MOLECULAR REPRODUCTION AND DEVELOPMENT 27:3-9 (1990)
Transforming Growth Factor-a RIK DERYNCK Genentech Corporation, South Sun Francisco, California
INTRODUCTION This paper introduces the reader to transforming growth factor-a (TGF-a) by reviewing the structure of its 50 amino acid form. The sequence of the human TGF-a was derived from a combination of protein biochemical data, which came from a collaboration by Hunkapiller, Todaro, and others, and from cDNA cloning done in the author’s laboratory. A striking similarity between epidermal growth factor (EGF) and TGF-a is t h a t they both have six cysteines. Indeed, TGF-a and EGF are structurally related, and that is why both molecules apparently bind to the same receptor. Based on the fact that they bind to the same receptor, they do a number of things quite similarly. There are two attitudes one can take about a result like this. One can say that they are the same molecule or very similar molecules and are merely agonists of each other. On the other hand, it should be pointed out that the similarity between EGF and TGF-a is mainly in the cysteines, with the other amino acids there is only about 30% identity. Thus TGF-a and EGF are really distinct molecules, and there may be some reason to believe that they are doing different things after binding to the same receptor. One new finding from the cDNA cloning was that the 50 amino acid form of TGF-a was part of a larger precursor sequence (Fig. 1). The initiator methionine precedes a quite hydrophobic region, which corresponds to the N-terminal signal peptide. The TGF-a precursor is cleaved in front of a leucine to release the signal peptide and produce the propeptide. The 50 amino acid form of TGF-a is released by a very specific cleavage a t both ends between a n ala and a Val-Val. This sequence has not been found in any other hormone precursors as far as the author knows, nor in any other growth factor precursor. Following the TGF-a sequence, there is a hydrophobic region flanked by two basic dipeptides. The cytoplasmic domain is cysteine-rich. These same features can also be found in the sequence of rat TGF-a. Because i t was found to contain a signal sequence and a long internal hydrophobic regions, we and the other workers have postulated that the internal hydrophobic region is a transmembrane region. This would mean that the TGF-a sequence and the remaining N-terminus would be outside the cell membrane, and the sequence C-terminal to the hydrophobic region would remain inside the cell.
0 1990 WILEY-LISS, INC.
THE TRANSMEMBRANE TGF-a PRECURSOR Processing of the TGF-cw Precursor Obviously if you want to convince people that TGF-a is processed from a transmembrane precursor, then you have to do some work. So, quite a while ago, we started to try to prove the model. We did it by raising antibodies against the 50 amino acid form of TGF-a. However, there is hardly any natural TGF-a around, so i t took us quite some time to generate recombinant, bioactive TGF-a using a n Escherichia coli expression system. Then i t took about 3 years to come up with a good antibody to TGF-a. We also prepared a n antibody against a peptide that corresponded with the putative cytoplasmic domain. We used these antibodies in a large number of experiments that are described briefly below. To test the model we first used cells that synthesize endogenous TGF-a endogenously, but we did not succeed. The main problem with these experiments was that we had to expose the gels for a month or longer, and then our results were not interpretable. So, following these disappointments, we made a n expression construct (Fig. 2) that contains a dihydrofolate reductase (DHFR) expression unit. We transfected DHFR- Chinese hamster ovary (CHO) cells with this expression construct and amplified the integrated sequences using methotrexate. An advantage of the CHO system is that CHO cells do not have EGF receptors. Therefore, we were not confronted with the complication of ligandinduced internalization of TGF-a and the corresponding receptor when working with those cells. Thus the CHO cell is a n attractive system for studying processing of the TGF-a, since the results of these types of experiments are expected to be quite clean. And they were, after a while a t least. The medium from these cells contains quite a lot of TGF-a. Using the antibody against TGF-a, we found two bands of TGF-a in the medium. One of these bands corresponded to the 50 amino acid form of TGF-a, but the other consisted of multiple bands of about 18K. The discovery of the larger form of TGF-a secreted by the TGF-a-transfected CHO cells was important, not only because we were looking here at different stages in the processing of TGF-a but also because i t has been reported many times that TGF-a made by various cell sources is 18K, 17K, or 20K. It has been proposed that this 18K form of TGF-a might respond to the entire precursor for TGF-a. Indeed, the calculated length of
4
R.DERYNCK Met Val
Dro
S e r A l a G l y Gin L e u A l a L e u Phe A l a Leu F l y lie V a l L e u A l a A,la
50
Cys G l n k i a Leu Glu Asn Ser Thr Ser Pro Leu Ser Ala
flsp P r o Pro U a l flla Rla
flla I$al U a l Ser His P h e Asn Asp Cys Pro Asp Ser His Thr Gln P h e Cys Phe His I
I
I
6r,
Gly Thr Cys Arg P h e Leu U a l Gln Glu Asp Lys P r o 80
HIS Cys Cys Gln U a l
H I S Cys Glu Trp Cys
I40
flrg flla Leu I l e Cys Arg His Glu Lys Pro Ser Ala Leu Leu Lys Gly Rrg Thr Ale '60
Cys Cys H I S Ser Glu Thr Val Val
Fig. 1. Amino acid sequence of the human TGF-a precursor. The signal peptide is cleaved from the rest of the precursor following amino acid 22. The N-glycosylation site in the precursor is overlined. The amino acid sequence of the 50 amino acid TGF-alpha is boxed. The hydrophobic transmembrane sequence is underlined and is flanked by the boxed basic dipeptide and precedes the cytoplasmic domain. The Cys-residues are printed in italics.
the 160 amino acid precursor sequence is about 18K. However, i t should be pointed out that in the presence of tunicamycin or with removal of the sugars with Nglycanase, the set of multiple bands is converted into two other, lower molecular weight bands. Thus, besides the 50 amino acid form, there are two other TGF-a polypeptides that are N-glycosylated. These peptides are obviously shorter than 18K. If, rather than precipitating from the medium, you immunoprecipitate the proteins (labeled with [35S]cysteine) from the cells, using the TGF-a antibody, you do not see anything. However, when you look a t the cells using the antibody against the TGF-(Yprecursor C-terminus, you can immunoprecipitate a protein band. So, TGF-a is found in the medium, and a segment of the precursor is located in the cells. You can also obtain much cleaner results from these experiments if you label the cells with radiolabeled palmitate rather than the cysteine. Following immunoprecipitation of the palmitate-labeled cells, there is one major band on the gel and a small amount of a somewhat larger form. We have shown that the C-terminal segment is associated with the membrane, so we are dealing with a palmitate-linked, C-terminal segment of the precursor t h a t is associated with the cell membrane. From the results of these and many other experiments, I propose the model shown in Figure 3. In this model, TGF-alpha is made as a transmembrane form, after which several cleavages can occur. The major cleavage in the system we studied resulted in release of the 50 amino acid form. Other cleavages result in the release of larger TGF-a forms that are glycosylated due to a n N-glycosylation site towards the N-terminus. In summary, several types of TGF-a molecules are found in the medium
of TGF-a-transfected cells: One molecule includes the 50 amino acid TGF-a sequence, and the other species extend toward the N-terminus and the C-terminus. It is important to realize t h a t TGF-a is found not only in the 50 amino acid form but the larger glycosylated forms are more common. There have been frequent literature reports of cells that make forms of TGF-a that are 17K, 18K, or 20K, whereas the 50 amino acid form is produced in only a few cell lines. This obviously opens up the possibility that there are even cells that do not have the ability to cleave the TGF-a precursor after synthesizing it.
Cell Surface Localization of the TGF-a Precursor We wanted to determine whether the TGF-a precursor molecule was biologically active. Before we investigated this question, we evaluated the presence of the transmembrane TGF-a in cells. In these experiments we looked for cells that contained the transmembrane protein. Similar experiments were done by David Lee and his group. If indeed there are no cells that made the transmembrane TGF-a, then there was little point in determining whether the TGF-a precursor was biologically active. Therefore, we performed immunof luorescence experiments with the TGF-a-transfected CHO cell line that secretes high levels of TGF-a in the medium. Using the anti-TGF-a antibody, we observed a n unambiguous surface staining, when the cells were kept on ice. Many controls showed that the staining was not due to TGF-a adsorbed to the membrane, especially since these cells do not express EGF receptors. Thus the staining was not due to TGF-alpha associated with the EGF receptors on the cell surface. Based on
TRANSFORMING GROWTH FACTOR-CY
5
2-
SV40 early promoter Sal I
3,
3'UT
C R
Fig. 2. Schematic diagram of the pMTE4E TGF-a expression vector. The simian virus 40 (SV40)promotor segments with the direction of early transcription are indicated by arrows. The coding sequences transcribed from the SV40 promoter elements are boxed. Within the TGF-a precursor coding sequence, the dotted box represents the signal sequence, and the dashed area corresponds to the 50 amino acid TGF-a. Some reference restriction sites are shown. The distances in kilobases are marked in the center. HBsAg3'UT, hepatitis B surface antigen gene 3' untranslated sequence; DHFR, dihydrofolate reductase.
the results of immunofluorescence studies and many other data, we are now quite comfortable in stating that this cell surface staining is due to the transmembrane form of TGF-a. The immunofluorescence pattern is quite different when the TGF-a-transfected CHO cells were incubated at 37"C, which allows internalization of the antibody-antigen complexes. The fluorescence is then distributed in bright spots that center around the nucleus, indicating that TGF-a goes to the lysosomes. The advantage of staining cells that have been incubated a t 37°C is th at the exposure times for the immunofluorescent photographs are much shorter and allow us to observe the TGF-a precursor in cells in which there is no overproduction. Thus we have examined a number of tumor cell lines for membrane-associated TGF-a precursor. One cell line that we found makes transmembrane TGF-a is the HT1080 cell. We expected to find that this cell line made TGF-a because we knew it contained TGF-a mRNA and secreted the protein into the medium. There are however, quite a few cells that make mRNA for TGF-a but for which there is no detectable TGF-a in the medium. This is a n important point to remember when studying TGF-a. One of these cells, for which we could never find anything in the medium, is a renal carcinoma cell line called 7860. These cells also show the fluorescence typical of membrane-associated TGF-a. Very long expo-
Fig. 3. Schematic representation of the transmembrane TGF-a precursor. The N-terminal signal peptide is shown as already cleaved from the precursor. The 50 amino acid TGF-a with its three disulfide bridges is shown as a bold line. The arrows indicate the potential proteolytic cleavage sites, two of them flanking the 50 amino acid TGF-a sequence and one following a basic dipeptide KK. The transmembrane segment is shown as a black box and precedes the C-terminal cytoplasmic domain rich in cysteines (C).
sures were needed to obtain photographs of anti-TGF-a antibody staining of the cells that naturally synthesize TGF-a even though they had been incubated at 37°C to consolidate the antigen. If we had not incubated them at 37"C, we would probably have considered the staining as background. These results show that there are cells that make transmembrane TGF-a but do not process the precursor to the soluble released form. All cells that synthesize TGF-a as determined by Northern analysis had detectable surface immunofluorescence, but they did not always release detectable TGF-alpha levels in the medium.
Biological Activity of the TGF-a Precursor The Purified TGF-a Precursor. Once we knew that some cells did express the TGF-a precursor on their surfaces, it then made sense to ask the question of whether the transmembrane form of TGF-a was biologically active. The way we approached the question was to eliminate the two cleavage sites following the TGF-a sequence by site-directed mutagenesis. We expected that the transfected cell then would be able to make only the transmembrane TGF-a. A similar approach was used by David Lee. We transfected a plasmid very similar to the one shown in Figure 2 into the CHO cells. Remember that the CHO cells do not make the receptors. We obtained a cell line that made a great
6
R.DERYNCK
deal of TGF-a precursor which could be stained on the cell surface by immunofluorescence. Using this cell line we took two approaches to ask the question of whether the TGF-a precursor was biologically active. One approach was biochemical and the other was to use cocultured cells. The first approach that we took to determine that the TGF-a precursor was active was to purify the TGF-a precursor and then look for biological activity. Using a n antibody affinity column as the main purification step, we purified the mutant “noncleavable” TGF-a precursor. We obtained two bands on a Coomassie-blue stained gel. We found that we could stain the larger of the two bands for carbohydrate moieties. The smaller band did not carry the carbohydrate. The two molecules are made in about equal amounts by the engineered cells. Our problem then was to identify these bands as the TGF-a precursor. The only way we could do this was by amino acid analysis or N-terminal analysis. We could not rely on the affinity of the molecule for the EGF receptor because the question we were asking was whether the precursor could bind the receptor. Thus we could not use any assays for activity to show that we had the TGF-a precursor, nor could we do any quantitation except by amino acid analysis. We obtained the N-terminal sequences of these bands and showed that they were the glycosylated and nonglycosylated forms of TGF-a precursor. There has been some discussion in the past of whether the N-terminal signal peptide is released from the precursor. From the amino acid analysis and the N-terminal sequencing of these proteins, we know that the signal peptide is cleaved off. It is difficult to demonstrate that the transmembrane precursor has TGF-a activity. How can one prove that there is no soluble TGF-a in the medium. We performed control experiments and were very stringent in their performance and in interpreting the results. One control experiment was to both stain and Western blot the same gel electrophoresis profile of a radiolabeled preparation of the secretions of the cells transfected with the “noncleavable” TGF-a sequence. We used a n antibody affinity column to purify the TGF-as before resolving them by electrophoresis. The Coomassie blue and Western stains coincided. When we exposed the gel to film until the higher molecular weight TGF-a precursor bands were black, we still did not find any smaller TGF-a forms. Now that we had these isolated forms of TGF-a we could ask whether they were biologically active. To assay for activity of the precursor, we determined whether i t could interact with the EGF receptor. If there is productive interaction, then one of the first things one would expect to occur is phosphorylation on tyrosine in the C-terminal cytoplasmic domain of the EGF receptor, as Stanley Cohen showed. We added the TGF-a precursors to A431 cells, which express many EGF receptors, then we immunoprecipitated the cells using anti-EGF receptor antibody. The immunoprecip-
itates were analyzed by Western blot using a n antiphosphotyrosine antibody. The only band one should pick up in this Western blot is the tyrosine phosphorylated EGF receptor. We did that assay using many different time points and conditions. In the absence of ligand, we did not detect the phosphorylated receptor. When we added soluble TGF-a (the 50 amino acid form) as a positive control, then the EGF receptor could be detected on the Western blot. The purified transmembrane form of TGF-a also stimulated tyrosine phosphorylation. One striking observation was that the concentration dependence of this reaction for the TGFa precursor had a very steep slope. We would very often observe a shift from no staining to maximum phosphorylation in a very narrow range of TGF-a precursor concentrations. Another assay for TGF-a with which we tested the activity of the TGF-a precursor is the assay by which TGF-a was discovered. In this assay, in the presence of TGF-P, TGF-alpha efficiently induces soft agar colony formation by a number of fibroblasts. To test the TGF-a precursor in this assay, we used Rat-1 fibroblasts. We found that the TGF-a precursor does induce soft agar colony formation. However, the effective concentrations of the precursor are about 100-fold higher than of the 50 amino acid form of TGF-a. This was true both in this assay and in the previously described assay in which the EGF receptor was activated. However, one has to be very careful in drawing conclusions from these quantitative observations. This is because we know from studies of many receptors and transmembrane molecules that purified receptors quite often demonstrate decreased affinity for their ligands. I am not yet prepared to state the affinity of the TGF-a precursor for its substrates when i t is in the membrane. Activity of the TGF-a Precursor Expressed on the Cell Surface. We also used a co-culture approach to determine whether the TGF-a precursor is active. We did these experiments as follows. We needed cells that express many EGF receptors. So we first tried to use A431 cells and we were more or less successful. However, we were more successful using engineered NIH 3T3 cells that make a lot of EGF receptor. The reason t h a t the engineered 3T3 cells provided the better system was probably because A431 make some TGF-a. Thus A431 cells do not provide a very clean system for these experiments. The experiment we did was first to plate cells that express a large amount of EGF receptor and that do not contain or make TGF-a. On top of these cells, we placed the CHO cells that do not make the EGF receptor but do make a lot of transmembrane TGF-a precursor. We then asked whether tyrosine phosphorylation of the EGF receptors was induced, and the answer was yes. Although we very often found some background phosphorylation in the negative control, the results were quite convincing. We also asked whether the TGF-a precursor stimulated more than just tyrosine phosphorylation of the
TRANSFORMING GROWTH FACTOR-CY EGF receptor. One example of what happens following the tyrosine phosphorylation of the EGF receptor in response to TGF-a is the induction of c-fos. We found that c-fos was induced in these cocultures to a much higher level than in control cocultures of engineered 3T3 cells with untransfected CHO cells. Thus the view of TGF-a is changing quite rapidly these days. When thinking about TGF-a one should think not about the 50 amino acid form, but very often about the larger glycosylated forms. These larger forms may have different properties. There are also a number of cells that express TGF-a as a transmembrane form, which is biologically active. This transmembrane form could be very important in the physiological context. An example is the mammary carcinoma cells which in real life is part of a tumor. There are also many fibroblasts in mammary carcinomas. One can easily imagine the carcinoma cells being in contact with the fibroblasts in the tissue and in this way stimulating the fibroblasts to grow. It is important to think about the physiological consequences of that type of interaction.
Comparisons With Other Growth Factors and Proteins Related to TGF-a It is also important to note that TGF-a is not the only growth factor with a transmembrane precursor. We know of other growth factors that, according to the structural predictions, are made a s transmembrane forms. Stanley Cohen mentioned similar questions about the EGF precursor, which is a very large precursor. Other growth factors that are made as large transmembrane precursors include vaccinia virus growth factor, amphiregulin, colony stimulating factor-1 (CSFl),and tumor necrosis factor (TNF). The same questions can be asked for these growth factors a s have been asked for TGF-a. Besides the mammalian growth factors just listed, i t has been shown by many people, especially more recently by those in the Drosophila field, that there are a large number of molecules that contain the EGF domain (Table 1). Today’s list is incomplete; the list of EGF-like molecules is growing fast. Many proteins containing EGF-like domains may be receptors or transmembrane molecules involved in differentiation that have not yet been discovered. One question that we are currently investigating is whether these larger molecules such as the MEL-14 adhesion receptor have a complete EGF domain that can interact with the EGF receptor. TISSUE LOCALIZATION OF TGF-cx In the past we have done a number of studies, which are not discussed here, on the localization of TGF-a in adult and fetal tissues. We found that TGF-a was expressed in the brain and during embryonic development. I will discuss some studies that we performed in collaboration with Bob Coffey, who was in Hal Moses’ laboratory at that time, and with Mark Pittelkow, who
7
TABLE 1. Proteins With “Complete” EGF Domain Cell adhesion MEL-14 ELAM a140 Cell differentiation Notch (Drosophila) Delta (Drosophila) 52D (silt) (Drosophila) 95F (Drosophilaj lin-12 (C. elegans) uEGF-1 (sea urchins)
Blood coagulation Factor IX Factor X Factor XI1 Protein C Protein Z Urokinase t-Plasminogen activator Receptors LDL receptor Thrombomodulin
is at the Mayo Clinic. We knew from past studies that a number of tumor cells made TGF-a. What was very striking was that squamous carcinomas or other carcinomas very often make TGF-a. Squamous carcinomas are very often derived from keratinocytes. So we looked at normal keratinocytes isolated from either breast or foreskin or from other sources for expression of TGF-a. We found that primary keratinocytes that were maintained in culture under defined conditions made TGF-a mRNA.
REGULATION OF THE EXPRESSION OF TGF-CY Expression Auto-Induction of TGF-CY A striking finding in our studies with keratinocytes t h a t was surprising a t the time was that these cells, which are normally grown in the presence of EGF, made much more TGF-a when grown in the presence of EGF than in its absence. This major differences in the level of TGF-a mRNA was also seen in the level of TGF-a protein. With time after the addition of EGF, we observed a n increase of TGF-a production. All of these assays were done using a n enzyme-linked immunosorbent assay (ELISA) that does not cross react with EGF. It is important to remember a t this point that both molecules bind to the same receptor. This result then suggested auto-induction of the TGF-a gene. We did indeed find that TGF-a increased TGF-a mRNA expression in these keratinocytes. If we could observe induction with exogenously added TGF-a, then it may also happen in response to endogenous TGF-a. The biological relevance of this type of auto-induction is puzzling. At the time that it was described, it was obviously fairly new, but now we know that it is not a n exception. Auto-induction also occurs with platelet-derived growth factor (PDGF), interleukin 1, melanoma growth stimulatory activity (MGSA), and TGF-P. On the other hand, it is not a general fact that all growth factors auto-induce the expression of their genes. Meanwhile, more recently it has been found that auto-induction is also a means of regulating the expres-
8
R.DERYNCK
sion of a number of transcription factors such a s myoD and fos. Is this endogenous TGF-a synthesis relevant for keratinocytes? To find the answer to that question i t is necessary to know whether TGF-a is made by these cells in vivo. Using in situ hybridization, we found that TGF-a mRNA is expressed in the skin. Similarly, with immunohistochemistry, we found the TGF-a protein is present in the epidermis. However, the levels of TGF-a that we found were fairly low. Therefore, one could question whether our observation of TGF-a expression in the epidermis has any physiological relevance. How much ‘I’GF-a do you really need for activation of mitosis in keratinocytes? Our estimates show that the presence of 10 pg/ml of TGF-a is sufficient to induce mitosis or colony formation by keratinocytes. Therefore, the levels of TGF-a made by keratinocytes are higher than needed to stimulate their mitosis.
Regulation of TGF-a Expression Through Protein Kinase C These studies were extended by Bob Coffey to show that the tumor promoter 12-0-tetradecanoyl phorbol acetate (TPA) also induces TGF-a. Compared with the induction by EGF and TGF-a, the increase in TGF-a mRNA in response to TPA was enormous. These results indicated that activation of protein kinase C (PK C) results in a higher level of TGF-a mRNA. The increased mRNA is reflected a t the protein level. I will not discuss here all the experiments that we did to establish that the regulation was through PK C, but I will mention a few key experiments. Tests of the effect of two analogs of TPA showed that TPA was likely to exert its effect on TGF-a mRNA level by binding to PK C. Cells that had been depleted of PK C showed a minimal change in TGF-a mRNA in response to TPA. Thus PK C appears to induce TGF-a synthesis. We also know, from other studies, that di-C8 or exogenous phospholipase C cause dramatic increases in TGF-a expression. Thus I feel quite comfortable in saying t h a t induction of PK-C pathway results in higher levels of TGF-a.
TGF-a EXPRESSION IN PSORIASIS The induction of TGF-a by PK C may be relevant to a skin disease called psoriasis. Psoriasis is a very common disease in which there are a number of aberrations. Depending on your background, you will consider psoriasis as a proliferation disease of keratinocytes or a n immune disorder in which there is a lot of inflammation. One can argue about which is the most important aspect of the disease, but both components are present in psoriasis. In collaboration with John Voorhees at Michigan, we examined a number of psoriatic skin samples for TGF-a expression. We found that psoriatic epidermis expresses high levels of TGF-a mRNA. You do not even have to make poly-A+ to pick up TGF-a mRNA. Although there was quite a bit of TGF-a mRNA
TABLE 2. Normal Cells Expressing TGF-a Keratinocytes Epithelial cells Activated macrophages Brain Pituitary Various tissues in embryonic development
in the psoriatic samples, there was hardly any TGF-a mRNA in the normal skin. Uninvolved skin from the same patient had the low level of TGF-alpha mRNA characteristic of normal skin. Thus psoriatic skin is quite different from normal skin with respect to TGF-a production. This is also reflected a t the protein level. TGF-a is very easily detected in extracts of psoriatic skin in contrast to TGF-a in normal or uninvolved skin. This leads us to propose that TGF-a could play a role in proliferation in psoriasis, although i t may not be the driving force in this skin disease.
SUMMARY In summary, although TGF-a was initially found in tumors, a number of later studies, some of them from the author’s laboratory, have shown that TGF-a should no longer be considered a tumor associated growth factor. Rather, TGF-a is a normal physiological ligand for the EGF receptor. Table 2 lists some of the normal cellular sources of TGF-a. Our list is incomplete, but we know that TGF-a is made in keratinocytes and a number of epithelial cells, including gut and breast epithelial cells. It seems very likely that TGF-a is a major growth factor secreted by cells of epithelial origin. Zena Werb‘s and Russell Ross’s groups have shown that activated macrophages make TGF-a. We have shown that brain makes TGF-a and Jeff Kudlow has found TGF-a made in the pituitary. Data from several sources, including David Lee, the author’s laboratory, and Zena Werb‘s laboratory has shown that TGF-a is made during embryonic development. Therefore, it is now important to look at TGF-a in its normal physiological context. Finally, i t should be stressed that, as was mentioned above, TGF-a is not necessarily a secreted growth factor 50 amino acids long. There is quite a bit of processing of the larger precursor that may or may not take place. This processing, which determines the ultimate size and location of the molecule, is also likely to influence its physiological action.
QUESTIONS AND ANSWERS Q: What form of TGF-a and what purity did you use in the soft agar colony formation assay for the preproTGF-a? A: We used the solubilized and purified transmembrane TGF-a. This was prepared by passing a solubi-
TRANSFORMING GROWTH FACTOR-a lized membrane preparation through a n anti-TGF-a antibody column. Q: Could the preparation have included some of the 50 amino acid form of TGF-a? A: No, i t was the larger precursor. However, in t h a t assay it takes quite a few days before the colonies are detected. Therefore, we cannot exclude the possibility that there was some cleavage of the precursor during the assay. Q: Is there any evidence that the precursor can be processed during contact with the EGF receptor? A: I do not know. We have tried to look a t a number of things relating to this question. For example, is the precursor internalized? We were quite unsuccessful in determining this. I do not think it is internalized as a complex with the receptor, but I cannot prove this. Q: Have you engineered the precursor in such a way as to add a polypeptide extension to the C-terminal end of the 50 amino acid form of TGF-a that could not be cleaved? In this way you could be sure that the protein remained in the membrane during your studies of the activity of the membrane-bound precursor. A: That would be another approach. On the other hand, I think that we did everything possible in our studies to exclude the possibility that the 50 amino acid form of TGF-a was released. We did a number of control experiments t h a t I did not mention, both with the co-cultures and with the single cell cultures of TGFa-expressing cells. We also used a very sensitive ELISA to try to look for leakage of TGF-a into the medium. Leakage from the membrane was a possibility, especially considering the results reported for the EGF precursor. We could not find any leakage of any form of TGF-a into the medium. The problem is you have to come up with a negative result to prove there is no leakage, but a negative result is never very satisfactory as evidence. Q: Do you know of anything that will decrease the expression of TGF-a in psoriatic skin. A: I a m not aware of anything that can decrease TGF-a production in psoriatic skin. The problem is you really have to do these things in vivo. Q: Could you please describe your ELISA assay. A: It is a double sandwich ELISA that uses a combination of two antibodies. One antibody is a monoclonal antibody directed against the 50 amino acid form, which seems to recognize a n epitope that corresponds
9
to a combination of the first and the second loops. TGFa has to be in the native conformation. It does not recoganize the denatured TGF-a. The second antibody is polyclonal and neutralizing. We do not know what epitope(s1 i t recognizes. Q: I noticed that you have two species of the TGF-a mRNA. Do you see both species expressed in all cells and tissues? A: The major mRNA species is 4.5-4.8 kb. Very often, though, especially at the higher levels of TGF-a mRNA expression, we do find a smaller, approximately 18S, form. We are not sure what this smaller mRNA species is. However, the sequence of the 3’ untranslated region of the TGF-a mRNA contains potential polyadenylation sites that would allow for premature polyadenylation. We have no evidence for alternative splicing, but we have not really looked for this.
REFERENCES Brachmann R, Lindquist PB, Nagashima M, Kohr W, Lipari T, Napier M, Derynck R (1989): Transmembrane TGF-alpha precursors activate EGFiTGF-alpha receptors. Cell 56:691-700. Bringman TS, Lindquist PB, Derynck R (1987): Different transforming growth factor-alpha species are derived from a glycosylated and palmitoylated transmembrane precurosr. Cell 48:429-440. Coffey RJ, Derynck R, Wilcox J N , Bringman TS, Goustin AS, Moses HL, Pittelkow MR (1987): Production and auto-induction of transforming growth factor-alpha in human keratinocytes. Nature 328: 817-820. Derynck R (1988): Transforming growth factor-alpha. Cell 54593595. Derynck R, Lindquist PB, Bringman TS, Wilcox JN, Elder JT, Fisher GL, Voorhees J J , Moses HL, Pittelkow M, Coffey RJ (1989):Expression of the transforming growth factor-alpha gene in tumor cells and normal cells. Cancer Cells [Cold Springs Harbor Laboratory] , 297-301. Derynck R, Roberts AB, Winkler ME, Chen EY, Goeddel DV: (1984): Human transforming growth factor-alpha: Precursor structure and expression in E . coli. Cell 38:287-297. Elder JT, Fisher GJ, Lindquist PB, Bennett GL, Pittelkow M, Coffey R J , Ellingsworth L, Derynck R, Voorhees JJ (1989):Overexpression of TGF-alpha in psoriatic skin. Science 243:811-814. Pittelkow MR, Lindquist PB, Abraham R, Graves-Deal R, Derynck R, Coffey RJ (1989): Induction of transforming growth factor-alpha expression in human keratinocytes by phorbol esters. J Biol Chem 2645164-5171. Wilcox JN, Derynck (1988a): Localization of cells synthesizing transforming growth factor-alpha mRNA in the mouse brain. J Neurosci 8:1901-1904. Wilcox JN, Derynck R (1988b): Developmental expression of transforming growth factor-alpha and beta in mouse fetus. Mol Cell Biol 8:3415-3422.
Transforming Growth Factor-a RIK DERYNCK Genentech Corporation, South Sun Francisco, California
INTRODUCTION This paper introduces the reader to transforming growth factor-a (TGF-a) by reviewing the structure of its 50 amino acid form. The sequence of the human TGF-a was derived from a combination of protein biochemical data, which came from a collaboration by Hunkapiller, Todaro, and others, and from cDNA cloning done in the author’s laboratory. A striking similarity between epidermal growth factor (EGF) and TGF-a is t h a t they both have six cysteines. Indeed, TGF-a and EGF are structurally related, and that is why both molecules apparently bind to the same receptor. Based on the fact that they bind to the same receptor, they do a number of things quite similarly. There are two attitudes one can take about a result like this. One can say that they are the same molecule or very similar molecules and are merely agonists of each other. On the other hand, it should be pointed out that the similarity between EGF and TGF-a is mainly in the cysteines, with the other amino acids there is only about 30% identity. Thus TGF-a and EGF are really distinct molecules, and there may be some reason to believe that they are doing different things after binding to the same receptor. One new finding from the cDNA cloning was that the 50 amino acid form of TGF-a was part of a larger precursor sequence (Fig. 1). The initiator methionine precedes a quite hydrophobic region, which corresponds to the N-terminal signal peptide. The TGF-a precursor is cleaved in front of a leucine to release the signal peptide and produce the propeptide. The 50 amino acid form of TGF-a is released by a very specific cleavage a t both ends between a n ala and a Val-Val. This sequence has not been found in any other hormone precursors as far as the author knows, nor in any other growth factor precursor. Following the TGF-a sequence, there is a hydrophobic region flanked by two basic dipeptides. The cytoplasmic domain is cysteine-rich. These same features can also be found in the sequence of rat TGF-a. Because i t was found to contain a signal sequence and a long internal hydrophobic regions, we and the other workers have postulated that the internal hydrophobic region is a transmembrane region. This would mean that the TGF-a sequence and the remaining N-terminus would be outside the cell membrane, and the sequence C-terminal to the hydrophobic region would remain inside the cell.
0 1990 WILEY-LISS, INC.
THE TRANSMEMBRANE TGF-a PRECURSOR Processing of the TGF-cw Precursor Obviously if you want to convince people that TGF-a is processed from a transmembrane precursor, then you have to do some work. So, quite a while ago, we started to try to prove the model. We did it by raising antibodies against the 50 amino acid form of TGF-a. However, there is hardly any natural TGF-a around, so i t took us quite some time to generate recombinant, bioactive TGF-a using a n Escherichia coli expression system. Then i t took about 3 years to come up with a good antibody to TGF-a. We also prepared a n antibody against a peptide that corresponded with the putative cytoplasmic domain. We used these antibodies in a large number of experiments that are described briefly below. To test the model we first used cells that synthesize endogenous TGF-a endogenously, but we did not succeed. The main problem with these experiments was that we had to expose the gels for a month or longer, and then our results were not interpretable. So, following these disappointments, we made a n expression construct (Fig. 2) that contains a dihydrofolate reductase (DHFR) expression unit. We transfected DHFR- Chinese hamster ovary (CHO) cells with this expression construct and amplified the integrated sequences using methotrexate. An advantage of the CHO system is that CHO cells do not have EGF receptors. Therefore, we were not confronted with the complication of ligandinduced internalization of TGF-a and the corresponding receptor when working with those cells. Thus the CHO cell is a n attractive system for studying processing of the TGF-a, since the results of these types of experiments are expected to be quite clean. And they were, after a while a t least. The medium from these cells contains quite a lot of TGF-a. Using the antibody against TGF-a, we found two bands of TGF-a in the medium. One of these bands corresponded to the 50 amino acid form of TGF-a, but the other consisted of multiple bands of about 18K. The discovery of the larger form of TGF-a secreted by the TGF-a-transfected CHO cells was important, not only because we were looking here at different stages in the processing of TGF-a but also because i t has been reported many times that TGF-a made by various cell sources is 18K, 17K, or 20K. It has been proposed that this 18K form of TGF-a might respond to the entire precursor for TGF-a. Indeed, the calculated length of
4
R.DERYNCK Met Val
Dro
S e r A l a G l y Gin L e u A l a L e u Phe A l a Leu F l y lie V a l L e u A l a A,la
50
Cys G l n k i a Leu Glu Asn Ser Thr Ser Pro Leu Ser Ala
flsp P r o Pro U a l flla Rla
flla I$al U a l Ser His P h e Asn Asp Cys Pro Asp Ser His Thr Gln P h e Cys Phe His I
I
I
6r,
Gly Thr Cys Arg P h e Leu U a l Gln Glu Asp Lys P r o 80
HIS Cys Cys Gln U a l
H I S Cys Glu Trp Cys
I40
flrg flla Leu I l e Cys Arg His Glu Lys Pro Ser Ala Leu Leu Lys Gly Rrg Thr Ale '60
Cys Cys H I S Ser Glu Thr Val Val
Fig. 1. Amino acid sequence of the human TGF-a precursor. The signal peptide is cleaved from the rest of the precursor following amino acid 22. The N-glycosylation site in the precursor is overlined. The amino acid sequence of the 50 amino acid TGF-alpha is boxed. The hydrophobic transmembrane sequence is underlined and is flanked by the boxed basic dipeptide and precedes the cytoplasmic domain. The Cys-residues are printed in italics.
the 160 amino acid precursor sequence is about 18K. However, i t should be pointed out that in the presence of tunicamycin or with removal of the sugars with Nglycanase, the set of multiple bands is converted into two other, lower molecular weight bands. Thus, besides the 50 amino acid form, there are two other TGF-a polypeptides that are N-glycosylated. These peptides are obviously shorter than 18K. If, rather than precipitating from the medium, you immunoprecipitate the proteins (labeled with [35S]cysteine) from the cells, using the TGF-a antibody, you do not see anything. However, when you look a t the cells using the antibody against the TGF-(Yprecursor C-terminus, you can immunoprecipitate a protein band. So, TGF-a is found in the medium, and a segment of the precursor is located in the cells. You can also obtain much cleaner results from these experiments if you label the cells with radiolabeled palmitate rather than the cysteine. Following immunoprecipitation of the palmitate-labeled cells, there is one major band on the gel and a small amount of a somewhat larger form. We have shown that the C-terminal segment is associated with the membrane, so we are dealing with a palmitate-linked, C-terminal segment of the precursor t h a t is associated with the cell membrane. From the results of these and many other experiments, I propose the model shown in Figure 3. In this model, TGF-alpha is made as a transmembrane form, after which several cleavages can occur. The major cleavage in the system we studied resulted in release of the 50 amino acid form. Other cleavages result in the release of larger TGF-a forms that are glycosylated due to a n N-glycosylation site towards the N-terminus. In summary, several types of TGF-a molecules are found in the medium
of TGF-a-transfected cells: One molecule includes the 50 amino acid TGF-a sequence, and the other species extend toward the N-terminus and the C-terminus. It is important to realize t h a t TGF-a is found not only in the 50 amino acid form but the larger glycosylated forms are more common. There have been frequent literature reports of cells that make forms of TGF-a that are 17K, 18K, or 20K, whereas the 50 amino acid form is produced in only a few cell lines. This obviously opens up the possibility that there are even cells that do not have the ability to cleave the TGF-a precursor after synthesizing it.
Cell Surface Localization of the TGF-a Precursor We wanted to determine whether the TGF-a precursor molecule was biologically active. Before we investigated this question, we evaluated the presence of the transmembrane TGF-a in cells. In these experiments we looked for cells that contained the transmembrane protein. Similar experiments were done by David Lee and his group. If indeed there are no cells that made the transmembrane TGF-a, then there was little point in determining whether the TGF-a precursor was biologically active. Therefore, we performed immunof luorescence experiments with the TGF-a-transfected CHO cell line that secretes high levels of TGF-a in the medium. Using the anti-TGF-a antibody, we observed a n unambiguous surface staining, when the cells were kept on ice. Many controls showed that the staining was not due to TGF-a adsorbed to the membrane, especially since these cells do not express EGF receptors. Thus the staining was not due to TGF-alpha associated with the EGF receptors on the cell surface. Based on
TRANSFORMING GROWTH FACTOR-CY
5
2-
SV40 early promoter Sal I
3,
3'UT
C R
Fig. 2. Schematic diagram of the pMTE4E TGF-a expression vector. The simian virus 40 (SV40)promotor segments with the direction of early transcription are indicated by arrows. The coding sequences transcribed from the SV40 promoter elements are boxed. Within the TGF-a precursor coding sequence, the dotted box represents the signal sequence, and the dashed area corresponds to the 50 amino acid TGF-a. Some reference restriction sites are shown. The distances in kilobases are marked in the center. HBsAg3'UT, hepatitis B surface antigen gene 3' untranslated sequence; DHFR, dihydrofolate reductase.
the results of immunofluorescence studies and many other data, we are now quite comfortable in stating that this cell surface staining is due to the transmembrane form of TGF-a. The immunofluorescence pattern is quite different when the TGF-a-transfected CHO cells were incubated at 37"C, which allows internalization of the antibody-antigen complexes. The fluorescence is then distributed in bright spots that center around the nucleus, indicating that TGF-a goes to the lysosomes. The advantage of staining cells that have been incubated a t 37°C is th at the exposure times for the immunofluorescent photographs are much shorter and allow us to observe the TGF-a precursor in cells in which there is no overproduction. Thus we have examined a number of tumor cell lines for membrane-associated TGF-a precursor. One cell line that we found makes transmembrane TGF-a is the HT1080 cell. We expected to find that this cell line made TGF-a because we knew it contained TGF-a mRNA and secreted the protein into the medium. There are however, quite a few cells that make mRNA for TGF-a but for which there is no detectable TGF-a in the medium. This is a n important point to remember when studying TGF-a. One of these cells, for which we could never find anything in the medium, is a renal carcinoma cell line called 7860. These cells also show the fluorescence typical of membrane-associated TGF-a. Very long expo-
Fig. 3. Schematic representation of the transmembrane TGF-a precursor. The N-terminal signal peptide is shown as already cleaved from the precursor. The 50 amino acid TGF-a with its three disulfide bridges is shown as a bold line. The arrows indicate the potential proteolytic cleavage sites, two of them flanking the 50 amino acid TGF-a sequence and one following a basic dipeptide KK. The transmembrane segment is shown as a black box and precedes the C-terminal cytoplasmic domain rich in cysteines (C).
sures were needed to obtain photographs of anti-TGF-a antibody staining of the cells that naturally synthesize TGF-a even though they had been incubated at 37°C to consolidate the antigen. If we had not incubated them at 37"C, we would probably have considered the staining as background. These results show that there are cells that make transmembrane TGF-a but do not process the precursor to the soluble released form. All cells that synthesize TGF-a as determined by Northern analysis had detectable surface immunofluorescence, but they did not always release detectable TGF-alpha levels in the medium.
Biological Activity of the TGF-a Precursor The Purified TGF-a Precursor. Once we knew that some cells did express the TGF-a precursor on their surfaces, it then made sense to ask the question of whether the transmembrane form of TGF-a was biologically active. The way we approached the question was to eliminate the two cleavage sites following the TGF-a sequence by site-directed mutagenesis. We expected that the transfected cell then would be able to make only the transmembrane TGF-a. A similar approach was used by David Lee. We transfected a plasmid very similar to the one shown in Figure 2 into the CHO cells. Remember that the CHO cells do not make the receptors. We obtained a cell line that made a great
6
R.DERYNCK
deal of TGF-a precursor which could be stained on the cell surface by immunofluorescence. Using this cell line we took two approaches to ask the question of whether the TGF-a precursor was biologically active. One approach was biochemical and the other was to use cocultured cells. The first approach that we took to determine that the TGF-a precursor was active was to purify the TGF-a precursor and then look for biological activity. Using a n antibody affinity column as the main purification step, we purified the mutant “noncleavable” TGF-a precursor. We obtained two bands on a Coomassie-blue stained gel. We found that we could stain the larger of the two bands for carbohydrate moieties. The smaller band did not carry the carbohydrate. The two molecules are made in about equal amounts by the engineered cells. Our problem then was to identify these bands as the TGF-a precursor. The only way we could do this was by amino acid analysis or N-terminal analysis. We could not rely on the affinity of the molecule for the EGF receptor because the question we were asking was whether the precursor could bind the receptor. Thus we could not use any assays for activity to show that we had the TGF-a precursor, nor could we do any quantitation except by amino acid analysis. We obtained the N-terminal sequences of these bands and showed that they were the glycosylated and nonglycosylated forms of TGF-a precursor. There has been some discussion in the past of whether the N-terminal signal peptide is released from the precursor. From the amino acid analysis and the N-terminal sequencing of these proteins, we know that the signal peptide is cleaved off. It is difficult to demonstrate that the transmembrane precursor has TGF-a activity. How can one prove that there is no soluble TGF-a in the medium. We performed control experiments and were very stringent in their performance and in interpreting the results. One control experiment was to both stain and Western blot the same gel electrophoresis profile of a radiolabeled preparation of the secretions of the cells transfected with the “noncleavable” TGF-a sequence. We used a n antibody affinity column to purify the TGF-as before resolving them by electrophoresis. The Coomassie blue and Western stains coincided. When we exposed the gel to film until the higher molecular weight TGF-a precursor bands were black, we still did not find any smaller TGF-a forms. Now that we had these isolated forms of TGF-a we could ask whether they were biologically active. To assay for activity of the precursor, we determined whether i t could interact with the EGF receptor. If there is productive interaction, then one of the first things one would expect to occur is phosphorylation on tyrosine in the C-terminal cytoplasmic domain of the EGF receptor, as Stanley Cohen showed. We added the TGF-a precursors to A431 cells, which express many EGF receptors, then we immunoprecipitated the cells using anti-EGF receptor antibody. The immunoprecip-
itates were analyzed by Western blot using a n antiphosphotyrosine antibody. The only band one should pick up in this Western blot is the tyrosine phosphorylated EGF receptor. We did that assay using many different time points and conditions. In the absence of ligand, we did not detect the phosphorylated receptor. When we added soluble TGF-a (the 50 amino acid form) as a positive control, then the EGF receptor could be detected on the Western blot. The purified transmembrane form of TGF-a also stimulated tyrosine phosphorylation. One striking observation was that the concentration dependence of this reaction for the TGFa precursor had a very steep slope. We would very often observe a shift from no staining to maximum phosphorylation in a very narrow range of TGF-a precursor concentrations. Another assay for TGF-a with which we tested the activity of the TGF-a precursor is the assay by which TGF-a was discovered. In this assay, in the presence of TGF-P, TGF-alpha efficiently induces soft agar colony formation by a number of fibroblasts. To test the TGF-a precursor in this assay, we used Rat-1 fibroblasts. We found that the TGF-a precursor does induce soft agar colony formation. However, the effective concentrations of the precursor are about 100-fold higher than of the 50 amino acid form of TGF-a. This was true both in this assay and in the previously described assay in which the EGF receptor was activated. However, one has to be very careful in drawing conclusions from these quantitative observations. This is because we know from studies of many receptors and transmembrane molecules that purified receptors quite often demonstrate decreased affinity for their ligands. I am not yet prepared to state the affinity of the TGF-a precursor for its substrates when i t is in the membrane. Activity of the TGF-a Precursor Expressed on the Cell Surface. We also used a co-culture approach to determine whether the TGF-a precursor is active. We did these experiments as follows. We needed cells that express many EGF receptors. So we first tried to use A431 cells and we were more or less successful. However, we were more successful using engineered NIH 3T3 cells that make a lot of EGF receptor. The reason t h a t the engineered 3T3 cells provided the better system was probably because A431 make some TGF-a. Thus A431 cells do not provide a very clean system for these experiments. The experiment we did was first to plate cells that express a large amount of EGF receptor and that do not contain or make TGF-a. On top of these cells, we placed the CHO cells that do not make the EGF receptor but do make a lot of transmembrane TGF-a precursor. We then asked whether tyrosine phosphorylation of the EGF receptors was induced, and the answer was yes. Although we very often found some background phosphorylation in the negative control, the results were quite convincing. We also asked whether the TGF-a precursor stimulated more than just tyrosine phosphorylation of the
TRANSFORMING GROWTH FACTOR-CY EGF receptor. One example of what happens following the tyrosine phosphorylation of the EGF receptor in response to TGF-a is the induction of c-fos. We found that c-fos was induced in these cocultures to a much higher level than in control cocultures of engineered 3T3 cells with untransfected CHO cells. Thus the view of TGF-a is changing quite rapidly these days. When thinking about TGF-a one should think not about the 50 amino acid form, but very often about the larger glycosylated forms. These larger forms may have different properties. There are also a number of cells that express TGF-a as a transmembrane form, which is biologically active. This transmembrane form could be very important in the physiological context. An example is the mammary carcinoma cells which in real life is part of a tumor. There are also many fibroblasts in mammary carcinomas. One can easily imagine the carcinoma cells being in contact with the fibroblasts in the tissue and in this way stimulating the fibroblasts to grow. It is important to think about the physiological consequences of that type of interaction.
Comparisons With Other Growth Factors and Proteins Related to TGF-a It is also important to note that TGF-a is not the only growth factor with a transmembrane precursor. We know of other growth factors that, according to the structural predictions, are made a s transmembrane forms. Stanley Cohen mentioned similar questions about the EGF precursor, which is a very large precursor. Other growth factors that are made as large transmembrane precursors include vaccinia virus growth factor, amphiregulin, colony stimulating factor-1 (CSFl),and tumor necrosis factor (TNF). The same questions can be asked for these growth factors a s have been asked for TGF-a. Besides the mammalian growth factors just listed, i t has been shown by many people, especially more recently by those in the Drosophila field, that there are a large number of molecules that contain the EGF domain (Table 1). Today’s list is incomplete; the list of EGF-like molecules is growing fast. Many proteins containing EGF-like domains may be receptors or transmembrane molecules involved in differentiation that have not yet been discovered. One question that we are currently investigating is whether these larger molecules such as the MEL-14 adhesion receptor have a complete EGF domain that can interact with the EGF receptor. TISSUE LOCALIZATION OF TGF-cx In the past we have done a number of studies, which are not discussed here, on the localization of TGF-a in adult and fetal tissues. We found that TGF-a was expressed in the brain and during embryonic development. I will discuss some studies that we performed in collaboration with Bob Coffey, who was in Hal Moses’ laboratory at that time, and with Mark Pittelkow, who
7
TABLE 1. Proteins With “Complete” EGF Domain Cell adhesion MEL-14 ELAM a140 Cell differentiation Notch (Drosophila) Delta (Drosophila) 52D (silt) (Drosophila) 95F (Drosophilaj lin-12 (C. elegans) uEGF-1 (sea urchins)
Blood coagulation Factor IX Factor X Factor XI1 Protein C Protein Z Urokinase t-Plasminogen activator Receptors LDL receptor Thrombomodulin
is at the Mayo Clinic. We knew from past studies that a number of tumor cells made TGF-a. What was very striking was that squamous carcinomas or other carcinomas very often make TGF-a. Squamous carcinomas are very often derived from keratinocytes. So we looked at normal keratinocytes isolated from either breast or foreskin or from other sources for expression of TGF-a. We found that primary keratinocytes that were maintained in culture under defined conditions made TGF-a mRNA.
REGULATION OF THE EXPRESSION OF TGF-CY Expression Auto-Induction of TGF-CY A striking finding in our studies with keratinocytes t h a t was surprising a t the time was that these cells, which are normally grown in the presence of EGF, made much more TGF-a when grown in the presence of EGF than in its absence. This major differences in the level of TGF-a mRNA was also seen in the level of TGF-a protein. With time after the addition of EGF, we observed a n increase of TGF-a production. All of these assays were done using a n enzyme-linked immunosorbent assay (ELISA) that does not cross react with EGF. It is important to remember a t this point that both molecules bind to the same receptor. This result then suggested auto-induction of the TGF-a gene. We did indeed find that TGF-a increased TGF-a mRNA expression in these keratinocytes. If we could observe induction with exogenously added TGF-a, then it may also happen in response to endogenous TGF-a. The biological relevance of this type of auto-induction is puzzling. At the time that it was described, it was obviously fairly new, but now we know that it is not a n exception. Auto-induction also occurs with platelet-derived growth factor (PDGF), interleukin 1, melanoma growth stimulatory activity (MGSA), and TGF-P. On the other hand, it is not a general fact that all growth factors auto-induce the expression of their genes. Meanwhile, more recently it has been found that auto-induction is also a means of regulating the expres-
8
R.DERYNCK
sion of a number of transcription factors such a s myoD and fos. Is this endogenous TGF-a synthesis relevant for keratinocytes? To find the answer to that question i t is necessary to know whether TGF-a is made by these cells in vivo. Using in situ hybridization, we found that TGF-a mRNA is expressed in the skin. Similarly, with immunohistochemistry, we found the TGF-a protein is present in the epidermis. However, the levels of TGF-a that we found were fairly low. Therefore, one could question whether our observation of TGF-a expression in the epidermis has any physiological relevance. How much ‘I’GF-a do you really need for activation of mitosis in keratinocytes? Our estimates show that the presence of 10 pg/ml of TGF-a is sufficient to induce mitosis or colony formation by keratinocytes. Therefore, the levels of TGF-a made by keratinocytes are higher than needed to stimulate their mitosis.
Regulation of TGF-a Expression Through Protein Kinase C These studies were extended by Bob Coffey to show that the tumor promoter 12-0-tetradecanoyl phorbol acetate (TPA) also induces TGF-a. Compared with the induction by EGF and TGF-a, the increase in TGF-a mRNA in response to TPA was enormous. These results indicated that activation of protein kinase C (PK C) results in a higher level of TGF-a mRNA. The increased mRNA is reflected a t the protein level. I will not discuss here all the experiments that we did to establish that the regulation was through PK C, but I will mention a few key experiments. Tests of the effect of two analogs of TPA showed that TPA was likely to exert its effect on TGF-a mRNA level by binding to PK C. Cells that had been depleted of PK C showed a minimal change in TGF-a mRNA in response to TPA. Thus PK C appears to induce TGF-a synthesis. We also know, from other studies, that di-C8 or exogenous phospholipase C cause dramatic increases in TGF-a expression. Thus I feel quite comfortable in saying t h a t induction of PK-C pathway results in higher levels of TGF-a.
TGF-a EXPRESSION IN PSORIASIS The induction of TGF-a by PK C may be relevant to a skin disease called psoriasis. Psoriasis is a very common disease in which there are a number of aberrations. Depending on your background, you will consider psoriasis as a proliferation disease of keratinocytes or a n immune disorder in which there is a lot of inflammation. One can argue about which is the most important aspect of the disease, but both components are present in psoriasis. In collaboration with John Voorhees at Michigan, we examined a number of psoriatic skin samples for TGF-a expression. We found that psoriatic epidermis expresses high levels of TGF-a mRNA. You do not even have to make poly-A+ to pick up TGF-a mRNA. Although there was quite a bit of TGF-a mRNA
TABLE 2. Normal Cells Expressing TGF-a Keratinocytes Epithelial cells Activated macrophages Brain Pituitary Various tissues in embryonic development
in the psoriatic samples, there was hardly any TGF-a mRNA in the normal skin. Uninvolved skin from the same patient had the low level of TGF-alpha mRNA characteristic of normal skin. Thus psoriatic skin is quite different from normal skin with respect to TGF-a production. This is also reflected a t the protein level. TGF-a is very easily detected in extracts of psoriatic skin in contrast to TGF-a in normal or uninvolved skin. This leads us to propose that TGF-a could play a role in proliferation in psoriasis, although i t may not be the driving force in this skin disease.
SUMMARY In summary, although TGF-a was initially found in tumors, a number of later studies, some of them from the author’s laboratory, have shown that TGF-a should no longer be considered a tumor associated growth factor. Rather, TGF-a is a normal physiological ligand for the EGF receptor. Table 2 lists some of the normal cellular sources of TGF-a. Our list is incomplete, but we know that TGF-a is made in keratinocytes and a number of epithelial cells, including gut and breast epithelial cells. It seems very likely that TGF-a is a major growth factor secreted by cells of epithelial origin. Zena Werb‘s and Russell Ross’s groups have shown that activated macrophages make TGF-a. We have shown that brain makes TGF-a and Jeff Kudlow has found TGF-a made in the pituitary. Data from several sources, including David Lee, the author’s laboratory, and Zena Werb‘s laboratory has shown that TGF-a is made during embryonic development. Therefore, it is now important to look at TGF-a in its normal physiological context. Finally, i t should be stressed that, as was mentioned above, TGF-a is not necessarily a secreted growth factor 50 amino acids long. There is quite a bit of processing of the larger precursor that may or may not take place. This processing, which determines the ultimate size and location of the molecule, is also likely to influence its physiological action.
QUESTIONS AND ANSWERS Q: What form of TGF-a and what purity did you use in the soft agar colony formation assay for the preproTGF-a? A: We used the solubilized and purified transmembrane TGF-a. This was prepared by passing a solubi-
TRANSFORMING GROWTH FACTOR-a lized membrane preparation through a n anti-TGF-a antibody column. Q: Could the preparation have included some of the 50 amino acid form of TGF-a? A: No, i t was the larger precursor. However, in t h a t assay it takes quite a few days before the colonies are detected. Therefore, we cannot exclude the possibility that there was some cleavage of the precursor during the assay. Q: Is there any evidence that the precursor can be processed during contact with the EGF receptor? A: I do not know. We have tried to look a t a number of things relating to this question. For example, is the precursor internalized? We were quite unsuccessful in determining this. I do not think it is internalized as a complex with the receptor, but I cannot prove this. Q: Have you engineered the precursor in such a way as to add a polypeptide extension to the C-terminal end of the 50 amino acid form of TGF-a that could not be cleaved? In this way you could be sure that the protein remained in the membrane during your studies of the activity of the membrane-bound precursor. A: That would be another approach. On the other hand, I think that we did everything possible in our studies to exclude the possibility that the 50 amino acid form of TGF-a was released. We did a number of control experiments t h a t I did not mention, both with the co-cultures and with the single cell cultures of TGFa-expressing cells. We also used a very sensitive ELISA to try to look for leakage of TGF-a into the medium. Leakage from the membrane was a possibility, especially considering the results reported for the EGF precursor. We could not find any leakage of any form of TGF-a into the medium. The problem is you have to come up with a negative result to prove there is no leakage, but a negative result is never very satisfactory as evidence. Q: Do you know of anything that will decrease the expression of TGF-a in psoriatic skin. A: I a m not aware of anything that can decrease TGF-a production in psoriatic skin. The problem is you really have to do these things in vivo. Q: Could you please describe your ELISA assay. A: It is a double sandwich ELISA that uses a combination of two antibodies. One antibody is a monoclonal antibody directed against the 50 amino acid form, which seems to recognize a n epitope that corresponds
9
to a combination of the first and the second loops. TGFa has to be in the native conformation. It does not recoganize the denatured TGF-a. The second antibody is polyclonal and neutralizing. We do not know what epitope(s1 i t recognizes. Q: I noticed that you have two species of the TGF-a mRNA. Do you see both species expressed in all cells and tissues? A: The major mRNA species is 4.5-4.8 kb. Very often, though, especially at the higher levels of TGF-a mRNA expression, we do find a smaller, approximately 18S, form. We are not sure what this smaller mRNA species is. However, the sequence of the 3’ untranslated region of the TGF-a mRNA contains potential polyadenylation sites that would allow for premature polyadenylation. We have no evidence for alternative splicing, but we have not really looked for this.
REFERENCES Brachmann R, Lindquist PB, Nagashima M, Kohr W, Lipari T, Napier M, Derynck R (1989): Transmembrane TGF-alpha precursors activate EGFiTGF-alpha receptors. Cell 56:691-700. Bringman TS, Lindquist PB, Derynck R (1987): Different transforming growth factor-alpha species are derived from a glycosylated and palmitoylated transmembrane precurosr. Cell 48:429-440. Coffey RJ, Derynck R, Wilcox J N , Bringman TS, Goustin AS, Moses HL, Pittelkow MR (1987): Production and auto-induction of transforming growth factor-alpha in human keratinocytes. Nature 328: 817-820. Derynck R (1988): Transforming growth factor-alpha. Cell 54593595. Derynck R, Lindquist PB, Bringman TS, Wilcox JN, Elder JT, Fisher GL, Voorhees J J , Moses HL, Pittelkow M, Coffey RJ (1989):Expression of the transforming growth factor-alpha gene in tumor cells and normal cells. Cancer Cells [Cold Springs Harbor Laboratory] , 297-301. Derynck R, Roberts AB, Winkler ME, Chen EY, Goeddel DV: (1984): Human transforming growth factor-alpha: Precursor structure and expression in E . coli. Cell 38:287-297. Elder JT, Fisher GJ, Lindquist PB, Bennett GL, Pittelkow M, Coffey R J , Ellingsworth L, Derynck R, Voorhees JJ (1989):Overexpression of TGF-alpha in psoriatic skin. Science 243:811-814. Pittelkow MR, Lindquist PB, Abraham R, Graves-Deal R, Derynck R, Coffey RJ (1989): Induction of transforming growth factor-alpha expression in human keratinocytes by phorbol esters. J Biol Chem 2645164-5171. Wilcox JN, Derynck (1988a): Localization of cells synthesizing transforming growth factor-alpha mRNA in the mouse brain. J Neurosci 8:1901-1904. Wilcox JN, Derynck R (1988b): Developmental expression of transforming growth factor-alpha and beta in mouse fetus. Mol Cell Biol 8:3415-3422.
