Mouse Models of Tumor Development in Neurofibromatosis Type 1

See allHide authors and affiliations

Science  10 Dec 1999:
Vol. 286, Issue 5447, pp. 2172-2176
DOI: 10.1126/science.286.5447.2172


Neurofibromatosis type 1 (NF1) is a prevalent familial cancer syndrome resulting from germ line mutations in the NF1 tumor suppressor gene. Hallmark features of the disease are the development of benign peripheral nerve sheath tumors (neurofibromas), which can progress to malignancy. Unlike humans, mice that are heterozygous for a mutation in Nf1 do not develop neurofibromas. However, as described here, chimeric mice composed in part ofNf1 −/− cells do, which demonstrates that loss of the wild-type Nf1 allele is rate-limiting in tumor formation. In addition, mice that carry linked germ line mutations inNf1 and p53 develop malignant peripheral nerve sheath tumors (MPNSTs), which supports a cooperative and causal role for p53 mutations in MPNST development. These two mouse models provide the means to address fundamental aspects of disease development and to test therapeutic strategies.

Neurofibromatosis type I (NF1) affects about 1 in 3500 individuals worldwide (1). The hallmark clinical feature of the disease is development of multiple benign neurofibromas, which can be debilitating, severely disfiguring, and, in some patients, progress to malignancy (2). NF1 patients are also predisposed to developing optic pathway gliomas, pheochromocytomas, and myeloid leukemia as well as several symptoms unrelated to cancer (3).

The NF1-encoded protein neurofibromin is a member of the GTPase-activating protein (GAP) family that includes mammalian p120GAP and the yeast IRA proteins (4). Like p120GAP, neurofibromin can stimulate the GTPase activity of Ras in vitro and in vivo (5). Because most mutations in the NF1 gene in patients are predicted to result in loss of function, deregulation of Ras-mediated signaling is likely to contribute to the pathology of NF1 (6).

Although NF1 appears to be a classic tumor suppressor gene, the molecular mechanism underlying tumor development in NF1 has been obscure. Although second hit mutations affecting the inherited wild-type NF1 allele have been clearly identified in the myeloid leukemias and pheochromocytomas in NF1 patients (7), such mutations have been reported for only a small number of neurofibromas (8). The difficulty in detecting mutations may be due in part to the complex nature of these lesions, which are composed of multiple cell types, not all of which are expected to develop a second mutation (2). However, it also has been suggested that NF1 heterozygosity may be sufficient for development of benign neurofibromas (haplo-insufficiency), with full loss of NF1 function being restricted to the progression to MPNSTs (9).

Heterozygous mutant (Nf1 +/ ) animals are predisposed to a number of tumor types; however, they do not develop peripheral nerve sheath tumors or other characteristic symptoms of human NF1 (10, 11). To test the possibility that a mutation in the wild-type Nf1 allele is required and rate-limiting in the formation of neurofibromas inNf1 +/ mice, we generated chimeric mice that were partially composed of Nf1 −/−cells. Germ line homozygosity for a Nf1 mutation (Nf1 −/−) results in embryonic lethality at about day 14 of gestation (10, 11). We also developed a model of MPNST formation by generating mice with combined mutations inNf1 and p53. These two models are described below.

We created two Nf1-deficient embryonic stem cell (ES) lines by successive rounds of gene targeting and injected the cells into C57BL/6 blastocysts to generate chimeric mice (12). We analyzed 18 chimeric adults, which fell into three phenotypic classes, in this study (13). The subset of chimeras (n = 4) that exhibited the highest degree of chimerism died by 1 month of age of unknown causes. The two animals that exhibited the lowest degree of chimerism (less than 15% by coat color) lived a typical life span and no unusual pathology was observed upon necropsy. Most animals (12/18) fell into the third phenotypic class: they exhibited a moderate degree of chimerism, frequent myelodysplasia, and progressive neuromotor defects. The life span of these animals varied from 2 to 26 months. Histological analysis of this subset of mice revealed the presence of neurofibromas in every animal (14). We detected multiple tumors (10 to 100 per mouse), which usually emanated from the dorsal root ganglia (Fig. 1, A and B) or peripheral nerves in the limbs (Fig. 1C). Notably, only plexiform neurofibromas (those growing along the plexus of internal peripheral nerves) were observed; dermal neurofibromas, which are more common in NF1 patients, were not detected. Histologically, the lesions exhibited the classic features of human neurofibromas (15). Most tumors were not visible macroscopically; thus, the number of tumors per animal was probably underestimated. In two chimeras, however, tumors on the trigeminal nerve and in the tongue were visible upon dissection.

Figure 1

Histological analysis of neurofibromas fromNf1 −/−:Nf1 +/+ chimeras. (A and B) Sections through the spinal cord (SC) and dorsal root (DR) or limb muscle (C) from differentNf1 −/−:Nf1 +/+ chimeric mice were stained with H+E. (B) Higher magnification of (A). Arrows in (A) and (B) indicate the neoplastic region. The entire field is composed of neoplastic tissue in (C). (D) Tissues fromNf1 +/+:Nf1 −/−;ROSA-26 +/ mice were stained with X-Gal as described in (31). Sections of muscle (E) and tongue (F) were stained with H+E. A nerve (N) is seen in the center of the neurofibroma in (E). The neurofibroma in the tongue shown in (F) is multinodular and a portion of it identified by the arrow is also shown in (H). (G andH) Sections adjacent to (E) and (F) (higher magnification) were stained with an antibody recognizing the S100 protein.

The contribution of Nf1 −/− cells in the neurofibromas in these mice was assessed by usingNf1 −/− ES cells containing a β-galactosidase (β-gal) expressing transgene (16). Five chimeras were produced with these cells. Two of the three highly chimeric animals developed multiple neurofibromas. In all cases, near uniform β-gal expression was observed in the tumors, demonstrating extensive contribution of Nf1 −/− cells (Fig. 1D). We have not detected significant recruitment of wild-type cells into these lesions as might have been expected given the multicellular nature of human neurofibromas (see below). However, the sensitivity of 5-bromo-4-chloro-3-indolyl β-d-galactopyranoside (X-Gal) staining is not sufficient to rule out a low-level contribution of wild-type cells in this model.

Human neurofibromas contain a variety of cell types found in normal peripheral nerve including Schwann cells, perineurial cells, fibroblasts, and neurons (15). We used electron microscopy (EM) to determine the cellular constituents of the lesions in this mouse model. Figure 2A illustrates the overall ultrastructural appearance of a neurofibroma in this model. As in human neurofibromas Schwann cells were the most common cell type present (Fig. 2, B and C). Importantly, Schwann cells (or their precursors) are believed to be the initiating cell type [the cell type that undergoes loss of heterozygosity (LOH)] in human tumors. Cells exhibiting features of perineurial cells were also observed by EM (Fig. 2D). Taken together with the histological analysis, these data demonstrate a close similarity between the lesions inNf1 −/−:Nf1 +/+chimeras and human neurofibromas.

Figure 2

Electron micrographs of murine neurofibromas. EM was performed on fixed tissue stained with lead citrate. (A) Low-power magnification of a neurofibroma containing myelinated axons (N), collagen bundles (C), and dissociated Schwann cells identified by arrows. Bar = 10 μm. Dissociated Schwann cells exhibit branching cytoplasmic processes (B) and surround collagen bundles (C). Note the continuous basal lamina (arrow) characteristic of Schwann cells. Bars = 1 μm. (D) Perineurial cells aligned in parallel arrays containing multiple pinocytotic vesicles and a basal lamina were also observed in murine neurofibromas. Cells exhibiting either discontinuous or continuous basal lamina (arrows) were observed in these tumors. Bar = 1 μm.

These lesions were further characterized by immunohistochemistry with an antibody recognizing S100, a protein expressed in mature Schwann cells and in most human neurofibromas and MPNSTs (17). As illustrated by the tumor shown in Fig. 1, E and G, S100-positive Schwann cells were observed exclusively associated with entrapped nerves (center) and were not found in the surrounding neoplastic tissue. In larger tumors in which the original nerve was disrupted (for example, Fig. 1, F and H), sparse S100 staining, associated with nerve remnants in the center of the lesion, were occasionally observed (Fig. 1H, right). In all cases (n= 25), only minimal S100 staining was observed in the lesion itself. The presence of S100-negative Schwann cells in the neurofibromas may be explained by their development within this model system, in whichNf1 −/− cells are introduced at an early developmental stage (equivalent to E3.5). It is possible that the absence of S100 expression in these neurofibromas reflects a requirement for Nf1 function in the differentiation of neural crest–derived precursors to S100-expressing Schwann cells in vivo. Indeed, neurofibromin is expressed early in Schwann cell differentiation in the mouse, 2 to 3 days before the onset of S100 expression (18). In contrast, in NF1 patients, neurofibromas arise from cells that are initially heterozygous for an NF1mutation. Cells within the human neurofibromas would be expected to become NF1 −/−, but in many cases this might occur after the onset of S100 expression, accounting for the high percentage of S100-positive neurofibromas in humans.

We next addressed the development of MPNSTs. Because these malignant tumors are often found emanating from primary plexiform benign lesions (2, 15), it is thought that additional (probably genetic) events are involved in the progression to malignancy. In fact, mutations in the p53 tumor suppressor gene have been detected in human MPNSTs and therefore have been implicated in this progression step (19). As a means of creating a model for MPNST formation and establishing a causal role for p53mutations in the malignant lesions, mice with germ line mutations inNf1 and p53 were generated on a mixed (129/sv × C57BL/6) genetic background. Because Nf1 andp53 are linked on chromosome 11 in mice (20), we initially generated animals carrying the Nf1 andp53 mutations on opposite chromosomes (NP trans) by crossingNf1 +/ andp53 +/ mice. We also crossed these NP trans animals to wild-type animals to generate mice with both mutations on chromosome 11 (NP cis), which arise after meiotic recombination. Because complete chromosomal loss is the most common mutational mechanism by which second-hit mutations occur in mice (21), we expected most of these NP trans mice to undergo LOH at either the Nf1 or the p53 locus (but not both); we also thought this population of mice would exhibit a tumor phenotype reminiscent of the Nf1 +/ and p53 +/ parental animals (10, 22). In the NP cis mice, however, a chromosomal loss would be expected to result in LOH at both tumor suppressor loci simultaneously (a phenomenon we term co-LOH), resulting in cells deficient for both Nf1 and p53.

As expected, NP trans animals succumbed to tumors more rapidly than mice heterozygous for a single mutation; they survived an average of 10 months and developed tumors similar to those found in mice with either single mutation (Fig. 3A). Southern blot analysis of tumor DNA (n = 6) revealed LOH of either the wild-type Nf1 locus or the wild-type p53locus (23). In contrast, mice carrying Nf1 andp53 mutations in cis survived an average of only 5 months and exhibited a significant increase in the percentage of soft tissue sarcomas compared with mice of other genotypes (Nf1 +/ , 5%; p53, 57%; NP trans, 36%; NP cis, 81%). Furthermore, although NP trans animals exclusively develop osteo-, fibro-, rhabdomyo-, and hemangiosarcomas, about 30% of tumors from the NP cis animals stained positively for S100 and exhibited classic histological features of MPNSTs (n = 28) (Fig. 3B). The percentage of MPNSTs identified by these criteria is likely to be an underestimate because only 50% of human MPNSTs are S100-positive and due to the characteristic heterogeneity of this tumor type (15, 24). Importantly, loss of both wild-type alleles in tumors from the NP cis mice was consistently observed (n = 12) (23), which suggests that the loss of both genes cooperates in formation of these lesions in mice and supports a causal role for p53 mutations in development of MPNSTs in NF1 patients. The presence of S100-positive cells in the tumors from the NP cis mice contrasts with the analysis of neurofibromas in the model described above, where the lesions were universally S100. Although this pattern was unexpected, it supports the hypothesis that the timing ofNf1 deficiency may be critical for expression of the S100 marker. Specifically, in contrast to the neurofibroma model, in the MPNSTs Nf1 −/− cells arise fromNf1 +/ cells, most likely at a later stage of gestation or in the adult mouse.

Figure 3

Analysis ofNf1 +/ :p53 +/ mice. (A) Survival curve of mice carrying Nf1 andp53 mutations on the same (NP cis) or opposing (NP trans) chromosomes. The Nf1 +/ andp53 +/ curves are identical to those published previously. We studied 51 NP trans animals, 34 NP cis animals, 19 wild-type animals, 37Nf1 +/ animals, and 41p53 +/ animals. (B) Sections of an MPNST arising in the rear flank of this animal were stained with H+E (upper) or S100 antibodies (lower). The nuclear S100 staining observed is also a characteristic of human MPNSTs and neurofibromas. MPNSTs were typically observed invading muscle (M) and were associated with nerves (N). Right panels are magnified views of left panels.

In summary, our data on the Nf1 −/−:Nf1 +/+ chimeras indicate that complete loss ofNf1 is an obligate step in neurofibroma development and suggest that the prevalence of neurofibromin-deficient cells in the developing nerve is the rate-limiting factor in formation of this tumor in the mouse. It remains unclear why fewer of these cells arise inNf1 +/ mice than in human NF1 patients. Possible explanations include interspecies differences in lifespan, target cell number, or proliferative properties or interspecies differences in the mutability of the NF1 locus. We have also demonstrated that mutations in Nf1 andp53 cooperate in the development of MPNSTs, which supports a causal role for p53 mutations in their formation. The development of these lesions might call into question the conclusion that a low rate of Nf1 LOH limits the development of neurofibromas inNf1 +/ mice (see above). We hypothesize that the concomitant loss of p53 andNf1 function that occurs in development of MPNSTs in the NP cis mice allows for the outgrowth of cells that otherwise would have undergone growth arrest or apoptosis were Nf1 mutated alone. Previous work has shown that dysregulation of the Ras pathway in rat Schwann cells leads to growth arrest that can be overcome by inhibition of p53 function (25); oncogenic ras alleles can also induce p53-dependent growth arrest or apoptosis in human and mouse fibroblasts (26).

The simultaneous homozygosing of linked mutations by the process described here as co-LOH could have more general importance in tumor development, which could be strongly influenced either positively or negatively by the arrangement of linked germ line or somatically acquired mutations. This effect may be more pronounced in mice than in humans, where LOH events are typically subchromosomal. Still, closely linked mutations that act synergistically or antagonistically could strongly affect the process of human tumor development through this mechanism.

Finally, we anticipate that both models will be critical in further characterization of tumorigenic mechanisms in NF1 and in the evaluation of potential therapies, such as compounds designed to inhibit Ras signaling, such as farnesyl transferase inhibitors (27).

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: tjacks{at}


View Abstract

Navigate This Article