JOURNAL OF CELLULAR PHYSIOLOGY 148235-244 (1991)

Differential Regulation of the Expression of Transforming Growth Factor-@ 1 and 2 by Retinoic Acid, Epidermal Growth Factor, and Dexamethasone in NRK-49F and A549 Cells DAVID DANIELPOUR,* KYUNG YOUNG KIM, THOMAS S. WINOKUR, AND MICHAEL B. SPORN Laboratory of Chernoprevention, National Cancer Institute, Bethesda, Maryland 20892

Although most biological activities of transforming growth factor-ps 1 and 2 (TGF-61 and TGF-P2) examined in vitro are similar or identical, recent studies suggest that each of these factors may be independently regulated in vivo. In this study we have used highly sensitive and specific sandwich enzyme-linked immunosorbent assays for TGF-P1 and TGF-P2 to examine the effects of a variety of treatments on expression of these two TGF-P isoforms. We show that epidermal growth factor (EGF) induces secretion of TGF-P1 and not TCF-P2, whereas retinoic acid (RA) induces secretion of TGF-p2 and not TGF-p1 in NRK-49F normal rat kidney fibroblasts and A549 human lung carcinoma cells. Moreover, treatment with EGF diminishesthe levels of TGF-P2, while RA decreasesthe levels of TGF-p1 in both cell lines. Dexamethasone (Dex),on the other hand, inhibits the secretion of both TGF-P1 and TGF-P2 in A549 cells, while selectively inhibiting TGF-Pl secretion in NRK-49F cells. The interactiveeffects of EGF, RA, and Dex on the production of TGF-PI and TGF-P2, which were studied on NRK-49F cells, demonstrate that EGF blocks the induction of TGF-P2 mRNA and peptide by RA, while Dex inhibits the induction of TGF-PI m R N A and peptide by EGF. These results demonstrate that RA, EGF and Dex are each unique, differential, and interactive regulators of the expression of TGF-6s 1 and 2. Numerous physiological processes, such as cell growth and differentiation, extracellular matrix formation, and immune function, are either directly or indirectly regulated by a structurally homologous family of 25 kDa homodimers, namely, transforming growth factor-ps (TGF-Ps) (reviewed by Roberts and Sporn, 1990). Of the five members of the TGF-P family identified to date only TGF-P1 and TGF-P2 have been characterized extensively. Although these two TGF-P isoforms share functional properties on many cells in vitro, such a s receptor binding and biological activity (Cheifez et al., 19871, important differences between them have been shown with respect to regulation of cell growth (Ottmann and Pelus, 1988; Jennings et al., 19881, mesoderm induction (Rosa et al., 1988; Roberts et al., 19901, binding to cell surface proteins (Segarini et al., 1987; Mackay and Danielpour, 19911, inactivation by a,-macroglobulin (Danielpour and Sporn, 19901, tissue-specific expression (Flanders e t al., 1989; Thompson et al., 1989; Danielpour et al., 19901, and most dramatically the regulation of their expression by growth modulators such a s retinoic acid and calcium ions in keratinocytes (Glick et al., 1989,19901, estradiol in human breast cancer cells (Arrick et al., 19901, tamoxifen in fibroblasts (Colletta et al., 19901, gestodene in breast cancer cells (Colletta et al., 19911, and autoinduction by TGF-Ps (Bascom et al., 1989). Such 0 1991 WILEY-LISS, INC.

differential regulation of expression may occur a t unique upstream regulatory elements (Kim e t al., 1989a; Malipiero et al., 1990; Noma e t al., 19911, a t the level of message stabilization (Kim et al., 1989b; Glick et al., 1989) and message translation (Colletta et al., 1990). Delineating the spectrum and mechanism of action of the regulators of TGF-P1 and TGF-P2 expression is important in understanding the differential roles of these modulators in mediating diverse cellular processes. To facilitate this process, we have recently developed sensitive and specific sandwich enzymelinked immunosorbent assays (SELISAs) for TGF-P1 and TGF-P2 that can rapidly quantitate these peptides in complex biological fluids (Danielpour e t al., 1989b). With these assays we have shown that TGF-P1 is the more abundant of the two peptides in nearly all tissues of adult Sprague Dawely rats (Danielpour et al., 1990) and that the pattern of TGF-P1 and TGF-P2 expression in these tissues correlates closely with their relative

Received February 19, 1991; accepted April 19, 1991. *To whom reprint requestsicorrespondence should be addressed. A preliminary report of these data was presented at the 18th annual UCLA Symposium on Molecular Biology held a t Taos, New Mexico, March 1990.

236

DANIELPOUR ET AL.

respective message levels (Thompson et al., 1989; Miller et al., 1989). In this report we have used both SELISA and Northern blot hybridization to demonstrate that under serum-free and growth factor-free conditions epidermal growth factor (EGF), retinoic acid (RA), and dexamethasone (Dex) are unique differential regulators of TGF-P1 and TGF-P2 expression in NRK-49F normal rat kidney fibroblasts and A549 human lung carcinoma cells. In both these cell lines, as occurs in mouse keratinocytes (Glick et al., 1989) RA up-regulates TGF-P2. RA also down-regulates TGF-P1 expression in both NRK-49F and A549 cells. Conversely, in both cell lines EGF up-regulates TGF-Pl and down-regulates TGF-P2 expression. Furthermore, we show that in NRK-49F cells the up-regulation of TGF-p2 expression by RA is blocked by EGF; likewise, in these cells the up-regulation of TGF-Pl expression by EGF is blocked by Dex. Our data suggest that such blocking of TGF-P expression is not mediated by down-regulation of the receptors for the inducers of these TGF-Ps. MATERIALS AND METHODS Materials TGF-Pl and TGF-P2 were purchased from R & D Systems, Inc. (Minneapolis, MN). Dexamethasone, retinoic acid, progesterone, estrogen, testosterone, and crystalline and fraction V bovine serum albumin (BSA) and dimethylsulfoxide (DMSO) were purchased from Sigma (St. Louis, MO). Human transferrin, epidermal growth factor, and sodium selenite were purchased from Collaborative Research, Inc. (Bedford, MD). Stock solutions (10 mM to 10 pM) of retinoic acid (RA) and all steroid hormones were prepared in DMSO and stored at -20°C. 1251-TGF-ps were prepared as described for TGF-P by Frolik et al. (1984) except 0.6 mCi instead of 0.75 mCi of NalZ5Iwas used and the concentration of chloramine T used in all three additions was reduced to 40% of the original. '251-EGF was obtained from Amersham (Arlington Heights, IL).

Culture conditions NRK-49F r a t kidney cells and A549 human lung carcinoma cells were maintained in the high glucose formulation of Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% calf serum and 10% fetal bovine serum, respectively. Cells were kept at 37°C a t 5% COz and passaged twice weekly at 80-90% confluence. Studies on modulation of expression of TGF-f3 peptide levels were conducted as follows. Subconfluent flasks were trypsinized and seeded in 35 mm-wells Costar dishes at 500,000 cells in 2 ml of DMEM supplemented with their respective sera. After an overnight incubation at 37°C for cell attachment, wells were washed twice with 4 ml of DMEM supplemented with 100 pgiml BSA fraction V. After a 2 - 4 h incubation a t 37°C in the second wash buffer the medium was replaced with SFM (DMEMiF12 supplemented with 10 pg/ml transferrin, 10 ngiml sodium selenite, 50 pM ethanolamine, 100 pgiml bovine serum albumin [fraction Vl) supplemented with either 0.1% DMSO (control), RA, 5 ngiml EGF, or various steroid hormones. All treatments were adjusted to the same amount of (DMSO),the RA vehicle. After every 24 h the

medium was replaced with SFM supplemented with their respective hormones and growth factors. Following collection of the conditioned media, medium was clarified by microfuging for 2 min and then stored for a week or less at -20°C long term a t -70°C after the addition of PMSF (120 kgiml) or 2 pgiml each of leupeptin, aprotinin, and pepstatin A.

Quantitation of TGF-p1 and TGF-P2 TGF-P1 and TGF-P2 in conditioned media were assayed by sensitive and specific SELISAs using both turkey and rabbit neutralizing polyclonal antibodies against native TGF-Ps, as described elsewhere (Danielpour e t al., 1989b). Samples used in the SELISA were prepared as follows: To 1.0 ml aliquots of conditioned media in 1.5 ml microfuge tubes, 55 pl of 100% (wiv) trichloroacetic acid were added, and tubes were vortexed and set on ice for 30 min. Samples were centrifuged at 13,OOOgfor 10 min, supernatant aspirated, and pellets were resuspended in 1.0 ml of cold ether-ethanol ( l : l , v/v). Following a 10 min, 13,0008 centrifugation, supernatants were aspirated and pellet were then lyophilized for 5 min. The resulting pellets were solubilized in 125 ~1 or 250 p1 of solubilization buffer (4 mM HC1, 150 mM NaC1, 0.5 mgiml BSA) by rocking overnight a t 4°C. Following a 2 min centrifugation the supernatants were added to the SELISAs a t 7 twofold serial dilutions. Fluorometeric quantitation of DNA Cells from 6 well dishes were lysed with 0.5 ml of 1 % SDS, and lysate sonicated for 20 sec. The DNA from 200 pl of this SDS extract was partially purified by extraction with 200 p1 phenol-chloroform-amyl alcohol (50:50:1) extraction. Aliquots, 100 pl, of the aqueous phase were transferred to fresh Eppendorf tubes, and the DNA was precipitated on dry ice with the addition of 10 p13M sodium acetate and 400 pl absolute ethanol. Then ethanol pellet was washed once with 80% cold ethanol, lyophilized for a minute, dissolved in 100 kl distilled water, and then DNA was assayed by fluorescence using the DNA specific dye 33258 Hoechst (Cesarone et al., 1979). Northern blot analysis Cytoplasmic RNA was isolated from cells using protocol described by Gough (1988) with the following modifications: 1) A final concentration of 5 mM vanadyl ribonucleoside complex (BRL, Bethesda, MD) was added to the lysis buffer; 2) the RNA was reextracted with 400 p1 of phenol-chloroform-isoamyl alcohol (50:50:1); and 3) the RNA was treated with 25 mM EDTA on ice for 15 min. RNA was electrophoresed through a 1% agarose-5.4% formaldehyde gel containing 30 pgiml ethidium bromide and transferred onto Nytran by capillary blotting. Ribosomal RNAs were visualized by UV exposure (4-5 min) after destaining gel and also after blotting to ensure equal loading and transfer of RNA. Message was then cross-linked by UV (Stratagene), and hybridized with a rat TGF-P1 or mouse TGF-P2 [32PlDNAprobe under high-stringency hybridization conditions (Church and Gilbert, 1984). The TGF-P1 probe was a PCR product from rat embryo and included the coding region from the second methio-

237

REGULATION OF THE EXPRESSION OF TGF-Bs 1 AND 2 8.0 ,RA -Dex

--

TABLE 1. Regulation of NRK-49F cell growth by RA, EGF, Dex, TGF-D1, and TGF-pZ1

+

EGF + EGF

+

RA

6.0

$ . 2

4.0

-EGF

2.0

2RA -RA Dex -Control - Dex

0 -Dex

0

+ EGF

+

0

48

24

72

Hours

Fig. 1. Regulation of NRK-49F cell growth by RA, EGF, and Dex in NRK-49F cells. Cells were seeded a t 5 x lo5 cellsi35 mm dishes in 2 ml of DMEM containing 10% calf serum. After an overnight incubation at 37°C wells were washed twice with 4 ml of DMEM and replaced with 1.5 mls of SFM (see Materials and Methods) containing either DMSO (control), 0.1 pM RA, 10 nM Dex, 5 ngiml EGF, or various combinations of the above. The concentration of DMSO in all treatments was adjusted to 0.3%.The medium was replaced every 24 h with fresh SFM containing these same additions. Cell growth was then monitored daily by DNA quantitation as described in Materials and Methods. Values represent the average of three determinations with SD < *lo% of the mean.

nine to the poly A addition site (Qian et al., 1990). The TGF-P2 probe was a PCR product from mouse embryo and included the entire coding region. Both probes were prepared by random priming (Feinberg and Vogelstein, 1984) with a kit from US Biochemicals.

RESULTS Growth of NRK-49F cells under serum-free conditions Although effects of EGF, TGF-P, Dex, and RA on proliferation of NRK-49F normal rat kidney fibroblasts have been reported, these studies have been conducted either in the presence of serum (Roberts et al., 1984) or serum-free medium enriched with a mixture of growth modulators (Nugent and Newman, 1989). Such conditions, however, are usually not optimal because of the presence of certain factors that may either suppress or mask the effect of test factors on gene expression, as will be demonstrated later in this report. To circumvent this difficulty in our present study we have formulated a serum-free and growth factor-free medium for NRK49F normal rat kidney cells. These cells were initially plated in DMEM supplemented with 10% calf serum; after incubation overnight a t 37°C and extensive washing with DMEM, the medium was changed to DMEMi F12 (l:l,v/v) supplemented with 5 pgiml human transferrin, 10 ngiml sodium selenite, 100 pgiml bovine serum albumin, and 50 pM ethanolamine (SFM). Under these conditions NRK-49F cells were growth arrested and could be maintained in culture for 5 days without decreased viability. The addition of EGF (5 ngiml) stimulated one round of cell division during the first 24 h after addition, after which no further cell growth occurred (Fig. 1).Dex and RA either alone or together had little to no effect on growth; however, RA potentiated and Dex inhibited the growth stimulation by EGF. Dex also partially inhibited the EGF + RA

Treatments Control RA (0.1 FM) EGF (5 ng/ml) Dex (10 nM) RA EGF RA Dex Dex + E G F Dex RA t EGF TGF-pl (100 pM) TGF-B2 (100 DM) T G F - ~ ~EGF TGF-p2 EGF

+

+ +

+ +

lZ5I-UdRincorporation (cpm) 355 f 23 277 28 1,816 52 211 2 4,120 f 322 237 33 421 k 30 1,190 53 891 i 43 832 30 26,600 f 87 26,590 5~ 120

*

+ +

* *

+

'NRK-49Fcellswereseededin24welldishesat5X104ce11s/we11in0.5m10fDMEM supplemented with lO%calfserum. After 24 h a t 37°C well were washed twice with 1 ml of DMEM and then replaced with 0.5 ml SFM. The various factors shown above were then added and after 24 h a t 37°C wells received 0.25 pCi of '251-UdR.Following a n additional 24 h incubation, cells were fixed and theincorporation of 12511-UdRinto DNA was measured a s described by Danielpour et al. (1989a). Values represent the mean of three determinations k 1 SD.

level of growth stimulation, but did not reduce growth to that of EGF alone. This anchorage-dependent responsiveness to EGF, RA, and Dex of NRK-49F cells in monolayer is similar to their response to these factors under anchorage-independent conditions in soft agar (Roberts et al., 1984). Growth stimulation by these factors was also followed by the incorporation of lZ5IUdR (5' -[1251]-Iodo-2'-deoxyuridine) into DNA after 48 h of treatment in the above serum-free medium (Table 1).Effects of these modulators on DNA synthesis a s measured by the incorporation of 1251-UdR were similar to those of total DNA synthesis.

Effect of RA, EGF, and steroid hormones on TGF-p1 and TGF-p2 levels in NRK-49F and A549 cells Recent evidence (Glick e t al., 1989, 1990; Arrick e t al., 1990; Colletta et al., 1990, 1991; Bascom et al., 1989) indicates that the expression of TGF-Ps 1and 2 is differentially regulated in many cultured mammalian cells. In this study we investigated the effects of RA and EGF on the expression of TGF-p1 and TGF-02 on NRK-49F cells and A549 human lung carcinoma cells under the above serum-free and growth factor-free conditions. EGF has been previously reported to induce TGF-p1 mRNA in NRK-49F cells (Van ObberghenSchilling et al., 1988). NRK-49F and A549 cells were treated with either 0.1 pM RA or 5 ngiml EGF for 2 days, with daily medium changes, and after 48 h of treatment the levels of TGF-Pl and TGF-p2 in their conditioned media were assayed by highly sensitive sandwich enzyme-linked immunosorbent assays (SELISAs) (Danielpour et al., 1989b). The rates of TGF-p1 and TGF-p2 production were normalized with respect to cell number or DNA content of the producer cells. For both cell lines, RA induced TGF-P2 expression by 3-4-fold, while it inhibited expression of TGFP l by 20 to 40%; conversely, EGF induced TGF-P1 expression by 2-4-fold, while it inhibited the levels of TGF-p2 by 40 to 60% (Table 2). Thus, RA and EGF are opposite-acting differential regulators of the expression

238

DANIELPOUR ET AL

TABLE 2. Effect of RA and EGF on production of TGF-pl and TGF-B2 by NRK-49F and A549 cells' Treatment

pg TGF-B/pg DNA/24 h TGF-01 TGF-BZ

A. NRK-49F cells Control RA EGF

92 f 6 52 i 2 384 19

*

26 f 2 85 f 4 14 k 0.3

TGF-pl/TGF-B2 3.5 0.61 27.4

up. TGF-B/106 cellsI24 h _

B. A549 cells Control RA EGF

Y

*

379 19 299 f 27 841 i 67

256 f 5 1,052 i 26 99 f 1

1.48 0.28 8.5

'Cells were seeded at 5 X lo" cells/35 mm dishes as described in Materials and M RA or 5 ng/ml EGF for 24 h, and then Methods, and treated with either medium was replaced with fresh SFM containing the same additions. Following an additional 24 h of culture,medium was assayed forTGF-ps 1 and 2 by SELISA. Data are expressed as rate of TGF-pproductionnormalizedto cell number or total DNA of producercells. Values representthe average of triplicatedeterminations 1 SD from the mean.

+

of TGF-p1 and TGF-P2 in both A549 and NRK-49F cells. Treatment of NRK-49F and A549 cells with various concentrations of RA for 48 h demonstrated that the induction of TGF-P2 by RA, measured by SELISA, was dose-dependent, with significant effects at 10 nM in both cell lines (Fig. 2a,b). To assess the specificity of this RA effect, the effects of several steroid hormones, each a t a single dose of 10 nM for 48 h, on expression of TGF-p1 and TGF-p2 levels in NRK-49F and A549 cells were studied under identical conditions as before. Estradiol, testosterone, and progesterone had slight inhibitory effects ( ~ 2 0 % inhibition) on TGF-p1 and TGFp2 levels, while Dex significantly inhibited TGF-P1 levels in both A549 and NRK-49F cells and inhibited TGF-P2 levels in only A549 cells (Fig. 3a,b).

Interaction of RA, EGF, and Dex in modulation of TGF-p1 and TGF-p2 levels Our results here and those of others (see Roberts and Sporn, 1990, for review) demonstrate that the expression of TGF-p1 and TGF-p2 in a variety of cell types is regulated by multiple modulators. Thus, in order to determine the overall regulation of TGF-P expression, it is important to understand the interactive effects of these modulators. NRK-49F cells were used as a model system to study the kinetics and interactive effects of RA, EGF, and Dex on TGF-p1 and TGF-P2 secreted protein and mRNA levels. These cells were treated with RA, EGF, and Dex either alone or together in all combinations for 1-3 days under the serum-free conditions described before. After harvesting the medium, total cellular DNA was extracted and quantified by fluorescent dye-binding. Following the measurement of TGF-ps 1 and 2 in conditioned media by SELISA, the rates of TGF-P production were normalized with respect to DNA of the producer cells (Fig. 4a,b). This study revealed several additional significant differences between the regulation of TGF-p1 and TGF-P2 expression by these three modulators. First, the induction of TGF-p1 and inhibition of TGF-P2 by EGF was maximal after 24 h of treatment, whereas the induction of TGF-P2 by RA was not detectable until the second 24

h of stimulation and the magnitude of such induction increased thereafter. Second, although Dex when used alone had a slight inhibitory effect on the level of TGF-p1, this steroid blocked the effects of EGF on TGF-p1 levels, but had no effect on TGF-P2 levels either under basal conditions or in the presence of EGF. However, Dex inhibited about 50% of the induction of TGF-p2 by RA over the entire period of such induction. Third, the induction of TGF-p2 by RA was completely blocked by EGF, bringing TGF-P2 level down to that of the non-treated control, whereas the induction of TGFp l by EGF was not significantly altered by RA. Fourth, although Dex abolished the ability of EGF to induce TGF-p1 either in the presence or absence of RA, this steroid did not block the ability of EGF to inhibit either basal or RA-induced level of TGF-p2, suggesting the intracellular signalb) elicited by EGF to inhibit TGFp2 is (are) different than those for TGF-p1 induction by EGF. We next sought to understand the mechanism by which Dex may block induction of TGF-p1 by EGF. Because Dex has been shown to down-regulate EGF receptors in NRK-49F cells (Roberts et al., 19841, it seemed probable that under these conditions Dex blocks the induction of TGF-p1 levels by EGF by down-regulating EGF receptor binding. To test this, NRK-49F cells were treated with Dex, RA, and Dex + RA for 48 h under serum-free conditions as before and the total specific binding of 1251-EGFto these cells was examined (Table 3). As shown, RA, 0.1 kM, increased the total specific binding by about twofold. Although Dex, 10 nM, partially reduced the RA-induced EGF receptor binding, Dex at this concentration without other additions decreased EGF receptor binding by under 20%. This small decrease in EGF binding suggests that Dex does not block the EGF-induction of TGF-p1 via down-regulation of EGF receptor binding. Thus, Dex may block EGF-stimulated TGF-p1 levels by acting a t a step distant from the binding of EGF to its receptor. The most dramatic result obtained was the ability of EGF to inhibit the induction of TGF-p2 levels by RA. We examined the ability of EGF to down-regulate RA-induced TGF-P2 as a function of EGF concentration after 48 h of treatment (Fig. 5). Our results demonstrate that EGF at relatively low doses (0.2 ngiml) blocked the ability of near optimal levels of RA (1 kM) to induce TGF-P2. This suggests that EGF overrides the induction of TGF-p2 by RA. The individual and interactive effects of RA, EGF, and Dex after 24 h of treatment under serum-free conditions on TGF-P1 and TGF-P2 message levels in NRK-49F cells were analyzed by Northern blot hybridization using total cytoplasmic RNA (Fig. 6a,b). The results of this study can be simply stated as follows: changes in message levels roughly followed those of the respective secreted TGF-P protein levels, with the exception that EGF and RA did not inhibit basal TGF-P2 and TGF-p1 mRNA levels, respectively. This suggest that EGF and RA inhibit TGF-p1 and TGF-p2 basal peptide levels, respectively, by translational control. For the other treatments, changes in the message levels preceded changes in protein levels. For example, RA induced TGF-p2 mRNA after 24 h treatment,

REGULATION OF THE EXPRESSION OF TGF-ps 1 AND 2

Fig. 2. Dose-dependence of RA on production of TGF-pl (solid bars) and TGF-PZ (shaded bars) by NRK-49F (A) and A549 cells (€3). Cells (plated at 5 x l o 5 cells./35 mm dish! were treated with various concentrations of RA in 1.5 ml SFM for 24 h (the DMSO vehicle concentration of all treatments were adjusted to 0.1%); medium was replaced with fresh SFM containing RA a t the same concentrations as before. Following an additional 24 h of culture, conditioned media were collected and assayed for TGF-pl and TGF-p2 by SELISA. Data shown were normalized to cell number. For NRK-49 cells the untreated control rates of TGF-p1 and TGF-p2 production were 85 pgipg DNAi24 h and 22 pgikg DNA124 h , respectively; for the A549 these values were 360 pg/1O6 cells324h and 240 pgi106 cellsi24 h, respectively. \'slues represent the average of three determinations with SD 95% latent, as determined by the

inability to assay these peptides in the SELISAs without transient acidification (data not shown). However, these data do not imply th a t these cells do not produce active TGF-(3s) since NRK-49F cells have previously been reported to rapidly degrade 1251-TGF-pl (Frolik et al., 1984). In order to confirm these findings in our system, NRK-49F cells, treated with various combinations of EGF, RA, and Dex, were given a single dose of 20 pM 1251-TGF-pl or lZ5I-TGF-p2,and after 24 h of culture their conditioned media were precipitated with 5% TCA (data not shown). The maximum recovery of 1251-TGF-pswas under 5%, a s measured by TCA precipitability of the labeled TGF-ps after this treatment. Approximately 90% of the total recoverable radioactivity was in the TCA soluble fraction. Furthermore, electrophoresis of the TCA precipitates indicated that a small fraction of the total precipitable activity migrated with intact TGF-ps. The addition of 100 FM chloroquine, which increases lysosomal pH, blocked essentially all degradation (data not shown). These results suggest that NRK-49F cells can rapidly degrade over 95% of the endogenous active TGF-P produced under these conditions. Thus, the absence of detectable (