Production of epidermal growth factor related ligands in tumorigenic and benign human lung epithelial cells

We recently demonstrated that human lung epithelial cells, overexpressing ErbB-2, formed tumors in nude mice only when high levels of transforming growth factor alpha (TGFalpha) were produced. Cells transfected with a TGFalpha antisense vector failed to form tumors in nude mice. In order to further evaluate the importance, for tumorigenicity, of TGFalpha and its stimulation of ErbB family signalling, the production of other EGF family growth factors by these human lung epithelial cells was studied. We demonstrate for the first time that both tumorigenic and non-tumorigenic human lung epithelial cells produced, in addition to TGFalpha, amphiregulin, betacellulin, heparin-binding EGF and heregulin. These data suggest that human lung epithelial cells have the potential for multifactorial modulation of ErbB receptor family signalling through control of ligand as well as receptor production. In this system, the probable importance of TGFalpha-stimulated signaling for tumorigenicity is supported by its 13-fold higher production in tumorigenic as compared with non-tumorigenic cells and the 2-fold or lower differences observed in production of the other epidermal growth factor (EGF) family ligands.


Introduction
Cancer cells produce peptide hormones that can be secreted, bind to cell surface receptors and stimulate their own growth in an autocrine manner. The role of autocrine growth factors in the pathogenesis of lung cancer has been the subject of intensive study [1,2]. Since autocrine-stimulated mitogenesis appears to contribute signi®cantly to disease progression, the clinical ef®cacy of agents that interrupt autocrine loops are currently being evaluated.
Our laboratory has been studying the role of epidermal growth factor (EGF) ligands and ErbB family receptors in the transformation of human lung epithelial cells. This growth factor family is comprised of at least six members including EGF, transforming growth factor a (TGFa), amphiregulin (AR), betacellulin (BTC), heparin-binding-EGF (HB-EGF), and several alternatively spliced isoforms of heregulin (HRG) [3]. TGFa, AR, HB-EGF and BTC compete with EGF for binding to ErbB-1 [4]. In addition, BTC also binds to ErbB-4 directly [5]. HRG binds ErbB-3 and ErbB-4 and stimulates EGFR indirectly [6]. Recent studies demonstrated that the speci®c patterns of downstream signaling by ErbB-1 and ErbB-2 were dependent on ligand-speci®ed homoor heterodimerization [7]. Thus, in understanding autocrine contribution to tumorigenic conversion, it is important to delineate the spectrum of ligands produced. Production of EGF family ligands has been associated with increased proliferation in cancer. It has been demonstrated that monoclonal antibodies directed against TGFa inhibited the growth of lung adenocarcinoma cell lines in vitro [8]. Rachwal et al. have demonstrated that both HRG and TGFa mRNAs were abundantly expressed in lung tumor cell lines expressing ErbB-1 [9], and in vivo production of TGFa has been associated with shortened survival of patients with lung cancer [10]. Similarly, AR protein is expressed in 40±80% of human lung cancers and is associated with a poor prognosis for patients with non-small cell lung cancer [11]. AR mRNA has been found in lung tumor tissue, but not in adjacent benign tissue [10,12]. These studies suggest that autocrine production of EGF family ligands participates in stimulating growth of lung cancer cells.
We have recently demonstrated the importance of TGFa in inducing a tumorigenic phenotype in immortalized human lung epithelial cells. We showed that human lung epithelial cells, overexpressing ErbB-2, formed tumors in nude mice only when high levels of TGFa were produced. Inhibition of TGFa expression using an antisense strategy resulted in inhibition of tumor growth in nude mice and blocked the induction of constitutive ErbB-1/ErbB-2 heterodimers in the malignant cells [13]. We hypothesized that induction of a tumorigenic phenotype depended on achieving a threshold level of TGFa secretion. However, the role that other ErbB-1 ligands played in induction of tumorigenicity in this model was not assessed.
In the current study, we compare the production of other EGF family ligands in the tumorigenic versus non-tumorigenic cells. We demonstrate for the ®rst time the production of a broad spectrum of EGF family ligands, HRG, AR, HB-EGF and BTC, in human lung epithelial cell lines.

Cell line derivation and culture
The BEAS-2B cell line is a non-tumorigenic, immortalized human bronchial epithelial cell line derived from the infection of normal human bronchial epithelial cells with SV40 Adeno 12 hybrid virus [14]. It was grown in serum-free LHC-8 medium (Bio¯uids, Rockville, MD) according to established protocols [14]. The BEAS-2B E6 cell line was derived by introducing the human c-ErbB-2 expression vector (pLTRERBB-2neo) into BEAS-2B cells, as previously described [15]. The BEAS-2B E6T cell line (referred to as E6T) was derived from BEAS-2B E6 cells that had been passaged once in nude mice and recultured in vitro. These cells were shown to be derived from BEAS-2B by karyotypic analysis [15]. They were grown in serum-free LHC-8 medium containing geneticin (200 mg/ml) (Gibco-BRL, Gaithersburg, MD). E6TA cells were prepared by introducing a TGFa antisense expression vector (pLTRTGF-aHYG) into E6T cells, as previously described [13]. These cells were grown in serumfree LHC-8 medium containing hygromycin B (200 mg/ml) (Boehringer Mannheim, Indianapolis, IN) as well as geneticin as for E6T. MDA-MB-453 breast cancer cells (kindly provided by Ruth Lupu, Berkeley, CA) were grown in IMEM medium (Bio¯uids) supplemented with 10% fetal bovine serum (Bio¯uids). MDA-MB-231 breast cancer cells (American Type Culture Collection, Rockville, MD) were grown in RPMI-1640 medium (Bio¯uids) supplemented with 10% fetal bovine serum.

RNA isolation
Cells grown to 70% con¯uence were lysed in Trizol Reagent (Gibco-BRL, Gaithersburg, MD), and RNA isolated according to the manufacturer's protocol.

Reverse transcriptase-polymerase chain reaction
One microgram of RNA was incubated with 3 mg random hexamers, 200 units M-MLV reverse transcriptase (RT), 0.1 mM dNTP mix, 1£ ®rst strand buffer, 0.01 M dithiothreitol (DTT) (all from Gibco-BRL) and 20 units of RNasin (Promega, Madison, WI) in a total volume of 20 ml. The reaction was incubated for 40 min at 378C, followed by 5 min at 958C. The reverse transcribed product was used in subsequent polymerase chain reaction (PCR) protocols.
For detection of the HRG b3a isoform, outer primers 5 H -GTACGTCCACTCCCTTTCTGTCTCT-3 H (10 pmol) and 5 H -ACAGCTTAGCAACACCCT-TTTCAG-3 H (10 pmol) were incubated with 10 ml of reverse transcribed product as above under the following conditions: one cycle of 958C for 5 min, 35 cycles of 948C for 40 s, 558C for 40 s, 728C for 90 s, followed by one cycle of 728C for 5 min. For the nested ampli®cation of b3a, 10 ml of PCR product from the ®rst PCR ampli®cation was incubated with primers 5 H -GTTGGTGCTGCTTTCTTGTTGC-3 H (10 pmol) and 5 H -ACAGCTTAGCAACACCCTTTT-CAG-3 H (10 pmol) and ampli®ed as above under the following conditions: one cycle of 958C for 5 min, 35 cycles of 948C for 40 s, 558C for 40 s, 728C for 90 s, followed by one cycle of 728C for 5 min. This nested PCR reaction yielded a product of 361 bp.
For detection of BTC and AR, primers (10 pmol) were incubated with 10 ml of reverse transcribed product, 1£ PCR buffer, 0.1 mM dNTP mix, and 1.5 mM MgCl 2 in a total volume of 50 ml as above under the following conditions: one cycle of 958C for 10 min, 35 cycles of 948C for 40 s, 558C for 30 s, 688C for 2 min, followed by one cycle of 728C for 5 min. PCR reactions yielded a BTC product of 345 bp and an AR product of 478 bp. BTC primer pairs consisted of 5 H -CAAGCATTACTGCATCAAAGGG-AG-3 H and 5 H -CAACCTGGAGGTAACTTCATAG-CC-3 H . AR primer pairs consisted of 5 H -TTGGACCT-CAATGACACCTACTGTG-3 H and 5 H -TGGACTTT-TCCCCACACCGTTC-3 H . PCR was performed with controls lacking template, primers, and template and primers to control for contamination.

HRG bioassay
Cells were grown to 70% con¯uence in LHC-8 medium (Bio¯uids) after which the medium was replaced with LHC Basal medium (Bio¯uids). After 48 h, the conditioned media from each cell line were loaded onto a column containing 0.5 ml heparinconjugated Sepharose (Sigma) equilibrated with 10 mM Tris. The column was washed once with 0.3 M NaCl in 10 mM Tris. Subsequently, 0.3 ml fractions were collected while eluting sequentially with 0.3 M NaCl, 0.6 M NaCl and 0.9 M NaCl in 10 mM Tris. The eluates were then run on NAP-5 desalting columns (Pharmacia Biotech). The resulting 1 ml fractions were concentrated to a volume of 300 ml. One hundred microlitres of each concentrated fraction was incubated for 30 min with 250 000 MDA453 cells which had been serum starved for 18 h. MDA 453 breast cancer cells express ErbB-2, ErbB-3 and ErbB-4, but not HRG [16]. The positive control for stimulation of phosphorylation was conditioned medium from MDA-MB-231 breast cancer cells which secrete four isoforms of HRG [16]. Serumfree IMEM medium (Bio¯uids) was used as a negative control. Following a 30 min incubation, cells were washed in serum-free media and lysed in sample buffer (Enprotech, Natick, MA). Lysates were electrophoresed on a 4±20% polyacrylamide gradient gel (Novex, San Diego, CA), transferred to Immobilon-P membranes (Millipore, Bedford, MA) and probed with anti-phosphotyrosine antibody (UBI, Lake Placid, NY) followed by anti-mouse horseradish peroxidase conjugated secondary antibody (Amersham). SuperSignal chemiluminescence reagent (Pierce, Rockford, IL) was used to develop blots according to manufacturer's protocols. Sixty microlitres of each biologically active fraction were pooled, concentrated to a volume of 7 ml and applied to a 4± 20% polyacrylamide gel for Western blot analysis. Quantitative comparisons of HRG production by different cell lines was not attempted due to the column puri®cation where non-quantitative elution introduces uncontrolled variability between samples since partial puri®cation eliminates any internal control marker.

Protein lysate preparation and immunoprecipitation
Cells were grown to 80% con¯uency in 100 mm tissue culture dishes in LHC-8 medium and then converted to basal conditions by incubation for 16 h in LHC Basal medium supplemented with insulin (5 mg/ml), transferrin (5 mg/ ml) and selenium (5 ng/ml) (ITS) (Sigma, St. Louis, MO). Cells were washed three times with cold HEPES-buffered saline (20mM HEPES in calcium-free phosphate-buffered saline with phenol red, (pH 7.5)) (Bio¯uids), lysed in RIPA buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% deoxycholic acid (sodium salt), 0.1% sodium dodecyl sulphate (SDS), 100 mg/ml phenyl-methylsuphonyl¯uoride, 1 mg/ml aprotinin, 1 mM DTT, 1 mM sodium orthovanadate) for 10 min and scraped. The extracts were centrifuged at 40 000 £ g for 30 min at 48C. Protein concentrations were measured using the BCA method (Pierce), according to the manufacturer's instructions. For Western analysis, lysates (100mg per sample) were resolved on 8% SDS gels. For immunoprecipitation, 1 mg of lysate was incubated with 1 mg of primary antibody overnight at 48C. Protein A/G beads (Oncogene Science, Cambridge, MA) were added to the lysate and incubated for 1 h at 48C. Immunoprecipitates were washed four times for 5 min each with cold PBS-0.05% Tween-20. Beads were resuspended in 2£ sample buffer, heated at 958C for 5 min and centrifuged. Supernatants were electrophoresed on 8% polyacrylamide gels. The gels were electrophoretically transferred overnight to Immobilon-P membranes. Membranes were blocked with 10% non fat dry milk in TBST (150 mM NaCl, 50 mM Tris (pH 7.5), 0.1% Tween-20) overnight followed by a 2 h incubation with primary antibody, three washes with TBST, and a 1 h incubation with horseradish peroxidase conjugated secondary antibody. Blots were developed as above. Anti-actin (Boehringer Mannheim, Indianapolis, IN), a-phosphotyrosine (UBI), a-BTC, a-AR, a-HB-EGF (R&D Systems, Minneapolis, MN), were purchased from commercial sources. Polyclonal a-HRG was a kind gift of Dr. Ruth Lupu.

Densitometry
Densitometric evaluation of Northern and Western analyses was performed on a Molecular Dynamics laser densitometer utilizing Image Quant analysis software. Multiple exposures were compared to establish linearity.

Expression of ErbB-1 ligands and HRG
We previously demonstrated the presence of constitutive ErbB-1/ErbB-2 heterodimers in tumorigenic E6T cells [13]. Since receptor dimerization is ligand-dependent [4], and determines which downstream signaling pathways will be activated [7], we examined the presence and relative abundance of ligands for ErbB-1 in both the tumorigenic E6T and the non-tumorigenic E6TA cells. Earlier studies had shown that EGF is not produced by BEAS-2B or its ErbB-2-overexpressing derivatives [15], while TGFa secretion by E6T cells (243 pg/ml per 10 6 cells) was reduced by 93% (17 pg/ml per 10 6 cells) by expression of an antisense construct in E6TA cells [13]. We initially examined the mRNA expression of the ErbB-1 ligands AR and BTC by RT-PCR. The data of Fig. 1 reveal that mRNA for both AR and BTC was expressed in both E6T and E6TA cell lines. In order to relate AR levels to the expression of TGFa in E6T cells, RNA from E6T cells was evaluated for TGFa, AR and actin expression by Northern analysis and densitometric evaluation (data not shown) revealing that TGFa and AR were expressed at equivalent levels (1.0:0.93).
In order to evaluate steady state expression levels of AR, BTC and heparin-binding EGF protein (HB-EGF), lysates (1 mg) of cells grown under basal conditions were immunoprecipitated with commercially available non-cross reacting antibodies to AR, BTC or HB-EGF. Supernatants from the immunoprecipitations were immunoblotted for actin as an internal control (Fig. 2). Densitometric evaluation of actin levels in each lysate was used to correct for unequal loading. Ratios of ligand density to actin density were determined for E6T and E6TA and these values were compared. After these corrections, steady state levels of AR were indistinguishable in E6T and E6TA cells (0.94:1.0), while levels of BTC and HB-EGF were 1.6-(BTC) or 2-(HB-EGF) fold higher in E6T than in E6TA cells.
To extend the evaluation of ligands for ErbB family receptors, expression of HRG was studied by RT-PCR ampli®cation. Only four of the possible forms of HRG generated by alternative splicing are detected in human breast cancer cells [17,18] and primers were designed to amplify these four species. HRG mRNA was not detected by a single ampli®cation in the E6T or E6TA cells. Therefore, one set of nested primers was designed to amplify isoforms a2a (277bp), b1a (292bp) and b2a (267bp) from RNA of E6T, E6TA and the parental BEAS-2B cell line. Only one secreted isoform (Fig. 3a) was expressed in both E6T and E6TA cell lines. The ampli®ed species observed, most closely approximates the 277bp a2a product found in MDA-MB 231 cells (Fig. 3a). In addition, a second set of nested primers was designed to evaluate expression of the non-secreted species, b3. Primers speci®c for the b3 isoform ampli®ed a product in BEAS-2B, E6T and E6TA cells (Fig. 3b). Cycle sequencing from both strands of products from the three lung cell lines con®rmed the ampli®cation of the a2a and b3 species in all cases (data not shown).

Biological activity of HRG in E6T and E6TA cells
Since detection of HRG mRNA required nested PCR ampli®cation, the biological relevance of the expression was evaluated by testing for HRG activity in conditioned medium from BEAS-2B, E6T and E6TA cells. Previous studies have shown that HRG induces tyrosine phosphorylation of a 180-kDa protein in MDA-453 breast cancer cells which endogenously express ErbB-2, ErbB-3, and ErbB-4 in the absence of ErbB-1 and do not express HRG [16]. This assay has been used for HRG puri®cation [16]. Therefore, MDA-453 cells were incubated with heparin± Sepharose fractions of conditioned medium of BEAS-2B, E6T and E6TA cell lines. Lysates of Fig. 2. Protein expression levels of AR, BTC and HB-EGF. Lysates (1 mg) of E6T and E6TA cells grown under basal conditions were immunoprecipitated with either antibodies to AR, BTC, or HB-EGF. Immunoprecipitates were analyzed by Western blotting with the indicated antibodies as described. In order to make quantitative comparisons between lysates, aliquots (50 ml) of supernatants of the ®rst precipitation of primary antibody and protein A/G beads were analyzed by Western blotting on 8% gels with antibody to actin as described. Ratios of expression in E6T and E6TA cells are corrected for loading based on densitometric evaluation of actin in each lysate. treated MDA-453 cells were prepared and analyzed by Western blotting with anti-phosphotyrosine (Fig.  4A). Active fractions of medium conditioned by MDA-231 breast cancer cells [16] served as positive controls, while serum-free medium served as a negative control. Fractions eluting at 0.9 M NaCl derived from conditioned media of E6T, E6TA, and BEAS-2B increased the phosphotyrosine content of a 180-kDa band from lysates of MDA-453 cells. While comparisons can only be qualitative, densitometric evaluation of active fractions indicated that the total activity secreted by BEAS-2B cells was less than half that observed for E6T and E6TA. The latter two cell lines were experimentally indistinguishable. Aliquots of active fractions were pooled, concentrated and analyzed by Western analysis using a polyclonal antibody to a-HRG. As shown in Fig. 4B, the expected 45-kDa band was detected in MDA-231-conditioned medium, while the biologically active species secreted by the lung epithelial cells migrated at 55 kDa. This migration pattern is within the range observed for HRG in other cell types [16]. Comparison of HRG expression in E6T and E6TA was qualitative due to the puri®cation steps required before evaluation of HRG protein. However, relative densities of ®nal bands indicate that E6TA produces at least as much HRG as tumorigenic E6T and, as expected for ErbB-2-transfected cells [19], more than parental BEAS-2B cells (Fig. 4B). These data, together with the activity data, suggest that overexpression of ErbB-2 [15] may stimulate HRG secretion in these cells.

Discussion
Peptide growth factors of the EGF family are important modulators of the proliferation of normal and transformed human lung epithelial cells. Lung cancer cells synthesize growth factors that can potentially regulate their proliferation though autocrine pathways. Indeed, EGF family ligands control the dimerization patterns of the four ErbB receptors and, thereby, their downstream signaling patterns [7]. We have previously demonstrated that high levels of TGFa were needed to induce a tumorigenic phenotype in ErbB-1 expressing and ErbB-2 overexpressing human lung epithelial cells [13]. To examine the role of other autocrine pathways in this system, we assessed the production of other EGF-related ligands in these cell lines.
We report here that both the tumorigenic E6T and the non-tumorigenic E6TA cells express AR, BTC, HB-EGF, and HRG a2a and b3. Previous studies determined the levels of TGFa and the absence of EGF production in these cells [13,15]. Northern blotting revealed that the endogenous level of AR expression is as high as that of TGFa in E6T cells. It has been shown that AR binds to ErbB-1 with a lower af®nity than TGFa and does not interact with ErbB 2/ 4 heterodimers as do other EGF family ligands [20,21]. Therefore, the biological effects of AR may not be comparable to those of TGFa, even when expression levels are similar.
Protein levels of AR, as determined by immunoblot analysis, were equivalent in E6T and E6TA cells. Similarly, Tsao et al. [22] demonstrated that parental and immortalized human bronchial epithelial cells express comparable, high levels of AR. Levels of BTC and HB-EGF were 2-fold higher in E6T as compared to E6TA cells. This difference may be insigni®cant in view of the 13-fold differential in TGFa production [13]. However, it is possible that higher production of these ligands may also contribute to tumorigenicity of E6T cells through increased signaling from activated ErbB family dimers.
BTC can initiate mitogenic signaling through ErbB-4 when it is expressed with ErbB-2 [5] and can also bind to ErbB-2 and ErbB-3 when these receptors are co-expressed [23,24]. While these ligand± receptor interactions would be expected to contribute to mitogenic signaling in this cell system, immunoprecipitation studies suggest that most of the ErbB-3 and ErbB-4 species present exist as heterodimers with ErbB-1 rather than ErbB-2 (Fernandes, submitted). The contribution of BTC and HB-EGF to tumorigenic conversion could be tested using an antisense approach as was used for TGFa [13].
We also examined the production of the ErbB-3/ ErbB4 ligand HRG in E6T and E6TA cells. Only four (a2, b1, b2 and b3) of the possible isoforms of HRG are expressed in human breast cancer cells [17,18]. The a2a and b3 species were detected in both E6T and E6TA, but detection required nested PCR ampli-®cation suggesting a low level of expression of these ligands. Our HRG bioassay data while qualitative, do not suggest signi®cant differences between E6T and E6TA cells in HRG levels. The apparent increase in immunologically detectable HRG in E6TA cells is puzzling. It is possible that the activity of immunologically detectable HRG derived from E6TA cells might have been masked in conditioned media by a copurifying inhibitory activity. However, since variability in elution during heparin column chromatography is expected, the experimental approach used here demonstrates the presence of HRG activity but must remain qualitative. With this caveat and the data showing similar activity levels in E6T and E6TA, we postulate that a differential expression of HRG did not contribute to the tumorigenic conversion of E6T cells.
The role that possible interactions among these growth factors play in inducing a tumorigenic phenotype has not yet been de®ned. It is known that different EGF ligands can synergize to enhance biological effects. For example, AR functions as an autocrine growth factor for neu-transformed human mammary epithelial cells [25]. AR has been postulated to be a mediator of EGF-induced proliferation, and AR antisense is able to inhibit the EGF induced proliferation of MCF-10A and ErbB2 transformed cells [25]. Thus, the AR expressed in our system may have potentiated the effects of TGFa in inducing a tumorigenic phenotype. However, AR is present at roughly comparable levels in tumorigenic E6T and non-tumorigenic E6TA with only HB-EGF and BTC showing a 1.6-or 2-fold differential, suggesting that any speci®c contribution to malignant proliferation would be dependent on TGFa.
In summary, this study reveals for the ®rst time that both non-tumorigenic and tumorigenic human lung epithelial cells express a variety of EGF family ligands, AR, BTC, HB-EGF, and HRG a2a and b3. The activation of multiple autocrine signaling pathways thus precedes neoplastic transformation. We previously showed that TGFa production is necessary to produce a tumorigenic phenotype. The fact that human lung epithelial cells produce a broad spectrum of EGF family ligands demonstrates the importance of future studies to delineate the relative and comparative importance of these ligands in tumorigenic conversion.
part by NIH grant F33CA63763 from the National Institutes of Health awarded to AWH.