Protein-tyrosine phosphatase epsilon. An isoform specifically expressed in mouse mammary tumors initiated by v-Ha-ras OR neu.

Transgenic mice that overexpress v-Ha-ras, c-myc, c-neu or int-2 proto-oncogenes in the mammary epithelium develop breast tumors with morphologies that are characteristic of each initiating oncogene. Since these morphological differences reflect distinctive patterns of tumor-specific gene expression, the identification of the products of these genes might shed light on the mechanisms of transformation and/or the identity of target cells that are transformed by specific classes of oncogenes. By focusing on the tyrosine phosphorylation pathway, we have found that the transmembranal protein-tyrosine phosphatase epsilon (PTP epsilon) is highly expressed in murine mammary tumors initiated by c-neu and v-Haras, but not in mammary tumors initiated by c-myc or int-2. This difference is striking and occurs both in primary tumors and in epithelial cells cultured from them. Moreover, PTP epsilon overexpression appears to be mammary tumor-specific in that it is not found in other ras-based tumors and cell lines. These observations suggest that PTP epsilon either plays a role in ras- and neu-mediated transformation of mammary epithelium or marks mammary epithelial cells particularly susceptible to transformation by these oncogenes. Because of its distinctive expression in these mammary tumors, we have further characterized murine PTP epsilon, cloning and determining the complete structures of its cDNAs and showing that it is a glycoprotein that is N-glycosylated in a tissue-specific manner.

Transgenic mice that overexpress v-Ha-ras, c-myc, cneu or int-2 proto-oncogenes in the mammary epithelium develop breast tumors with morphologies that are characteristic of each initiating oncogene. Since these morphological differences reflect distinctive patterns of tumor-specific gene expression, the identification of the products of these genes might shed light on the mechanisms of transformation and/or the identity of target cells that are transformed by specific classes of oncogenes. By focusing on the tyrosine phosphorylation pathway, we have found that the transmembranal protein-tyrosine phosphatase ⑀ (PTP⑀) is highly expressed in murine mammary tumors initiated by c-neu and v-Haras, but not in mammary tumors initiated by c-myc or int-2. This difference is striking and occurs both in primary tumors and in epithelial cells cultured from them. Moreover, PTP⑀ overexpression appears to be mammary tumor-specific in that it is not found in other ras-based tumors and cell lines. These observations suggest that PTP⑀ either plays a role in ras-and neu-mediated transformation of mammary epithelium or marks mammary epithelial cells particularly susceptible to transformation by these oncogenes. Because of its distinctive expression in these mammary tumors, we have further characterized murine PTP⑀, cloning and determining the complete structures of its cDNAs and showing that it is a glycoprotein that is N-glycosylated in a tissue-specific manner.
Transgenic mice in which the expression of a specific oncogene is targeted to the mammary epithelium are useful models for studying the origin of breast cancer. In recent years we have developed several such paradigms in which expression of various oncogenes is directed primarily to the mammary gland by the action of a mouse mammary tumor virus (MMTV) 1 promoter/enhancer. Thus, mice transgenic for MMTV-c-myc, -v-Haras, -c-neu, or -int-2 develop spontaneous, generally stochastic, adenocarcinomas of the mammary gland and occasionally of other tissues as well (1)(2)(3)(4)(5)(6).
Close examination of the mammary tumors obtained from MMTV-myc, -ras, and -neu transgenic mice revealed significant correlation between the distinctive histologic appearance of each tumor and the identity of its transforming oncogene (6). This correlation can be thought of as having arisen through either of two mechanisms (6,7). One stipulates that the mammary gland contains several cell types that differ in their susceptibility to transformation by individual oncogenes; that is, each transforming oncogene targets a particular cell type that is then expanded to give rise to a distinct tumor type. An alternative hypothesis suggests that one or few mammary cell types are susceptible to oncogenic transformation, but that each oncogene drives the transformed cell along a distinct developmental pathway, imparting to it a distinctive histologic phenotype. Both mechanisms could certainly operate in whole or in part simultaneously. The existence of oncogene-specific morphology also suggests that each type of mammary tumor expresses a somewhat different cohort of genes. The characterization of these marker genes might shed light on the details of transformation. Accordingly, we undertook to define genes expressed in particular types of mammary tumors. To date we have isolated several mRNA species whose expression pattern parallels the morphological differences found between tumor types (7), 2 In general, we have found that tumors arising from ras-and neu-bearing transgenic mice fall into a common class at the molecular level and are distinguished from tumors that arise from transgenic mice bearing the c-myc transgene. The central role tyrosine phosphorylation plays in tumorigenesis has now prompted us to look specifically at genes whose products might control this process.
Tyrosine phosphorylation is a reversible process that is controlled by the opposing actions of protein-tyrosine kinases and phosphatases (PTPases) (reviewed in Refs. 8 and 9). PTPases comprise a large and complex family of related molecules that can broadly be divided into transmembranal, receptor-like molecules and non-transmembranal, intracellular ones (reviewed in Refs. 10 -12). Transmembranal PTPases generally contain two PTPase catalytic domains and are thought to be capable of responding to extracellular signals. Intracellular PTPases contain one PTPase domain flanked by protein sequences that regulate the activity or localization of the entire molecule. Dual specificity phosphatases, i.e. molecules that can dephosphorylate both phosphoserine/threonine and -tyrosine exist as well. The best characterized of these, MKP1, can dephosphorylate mitogen-activated protein kinase in vivo, thereby linking this phosphatase to the latter stages of signal transduction to the nucleus (13).
Being the antithesis of tyrosine kinases, PTPases were originally thought to suppress transformation. Indeed, inhibition of PTPase activity was shown to result in transformation of cul-tured rat kidney cells (14). This interpretation was strengthened by the ability of the cytoplasmic PTPases, TCPTP (15) and PTP1B (16), to suppress oncogenic transformation in cultured cells. PTP1B has also been shown to be up-regulated in a mammary cell line transformed by neu (17). Furthermore, overexpression of the dual specificity phosphatase MKP-1 has been shown to counteract some of the actions of activated ras, suggesting that it can function as an anti-oncogene (18). Nonetheless, overexpression of PTPases can in itself be transforming, as in the case of PTP␣ (19), suggesting that various PTPases may be linked to transformation in different ways.
In an effort to focus on the role of tyrosine phosphorylation in mammary cell transformation, we sought PTPases that showed distinctive expression patterns in each tumor type. Here we report that the transmembranal PTPase ⑀ (20) is highly expressed in mammary tumors initiated by ras or neu, but is absent or nearly absent from mammary tumors initiated by myc or int-2. Moreover, PTP⑀ expression appears to be mammary tumor-specific and is not found in ras-based tumors that develop in several other organs. In addition, the PTP⑀ protein displays a tissue-specific pattern of N-glycosylation.  (23). Amplifications were performed using Taq DNA polymerase (Boehringer Mannheim) and 100 nM of each oligo mixture. Cycle parameters were: 1 min, 94°C; 1.5 min, 37°C; 2 min, 72°C for three cycles, followed by 30 additional cycles with annealing at 42°C and a final elongation step of 10 min, 72°C. Following a second round of amplification using the same conditions, the products of the PCR amplifications were subcloned and fragments of individual PTPases were identified by sequencing. The fragment of PTP⑀ used here included nucleotides 993-1323 (Fig. 5).
RNA Analysis-RNA was isolated from cultured cells and tissues using the guanidium-isothiocyanate/CsCl gradient technique (24). Poly(A) ϩ RNA was purified by oligo(dT) column chromatography (mRNA Purification Kit, Pharmacia Biotech Inc.). Electrophoresis of RNA, blotting onto GeneScreen membranes (DuPont), probe labeling, hybridization, and washing were all done using standard procedures (25). Probes used in this work were the PTPase cDNA fragments isolated by PCR and a rat 28 S ribosomal RNA probe (26).
cDNA Library Construction, Screening, and 5Ј RACE-An oligo(dT)primed cDNA library was constructed from the AC-ras-based mammary tumor cell line AC816 in the -ZAP vector (Stratagene). A -ZAP random hexamer-primed adult mouse brain cDNA library was a generous gift from Dr. David Chan, Harvard Medical School. Both strands of a series of overlapping PTP⑀ clones isolated from these libraries were sequenced.
5Ј-RACE was performed using the 5Ј-amplifinder RACE Kit (Clontech). Briefly, poly(A) ϩ RNA from the AC-ras mammary tumor cell line AC204 was reverse-transcribed using oligomer E1225 (CTCTGCT-GCTCTTGCTCTTCT, antisense to bases 613-593 in Fig. 5A) as primer. The 3Ј end of the first-strand cDNA was ligated to the Amplifinder Anchor oligomer (Clontech) and two full rounds of PCR amplification were performed using this oligomer and oligomer PREX1 (AGG-GAGGGTCGTGGGGACG, antisense to bases 181-163 in Fig. 5A). The amplification product was subcloned and sequenced. Antibody Production, Immunoprecipitations, and Protein Blot Analysis-The peptide NKEENREKNRYPNI (amino acids 154 -167 in Fig.  5A), to which a cysteine residue was added at the amino-terminal, was coupled to keyhole limpet hemocyanin (Imject, Pierce) and was used to immunize a single rabbit (Pocono Rabbit Farm & Laboratory, Canadensis, PA). 48 h after transfection of COS7 cells with PSVK/⑀-transmembranal, cells were lysed in 1 ml/plate of cold RIPA buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris-Cl, pH 8.0) and spun at 10,000 ϫ g, 4°C for 10 min. Following preclearing with protein-A-Sepharose beads (Pharmacia), fresh beads were added and the lysate divided into two aliquots: immune serum at a final dilution of 1:150 was added to one aliquot and preimmune serum from the same rabbit and at the same dilution was added to the other. The reactions were gently shaken at 4°C for 2 h after which the beads were pelleted and washed three times in cold RIPA buffer. Bound proteins were eluted by boiling in SDS loading buffer and were analyzed on a 7.5% SDS-polyacrylamide gel followed by fluorography or blotting onto Immobilon-P membranes (DuPont). Blocking, hybridization to the anti-PTP⑀ antiserum, and detection were done using standard protocols (28). Mammary tumor cells (lines AC816, SMF, and 16MB9A) and organs were lysed in RIPA buffer and were electrophoresed, blotted, and analyzed in the same manner as described above.

Transient Expression of PTP⑀ in COS Cells
Enzymatic Deglycosylation of PTP⑀-Deglycosylation was performed essentially as described in Ref. 29, using deglycosylating enzymes and proteinase inhibitors from Boehringer Mannheim. Briefly, ϳ10 7 cells were lysed in 1 ml of buffer A (10 mM sodium phosphate, pH 7.0, 2 mM EDTA, 1% Triton X-100, 5% glycerol, 0.1% ␤-mercaptoethanol, 2 g/ml aprotinin, 200 g/ml Pefabloc, 0.5 g/ml leupeptin, and 0.7 g/ml pepstatin) and spun at 10,000 ϫ g, 4°C. Organs were homogenized in buffer A and spun as above. 50 l of extract supernatant were incubated at 37°C for 3 h to overnight with 20 milliunits of neuraminidase, 300 milliunits of endoglycosidase-F, 1.3 milliunits of endo-␣-N-acetylgalactosaminidase, or combinations thereof as indicated in the legend to Fig. 8. Reactions were stopped by adding SDS-PAGE loading buffer and boiling. The products were analyzed by SDS-PAGE followed by Western blot analysis with anti-PTP⑀ antiserum.

Identification of PTPases Expressed in Murine Mammary
Tumor Cell Lines-A preliminary screen for tumor-specific PTPases was conducted in three representative epithelial cell lines previously established from mammary tumors initiated in transgenic mice by myc, neu, or ras. Samples of first strand cDNA from these lines were amplified by PCR using pairs of degenerate oligomer mixtures targeted at conserved sequence motifs of the PTPase catalytic domain. PCR products were then subcloned and identified by sequencing. Fragments of catalytic domains from the following seven PTPases were identified: PTP␣ (30), PTP⑀ (20), PTP (31), LAR (32), MPTP H3/70.15 (33), PTPH1 (34), and musCPTP (35).
PTP⑀ mRNA Is Preferentially Expressed in neu-and ras-, but Not in myc-or int-2-based Murine Mammary Tumors-The cloned PTPase fragments were used to estimate the steadystate levels of PTPase mRNAs in mammary tumor cell lines. Each fragment was hybridized to blots containing RNA samples isolated from a panel of 18 epithelial cell lines previously established from mammary tumors initiated by neu, myc, ras, or int-2. As seen in Fig. 1, PTPase ⑀ mRNA levels varied among the cell lines in a manner that correlated with the identity of the transforming oncogene. The six other PTPases listed above were uniformly expressed in all the tumor types examined and most were not studied further (results not shown).
PTP⑀ mRNA was relatively abundant and was expressed at similar levels in all the neu-and ras-based cell lines examined. On the other hand, PTP⑀ mRNA was either absent or present at very low levels in the MMTV-myc and -int-2 cell lines (Band A, Fig. 1). This heterogeneity of expression is characteristic of the myc-based mammary tumors and has been noted previously (7). The three additional bands present in the blot depicted in Fig. 1 (Bands B, C, and D) were subsequently found to be other forms of PTP⑀ mRNA (results not shown). 3 Examination of primary tumor samples revealed that PTP⑀ mRNA was expressed preferentially in MMTV-ras and MMTV-neu mammary tumors (Fig. 2, right panel, and not shown). This last result argues strongly that differential expression of PTP⑀ mRNA is a characteristic of the original tumors and did not result from their adaptation to tissue culture conditions.
Up-regulation of PTP⑀ in ras Tumors Is Mammary Glandspecific-If up-regulation of PTP⑀ were part of a general mechanism of transformation by ras or neu, then PTP⑀ mRNA should be up-regulated in all tumors initiated by these oncogenes. To address this point, levels of PTP⑀ mRNA were examined in a number of ras-based tumors and cell lines derived from various tissues of mice transgenic for ras (2,5). Significantly lower levels of PTP⑀ RNA were found in cell lines derived from fibrosarcomas and from a uterine tumor when compared with a mammary tumor cell line (Fig. 2, left panel). Comparison of primary epithelial tumors from the salivary and mammary glands of mice transgenic for ras yielded similar results (Fig. 2, right panel). This suggests that increased levels of PTP⑀ mRNA are not seen in all epithelial tumors. Due to lack of suitable material, similar analyses were not performed in the case of neu-initiated tumors.
Two Major PTP⑀ mRNAs of Distinct Tissue Specificity Exist-Examination of poly(A) ϩ RNA from several organs revealed that the PTP⑀ mRNA species overexpressed in mammary tumors was present in brain, heart, testes, and lung (Fig.  3, Band A). Lower levels of a slightly smaller PTP⑀ mRNA were present in lymph nodes, thymus, lung (Fig. 3, Band B), and spleen (not shown). Virtually no expression of either PTP⑀ mRNA was detected in liver and kidney (not shown). There was little overlap in the expression patterns of the two mRNAs, suggesting each has a specific physiological role. As will be demonstrated below, the longer mRNA species encodes the transmembranal form of PTP⑀ (⑀-transmembranal). The smaller mRNA has been shown to encode a novel, non-trans-membranal isoform of PTP⑀ that is inducible by phorbol ester. 3 PTP⑀ mRNA Is Expressed in Mammary Glands during Pregnancy and Regression-Examination of the levels of PTP⑀ expression in normal, untransformed mammary glands revealed relatively low expression of PTP⑀ mRNA (Fig. 4). Expression was low in the virgin gland, but was somewhat increased during pregnancy. PTP⑀ mRNA was nearly undetectable in glands from lactating mice, due either to its down-regulation or dilution resulting from massive expression of milk protein mRNAs at this stage. PTP⑀ mRNA was again detected starting at the beginning of mammary gland regression. In accordance with there being distinct roles for the two PTP⑀ isoforms, glands from pregnant mice preferentially expressed ⑀-trans-3 A. Elson and P. Leder, manuscript submitted for publication. membranal mRNA while the shorter form predominated in regressing glands. In all cases, expression levels in the mammary gland and other organs were clearly well below those found in ras-and neu-based mammary tumor cell lines (Figs. 3  and 4). This indicates that the high levels of PTP⑀ mRNA in ras and neu mammary tumors, rather than the low levels in myc tumors, were the abnormal finding.
The Longer PTP⑀ mRNA Is the Murine Homologue of the Human Transmembranal ⑀ Phosphatase-In order to provide additional tools for understanding the role of PTP⑀ in tumorigenesis, we cloned the ⑀-transmembranal cDNA from libraries prepared from the AC-816 ras-based mammary tumor cell line and from adult mouse brain. Probing was initiated with the PTP⑀ cDNA fragment originally isolated by PCR. Sequence analysis of a number of overlapping PTP⑀ cDNA clones revealed that the ⑀-transmembranal cDNA is 5412 base pairs in length, in good agreement with the position of the ⑀-transmembranal mRNA signal on RNA blots (Fig. 5).
The 5Ј-untranslated region of the ⑀-transmembranal cDNA is 323 base pairs long and, as in the case of the human PTP⑀ cDNA, lacks an in-frame stop codon. The authenticity of the 5Ј end was verified by performing 5Ј-RACE on mRNA obtained from another ras-based mammary tumor. Sequencing of several independent 5Ј-RACE products revealed that the 5Ј ends of the ⑀-transmembranal cDNA were essentially as cloned from the cDNA libraries (not shown). No additional 5Ј sequences were found despite extensive screening of both libraries used here. Both items, as well as cDNA expression and protein deglycosylation studies (see below), are consistent with the 5Ј end of the ⑀-transmembranal cDNA described here being the true one.
The coding portion of the ⑀-transmembranal cDNA begins at an ATG codon which is part of a near perfect Kozak consensus sequence (36) and is conserved in the human PTP⑀ (Fig. 5A and Ref. 20). The putative product of this mRNA is a 699-amino acid-long protein that has leader, extracellular, and transmembranal domains characteristic of a transmembranal protein (Fig. 5A). Hydrophobicity and sequence analyses (37) suggests that the extracellular domain of murine PTP⑀ is 28 amino acids long. Overall, 87% of the nucleic acid sequence and 93% of amino acid residues of the murine PTP⑀ are identical to the human form described in Ref. 20.
Examination of the relatively long 3Ј-untranslated sequence of ⑀-transmembranal mRNA revealed the presence of seven repeats of the sequence ATTTA which is found in short-lived mRNAs (reviewed in Ref. 38). Two polyadenylation signals were also found, the distal of which was used in the set of clones examined here. This signal directed polyadenylation at either of two distinct sites, located at nucleotides 5405 and 5412 (Fig.  5, A and B).
Production and Characterization of Anti-PTP⑀ Polyclonal Antibodies-In order to characterize PTP⑀ expression at the protein level, polyclonal antibodies were raised against a peptide derived from the intracellular region of murine PTP⑀. Extracts of COS cells transiently transfected with the coding region of the ⑀-transmembranal cDNA served as positive controls in all the subsequent work. The anti-PTP⑀ immune serum recognized a protein of ϳ105 kDa in extracts of transfected cells in immunoprecipitation and in protein blotting experiments (Fig. 6, lane 3, and results not shown). This protein was not observed in mock-transfected cells (Fig. 6, lane 1), nor when the experiments were repeated with preimmune serum or with immune serum in the presence of an excess of the immunizing peptide (Fig. 6, lane 2; Fig. 7, right panel, and not shown). Taken together, these results show that the antiserum was capable of recognizing the ⑀-transmembranal protein. The observed size of PTP⑀ in these experiments was larger than its predicted molecular mass of ϳ80 kDa. This fact as well as its migration as a broad band indicated that the ⑀-transmembranal protein is glycosylated, as is demonstrated below.
Mammary Tumor Type-specific Expression of the PTP⑀ Protein-Protein blots prepared from homogenates of several representative mammary tumor cell lines were probed with anti-PTP⑀ serum (Fig. 7). The ⑀-transmembranal protein was detected as a broad band of ϳ105 kDa in extracts from MMTVras and MMTV-neu cell lines. In agreement with RNA data presented in Fig. 1, the ⑀-transmembranal protein was absent from the representative MMTV-myc cell line used here (Fig. 7,  left panel). This result indicates that the expression of the transmembranal form of PTP⑀ is indeed limited to ras-and neu-based murine mammary tumors. Interestingly, a protein band corresponding to the shorter, non-transmembranal isoform of PTP⑀ 3 was present at relatively low levels in extracts from the MMTV-myc cell line (Fig. 7, left panel). The particular myc cell line used here does indeed express low levels of the mRNAs of both PTP⑀ isoforms (Fig. 1, lane 10). Note that despite its presence, the ⑀-transmembranal mRNA present at low levels in this myc cell line does not give rise to mature ⑀-transmembranal protein.
Expression of Electrophoretically Distinct Forms of PTP⑀ in Brain and Lungs-We next examined the forms of the PTP⑀ protein present in normal mouse tissues. In agreement with RNA data (Fig. 3), the ⑀-transmembranal protein was detected as a broad band in protein extracts of brain and lung (Fig. 7, left  panel). Interestingly, ⑀-transmembranal protein from lung and brain were of distinct sizes. An additional broad band corresponding to a protein of ϳ140 -160 kDa was specifically recognized by the anti-PTP⑀ serum in both of these tissues (Fig. 7,  left panel, upper-most band). This band was not as prominent in the mammary tumor cell lines. Cellular fractionation experiments reveal that this protein is membrane-associated (not shown), and its migration as a broad band suggests it is glycosylated. This protein may represent yet another form of the ⑀-transmembranal protein or a cross-reactive form of another protein.
The PTP⑀ Protein Is N-Glycosylated in a Tissue-specific Manner-Detection of the ⑀-transmembranal protein as a broad band of larger than expected size in protein blots suggested that it is glycosylated. In order to address this point protein extracts of a ras-based mammary tumor cell line were treated with a variety of glycosidases (Fig. 8). Treatment with neuraminidase or with endo-␣-N-acetylgalactosaminidase did not appear to significantly affect the mobility of the ⑀-transmem- branal protein (Fig. 8, lanes 2 and 5). On the other hand, removal of N-linked carbohydrates by N-glycosidase-F resulted in a significant reduction in the apparent size of the molecule (Fig. 8, lanes 3 and 4). Most of the protein was now found in two sharp, closely spaced bands of ϳ80 kDa, in good agreement with the predicted mass of the ⑀-transmembranal protein. The same results were obtained when the extracts were treated with a mixture of all three deglycosylating enzymes (Fig. 8,  lane 6). The heavier ϳ140-kDa band present in the extracts was virtually unaffected by neuraminidase and endo-␣-Nacetylgalactosaminidase; it exhibited only a slight decrease in size after treatment with N-glycosidase-F. We believe that it is glycosylated in a way that is resistant to the actions of the deglycosylating enzymes used here.
Similar experiments were repeated with protein extracts from lung and brain (Fig. 8, right panel). As in the case of the ras mammary tumor, removal of N-linked carbohydrates by N-glycosidase-F had the most significant effect on the mobility of the ⑀-transmembranal protein. Note that the significant differences in the apparent molecular mass of the glycosylated ⑀-transmembranal proteins of lung and brain disappeared after deglycosylation (compare Fig. 7, left panel, with Fig. 8, lanes 10  and 12)). This indicates that the differences in the electrophoretic mobility of the ⑀-transmembranal proteins of lung and brain reflect tissue-specific glycosylation patterns. Note also that the apparent masses of the deglycosylated forms of the ⑀-transmembranal protein from lung and brain were slightly different from that of the mammary tumor (Fig. 8, lanes 10 -12). This raises the possibility that other post-translational modifications, such as phosphorylation, are still present on the deglycosylated ⑀-transmembranal protein. DISCUSSION Our results indicate that there is a dramatic increase in expression of PTP⑀ in ras-and neu-based mammary tumors, both at the mRNA and protein levels. Mammary tumors initiated by these two oncogenes in fact form a class that specifically expresses a cohort of marker genes which are underexpressed or silent in mammary tumors initiated by myc and int-2 (7). 2 This finding can be accounted for by assuming that the ras and neu oncogenes, acting in the context of the mammary epithelial cell, share a common transformation mechanism and/or transform a common susceptible target cell. Interestingly, neu and ras have been suggested to act through a common pathway (7, 39 -41) which may serve as the basis for up-regulating PTP⑀ and other marker genes.
The possibility that PTP⑀ is an invariant link in the ras transformation pathway appears less likely in view of the fact that relatively low levels of PTP⑀ mRNA are found in several ras-based tumors that arise in non-mammary tissues, including some of epithelial origin. Thus it seems that overexpression of PTP⑀ in ras-based tumors is at least partially specific to mammary tumors and is not part of a general mechanism by which ras and neu transform. Nonetheless, it remains possible that PTP⑀ is part of a more restricted mammary-specific transformation pathway.
Bearing this point in mind, it is important to note that the isoform of PTP⑀ up-regulated in these mammary tumor cells is transmembranal and therefore can be conceptually integrated into the mechanisms by which neu and other tyrosine kinases act. A similar integration is possible in the case of ras, whose association with the cellular membrane is critical for its transforming ability (42,43). Moreover, ras influences the expression of genes whose products are thought to act upstream of itself in signal transduction pathways. For example, ras-mediated transformation has recently been shown to affect expression of the transcription factor AP2 (44) which in turn regulates neu expression (45).
Examination of the extracellular domain of transmembranal PTP⑀ protein reveals two sites for N-linked glycosylation and a number of serine and threonine residues that could be substrates for O-linked glycosylation. Indeed, the transmembranal isoform of PTP⑀ is present in various cell types as an N-linked glycoprotein of 100 -110 kDa that is analogous to a 100-kDa form of the closely-related ␣ phosphatase which is also predominantly N-glycosylated (29).
In a way that creates an opportunity for differential function, each tissue or tumor examined displays a unique pattern of PTP⑀ glycosylation. Tissue-specific glycosylation arises from differences in the glycosylating capabilities of tissues rather than from a particular property of the glycosylated protein (46). Nonetheless, the precise glycosylation pattern of PTP⑀ could affect its function by, for example, influencing proteins that interact with its extracellular domain. The electrophoretic mobilities of PTP⑀ from the mammary tumors and from normal mammary tissue are identical (not shown), indicating that their glycosylation patterns are similar. The precise pattern of glycosylation of PTP⑀ in the mammary tumors may therefore be mammary gland-rather than tumor-specific.
Finally, it is worth noting that the structurally related PTPase, PTP␣, is the only PTPase known to have transforming ability when overexpressed (19). PTP⑀ and PTP␣ are both type IV PTPases and are the only known members of this class (10,12). Given that PTP␣ displays the highest degree of sequence and structural similarity to PTP⑀, it is possible that PTP⑀ may share some of the transforming abilities of PTP␣ and that up-regulation of PTP⑀ in mammary tumors is indeed linked to the transformation process. Subsequent efforts, involving the creation of transgenic mice, will be directed toward testing this possibility.