survival. Such protein complexing with membrane-anchored HB-EGF is supported by our findings from the crosslinking experiments; however, the specific protein partners of noncleavable HB-EGF in our model system remain to be identified.
The juxtacrine activation of EGFR by the noncleavable HB-EGF suggests physical interaction between HB-EGF and EGFR. Our data from co-immunoprecipitation and co-immunolocalization also suggest direct physical association between the membrane-anchored HB-EGF and EGFR. These observations are supported by findings from a recent report that substitution of the transmembrane domain of human pro-EGF with the respective domain of pro-HB-EGF not only induced the ability to activate EGFR in a juxtacrine fashion but also resulted in physical association between these molecules (74). Of note, pro-EGF does not function in a juxtacrine fashion. It is noteworthy that these researchers noted that the interaction between EGFR and the pro-EGF construct containing the transmembrane domain of HB-EGF was most robust when cells were establishing cell-cell contact (74).
Taken together, in the present study, we demonstrate that in response to interruption of cell-extracellular matrix interactions, expression of membrane-anchored HB-EGF provides cytoprotection while maintaining epithelial characteristics. The cytoprotection was the result of juxtacrine activation of EGFR as well as possible partnering of membrane-anchored HB-EGF with other proteins. Our findings may partially explain why some epithelia are relatively protected against epithelial injury. In this regard, it is noteworthy that in the kidney the distal nephron, which is relatively protected against toxic or ischemic injuries, is also the principal site of expression of membrane-anchored HB-EGF. To our knowledge, this is first report describing a cytoprotective role in response to the juxtacrine activation of EGFR by membrane-anchored HB-EGF.