Epidermal Growth Factor Receptor Signaling Pathways Are Associated with Tumorigenesis in the Nf1:p53 Mouse Tumor Model

NF1² is a dominantly inherited human disease affecting ∼1:3500 individuals worldwide (1, 2). The predominant lesion associated with NF1 is the development of benign neurofibromas of the peripheral nervous system, comprised primarily of Schwann cells (60–80%) but also containing fibroblasts and mast cells among others (3, 4). In some patients these benign neurofibromas progress to form MPNSTs or MTTs, and NF1 patients are also at increased risk for the development of pheochromocytomas and certain forms of myeloid leukemia (5). All of the known NF1 phenotypes result from the inheritance or appearance of a mutant allele of the NF1 gene (6). NF1 encodes a large protein designated neurofibromin, which functions as a negative regulator of the cellular Ras proteins (7). The mutational inactivation of NF1 results in the hyperactivation of Ras, leading to alterations in cell growth and differentiation that result in the abnormal growths that characterize NF1 disease.

The properties of NF1 disease and the role of mutational inactivation of both NF1 alleles in the disease process place NF1 in the class of tumor suppressor genes, as originally postulated by Knudson (8). Another member of the tumor suppressor gene family is the p53 gene, which is the most frequently mutated gene in human cancers (9). Interestingly, mutation or loss of p53 has been identified as a common alteration in NF1-related MPNSTs and MTTs (10, 11, 12). Furthermore, the NF1 and p53 genes are linked on the same chromosome in both humans and mice, raising the possibility that coordinate loss of these two tumor suppressors may contribute to the development of certain malignancies, such as those associated with NF1 disease.

The development of most human malignancies is thought to involve multiple genetic and/or epigenetic events, including the loss of tumor suppressors such as NF1 and p53, but also frequently involving the mutational activation, inappropriate (ectopic) expression, or overexpression of another class of genes known as proto-oncogenes. These genes generally encode proteins that function in signal transduction pathways that activate cellular growth, and their overexpression or mutational activation leads to the constitutive activation of growth-signaling pathways (13). One class of alterations involving proto-oncogenes that has been implicated in the development of NF1-related lesions is the signaling pathways mediated by growth factor RTK (14). In the case of myeloid leukemia, both cells from NF1 patients and explanted cells from mice lacking an intact Nf1 gene exhibit hypersensitivity to the RTK ligand granulocyte-macrophage colony stimulating factor (15, 16, 17). Similarly, mast cells from Nf1-deficient mouse embryos have been shown to exhibit altered signaling involving the c-kit (steel factor) RTK (18). Finally, recent observations from our laboratory have identified an involvement of the EGFR in the growth of human NF1 (and non-NF1) MPNSTs, in a chemically induced rat Schwannoma-derived cell line, and in transformed Schwann cells derived from mouse embryos lacking an intact copy of Nf1 (19). The involvement of the EGFR was surprising, as normal Schwann cells develop from the neural crest lineage and lack EGFR expression (20). Their growth is regulated by contact with neurons, cyclic AMP, and by the EGF-related heregulin/neuregulins, including GGF, which activate other members of EGFR family, such as heterodimers of erbB2 and erbB3 (21, 22, 23). Our results suggested that expression of the EGFR might play an important role in NF1 tumorigenesis.

Animal models of NF1 have been developed to further the understanding of disease mechanisms and to provide a setting for the evaluation of potential therapeutic strategies. Several years ago, the creation and characterization of mouse lines with targeted disruption of Nf1 were reported (24, 25). The usefulness of these mouse models was limited by the fact that homozygous mutant (−/−) embryos are inviable beyond day E13.5 and that heterozygous (+/−) mice fail to develop neurofibromas or MPNSTs (although an increase in pheochromocytomas and myeloid leukemia was observed). Recently, a second-generation mouse model for NF1 was developed by brother-sister mating of doubly heterozygous mice harboring null Nf1 and p53 alleles linked in cis (26, 27). These mice developed soft tissue sarcomas between 3 and 7 months of age, associated with concomitant loss of the normal Nf1 and p53 alleles. Furthermore, the cells comprising these tumors were found to express phenotypic traits characteristic of neural crest derivatives and human NF1 malignancies. Many of these tumors were grown in vitro to produce established cell lines (27). Here, we have investigated 24 of the Nf1:p53 tumor cell lines for their expression of EGFR family members, growth response to treatment with EGF, and sensitivity to growth inhibition by agents that antagonize the EGFR or its downstream signaling pathways. Our results suggest that expression of the EGFR is a consistent feature of tumorigenesis in this system and that inhibition of the EGFR or of the PI3k pathway may represent a novel approach in the treatment of NF1 neural crest-derived malignancies.

Nf1:p53 tumor-derived cell lines were isolated as described (27) by the following procedure: overlying skin and hair were removed from the tumor mass, which was then immersed briefly in sterile solution of Dulbecco’s PBS and penicillin/streptomycin (Life Technologies, Inc., Grand Island, NY). A few small pieces of the tumor mass were minced in DMEM (supplemented with 10% heat-inactivated FCS, penicillin/streptomycin, and nonessential amino acids, all from Life Technologies, Inc.) using watchmaker forceps and fine curved scissors. Tumor pieces were allowed to attach to 60-mm tissue culture plastic dishes, and clonal cell lines were established from tumor outgrowths after 4–6 passages.

Nf1:p53 mouse tumor cell lines were cultured in DMEM with 10% FBS. The monolayer cell growth assays were carried out as described (Ref. 19; Fig. 3). Cells were plated at 2 × 10⁴/well in six-well plates and grown in serum-limiting condition (0.05–0.3% FBS), in the presence or absence of the indicated concentration of inhibitors: tyrphostin AG1478, PI3k inhibitor LY 294002, or MEK inhibitor U0126. All of the inhibitors were obtained from BIOMOL Research Laboratories, Inc. (Plymouth Meeting, PA). Agar colony formation assays were carried out as described (28). Agar was obtained from Sigma Chemical Corp. (St. Louis, MO).

Cells were grown until confluent and lysed in lysis buffer [20 mm Tris-HCl (pH 7.5), 150 mm NaCl, 5 mm MgCl₂, 2 mm EDTA, 1% NP40, 0.5% sodium deoxycholate, 0.05% SDS, 1 mm sodium orthovanadate, 10 mm sodium PP_i, and 1 mm DTT]. Portions of the lysates containing 50 μg (Fig. 2, line 6) or 100 μg (Fig. 1) of cell protein were subjected to analysis by SDS-PAGE, and immunoblotting was carried out as described (19) using antibodies from Santa Cruz Biotechnology (Santa Cruz, CA). For the EGFR, antibody sc03 was used; for erbB2, sc284; for erb3, sc285; and for erbB4, sc283. The antibody against phospho-Akt (Ser 473) and against Akt were purchased from New England Biolabs, Inc. (Beverly, MA). The antibodies against activated MAPK and against ERK1/2 were purchased from Promega Corp. (Madison, WI).

The pattern of expression of the erbB family of receptors and EGFR ligands in the cell lines was analyzed by RT-PCR analysis using gene-specific primers. Seventeen different lines were analyzed, as depicted in Fig. 6 and in Table 1. Total cellular RNA was extracted from cultured cells or mouse tissues according to standard procedures using cesium chloride gradients. Total RNA (0.2–2 μg) was used for cDNA synthesis using a mixture of hexamers and oligodeoxythymidylic acid, and Superscript II Reverse Transcriptase (Life Technologies, Inc.). After this, each reaction was purified using the PCR-clean protocol (Qiagen Inc., Valencia, CA). cDNAs were subsequently amplified with Taq polymerase in the presence of 10 pmol of transcript-specific primers. The PCR conditions were established such that amplification of the cDNAs was dependent on the concentration of the corresponding mRNA. This was achieved by testing the PCR reaction using different annealing temperatures (42–55°C), two different numbers of cycles (32 and 35), and by varying the concentration of template until linear conditions were achieved. One-tenth of the PCR products were separated in a 1.2% agarose gel and stained with ethidium bromide. In addition, amplified cDNAs were blotted onto Nytran membranes (Schleicher & Schuell) and detected by hybridization with a randomly primed [³²P]dCTP gene-specific probe generated by RT-PCR. The sequences of the primers used for RT-PCR in Fig. 6 are as follows: Amphiregulin, 5′: TCAGTGCTGTTGCTGCTGGTCT/3′: ACAACTGGGCATCTGGAACCAT; HB-EGF, 5′: GACTCTGAACAGACAGACGAA/3′: ATGGGAGACCAAGTGCTGATGA; TGF-A, 5′: TATCCTGTTAGCTGTGTGCCAG/3′: TAAGACACGCATCCTGACGACA; and Epiregulin, 5′: GTCTAGGTTCCCACCTTCTACA/3′: CTGAAGAGACATCTTGTCCAGG; GAPDH, 5′: TCTTCTTGTGCAGTGCCAGC/3′: CAGTAGACTCCACGACATAC.

To assess the expression of EGFR family members in tumor cell lines derived from Nf1:p53 mice, a series of 24 lines generated as described (27) were grown until just confluent and lysed, and portions of the lysates were subjected to immunoblotting with antisera specific for EGFR, erbB2/neu, erbB3, or erbB4 (Fig. 1). As a positive control, an equal portion of a lysate from the human embryonic kidney 293, which we have found to express all of these proteins (19), was compared (Fig. 1, far right lane). Twenty-three of 24 tumor cell lines examined were found to express the EGFR. Fourteen of these 22 (Fig. 1, lines 1, 3–6, 8, 10–12, 14, 15, and 17–19) expressed high levels of the EGFR, whereas 8 others (Fig. 1, lines 7, 9, 20–24) expressed significant but more moderate levels, and one (Fig. 1, line 13) expressed only a trace amount. Only one of the lines (Fig. 1, line 2) expressed no detectable EGFR. Consistent with what we have observed in our previous studies, all of the 24 lines expressed robust levels of the erbB2 product (Fig. 1). In contrast, when we examined the expression of erbB3, which normally is expressed in Schwann cells and pairs with erbB2 to form a receptor for GGF (20, 23), only 6 of the 24 lines were found to express detectable erbB3, and only 2 of the 6 exhibited robust expression (Fig. 1). None of the Nf1:p53 tumor cell lines expressed significant levels of erbB4. This negative result was not surprising as Schwann cells and other neural crest derivatives lack expression of erbB4 (29). We conclude that virtually all of the Nf1:p53 mouse tumor-derived cell lines express the EGFR and that only a limited subset expresses erbB3, whereas all express erbB2. On the basis of these results, these lines strongly resemble the human NF1 MPNST-derived cell lines we have examined previously (19).

As a complement to these studies, additional analyses were performed to determine the expression of mRNA encoding EGFR family members in a series of 17 tumor cell lines, 13 of which were members of the group examined in Fig. 1. Semiquantitative RT-PCR was carried out, and 17 of 17 were found to express EGFR mRNA as well as erbB2 mRNA. erbB3 mRNA expression was observed in 10 of 17 of the lines, whereas erbB4 was not expressed in these tumor-derived cell lines. On the basis of the perfect correlation between EGFR mRNA and protein expression seen for the other lines, these data suggest that the 4 additional lines also are likely express the EGFR protein, raising the total to 27 of 28 of the entire group of cell lines examined.

Treatment of cells expressing the EGFR with a ligand that can activate this receptor is known to activate a variety of mitogenic signaling pathways (30, 31). To assess the activation of specific pathways after the addition of EGF to the Nf1:p53 mouse tumor-derived cell lines, we treated cells grown in vitro with EGF or left them unstimulated. We then prepared lysates from the cells and subjected the lysates to immunoblotting with antibodies specific for activated forms of the AKT protein (32), which is regulated by products produced by PI3k and antibodies to activated MAPK, which lies downstream of the MEK protein kinase (Fig. 2, A and B, top panels). As a control for protein expression and gel loading, immunoblotting was also carried out with antibodies that recognize the total cellular pool of AKT or MAPK (Fig. 2, A and B, bottom panels). These analyses demonstrated that treatment of the cell lines with EGF resulted in the activation of AKT and MAPK in virtually every line tested. One exception was line 2, which was the only line lacking detectable EGFR expression (Fig. 1). Line 13, which exhibited only a trace expression of the EGFR, nevertheless displayed clear MAPK activation after EGF treatment (Fig. 2), indicating that this line expresses functional levels of the EGFR. Addition of EGF to line 22, which expresses EGFR (Fig. 1), failed to exhibit significant activation of AKT or MAPK; the reason for this lack of activation is not known. Lines 9, 13, and 16, which exhibited clear EGF-induced MAPK activation, had only minimal activation of AKT, as judged by the immunoblot (Fig. 2). We conclude that activation of the EGFR in the Nf1:p53 mouse tumor-derived cell lines results in the activation of downstream signaling pathways that are associated with mitogenesis and tumorigenesis.

The analyses demonstrating that treatment of the Nf1:p53 tumor-derived lines with EGF led to the activation of mitogenic signaling pathway suggested that EGF might also stimulate the growth of these cells in vitro. To test this idea, a representative group of the lines was grown in limiting amounts of FBS (0.05–0.3%) in the presence or absence of EGF (10 ng/ml), and the number of cells per well was determined every 2 days (Fig. 3; data not shown). The lines tested were numbers 1, 3, 4, 5, 8, 10, and 11 (see also Table 1). We found that 8 of 8 of the lines grew poorly or not at all without exogenous EGF, whereas inclusion of EGF in each case significantly stimulated the growth of the lines. Two examples of this are shown in Fig. 3; note the increased growth of line 11 and line 1 in the presence of EGF (Fig. 3, A and B, respectively). Treatment of the same cells under identical conditions with other known growth factors such as platelet-derived growth factor and insulin failed to elicit a similar stimulation of growth (data not shown), indicating a specific sensitivity of these tumor cell lines for EGF.

To confirm and extend these observations, we tested two representative lines for their growth in the presence of EGF, and inhibitors of the EGFR, and of the MEK/MAPK and PI3k/AKT mitogenic signaling pathways. Line 11 exhibited EGF-dependent growth, and the stimulation of growth by EGF was completely blocked by treatment with the EGFR antagonist AG1478 (Fig. 3,A). This compound, a member of the tyrphostin family, is a cell-permeable inhibitor that binds tightly to the catalytic domain of the EGFR and inhibits the associated tyrosine-specific protein kinase activity of the receptor (33). Furthermore, treatment with the compound LY294002, a specific inhibitor of PI3k (34), also dramatically inhibited the growth of this line. A third inhibitor, U0126, which selectively inhibits MEK (35), had only a modest inhibitory effect on the EGF-stimulated growth of this line. A second line (number 1) was analyzed in a similar set of experiments (Fig. 3 B). Again, we found that the growth of this line was stimulated by EGF, although the growth in the absence of EGF was more robust than for line 11. Treatment with AG1478 strongly inhibited the growth of this line, as did LY294002. Unlike line 11, treatment of line 1 with the MEK inhibitor U0126 did inhibit the growth of this line. This line might display more EGF-independent, MEK-dependent growth because of other alterations in cellular signaling pathways.

As a second assay for the growth of the tumor cell lines under various conditions, we carried out assays of anchorage-independent growth in soft agar (Fig. 4). When plated in moderate levels (7%) of serum, no significant colony formation by line 11 was observed. However, the presence of 100 ng/ml EGF supported the formation of small-to-large colonies under the same conditions (Fig. 4,A). Furthermore, the presence of tyrphostin AG-1478 or PI3k inhibitor LY294002 strongly inhibited the formation of such colonies, whereas the MEK inhibitor U0126 displayed a more limited ability to inhibit colony formation (Fig. 4,A). Colony formation by line 1 under similar conditions (except with 1% serum) was also assayed (Fig. 4,B). The results obtained with this line resembled those seen in Fig. 3, as EGF stimulated the formation of agar colonies, and this effect was blocked by the addition of AG1478, LY294002, and U0126 (Fig. 4 B). We conclude from these studies that expression of the EGFR in the Nf1:p53 lines renders their growth responsive to stimulation by EGF and that inhibition of the EGFR or of downstream signaling, especially via the PI3k pathway, blocks this effect.

The results obtained with inhibitors of specific arms of downstream signaling elicited by EGFR activation led us to test the biochemical effects of these inhibitors on the activation of AKT and MAPK. Therefore, we examined the appearance of activated forms of AKT and MAPK at various times after EGF addition, after pretreatment of the cells with the LY294002 or U0126 (Fig. 5). Treatment of line 11 with EGF led to activation of both AKT and MAPK, consistent with the results of Fig. 2. In both cases, the activation was visible throughout the time course tested, from 5 min to 24 h. Pretreatment with LY294002 (Fig. 5, L Lanes) blocked the EGF-induced activation of AKT throughout the time course, whereas pretreatment with the MEK inhibitor U0126 (Fig. 5, U Lanes) blocked EGF-dependent activation of MAPK in a similar manner. Significantly, the inhibition lasted for up to 24 h, which is likely sufficient to produce the growth inhibition seen in Figs. 3 and 4. Furthermore, the expression levels of AKT and MAPK were not affected by inhibitor treatment, and no cross-inhibition of AKT by U0126 or of MAPK by LY294002 was observed (Fig. 5). When the same analysis was carried out using line 1, similar results were observed (data not shown). These results demonstrate that the biological effects produced by treatment of cells with specific inhibitor reflect the blocking of specific arms of the EGFR signaling network.

Like other growth factor receptors, activation of the EGFR is known to be strictly dependent on the stimulation of cells by a specific ligand for the receptor (36). The aberrant expression of the EGFR in cells of the Schwann/neural crest lineage described here and in our previous studies raises the question of how this receptor interacts with such ligands as tumors develop in vivo. One common mechanism that has been observed is the presence of an autocrine loop, in which cells expressing the receptor also produce one or more ligands that, when released from the cell, can in turn activate the receptor at the cell surface (reviewed in Ref. 37). To examine whether the Nf1:p53 tumor-derived lines also express specific ligands for the EGFR, we carried out RT-PCR on 17 of these lines, using oligonucleotide primers specific for different EGFR ligands (Fig. 6). Amphiregulin is a heparin-binding glycoprotein that is structurally related to EGF and acts only through binding to the EGFR, although it can induce transphosphorylation of erbB2. Sixteen of the 17 cell lines expressed abundant levels of amphiregulin (Fig. 6). We also found that 16 of the 17 cell lines expressed HB-EGF, which also binds to and causes activation of the EGFR. Additionally, 17 of the 17 cell lines were found to express epiregulin, whereas only 2 cell lines expressed TGF-α. These results demonstrate that these tumor cell lines express mRNAs encoding multiple EGF family members. In an attempt to identify potential soluble EGF-related peptides being released by the tumor cells, we collected supernatants from cultures of the cells and tested them using an a very sensitive indicator NIH3T3-derived line that overexpresses the EGFR. No activity was detected in the supernatants (data not shown). However, it may be the case that only low levels of the growth factors are produced by the cells or that they remain cell surface-bound. Thus, it is still possible that autocrine loops contribute to the formation of tumors in vivo in the Nf1:p53 mice.

Here we have demonstrated that tumor cell lines derived from compound heterozygous mice bearing disrupted alleles of Nf1 and p53 express the EGFR. The proportion of lines expressing the EGFR protein was 23 of 24, whereas an additional 4 of 4 lines examined only for the EGFR mRNA were also positive, indicating that this is virtually a universal feature of these lines. These results parallel the findings of our recent report in which 3 of 3 human NF1 MPNST-derived lines, 1 of 1 non-NF1 MPNST-derived lines, 1 of 1 rat chemically induced Schwannoma lines, and every isolate tested (n > 20) of transformed Schwann cells derived from Nf1−/− embryos expressed the EGFR (19). On the basis of these studies it is clear that expression of the EGFR in the neural crest-derived cells and Schwann cells that give rise to the tumors characteristic of human NF1 disease and animal models for NF1 is a strikingly common feature of these abnormal cells. In light of the widespread role of the EGFR in carcinogenesis in a variety of systems (36), these findings implicate its aberrant expression in the neural crest/Schwann cells as potentially playing a causative role in the formation of the transformation of cells that leads too the formation of lesions. Whereas the studies presented here focused on EGFR expression in tumor cell lines, we can rule out the possibility that EGFR expression is a secondary effect because of in vitro culture of the tumor cells. An examination of primary tumors from the Nf1:p53 mice by Northern blotting revealed that 13 of 19 (68%) tumors expressed EGFR mRNA (data not shown). Expression was variable but was particularly robust in 10 of 19 of the tumors. This is similar to our previous analyses in which we found that 7 of 7 primary human MPNST expressed the EGFR, and 9 of 9 benign neurofibromas from NF1 patients contained EGFR-positive/S-100-positive cells (19).

The development of the Nf1:p53 compound heterozygote model for NF1 has provided a clear advancement in the study of this disorder. This model reflects the documented identification of alterations in p53 in NF1-related soft tissue sarcomas (10, 11, 12). The location of p53 and Nf1 on the same chromosome in both humans and mice suggests that mechanisms such as large deletions, rearrangements, or chromosomal loss may lead to the somatic inactivation of one allele of both of these genes during tumor development. Finally, in vitro studies of primary cell transformation, including a study involving primary Schwann cells, suggest that a signal involving Ras (such as that resulting from inactivation of NF1) can cooperate with the functional inactivation of p53 to transform primary cells (38, 39, 40). In these ways the Nf1:p53 mouse model reflects the likely events leading to tumorigenesis in human NF1 patients. The vast majority of tumors identified by Vogel et al. (27) were either MTT or MPNST. Each of these tumor types arises at elevated frequency in NF1 patients, and is characterized by expression of Schwann cell and neural crest markers. Thus, it is likely that in both human NF1 patients and in this mouse model the genesis of the tumor reflects transformation of a pluripotent stem cell of the neural crest. The results described here provide direct experimental validation of another aspect of the Nf1:p53 mouse model that resembles the human situation, namely, the involvement of EGFR expression in the cells of the developing tumor. The documentation of this pathogenetic similarity between the mouse model and the human disease not only makes the Nf1:p53 mice suitable for use in tumor inhibition studies targeting the EGFR but also suggests that other changes found in the tumors of the Nf1:p53 mice will also reflect those present in human NF1 patients.

The consistent observation of EGFR expression in NF1-related tumor cells of neural crest/Schwann cell origin raises several important issues. The first is whether there are functional biological consequences for the cells because of the presence of the EGFR. These studies clearly demonstrate that treatment of the Nf1:p53 tumor cell lines with EGF enhances the growth of these cells in both monolayer cultures and in agar suspension, as we also found for human NF1 patient tumor cell lines. This suggests that EGFR expression also provides a growth advantage for these cells during tumor development in vivo. These observations then raise the issue of the nature of the growth advantage provided by EGFR expression in cells of neural crest/Schwann cell origin. One possibility is that by expressing the EGFR, the cells are susceptible to growth stimulation by ligands that bind specifically to the EGFR but not to other members of the erbB family (37). Such ligands may be available in the environment of the developing tumor, perhaps released by the nearby heterozygous (NF1 +/−) fibroblasts, which likely have some deregulation of Ras signaling (18). Consistent with this, TGF-α was originally identified as a factor released from Ras-transformed fibroblasts (41). Alternatively, the formation of autocrine signaling loops by the EGFR-expressing cells, for which we have provided evidence (Fig. 6), may obviate the need for exogenous ligands by providing continual stimulation of the EGFR in the cells of the developing tumor. The activation of Ras signaling observed in (NF1 −/−) human MPNST and (Nf1−/−) mouse Schwann cells (28, 42, 43) may stimulate the production of autocrine EGFR ligands, leading to a situation in which codependent epigenetic changes involving the EGFR and its ligands drive NF1 tumor formation. Another possibility is that the EGFR contributes to growth activation by providing an alternative set of docking sites for the binding, phosphorylation, and activation of downstream signaling molecules. Although similar in overall structure to the other erbB family members (two of which, erbB2 and erbB3, are expressed in normal Schwann cells), the EGFR does exhibit distinct kinetic properties, substrate preferences, and binding sites for substrates that enable it to direct a different signaling response from that of the other family members (29). It has been shown that the EGFR can form heterodimers with erbB2 (which was expressed in all of the lines examined here) that elicit distinct signaling responses from those in cells expressing only the EGFR (44, 45).

The identification of EGFR expression as a common feature of malignant growths associated with human NF1 disease and mouse models of this disorder suggests that therapeutic strategies designed to inhibit EGFR function may be beneficial in the treatment of these lesions. The most promising of these currently involve the use of small, cell-permeable molecules (46) and monoclonal antibodies (47). In addition, the dramatic inhibitory effect of the PI3k inhibitor on these lines suggests that this molecule may also represent a target for intervention. Indeed, we have found that LY294002 also had a potent inhibitory effect on the human 88–14 MPNST cell line.³ Taken together, these results raise the possibility that the EGFR may function in multiple ways to drive tumor development: by providing a mitogenic signal through the activation of the MEK/MAPK and PI3k pathways; and also by serving as a survival factor for a subset of neural crest cells that otherwise would be programmed for destruction through apoptotic mechanisms. In this regard, the activation of AKT signaling may be particularly relevant, as this protein is known to transmit an antiapoptotic signal (48). The relative contribution of mitogenic and survival signaling by the EGFR in this system is deserving of additional investigation.

Another issue raised by the studies described here using the Nf1:p53 mouse model system regards the natural history of the tumors: specifically, the correlation between the appearance of specific genetic and epigenetic changes such as EGFR expression; the development of hyperproliferating cells that characterize the benign lesions; and the appearance of cells of a more advanced tumorigenic phenotype that correlate with malignancy. The natural history of the malignant lesions in human NF1 patients suggests a multistep cascade that occurs over years, with the loss of both NF1 alleles required for the appearance of a benign lesion (49, 50, 51, 52, 53, 54), EGFR overexpression associated with a subset of cells in benign lesions and larger clusters of cells (or all of the cells) in MPNSTs (19), and p53 alterations in a proportion of MPNSTs (10, 11, 12). In contrast, in mouse models, EGFR expression can be detected in transformed Schwann cells that arise in embryos by day E12.5 and appears to be a nearly universal feature of the malignancies in the p53:Nf1 model that appear by 3–6 months of life. Another observation we have noted is the rarity of expression of erbB3 in the NF1-related tumors. This may reflect the possibility that Schwann cells arrested in an early stage of development may comprise the majority of the malignant cells in these tumors (55). In our earlier studies, 3 of 4 human MPNST lines examined lacked erbB3 expression, and a similar majority of the Nf1:p53 tumor cell lines examined here failed to express erbB3, although this receptor is expressed in normal Schwann cells and forms part of the functional receptor for neuregulins such as GGF (21, 23). Considering the sum of these observations, it is possible that there is a subpopulation of neural crest derivatives that appears in both human and mouse embryos bearing a disrupted allele of NF1, which lack erbB3 and instead express the EGFR. Altered signaling in these cells, perhaps in cooperation with the presence of altered fibroblasts and/or mast cells in the tumor environment, may lead to the formation of abnormal growths in synergy with the presence of alterations in p53.

We thank Doug Lowy for helpful suggestions regarding this work and for help with preparation of the manuscript. We also thank Russ Daniel for technical assistance.