Transcriptional upregulation of TGF-α by phenylacetate and phenylbutyrate is associated with differentiation of human melanoma cells

TRANSCRIPTIONAL UPREGULATION OF TGF-α BY PHENYLACETATE AND PHENYLBUTYRATE IS ASSOCIATED WITH DIFFERENTIATION OF HUMAN MELANOMA CELLS Lei Liu,1 W. Robert Hudgins,1 Alexandra C. Miller,2 Li-Chaun Chen,1 Dvorit Samid1

The aromatic fatty acids phenylacetate (PA) and phenylbutyrate (PB) induce tumour cell differentiation in experimental models and both are currently in clinical trials. The purpose of this study was to determine the effect of these antitumour agents on the expression of transforming growth factor-alpha (TGF-α) in neoplastic cells. Treatment of human melanoma 1011 cultures with either PA or PB caused over 40-fold increase in TGF-α biosynthesis and secretion into the media. Whereas elevation in TGF-α mRNA steady-state levels became evident within 6–12 h and reached peak quantities the following day, the amounts of its coded protein increased gradually over a period of 5 days of treatment. Further molecular analysis revealed that regulation of TGF-α expression occurred at the transcriptional level. In contrast to TGF-α, expression of its receptor remained below detectable levels, indicating that an autocrine loop involving this growth factor is unlikely. Interestingly, the increase in TGF-α production paralleled drug-induced cytostasis and differentiation defined by morphological changes and increased melanogenesis. Like PA and PB, other differentiation inducers such as all-trans-retinoic acid, dimethyl sulfoxide, and 5-aza-29-deoxycytidine, all induced TGF-α expression in the melanoma cells. The close association between enhanced TGF-α production and melanoma cell differentiation suggests that this growth factor, often linked to mitogenesis, may play a novel role in tumour differentiation by PA and PB. © 1995 Academic Press Limited.

Phenylacetate (PA), a common metabolite of phenylalanine, regulates cell growth and differentiation in diverse organisms throughout phylogeny.l At millimolar concentrations, it selectively suppresses the growth of poorly differentiated embryonic plant, rodent and human tissues.2–5 PA also induces differentiation of human leukaemic cell lines2,6 and reverses the malignancy of various solid tumour cell lines, including hormone-refractory prostatic carcinoma,7 glioblastoma,8 neuroblastoma,9 and rhabdomyosarcoma.10 Changes in tumour biology induced by PA are accompanied by alterations in the expression of genes impliFrom the 1Clinical Pharmacology Branch, National Cancer Institute, and 2Radiation Biochemistry Department, Armed Forces of Radiation Research Institute, Bethesda, Maryland, USA Correspondence to: Dvorit Samid, Ph.D., Clinical Pharmacology Branch, National Cancer Institute, Bldg 10, Room 12C103, Bethesda, MD 20892, USA Received 17 November 1994; accepted for publication 25 January 1995 © 1995 Academic Press Limited 1043-4666/95/05044918 $12.00/0 KEY WORDS: aromatic fatty acids/human melanoma/TGF-α/cell differentiation CYTOKINE, Vol. 7, No. 5 (July), 1995: pp 449–456

cated in neoplastic transformation and evasion of the immune system.7,11 Phenylbutyrate (PB), an analog of PA, shares some similar effects on tumour cell and molecular biology.11 The concentrations of PA and PB effective in experimental models, although high (millimolar), have been achieved in humans without significant toxicity.12,13 Moreover, clinical studies with PA have documented antitumour activity at well tolerated doses.14 The promise of PA and PB as antitumour agents with low toxicity prompted further studies of their effect on gene expression. Here we focus on TGF-α. TGF-α is an acid- and heat-stable polypeptide of 50 amino acid length, which is structurally and functionally related to EGF.15–17 A variety of solid tumours, such as squamous and renal carcinomas, hepatomas, melanomas and glioblastomas, produce larger amounts of TGF-α compared to the respective normal tissues.18–21 In tumours, increased TGF-α expression is often accompanied by elevated expression of its receptor, EGF-R.16 High EGF-R levels are thought to increase the sensitivity of tumour cells to mitogenetic

449

450 / Liu et al.

CYTOKINE, Vol. 7, No. 5 (July 1995: 449–456)

1

2

3

4

5

A

Phenylacetate

Time (h)

Phenylbutyrate

0 12 24 48 72 96 120 0 12 24 48 72 96 120

TGF-α

β-Actin

30

B

Figure 1. Dose-dependent induction of TGF-α mRNA levels by sodium phenylacetate (PA) and sodium phenylbutyrate (PB). Northern blot analysis of mRNA isolated from 1011 cells treated with PA or PB at various doses for 3 days. Lane 1, untreated control; lane 2, 5 mM PA; lane 3, 10 mM PA; lane 4, 1.5 mM PB; lane 5, 3.0 mM PB. Actin levels indicate the relative amounts of mRNA loaded in each lane.

TGF-α mRNA levels (fold increase)

25 20 15 10 5 0

stimulation by TGF-α, amplifying the autocrine effects of this growth factor.22–25 However, TGF-α is not limited to malignant cells; it is also produced and secreted by embryogenic tissues,26–28 as well as by some mature normal cells.29–31 Despite the prevalence of TGF-α, its regulation of expression and biological role are not clearly understood. Increased TGF-α expression is characteristic of malignant melanomas.32 We have previously shown that treatment of cultured human melanoma cells with PA and PB at pharmacologically attainable concentrations leads to cytostasis with enhanced expression of several differentiation markers.11 We postulated that phenotypic reversion by these drugs may be accompanied by a decline in TGF-α expression. We report here that, surprisingly, both PA and PB significantly increased TGF-α expression concomitant with induction of cytostasis and differentiation of human melanoma 1011 cells.

RESULTS Selective increase in TGF-α mRNA levels by PA and PB Melanoma 1011 cells had very low TGF-α mRNA levels, undetectable under the experimental conditions used. Treatment with either PA or PB resulted in a dosedependent increase in the amounts of TGF-α transcripts (Fig. 1). Subsequent kinetic studies showed that

0

12

24

48 72 Time (h)

96

120

Figure 2. Time-course changes of TGF-α mRNA levels by sodium phenylacetate (PA) and sodium phenylbutyrate (PB) (A) Northern blot analysis of mRNA extracted from 1011 cells treated with PA (7.5 mM) or PB (1.25 mM) for various time durations. Actin levels indicate the relative amounts of mRNA loaded in each lane. (B) Following densitometry analysis, the relative changes in TGF-α mRNA levels were quantitated. PA (s); PB (d).

induction of TCF-α occurred as early as 12 h following PA treatment, reached peak levels by 48 h, and remained elevated for at least 96 h (Fig. 2). In PBtreated cells, induction was detected as early as 6 h (data not shown), reached its peak by 24 h, and declined by 48 h despite continuous treatment (Fig. 2). Since TGF-α is homologous to EGF, and both share the same receptor (EGF-R) to mediate biological activities, we examined whether expression of EGF and EGF-R was affected by PA or PB. Northern blot analysis revealed that, in contrast to TGF-α, EGF and EGF-R mRNA levels remained undetectable following treatment (data not shown). The relation between increased TGF-α gene expression and melanocytic differentiation induced by PA and PB Consistent with our previous observation,11 treatment of 1011 melanoma cells with either PA or PB resulted in a dose-dependent growth arrest, with the

TGF-α upregulation by PA and PB / 451

100

40

80

30

60

20

40

10

20

0

10

5 mM

0

80

120 B

Percent cell growth

100 60 80 60

40

40 20 20 0

1.5 mM

0

3.0

TGF-α mRNA levels (fold increase)

Percent cell growth

A

TGF-α mRNA levels (fold increase)

50

120

0

IC50 values being 7.5 and 1.25 mM, respectively. Moreover, cells treated for 3 days (or longer) underwent melanocytic differentiation characterized by increased cytoplasm/nuclear ratio, development of multiple dendritic projections, and enhanced melanin production. Interestingly, the increase in TGF-α mRNA levels was directly proportional to the degree of growth arrest and melanogenesis induced by PA and PB (Fig. 3). Release of TGF-α protein into the media To evaluate whether TGF-α mRNA induced by drugs was translated into its coded protein and subsequently secreted into media, we further measured TGF-α protein levels using ELISA technique. Data summarized in Figure 4 show increased amounts of TGF-α protein released into the media by melanoma cells following 5 days treatment with either PA or PB. The effect was dose- and time-dependent. Significantly higher levels of TGF-α protein were observed after 72 h of treatment with PA (7.5 mM), reaching peak levels of approximately 70 pg/106 cells/ml at 120 h (Fig. 5). PB (1.25 mM) had a similar induction pattern, although the amount of TGF-α released at 120 h of treatment was significantly higher (280 pg/106/ml) than that induced by PA. PAG (5 mM), the end-metabolite of PA and PB, had no effect on TGF-α release (Fig. 4). Consistent with the RNA data, the number of EGF receptors on the cell surface was low and remained unchanged following treatment with PA or PB (106 6 26 receptors per cell were occupied when the 125I-TGF-

30

50

40 20 30

20 10 10

TGF-α mRNA levels (fold increase)

Melanin production (µg/ml/106)

C

PB 3.0 mM PB 1.5 mM PB 0.75 mM PA 10 mM PA 5.0 mM PA 2.5 mM PAG 5 mM

0

0

2

4 Days

6

8

Undetectable

0 Control

Undetectable 0

Figure 3. Relationship between induction of TGF-α mRNA, rate of cell growth and melanin production in treated 1011 cells. The relative changes in TGF-α mRNA levels (determined by densitometric analysis of the Northern blot presented in Fig. 1) were compared to cell growth rate at 3 days of treatment with PA (Panel A) or PB (panel B). (h· ) Percent cell growth; (r) TGF-α mRNA. Panel C shows the changes in TGF-α gene expression (data from Fig. 2) over time compared to melanin production in cells exposed to 1.25 mM PB. (s) Melanin production; (d) TGF-α mRNA.

100 200 300 TGF-α, pg/106 cells/ml

400

Figure 4. Release of TGF-α protein into the media by differentiating melanoma cells treated with phenylacetate (PA), phenylbutyrate (PB), or phenylacetylglutamine (PAG). Dose-response of TGF-α release by 1011 cells following PA, PB or PAG treatment for 5 days. Data are means of duplicates. TGF-α protein was quantitated by ELISA as described in Materials and Methods.

452 / Liu et al.

CYTOKINE, Vol. 7, No. 5 (July 1995: 449–456)

80

A

1

2

3

70 TGF-α

50

20 10 0

Figure 5. cells.

1

2

β-Actin

B 3 Day

4

4

5

Time-course of TGF-α protein secretion by treated 1011

Supernatants from 1011 cultures treated with PA at 7.5 mM were collected at the indicated time periods. TGF-α protein was quantitated by ELISA techniques as described in Materials and Methods. Data are given as means of duplicates.

TGF-α levels TGF-α mRNA nRNA levels (fold (fold increase) increase)

30

Undetectable

40 Undetectable

TGF-α, pg/106 cells

60

3

2

1

α concentrations were 3.1 3 10210 M). Assuming a dissociation constant of 1 3 1029, maximum total number of receptors would be approximately 700 per cell, i.e. too few to stimulate cell proliferation.33 Transcriptional regulation of TGF-α by PA To determine the molecular level at which PA regulates TGF-α expression, we measured the rate of TGF-α transcription. Nuclear run-on experiments showed a dose-dependent increase in the amount of TGF-α transcripts in melanoma cells following treatment. The effect on gene transcription was selective, as beta-actin expression was unchanged (Fig. 6). Induction of TGF-α by other differentiation inducers To evaluate the effect of other differentiationinducing agents on TGF-α expression in melanoma 1011 cells, we treated cultures for 3 days with all-transretinoic acid (ATRA), dimethyl sulfoxide (DMSO), and 5-aza-29-deoxycytidine (5AzadC). In all cases, TGF-α mRNA levels were significantly increased following 3 days of treatment (Fig. 7).

DISCUSSION TGF-α is thought to be a growth-stimulating factor for various malignant cells, including melanomas. As such, its expression would be expected to decline upon pharmacological manipulations leading to tumour growth arrest and differentiation. We show here

0

Figure 6.

Control

PA 5 mM Treatment

PA 10 mM

Effect of PA on the rate of TGF-α gene transcription.

The run-on in vitro transcription assay was used to analyse nuclear TGF-α RNA transcripts labelled with [32P] UTP. (A) Upper row, TGF-α: Lane 1, untreated control; lane 2, 5 mM PA; lane 3, 10 mM PA. Lower row, β-Actin for samples loading control. (B) Following a densitometry analysis, relative changes of transcripts were quantitated.

that, paradoxically, differentiation of human melanoma cells by PA and PB is associated with a dramatic increase in TGF-α production and secretion. This unexpected link between TGF-α and differentiation led us to further characterize the changes in melanoma cell and molecular biology. In agreement with previous observations, 1011 melanoma cells treated with either PA or PB had reduced growth rate, increased cytoplasm/nuclear ratio, extended dendritic processes, and enhanced melanogenesis, all characteristic of melanocytic maturation. The profound changes in tumour biology, first evident 2–3 days after cell exposure to PA/PB, were preceded by an increase in TGF-α expression that occurred during the first day of treatment. Changes in the steady-state mRNA levels could be detected within 6 h of treatment, while the amounts of TGF-α protein secreted to the medium increased gradually during 5 days of continuous treatment. Run-on experiments revealed that TGF-α was

TGF-α upregulation by PA and PB / 453 A

1

2

3

4

TGF-α

β-Actin

B

1

2

3

4

TGF-α

β-Actin

C PB 3 mM

Treatment

PA 7.5 mM 5AzadC 1 µM DMSO 3 µM ATRA 2 µM Control 0

20

40 60 TGF-α mRNA levels (fold increase)

80

Figure 7. Induction of TGF-α by the differentiating-inducers 5-aza29-deoxycytidine (5AzadC), all-trans retinoic acid (ATRA), and dimethyl sulfoxide (DMSO). (A) Northern blot analysis of the dose-response in 1011 cells treated with 5AzadC for 3 days. Lane 1, untreated control; lane 2,4,5AzadC treatment at 0.3, 1, and 3 µM, respectively. (B) Northern blot analysis of TGF-α mRNA in 1011 cells treated with other differentiation inducers for 3 days. Lane 1, control; lane 2, 2 µM ATRA; lane 3, 3 mM DMSO. For comparison, the effect of 7.5 PA is shown in lane 4. β-Actin level indicates the relative amount of mRNA loaded in each lane. (C) Summary of the relative changes in TGF-α mRNA levels in treated 1011 cells as determined by densitometric analysis of Northern blots.

regulated at the transcriptional level. We cannot exclude, however, the possibility of additional posttranscriptional regulation related to mRNA and/or protein stabilization. Since the matured TGF-α protein is approximately 30% homologous to EGF in amino acid sequence and shares EGF-Rs,34 we further examined the expression of EGF and EGF-R mRNAs in melanoma 1011 cells, and determined the ligand binding capacity of EGF-R using 125I-labelled TGF-α. Neither EGF nor EGF-R mRNA was detectable, and very low TGF-α binding counts were measured in both control and treated cells. Furthermore, treatment of the cultures with exogenous TGF-α or EGF did not alter the rate of cell growth (data not shown). It appears, therefore, that an autocrine loop involving TGF-α in melanoma 1011 cells may not be present. The mechanism(s) by which PA and PB selectively upregulate TGF-α gene transcription are not known. Expression of TGF-α appears to depend in part on the degree of cytosine methylation at its DNA regulatory sites.35 Both PA and PB have been shown to inhibit DNA methylation and upregulate the expression of methylation-dependent genes.36 Our recent findings indicate that these drugs also activate a transcriptional factor, the peroxisome proliferator-activated receptor (PPAR), a member of the nuclear steroid receptors known to control gene expression and cell phenotype.37 Thus PA and PB could induce TGF-α through DNA hypomethylation and/or PPAR activation. This hypothesis is currently under investigation. The increase in TGF-α was closely associated with melanoma cell differentiation. While differentiation of 1011 cells by PA and PB was accompanied by over 40fold increase in TGF-α production and secretion, the drugs were less effective in inducing TGF-α in other melanoma cell lines which failed to undergo terminal differentiation (Liu et al., unpublished data). The effect was not limited to PA and PB, as other differentiating agents including ATRA, DMSO and the hypomethylating cytosine analog, 5AzadC, all increased TGF-α expression. Consistent with our findings, Walz et al. have recently shown that TGF-α was induced in differentiating leukaemic HL-60 cells treated with ATRA or DMSO.18 Mehler et al. reported that TGF-α treatment of hippocampal progenitor cells with other cytokines such as basic FGF and interleukines resulted in progressive neuronal differentiation.38 The differentiation inducer, N,N-Dimethylformamide, was shown to increase TGF-α mRNA (3.8-fold) and protein (4.4fold) in poorly differentiated colon carcinoma HCT116 cell line.39 More recently, it was documented that TGF-α acts as morphogen in a colon cancer SW1222 cell line.40 It appears therefore that, in addition to its role in autocrine growth of neoplastic cells, TGF-α may also be involved in restoration of growth control and

454 / Liu et al.

tumour cell maturation. In conclusion, it now appears that drugs inducing tumour differentiation and cytostasis may not necessarily bring about a decrease in TGF-α production. PA and PB are currently in clinical trials at the National Cancer Institute. We have shown here, for the first time, that both drugs markedly induce TGF-α production and secretion in growth-inhibited melanoma cells in vitro. A selective increase in TGF-α expression associated with differentiation may reflect on the contribution of this growth factor to novel regulatory pathways, yet to be identified.

CYTOKINE, Vol. 7, No. 5 (July 1995: 449–456)

Gaithersburg, MD) was used as internal control to ensure equal loading of the samples. cDNA probes were labelled with 32P dCTP (NEN, Boston, MA) using a random primed DNA labelling kit (Ready-To-Go) from Pharmacia P-L Biochemicals Inc. (Piscataway, NJ). Oligonucleotides were labelled with 32P gamma-ATP (NEN, Boston, MA) using a NDA 5 Endlabeling Kit (Boehringer Mannheim Corp., Indianapolis, IN). Membranes were hybridized with probes at 68°C for 1 h, and washed twice for 15 min each at room temperature with 2X standard saline-citrate/0.1X sodium dodecyl sulfate followed by a final wash at 60°C for 30 min with 0.1X standard saline-citrate/0.1X sodium dodecyl sulfate. Membranes were autoradiographed with Kodak XAR5 films at 270°C with intensifying screens.

MATERIALS AND METHODS Quantitation of melanin production Cell cultures and reagents The human melanoma 1011 cells established from a patient with a melanoma were kindly provided by Dr J. Weber (Surgery Branch, National Cancer Institute, Bethesda, MD) and maintained in RPMI 1640 medium, supplemented with 10% heat-inactivated fetal calf serum, antibiotics and 2 mM L-glutamine (BioFluids, Inc, Rockville, MD), unless otherwise specified. The cultures were incubated in humidified air with 5% CO2 at 37°C and maintained subconfluent with one passage each week. The sodium salts of phenylacetic acid and phenylbutyric acid were provided by Elan Pharmaceutical Research Corp. (Gainesville, GA). 5AzadC, ATRA, butyric acid, DMSO were purchased from Sigma Chemical Co. (St. Louis, MO).

Quantitation of cell proliferation and viability, and collection of conditioned media Melanoma 1011 cells were seeded at the density of 5 3 104 per well in 6-well plates (Costar, Cambridge, MA). Treatment usually started 24 h after seeding. On day 3, media were collected and stored at 220°C in siliconized micro-tubes (PGC Scientific, Gaithersburg, MD) with 5% BSA, aprotinin 1 µg/ml, pepstatin 1 µg/ml, and 120 µg/ml phenylmethylsulfonyl fluoride (PMSF). In parallel, cells numbers were determined by a Cell Counter following detachment with trypsin/EDTA. Cell viability was determined by trypan blue exclusion assay.

RNA extraction and Northern blot analysis Poly A1RNA was extracted from various samples by Invitrogen Fastrack mRNA isolation kit (San Diego, CA). Samples (5.0 µg per lane) were denatured at 55°C for 15 min before being electrophoresed through 1% agarose/formaldehyde gels. Message RNAs were then blotted onto DuralonUV membrane (Stratagene, La Jolla, CA), cross-linked with UV, and hybridized with 32P-labeled specific probes at 68°C according to the Quikhyb protocol (Stratagene, La Jolla, CA). TGF-α cDNA probe was from Oncor Inc. (Gaithersburg, MD). EGF/EGF-R probes were single-stranded synthetic oligonucleotides of 40 bases purchased from Oncogene Sciences (Uniondale, NY). The beta-actin probe (Oncor Inc.

Melanin content was measured by the colorimetric method described by Whittaker.41 Briefly, tumour cells were plated at 1–2 3 106 in 225 mm2 flasks, and treated for a period of 1 to 14 days. To determine the amount of melanin produced, cells were detached at different time points with trypsin/EDTA and lysed (5 3 106 cells per point) by the addition of 0.5 ml of deionized water with two cycles of freezing and thawing. Perichloric acid was added to a final concentration of 0.5 N and the suspension was kept on the ice for 10 min, then centrifuged at 15 000 rpm for 5 min. The pellets were extracted twice more with 0.5 N HClO4 followed by two extractions with a cold mixture of ethyl alcohol:ethyl ether (3:1, v/v) and final extraction with ethyl ether. The pellets were air dried, 1 ml 0.85 N KOH was added, and then dissolved by heating to 100°C for 10 min. After insoluble residue was pelleted and the supernatant was cooled to room temperature and the absorbance at 400 nm was read in a double beam spectrophotometer, (UV-160A) from Shimadzu Corporation (Kyoto, Japan). A standard curve was constructed by using synthetic melanin (Sigma) dissolved in hot KOH at concentrations ranging from 5 to 150 µg/ml. The relative melanin content is expressed as the absorbance at 400 nm per 5 3 106 cells/ml.

Quantitation of TGF-α protein TGF-α protein was measured by a quantitative enzymelinked immunosorbent assay kit (Oncogene Science. Inc., Uniondale, NY). This kit utilizes affinity-purified goat polyclonal antibodies specific for mammalian TGF-α which are highly specific and show no cross-reactivity with human EGF, human platelet-derived factor and many other cytokines. The assay can detect less than 10 pg/ml of human TGF-α in a variety of biological fluids and tissue samples. Since PA and PB inhibit melanoma cell proliferation, the amount of TGF-α protein was normalized by cell number and expressed as pg/106 cells/ml.

Isolation of nuclei and preparation of nuclear RNA transcripts Nuclei were isolated as previously described.42,43 Briefly, cells were collected, washed with buffer A (10 mM TRIS at

TGF-α upregulation by PA and PB / 455

pH 8.4 with 14 mM NaCl and 1.4 mM MgCl2), and lysed with buffer B (10 mM TRIS at pH 8.4 with 14 mM NaCl, 1.4 mM MgCl2, and 1% Nonidet P-40). The nuclei were pelleted by centrifugation at 1000 3 g for 5 min at 4°C. Nuclei, sheared with a 21-gauge needle, were isolated on a 1 M sucrose cushion in buffer A, resuspended in buffer C (20 mM TRIS at pH 8.4 with 20% glycerol, 0.14 M KCl, 0.01 M MgCl2, and 14 mM beta-mercaptoethanol), and stored at 280°C until further use.

Nuclear run-on assay Purified nuclei, 1–10 3 107 per point, were pelleted at 1000 3 g for 3 min, and the rate of gene transcription was determined as previously described.43 Briefly, nuclei were incubated at 30°C for 15 min with labelling solution as follows: 10% buffer C containing 0.033 M of ATP, CTP and GTP, 10 mg/ml phosphocreatine, and 300 uCi [alpha32P] UTP (NEN, Boston, MA; 760 mCi/mol). The reaction was stopped by addition of buffer D (0.05 M TRIS at pH 7.5 with 0.8% SDS and 0.02 M EDTA). RNA transcripts were purified as described above, and recovery of trichloroacetic acid-precipitable radioactivity was determined. The labelled RNA was then hybridized to excess TGF-α and β-actin cDNA probes immobilized onto nitrocellulose membrane.

Ligand binding assay of TGF-α Melanoma cells were seeded in 6-well plates at a density of 5 3 104 per well. Triplicated wells were given 5 mM, 10 mM phenylacetate and no treatment for control cells in RPMI 1640 plus 10% FBS at 37°C with 5% CO2 for seven days. At the end of treatment period, the viable cells from each group were collected, counted and adjusted to 1 3 106 cells per well. 3.1 3 10210 M 125I-TGF-α obtained from Amersham in RPMI with FBS was added to each well. After 6 h incubation at 25°C, the cells were then chilled on ice, washed four times with cold PBS and lysed. The lysate was counted on a BeckmanClinigarmma Counter.

Acknowledgement The authors thank J. Weber for providing 1011 melanoma cell cultures. This study was supported by funds from Elan Pharmaceutical Research Corporation through a Cooperative Research and Development Agreement (CRADA-0139).

REFERENCES 1. Kishore G, Sugumaran M, Vaidyanathan CS (1976) Metabolism of DL-phenylalanine by Aspergillus niger. J Bacteriol 128:182–191. 2. Samid D, Shack S, Sherman LJ (1992) Phenylacetate: a novel nontoxic inducer of tumor cell differentiation. Cancer Res 52:1988–1992. 3. Wightman F, Lighty DL (1982) Identification of phenylacetic acid as a natural auxin in the shoots of higher plants. Physiol Plant 55:17–24. 4. Wem GY, Wisniewski HM, Shek JW, Loo YH, Fulton TR (1980) Neuropathology of phenylacetate poisoning in rats: an exper-

imental model of phenylketonuria. Ann Neurol 7:557–566. 5. Lenke RR, Levy HL (1980) Maternal phenylketonuria and hyperphenylalaninemia, an international survey of outcome of untreated and treated pregnancies. N Engl J Med 303:1202–1208. 6. Samid D, Yeh A, Prasanna P (1992) Induction of erythroid differentiation and fetal hemoglobin production in human leukemic cells treated with phenylacetate. Blood 80:1576–1581. 7. Samid D, Shack S, Myers CE (1993) Selective growth arrest and phenotypic reversion of prostate cancer cells in vitro by nontoxic pharmacological concentrations of phenylacetate. J Clin Invest 91:2288–2295. 8. Samid D, Ram Z, Hudgins, WR, Shack S, Liu L, Walbridge S, Oldfield EH, Myers CE (1993) Selective growth arrest of malignant gliomas by phenylacetate: similarity to fetal brain damage in phenylketonuria. Cancer Res 54:891–895. 9. Cinatl J, Mainke M, Weibflog A, Rabenau H, Kornhuber B, Doerr HW (1993) In vitro differentiation of human neuroblastoma cells induced by sodium phenylacetate. Cancer Lett 70:15–24. 10. Gorski GK, McMorrow ME, Donaldson MH (1993) Letter to the editor: Synergistic inhibition of human rhabdomyosarcoma cells by sodium phenylacetate and tretinoin. In Vitro Cell Dev Biol 29:189–1993. 11. Liu L, Shack S, Stetler-Stevenson WG, Hudgins WR, Samid D (1994) Differentiation of cultured human melanoma cells induced by the aromatic fatty acids phenylacetate and phenylbutyrate. J Invest Dermatol 103:335–340. 12. Brusilow SW, Danney M, Waber LJ, Batshaw M, Burton B, Levitsky L, Roth K, McKeethren C, Ward J (1984) Treatment of episodic hyperammonemia in children with inborn errors of urea synthesis. N Engl J Med 310:1630–1634. 13. Simell O, Sipila I, Rajantie J, Valle DL, Brusilow SW (1986) Waste nitrogen excretion via amino acid acylation: benzoate and phenylacetate in lysinuric protein intolerance. Pediatric Res 20:1117–1121. 14. Thibault A, Cooper MR, Figg WD, Venzon DJ, Sartor O, Tompkins AC, Weinberger MS, Headlee DJ, McCall NA, Samid D, Myers CE (1994) A phase I and pharmacokinetics study of intravenous phenylacetate in patients with cancer. Cancer Res 54:1690–1694. 15. Derynck R (1986) Transforming growth factor-α: Structure and biological activities. J Cell Biochem 32:203–204. 16. Derynck R (1988) Transforming growth factor. Cell 54:593–595. 17. Derynck R (1992) The physiology of transforming growth factor-alpha. Advances Cancer Res 50:27–52. 18. Walz TM, Malm C, Wasteson A (1993) Expression of the transforming growth factor-alpha protooncogene in differentiating human promyelocyte leukemia (HL-60) cells. Cancer Res 53:191–196. 19. Matsui Y, Halter SA, Holt JT, Hogan BM, Coffey RJ (1990) Development of mammary hyperplasia and neoplasia in MMTVTGF-α transgenic mice. Cell 61:1147–1155. 20. Rappole DA, Banda MJ, Werb Z (1988) Wound macrophages express TGF-α and other growth factors in vivo: analysis by mRNA phenotyping. Science (Washington DC) 241:708–712. 21. Nicholson RI, McClelland RA, Gee JM, Manning DL, Cannon P, Robetson SF, Ellis IO (1994) Transforming growth factor-α and endocrine sensitivity in breast cancer. Cancer Res 54(7):1684–1688. 22. Anklesaria P, Teixido J, Laiho M, Pierce JH, Greenberger JS, Massague J (1990) Cell–cell adhesion mediated by binding of membrane-anchored transforming growth factor-α to epidermal growth factor receptors promotes cell proliferation. Proc Natl Acad Sci USA 87:3289–3293. 23. Brachmann R, Lindquist PB, Nagashima M, Kohr W, Lipari T, Napier M, Derynck R (1989) Transmembrane TGF-α precursors activate EGF/TGF-α receptors. Cell 56:691–700. 24. Robert J, Coffey RJ Jr, Derynck R, Wilcox JN, Bringman

456 / Liu et al. TS, Goustin AS, Moses HL, Pittelkow MR (1987) Production and auto-induction of transforming growth factor-α in human keratinocytes. Nature (Lond.) 328: 817–820. 25. Reynolds FH Jr, Todaro GJ, Fryling C, Stephenson JR (1983) Human transforming growth factors induced tyrosine phosphorylation of EGF receptors. Nature (Lond.) 292:259–262. 26. Lee DC, Rockford R, Todaro GJ, Villarreal LP (1985) Developmental expression of rat transforming growth factor-α mRNA. Mol Cell Biol 8:3644–3646. 27. Han VK, Hunter ES, Pratt RM, Zendegui JG, Lee DC (1987) Expression of rat transforming growth factor-α mRNA during development occurs predominantly in the maternal decidua. Mol Cell Biol 7:2335–2343. 28. Wilcox JN, Derynck R (1988) Developmental expression of transforming growth factor-α and β in mouse fetus. Mol Cell Biol 8:3415–3422. 29. Pittelkow MR, Lindquist PB, Abraham RT, Graves-Deal R, Derynck R, Coffey RJ Jr (1989) Induction of transforming growth factor-α expression in human keratinocytes by phorbol esters. J Biol Chem 264:5164–5171. 30. Travers MT, Barrett-Lee PJ, Berger U, Luqmani YA, Gazet J-C, Powles TJ, Coombes RC (1988) Growth factor expression in normal, benign, and malignant breast tissue. Br Med J 296:1621–1624. 31. Beauchamp RD, Bernard JA, McCutchen CM, Cherner JA, Coffey RJ (1989) Localization of TGF-α and its receptor in gastric mucosal cells: implications for a regulatory role in acid secretion and mucosal renewal. J Clin Invest 84:1017–1023. 32. Herlyn M (1993) Structure and function of molecules expressed by melanoma cells in Chapter 5, Molecular and Cellular Biology of Melanoma. R G Landes Co, Austin, TX, pp. 44–92. 33. Carpenter G (1987) Receptors for epidermal growth factor and other polypeptide mitogens. Ann Rev Biochem 56:881–914. 34. Marquardt H, Hunkapiller MW, Hood LE, Twardzik DR, DeLarco JR, Stephenson JR, Todaro GJ (1983) Transforming growth

CYTOKINE, Vol. 7, No. 5 (July 1995: 449–456) factors produced by retrovirus-transformed rodent fibroblasts and human melanoma cells: amino-acid sequence homology with epidermal growth factor. Proc Natl Acad Sci USA 80:4684–4688. 35. Shin TH, Paterson AJ, Grant III JH, Meluch AA, Kudlow EJ (1992) 5-Azacytidine treatment of HA-A melanoma cells induces Sp1 activity and concomitant transforming growth factor-α expression. Mol Cell Biol 12:3998–4006. 36. Parsanna P, Shack S, Wilson VL, Samid D (1994) Phenylacetate and derivatives as novel, nontoxic chemopreventive agents. Proceedings AACR 35:325. 37. Gearing KL, Gottlicher M, Widmark E, Banner CD, Tollet P, Stromstedt M, Rafter JJ, Berger RK, Gustafsson JA (1994) Fatty acid activation of the peroxisome proliferator activated receptor, a member of the nuclear receptor gene superfamily. J Nutr 124:1284S–1288S. 38. Mehler MF, Rozental R, Dougherty M, Spray DC, Kessler JA (1993) Cytokine regulation of neuronal differentiation of hippocampal progenitor cells. Nature (Lond.) 362:62–65. 39. Zipfel PA, Ziober BL, Morris SL, Mulder KM (1993) UpRegulation of transforming growth factor-α expression by transforming growth factor β1, epidermal growth factor, and N,N-Dimethylformamide in human colon carcinoma cells. Cell Growth & Differ 4:637–645. 40. Liu D, Gagliardi G, Nsaim MM, Allson R, Oates T, Lalani E-N, Stamp GW, Pignatell M (1994) TGF-α can act as morphogen and/or mitogen in a colon cancer cell line. Int J Cancer 65:603–608. 41. Whittaker JR (1963) Changes in melanogenesis during the dedifferentiation of chick retinal pigment cells in cell culture. Dev Biol 8:99–127. 42. Ciccarelli C, Philipson L, Sorrentino V (1990) Regulation of expression of growth arrest-specific genes in mouse fibroblasts. Mol Cell Biol 10:1525–1529. 43. Miller AC, Gafner J, Clark EP, Samid D (1993) Posttransscriptional down-regulation of ras oncogene expression by