A R T I C L E
Therapy-induced malignant neoplasms in Nf1 mutant mice
Richard C. Chao,1,8,11Urszula Pyzel,1Jane Fridlyand,2Yien-Ming Kuo,1,3Lewis Teel,1Jennifer Haaga,1
Alexander Borowsky,9Andrew Horvai,4Scott C. Kogan,5,6Jeannette Bonifas,1Bing Huey,6Tyler E. Jacks,10
Donna G. Albertson,5,6,7and Kevin M. Shannon1,6,*
1Department of Pediatrics, University of California, San Francisco, San Francisco, California 94143
2Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, California 94143
3Department of Medicine, University of California, San Francisco, San Francisco, California 94143
4Department of Pathology, University of California, San Francisco, San Francisco, California 94143
5Department of Laboratory Medicine, University of California, San Francisco, San Francisco, California 94143
6Comprehensive Cancer Center, University of California, San Francisco, San Francisco, California 94143
7Cancer Research Institute, University of California, San Francisco, San Francisco, California 94143
8Department of Medicine, Division of Hematology/Oncology, Department of Veterans’ Affairs Medical Center, San Francisco,
9Medical Pathology Center for Comparative Medicine, University of California, Davis, Davis, California 95616
10Center for Cancer Research, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139
11Present address: Pfizer Global Research and Development, 10578 Science Center Drive (B95), San Diego, California 92121
Therapy-induced cancers are a severe complication of genotoxic therapies. We used heterozygous Nf1 mutant mice as a
sensitized genetic background to investigate tumor induction by radiation (RAD) and cyclophosphamide (CY). Mutagen-
exposed Nf1+/−mice developed secondary cancers that are common in humans, including myeloid malignancies, sarco-
mas, and breast cancers. RAD cooperated strongly with heterozygous Nf1 inactivation in tumorigenesis. Most of the
solid tumors showed loss of the wild-type Nf1 allele but retained two Trp53 alleles. Comparative genomic hybridization
demonstrated distinct patterns of copy number aberrations in sarcomas and breast cancers from Nf1 mutant mice, and
tumor cell lines showed deregulated Ras signaling. Nf1+/−mice provide a tractable model for investigating the pathogene-
sis of common mutagen-induced cancers and for testing preventive strategies.
Therapy-induced malignant neoplasms, also known as second
malignant neoplasms (SMNs), are a severe complication of
genotoxic cancer treatments including radiation (RAD) and
chemotherapeutic agents (Bhatia and Sklar, 2002; Matesich
and Shapiro, 2003; Smith et al., 2003). SMNs have a substan-
tial public health impact, as they account for most of the
w90,000 new cancers that are diagnosed in the United States
each year in persons who had a previous histologically distinct
malignancy (Bhatia and Sklar, 2002). Whereas early reports
emphasized the risk of SMNs in patients with Hodgkin’s dis-
ease and other hematopoietic malignancies (Le Beau et al.,
1986; Rowley et al., 1977; Tucker et al., 1988), these cancers
are increasingly recognized after intensive treatment for breast
cancer and other solid tumors (Matesich and Shapiro, 2003;
S I G N I F I C A N C E
Cancers induced by genotoxic treatments are a major clinical problem; however, the long-term mutagenic potential of specific
therapeutic regimens may not be known until many years from now. We used Nf1 mutant mice to recapitulate the dynamic interac-
tion between mutagen exposure and tumorigenesis that underlies the development of human therapy-induced malignancies. These
animals develop a similar spectrum of malignancies as human patients who are treated with radiation and alkylating agents, and
provide a tractable system for performing mechanistic studies, for comparing the mutagenic potential of different regimens, and
for testing preventive strategies. Our data also have translational implications for assessing the potential risks of genotoxic modalities,
particularly in persons with neurofibromatosis type 1.
CANCER CELL : OCTOBER 2005 · VOL. 8 · COPYRIGHT © 2005 ELSEVIER INC.DOI 10.1016/j.ccr.2005.08.011337
Smith et al., 2003). Myeloid leukemia, lymphoma, and sarcoma
are the most common SMNs found in survivors of hematologic
and nonhematologic cancers (Bhatia and Sklar, 2002; Matesich
and Shapiro, 2003; Smith et al., 2003). With prolonged follow-
up, cancer survivors are also at elevated risk of developing epi-
thelial tumors of the breast, uterus, and gastrointestinal tract
(de Vathaire et al., 1989a, 1989b; Hawkins et al., 1987; Tucker
et al., 1988). Importantly, many SMNs are resistant to treat-
ment, and previous exposure to cytotoxic agents may limit the
use of intensive salvage regimens. The lack of relevant animal
models of SMNs has impeded efforts to understand how muta-
genic cancer therapeutics induce tumors in vivo, and to test
Studies of familial cancer syndromes have provided funda-
mental insights into mechanisms that underlie tumorigenesis.
In addition to markedly increasing the incidence of primary ma-
A R T I C L E
lignancies, some inherited cancer predispositions confer a high
risk of SMNs. For example, persons with the Li-Fraumeni syn-
drome carry germline TP53 mutations, which leads to an ele-
vated risk of both primary malignancies and SMNs (Li and
Fraumeni, 1982; Malkin et al., 1992; Nichols et al., 2001). Simi-
larly, most children with a germline RB1 mutation develop one
or more retinoblastoma tumors and are predisposed to osteo-
sarcoma and other cancers later in life (Wong et al., 1997). The
incidence of osteosarcoma is dramatically increased in areas
exposed to RAD in the course of treating the initial retinoblas-
toma (Wong et al., 1997). As in de novo cancers, investigating
how heritable mutations cooperate with genotoxic cancer ther-
apies will likely provide mechanistic insights that are relevant
to tumorigenesis in individuals who develop SMNs without a
known genetic predisposition.
Mutations in the NF1 tumor suppressor gene cause neurofi-
bromatosis type 1 (NF1), an inherited cancer syndrome that
affects 1 in 3500 persons (Cichowski and Jacks, 2001; Das-
gupta and Gutmann, 2003). NF1 encodes neurofibromin, a
GTPase-activating protein that negatively regulates Ras signal-
ing (Boguski and McCormick, 1993; Donovan et al., 2002b).
Affected individuals are predisposed to specific benign and
malignant tumors, particularly in tissues derived from the em-
bryonic neural crest (Side and Shannon, 1998). In addition, the
incidence of juvenile myelomonocytic leukemia (JMML) and
other myeloid malignancies is increased 200- to 500-fold in
children with NF1 (Stiller et al., 1994). Clinical data suggest that
persons with NF1 are also predisposed to SMNs. Maris and
coworkers (Maris et al., 1997) reported five children with NF1
who developed myeloid malignancies and performed a sys-
tematic review of 64 children with NF1 who received chemo-
therapy and/or RAD to treat a primary cancer. This study re-
vealed an 11% incidence of SMNs, with an especially high risk
in children who had a primary embryonal cancer. Two adults
with NF1 also developed therapy-related myelodysplastic syn-
drome (MDS) after treatment for de novo acute myeloid leuke-
mia (AML) (Papageorgio et al., 1999). These reports suggested
that therapeutic exposure to genotoxic agents might cooperate
with germline NF1 mutations in the genesis of common SMNs
found in the general population, namely myeloid leukemia
Based on these clinical observations, we reasoned that het-
erozygous Nf1 mutant mice (Nf1+/−) might be harnessed to
investigate the pathogenesis of SMNs in vivo. Nf1+/−mice
spontaneously develop pheochromocytoma and a myeloprolif-
erative disorder (MPD) that resembles JMML with incomplete
penetrance (Jacks et al., 1994b). In a previous study, exposing
these mice to the alkylating agent cyclophosphamide (CY) in-
creased the incidence of MPD and reduced the latency (Mah-
goub et al., 1999). Here, we show that RAD alone or in combi-
nation with CY induces a spectrum of SMNs in Nf1+/−mice
that includes soft tissue sarcomas and breast carcinomas. The
normal Nf1 allele is inactivated in most of these solid tumors,
and some also demonstrate loss of heterozygosity (LOH) at the
Trp53 locus. Comparative genomic hybridization (CGH) uncov-
ered tumor-specific patterns of copy number aberrations,
which implies the existence of distinct pathways of cooperat-
ing genetic lesions in different cancers. Biochemical investiga-
tion of cell lines developed from a subset of these malignant
tumors revealed deregulated Ras signaling. Nf1+/−mice pro-
vide a tractable in vivo model for understanding how RAD and
CANCER CELL : OCTOBER 2005
alkylating agents induce cancer, and for testing preventive stra-
CY and RAD induce reversible myelosuppression
We selected CY for investigation because this alkylating agent
is a component of many front-line therapeutic regimens. To
model the myelosuppression that occurs in human patients,
we intercrossed wild-type C57Bl/6 and 129/Sv mice to gener-
ate cohorts of five to ten F1 animals that were exposed to dif-
ferent CY doses. Mice injected with a weekly intraperitoneal
CY dose of 200 mg/kg for 6 consecutive weeks reproducibly
developed anemia and leukopenia that resolved after the drug
was discontinued (data not shown). This regimen was not
otherwise associated with obvious morbidity. A single RAD
dose of 3 Gy, which was selected on the basis of previous data
showing that this dose was leukemogenic in CBA mice (Major,
1979; Major and Mole, 1978; Mole et al., 1983), was adminis-
tered 2 weeks after the last dose of CY. In a pilot experiment
that assessed the combination of CY at 200 mg/kg/week for 6
weeks followed by 3 Gy of RAD, we found that mice tolerated
sequential treatment without significant toxicity (data not
Based on these preliminary data, 192 wild-type and Nf1+/−
mice were assigned to one of four groups at 8–17 weeks of
age: no treatment, CY only, RAD only, or CY followed by RAD.
Treatment with CY, RAD, or the combination resulted in revers-
ible myelosuppression (Figure S1 in the Supplemental Data
available with this article online). Mice that received six weekly
injections of CY developed anemia with decreases in hemoglo-
bin concentration (from 16.2 ± 1.0 g/dl to 12.3 ± 1.3 g/dl; p <
0.00001) and white blood cell count (from 7.4 ± 3.0 × 103to
1.5 ± 0.7 × 103cells/?l; p < 0.00001). Animals assigned to
receive RAD alone entered the study cohort concurrently and
were irradiated at the same time as mice that had been treated
with CY. RAD induced a significant reduction in leukocyte
counts and a modest fall in the hemoglobin concentration that
was not statistically significant (Figure S1). Myelosuppression
was similar in wild-type and Nf1+/−mice, and peripheral blood
cell counts recovered quickly after cessation of CY and/or
RAD. The only early treatment-related deaths occurred 19 and
20 days after RAD in two wild-type mice that received both CY
and RAD (w1% of the cohort).
Survival and tumorigenesis in wild-type and Nf1+/−mice
Pathologic analysis was performed on 91% of the study co-
hort, including 97 of 104 wild-type mice and 77 of 86 Nf1+/−
mice. Three wild-type and four Nf1+/−mice were considered
evaluable without complete pathologic analysis, including two
mice with treatment-related mortality, two mice with massive
splenomegaly in which there was no histologic analysis of
other organs, and three mice with tumors of the Harderian
gland, which secretes lipid and porphyrins over the eye. Eleven
mice that died unexpectedly could not be analyzed for tumor
formation. Heterozygous inactivation of Nf1 was strongly asso-
ciated with an increased risk of premature death following
treatment with a survival rate of only 30% after 15 months in
Nf1+/−mice compared with 78% in wild-type littermates (Figure
1, left panel; p < 0.001). Death was due to cancer in 96% of
evaluable mice. We identified 51 malignancies in 81 Nf1+/−ani-
A R T I C L E
Figure 1. Reduced survival and increased inci-
dence of cancer in Nf1+/−mice
Kaplan-Meier analysis was performed to ana-
lyze survival and tumor formation in the entire
cohort from the date that mice assigned to re-
ceive radiation were treated. Left panel: analy-
sis of overall survival demonstrates an increased
risk of death in Nf1+/−versus wild-type (WT) mice
with a hazard ratio of death of 3.83 (confidence
interval: 2.52–6.40). The survival plots include
mice of each genotype. Right panel: Nf1+/−
mice that were exposed to RAD alone (p =
0.0043) or RAD in combination with CY displayed a significantly higher risk of cancer than Nf1+/−mice that were not irradiated. The incidence of cancer
was not significantly different in control and CY-treated Nf1+/−mice. In these plots, the latency to death or tumor was measured from the date of
mals that were available for pathologic examination compared
with 17 cancers in 100 wild-type mice (Table 1; p < 0.0001).
Tumor types that occurred significantly more often in Nf1+/−
mice included soft tissue sarcomas (p = 0.001), myeloid malig-
nancies (p = 0.0005), and pheochromocytomas (p = 0.0007).
Myeloid malignancies and sarcomas, which are the most com-
mon SMNs in human patients, accounted for 24 of the 52 can-
cers found in Nf1+/−mice. We also unexpectedly detected
breast cancers in four female Nf1+/−mice but observed none
in the wild-type animals (p = 0.04, Fisher’s exact test).
RAD cooperated strongly with heterozygous Nf1 inactivation
in tumorigenesis whether administered alone or after CY (Fig-
ure 1, right panel; p = 0.002). By contrast, the incidence of
cancer was similar in untreated Nf1+/−mice and in animals that
received CY only (Figure 1, right panel). Interestingly, the nature
of the mutagen exposure influenced the tumor spectrum in
Nf1+/−mice. In particular, RAD alone resulted in a greater fre-
quency of myeloid malignancies, whereas exposure to RAD in
combination with CY resulted in a high incidence of solid tu-
mors, including breast cancer (Table 2).
Sixteen Nf1+/−mice developed tumors of presumed neural
crest origin, which included soft tissue sarcomas, pheochro-
Table 1. Tumor spectra in untreated and genotoxin-exposed mice according
Type of neoplasm Wild-type (n = 100)
Nf1+/−(n = 81)
Neural crest tumors
Soft tissue sarcomas
Poorly differentiated malignant
Tumors identified in mice that were examined pathologically. The neural crest
tumors were further classified as soft tissue sarcoma, pheochromocytoma, or
neuroblastoma/paraganglioma. *The myeloid disorders identified included MPD
(n = 9), MDS (n = 1), AML (n = 3), histiocytic sarcoma (n = 1), and refractory
cytopenia with splenomegaly (n = 3). Two additional Nf1+/−mice with massive
splenomegaly (>1 g) but no other gross abnormalities were not classified as
having a myeloid malignancy as the cause of death.
CANCER CELL : OCTOBER 2005339
mocytoma, neuroblastoma, and paraganglioma (Table 1).
Pheochromocytomas occurred in both untreated Nf1+/−mice
and in animals that were exposed to genotoxins, whereas sar-
comas, neuroblastomas, and paragangliomas only developed
after exposure to CY and/or RAD. None of these tumors oc-
curred in wild-type animals (p = 0.000001). Soft tissue sarco-
mas from Nf1+/−mice demonstrated a spindle cell pattern, and
some were positive by S100 staining (Figures 2A–2F), which is
typical of the malignant peripheral nerve sheath tumors
(MPNSTs) that occur in NF1 patients (Weiss et al., 1983; Wick
et al., 1987). Most of these tumors arose in close proximity to
peripheral nerves, were deep-seated, and ranged from mor-
phologically low-grade spindle cell proliferations with %1 mi-
totic figure (mf) per 10 high-power fields (hpf) and minimal
nuclear atypia to highly malignant tumors with >50 mf per 10
hpf and highly atypical nuclei. The low-grade tumors were uni-
formly S100-positive, whereas higher-grade lesions were
S100-negative, which is consistent with the pattern of S100
expression in human MPNSTs. Interestingly, areas that were
architecturally and cytologically reminiscent of benign Schwann
cell proliferations were found immediately adjacent to some of
the high-grade sarcomas. We also assessed these tumors for
epidermal growth factor receptor (EGFR), p16, and p21 expres-
sion. Interestingly, none expressed detectable surface EGFR or
nuclear p16, whereas seven of eight were positive for nuclear
p21 (Figure 3).
Consistent with previous reports (Jacks et al., 1994b; Mah-
goub et al., 1999), Nf1+/−mice were predisposed to myeloid
malignancies (p = 0.0065 versus wild-type littermates). Al-
though the overall incidence of these cancers did not differ in
Nf1+/−mice according to treatment group, exposure to RAD
was associated with an unexpectedly diverse spectrum of my-
eloid diseases, which included cases that were classified as
AML, MDS, or cytopenias without definitive criteria for AML or
MDS (Kogan et al., 2002). In addition, two Nf1+/−mice that
died suddenly without a confirmed cause of death had massive
splenomegaly, which suggests an underlying myeloid malig-
nancy (Table 1). All of the wild-type mice with myeloid malig-
nancies were from the RAD treatment group. Interestingly, se-
quential administration of CY and RAD was associated with
fewer myeloid malignancies in heterozygous Nf1 mutant mice
than exposure to either modality alone (Table 1).
Four of twenty-one female Nf1+/−mice that received CY +
RAD were diagnosed with breast cancers. These tumors were
characterized by cytokeratin (Figures 2G–2I), which is consis-
A R T I C L E
Table 2. Tumors in Nf1+/−mice according to treatment group
Tumor type Untreated (n = 12)CY (n = 14) RAD (n = 31)CY + RAD (n = 29) Latency (range in days)
Neural crest tumors
Soft tissue sarcomas
Poorly differentiated tumors
Tumors developing in Nf1 mutant mice were subdivided according to exposure to radiation (RAD), cyclophosphamide (CY), or both modalities (RAD + CY), and the
number of mice assigned to each group is also shown. The corresponding numbers of wild-type mice were as follows: control (n = 20), RAD (n = 32), CY (n = 17), and
RAD + CY (n = 37). The neural crest tumors were further subclassified as soft tissue sarcoma, pheochromocytoma, or neuroblastoma/paraganglioma. Latency refers
to the number of days from RAD treatment to when a tumor was detected. The mean latency was the same in mice that developed soft tissue sarcomas (359 days),
myeloid malignancies (347 days), or breast cancer (341 days).
tent with epithelial origin. The histology of these breast cancers
was remarkably uniform between cases. In general, areas of
adenocarcinoma in situ were identified adjacent to infiltrating
ductal carcinoma with both glandular and more solid growth
patterns. The invasive component demonstrated high-grade
nuclear features and abundant mitotic activity (>20 mf/10 hpf).
Most of the breast cancers invaded into the underlying skeletal
muscle. Overall, the pathologic features of these murine breast
cancers were very similar to those of high-grade adenocarci-
noma of the breast in humans.
Three Nf1+/−mice developed poorly differentiated malignant
tumors that could not be characterized with precision. In addi-
tion, we incidentally detected benign bronchiolo/alveolar ade-
nomas in the lungs of many older wild-type and Nf1+/−mice.
Similar findings have been reported in other mouse strains
Figure 2. Histologic and immunocytochemical
analysis of malignant tumors from Nf1+/−mice
Tumor tissues were stained with hematoxylin
and eosin (A, D, and G), cytokeratin (CK; B, E,
and H), or S100 (C, F, and I). A–C depict a spin-
dle cell tumor from an Nf1 mutant mouse that
was exposed to CY and RAD. Immunohisto-
chemical staining for CK was negative, while
staining for S100 revealed intense cytoplasmic
staining of a large proportion of spindle cells. D–
F are sections of a spindle cell neoplasm from
an Nf1+/−that was exposed to RAD. This tumor
displayed positive staining for both CK and
S100, with islands of CK-positive cells found in a
cells. G–I are from a breast adenocarcinoma
that developed in a female Nf1+/−mouse ex-
posed to both CY and RAD. This tumor dis-
played staining for CK but not S100. The scale
bars are 50 ?m in length.
CANCER CELL : OCTOBER 2005
(Dixon and Maronpot, 1991). These lesions were excluded from
the analysis of malignancies, with the exception that six very
large lung tumors (four in treated wild-type mice and two in
treated Nf1+/−mice) that resulted in premature death were clas-
sified as malignant. Other tumor types noted included lym-
phoma (four wild-type, five Nf1+/−), Harderian gland (three wild-
type, two Nf1+/−), osteosarcoma (two wild-type), and liver (one
Nf1+/−). Most of the wild-type and Nf1+/−mice that developed
these tumors were exposed to CY, RAD, or both.
CGH analysis of primary tumors and tumor-derived
We performed CGH on primary sarcomas and breast cancers
to identify copy number changes that might contribute to tu-
morigenesis. These studies uncovered multiple genomic changes
A R T I C L E
Figure 3. EGFR, p21, and p16 staining of geno-
A: A sarcoma section stained with an anti-EGFR
antibody is negative for surface EGFR expres-
sion, while skin from the same mouse shows posi-
tive staining in the epidermis (insert).
B: Staining with an antibody against p21 de-
monstrates nuclear staining of the tumor cells.
C: A section stained with anti-p16 antibody is
negative for staining in tumor cells. Note the
positive staining seen in rare infiltrating leuko-
The scale bars are 50 ?m in length.
in four of six sarcomas and in all of the breast cancers (see
Table S1). Two independent clustering methodologies were
employed to classify four sarcomas and four breast cancers
that developed in mice exposed to CY + RAD based on their
genomic profiles (see the Experimental Procedures). Both ap-
proaches led to a perfect separation of the sarcomas from the
breast tumors (Figure 4). To evaluate the relative similarities of
individual tumors grouped within each cluster, we performed
within sum of squares analyses and found that this difference
for breast tumors was less than half of that for the sarcomas.
These results demonstrate that the genomic profiles of the
breast cancers and sarcomas are distinct and that unsuper-
vised approaches are capable of separating them into two
groups, with the breast tumors showing a higher degree of sim-
ilarity than the sarcomas. We were able to establish stable cell
lines from four genotoxin-induced tumors that developed in
Nf1+/−mice. Three were derived from breast cancers, and the
fourth was from a soft tissue sarcoma. When we included
these cell lines and performed hierarchical clustering analysis
as before, the separation between breast cancers and sarco-
mas was preserved, and each cell line clustered together with
the corresponding primary tumor with the exception of breast
cancer B1 and cell line B1CL (Figure 4).
LOH at the Nf1 and Trp53 loci in tumors
Molecular investigation of tumors that develop spontaneously
in persons with NF1 and in strains of Nf1 mutant mice fre-
quently demonstrates somatic loss of constitutional heterozy-
gosity (Jacks et al., 1994b; Side and Shannon, 1998). Similarly,
Southern blot analysis showed loss of the wild-type C57Bl/6
Nf1 allele in six of eight soft tissue sarcomas, confirming the
important role of biallelic inactivation of Nf1 in tumorigenesis
(Figure 5A). We also observed LOH in three of three pheochro-
mocytomas and in four of four breast cancers. By contrast,
polymerase chain reaction (PCR) genotyping did not reveal
LOH in any of the myeloid malignancies.
Inactivation of Trp53 and Nf1 cooperate in tumorigenesis,
particularly sarcoma development (Cichowski et al., 1999; Vo-
gel et al., 1999). The loci for these two genes are approximately
7 cM apart on mouse chromosome 11, whereas the human
homologs are separated by approximately 22 Mb on chromo-
some 17. We assayed the D11Mit29 polymorphic microsatellite
marker, which is tightly linked to the Trp53 locus, to determine
if Trp53 and Nf1 are coordinately lost in malignancies that arise
after mutagen exposure. Surprisingly, while LOH occurred in
each of the breast tumors and in some pheochromocytomas,
all of the sarcomas that developed in Nf1+/−mice retained both
Trp53 alleles (Figure 5B).
CANCER CELL : OCTOBER 2005341
Chromosome 11 demonstrated copy number losses span-
ning the Nf1 locus in all of the breast cancers and in two of
the sarcomas (Figure 5C). Each tumor with copy number loss
showed LOH at Nf1. While these data exclude amplification of
Figure 4. Hierarchical clustering of CGH data generated from SMNs and
tumor cell lines in Nf1 mice
The sarcomas are labeled “S1,” “S2,” “S3,” and “S4” and are shown in dark
blue; the sarcoma cell line is labeled “S4CL” and is shown in light blue; the
breast cancers are labeled “B1,” “B2,” “B3,” and “B4” and are shown in
dark green; and the corresponding breast cancer cell lines are labeled
“B1CL,” “B2CL,” and “B3CL” and are shown in light green. A red-to-green
scale indicates the relative change in copy number, with losses shown in
red and gains shown in green. The sarcomas and breast cancers form
distinct clusters, and the cell lines are closely related to the primary tumors
from which they were derived. Note the shared pattern of genomic
changes within each tumor type (e.g., all of the breast cancers and cell
lines show copy number reductions spanning chromosome 12).
A R T I C L E
Figure 5. Molecular analysis at the Nf1 and Trp53 loci
A: Representative data from Southern blot analysis of tumors from Nf1 mu-
tant mice using a probe that distinguishes the wild-type (WT) and mutant
Nf1 alleles. Lanes 1 and 2 show analysis of normal tissues from wild-type
and heterozygous Nf1 mutant (HET) mice. The remaining lanes show paired
normal (N) and tumor tissues from Nf1+/−mice that developed neuro-
blastoma (Nb), pheochromocytoma (Ph), sarcoma (Sa), or breast cancer
(Br). Asterisks denote tumors with loss of the wild-type C57Bl/6 Nf1 allele.
Overall, 14 of 18 solid tumors that developed in Nf1+/−mutant mice showed
loss of the normal Nf1 allele (includes data not shown).
B: Representative data from analysis at the Trp53 locus with the microsatel-
lite marker D11Mit29 showing paired normal tissue (N) with tumor speci-
mens from mice with breast cancer (Br) or pheochromocytoma (Ph). Tu-
mors with LOH are marked with an asterisk. Whereas LOH was detected
consistently in the breast cancers and was seen in most of the pheochro-
mocytomas, it was uncommon in the sarcomas (data not shown). In each
tumor with LOH, the Trp53 allele was deleted from the C57Bl/6 chromo-
some, which carries a normal Nf1 allele in cis.
C: Copy number gains and losses across chromosome 11 in sarcoma S4
that showed LOH at Nf1 but retained two alleles at Trp53. The CGH data
reveal reduced copy number spanning a DNA segment that includes Nf1,
but not Trp53.
the mutant Nf1 allele in tumor tissues, none of the copy num-
ber changes in the Nf1 or Trp53 regions is equivalent to a sin-
gle-copy loss in a diploid cell. This could be due to the pres-
ence of some normal cells admixed with the tumor samples or,
alternatively, to loss of the normal C57Bl/6 allele with duplica-
tion of the 129/Sv chromosome in a polyploid tumor genome
with duplication of the 129/Sv chromosome.
Biochemical characterization of tumor cell lines
Based on previous data implicating aberrant EGFR signaling in
NF1-associated sarcomagenesis (DeClue et al., 2000; Li et al.,
2002), we compared EGF-induced activation of Ras and phos-
phorylation of the downstream effectors MEK and Akt in the
four cell lines derived from genotoxin-induced cancers. Mouse
embryonic fibroblasts (MEFs) were assessed in parallel. Nf1+/−
MEFs that were plated, serum starved, and stimulated with
EGF displayed a marked increase in Ras-GTP levels with a
CANCER CELL : OCTOBER 2005
Figure 6. Biochemical analysis of MEFs and tumor cell lines developed from
A: EGF induces a transient spike in Ras-GTP levels in Nf1+/−MEFs, which is
followed by a rapid return to baseline values. By contrast, Nf1−/−MEFs show
a sustained Ras-GTP response and prolonged ERK phosphorylation. Three
breast cancer cell lines also show prolonged Ras-GTP levels as well as pro-
longed ERK and Akt phosphorylation.
B: Doxorubicin induces p21 expression in heterozygous and homozygous
Nf1 mutant MEFs. By contrast, p21 was not detected in any of the cell lines
before or after doxorubicin exposure.
rapid return to baseline levels (Figure 6A). By contrast, Nf1−/−
MEFs showed slightly higher peak levels of Ras-GTP in re-
sponse to EGF and demonstrated prolonged Ras activation,
which is consistent with a previous report (Cichowski et al.,
2003). Levels of phosphorylated ERK and Akt changed in par-
allel with Ras-GTP. While there was some variability among the
three breast cancer cell lines with respect to the degree and
duration of Ras-GTP, ERK, and Akt activation, all of these cell
lines showed elevated Ras-GTP levels for at least 30 min after
EGF stimulation and persistence of phosphorylated Akt and
ERK above baseline levels for 60 min (Figure 6A). By contrast,
basal and EGF-induced Ras activation showed no consistent
differences between the one sarcoma cell line and control Nf1+/−
MEFS (Figure 6A).
We also compared the functional status of the p53 damage
response pathway in MEFs and in tumor cell lines. Nf1 mutant
MEFs induced p21 expression in response to either doxoru-
bicin (Figure 6B) or 5 Gy of RAD, which is expected in cells
with an intact p53 pathway. The three breast cancer cell lines
demonstrated LOH at the Trp53 locus, whereas the sarcoma
line retained both Trp53 alleles. Each tumor cell line had high
basal p53 expression, which is commonly associated with ex-
pression of mutant proteins, and failed to induce p21 in re-
sponse to either doxorubicin (Figure 6B) or RAD (data not
Therapy-induced cancers are an increasing concern as more
patients are surviving after receiving intensive regimens to treat
a primary malignancy (Armitage et al., 2003; Bhatia and Sklar,
2002; Curtis et al., 1992; Kushner et al., 1998; Matesich and
A R T I C L E
Shapiro, 2003; Neglia et al., 1991; Smith et al., 2003; Tucker
et al., 1988; van Leeuwen et al., 1994). Molecularly targeted
therapeutics are appealing in principle as an alternative to gen-
otoxic agents; however, the paucity of tractable biochemical
targets and the frequent occurrence of de novo and acquired
resistance imply that RAD and conventional chemotherapy will
remain mainstays of therapeutic protocols for the foreseeable
future. The experience in childhood cancers, where survival
rates of w75% have been achieved through the use of inten-
sive chemotherapy and/or RAD, illustrates that success in cur-
ing primary cancers is associated with an increased incidence
of SMNs (Bhatia and Sklar, 2002; de Vathaire et al., 1989a,
1989b). Unfortunately, the long-term toxicity of a therapeutic
regimen may only become apparent years after many patients
have been treated. Given these considerations, tractable ani-
mal models of SMN would be valuable for comparing the mu-
tagenic potential of specific agents, for uncovering mecha-
nisms of DNA damage in vivo, and for evaluating prevention
SMNs are problematic to model accurately in animals be-
cause the causative mutations result from exposure to geno-
toxins. By contrast, conventional strategies for generating
tumor-prone mice involve engineering cancer-associated mu-
tations into the germline. Although this is a powerful approach
for investigating biologic effects of known therapy-related mu-
tations such as MLL gene fusions in epipodophyllotoxin-
induced leukemias (Dobson et al., 1999), introducing these
cancer-associated mutations into the germline does not reca-
pitulate the dynamic interaction between mutagen exposure
and the host genome that underlies tumorigenesis. However,
exposing wild-type mice to various doses and combinations of
genotoxins is an inefficient and nonselective strategy for gener-
ating SMNs. The sensitized genetic background of Nf1+/−mice
allowed us to develop a tractable and penetrant model of com-
mon SMNs that affect human patients. Moreover, the highly
significant increase in tumor formation in Nf1 mutant mice and
the frequent loss of the normal Nf1 allele in genotoxin-induced
tumors reveals a direct mechanistic interaction between muta-
gen exposure and somatic mutations at this locus.
Myeloid malignancies and sarcomas were the most common
SMNs detected in Nf1+/−mice. These data are consistent with
observations in NF1 patients and, importantly, represent two of
the most common sporadic SMNs (Armitage et al., 2003;
Bhatia and Sklar, 2002; Matesich and Shapiro, 2003; Smith et
al., 2003). We previously found that CY cooperates with hetero-
zygous loss of Nf1 to increase the incidence of MPD, with a
substantially higher penetrance in inbred 129/Sv strain mice
than in F1 129/Sv × C57Bl/6 animals (Mahgoub et al., 1999).
Genetic analysis of bone marrow and splenic DNA from mice
with MPD revealed a higher frequency of LOH at Nf1 in the
129/Sv background. The genetic mechanism, which involves
deletion of the normal Nf1 allele and duplication of the mutant
allele (Mahgoub et al., 1999), is also seen in most of the my-
eloid malignancies that arise in patients with NF1 (K. Stephens,
M.M. Le Beau, and K.M.S., unpublished data). Based on these
results, we speculate that the low rate of myeloid malignancies
that we observed in CY-treated mice is due to the F1 strain
background and might be related to a reduced rate of mitotic
recombination between 129/Sv and C57Bl/6 chromosome 11
homologs. In the current study, we investigated F1 mice to in-
crease the probability of obtaining nonhematologic cancers
CANCER CELL : OCTOBER 2005 343
and to induce tumors that were heterozygous at multiple ge-
netic loci. It is striking that exposing Nf1+/−mice to RAD or to
the combination of RAD + CY induced a wide spectrum of my-
eloid disorders, including refractory cytopenias, MDS, and
AML, whereas control and CY-treated Nf1+/−mice almost al-
ways develop MPD (Mahgoub et al., 1999). Similarly, Japanese
atomic bomb survivors are predisposed to chronic myeloid leu-
kemia (CML), AML, and MDS (Imamura et al., 2002; Preston et
al., 1994), and patients who are treated with alkylating agents
and/or RAD develop both MDS and AML (Armitage et al., 2003;
Le Beau et al., 1986; Thirman and Larson, 1996). Because
MDS can be difficult to diagnose in both humans and mice, we
may have underestimated the incidence of myeloid malignan-
cies by excluding two mice that died with splenomegaly with-
out a definitive diagnosis. Although the number of animals as-
signed to each treatment group is too small to compare
individual regimens, there is no evidence that RAD and CY co-
operate in this model, and our data raise the intriguing possi-
bility that RAD may be more leukemogenic than RAD + CY.
One potential explanation for this apparent paradox is that the
probability of a mutant leukemia-initiating cell surviving and in-
ducing disease in vivo might be higher after RAD only if the
combination of CY + RAD induces severe DNA damage that
results in cell death or senescence. It is interesting that we did
not detect LOH at Nf1 in bone marrow from mice that devel-
oped myeloid malignancies, which contrasts with the solid tu-
mors. Similarly, five therapy-induced human leukemias from
NF1 patients retained both alleles (Maris et al., 1997). It is pos-
sible that the wild-type Nf1 allele is inactivated by point mu-
tations or other subtle alterations in humans and mice; alterna-
tively, genotoxin-induced mutations may cooperate with
haploinsufficiency at the Nf1 locus in the F1 genetic back-
ground. Consistent with this hypothesis, heterozygous inacti-
vation of Nf1 has phenotypic consequences in mast cells and
melanocytes (Ingram et al., 2000), and therapy-related myeloid
malignancies from NF1 patients demonstrate acquired chro-
mosomal abnormalities such as monosomy 7 (Maris et al.,
1997; Papageorgio et al., 1999).
In addition to inducing myeloid malignancies, RAD is
strongly associated with subsequent development of sarcomas
within the RAD portals of patients treated for nonmalignant dis-
eases and for a variety of cancers (Beck, 1922; Boice et al.,
1988; Darby et al., 1987; Doherty et al., 1986; Halperin et al.,
1984; Marsche, 1922; Murray et al., 1999; Pendlebury et
al., 1995; Smith and Doll, 1982). Many of these tumors are
high-grade osteosarcomas or malignant fibrous histiocytomas
that have a poor prognosis (Huvos et al., 1985; Inwards and
Unni, 1995; Laskin et al., 1988; Murray et al., 1999). Although
TP53 inactivation is common in RAD-induced sarcomas
(Brachman et al., 1991; Nakanishi et al., 1998), the other mo-
lecular lesions are largely unknown (Beech et al., 1998; Cowan
et al., 1990; Remmelink et al., 1998; Sekyi-Otu et al., 1995;
Wang et al., 1994). MPNST, the most common sarcoma that
arises de novo in persons with NF1, has also been reported
as a SMN (Li et al., 2002). MPNSTs typically display complex
karyotypes and may also show loss of the normal NF1 allele,
mutations that deregulate the pRb and p53 pathways, and ab-
errant expression of EGFR (Birindelli et al., 2001; DeClue et al.,
2000; Halling et al., 1996; Liapis et al., 1999; Plaat et al., 1999).
A few small studies of human RAD-induced sarcomas have
suggested the existence of nonrandom cytogenetic aberr-
A R T I C L E
ations including translocation breakpoints involving chromo-
some band 1p13 and 7p22, gain of 7q1, and losses of 9p2,
1p3, and 21p1-q2 (Bridge et al., 2004; Mertens et al., 1995;
Plaat et al., 1999). Human chromosome band 7q1 is ortholo-
gous to two segments of mouse chromosome 5. Interestingly,
our CGH analysis revealed a discrete region of amplification in
four of six sarcomas and in the four breast cancers at BAC
RP23-281N22, which is within the segment of mouse chromo-
some 5 that is syntenic to human 7q1. This BAC includes a
number of putative genes including a homolog of the Drosoph-
ila frizzled gene, a FK506 binding protein, and a putative zinc
finger transcription factor. We also identified a region of amplifi-
cation on mouse chromosome 8 in breast cancer samples that
corresponds to an interval of human chromosome 13q34 that
is amplified in hepatocellular carcinoma and some other epi-
thelial cancers (Yasui et al., 2002). This segment includes the
TFDP1, CUL4A, and CDC16 genes.
Two groups crossed Nf1 and Trp53 mutant mice to generate
recombinant founders that carried both mutant alleles on the
same chromosomal homolog (i.e., in cis configuration) in order
to recapitulate the simultaneous loss of NF1 and TP53 fre-
quently seen in NF1-associated MPNSTs. Whereas heterozy-
gous and homozygous Trp53 mutant mice spontaneously de-
velop tumors (primarily lymphomas) at 30–50 weeks of age
(Donehower et al., 1992; Jacks et al., 1994a), many cis and
trans Nf1+/−;Trp53+/−mice developed early-onset sarcomas
that resembled human MPNSTs. Molecular analysis revealed
LOH at both Nf1 and Trp53 in at least 70% of the soft tissue
tumors from cis Nf1+/−;Trp53+/−mice. Based on these data, we
were surprised that RAD- and CY-induced sarcomas with LOH
at Nf1 generally retained both Trp53 alleles. Residual p53 func-
tion may underlie the failure of most of these tumors to grow
as permanent cell lines, which is in contrast to sarcomas from
cis Nf1+/−;Trp53+/−mice (L. Parada, personal communication).
In this regard, it is of interest that the sarcoma cell line that we
obtained had abnormal p53 pathway function as assessed by
elevated p53 expression and failure to induce p21 in response
to RAD and doxorubicin. Although we cannot exclude similar
defects in other RAD- and CY-induced tumors, our data sug-
gest the existence of one or more p53-independent pathways
to sarcoma development. This hypothesis is consistent with
studies of NF1-associated de novo MPNSTs, which uncovered
either LOH at TP53 or a somatic mutation in less than half of
the tumors (Birindelli et al., 2001; Halling et al., 1996; Lothe et
al., 2001; Menon et al., 1990). Aberrant expression of the EGF
receptor has also been implicated in the development of
MPNST in humans and in mouse models (DeClue et al., 2000;
Li et al., 2002; Ling et al., 2005), and a recent study demon-
strated that reducing EGFR signaling by w90% dramatically
reduced the incidence of sarcoma in cis Nf1+/−;Trp53+/−mice
(Ling et al., 2005). By contrast, none of the genotoxin-induced
soft tissue tumors that developed in Nf1+/−mice expressed de-
tectable level of EGFR. These data are consistent with the ob-
servation that w30% of MPNSTs from NF1 patients do not
express EGFR mRNA (Watson et al., 2004). The expression
profiles of this subset of tumors were characterized by elevated
levels of neuroglial markers and by relatively low expression of
transcripts that are associated with proliferation. Together, our
molecular and immunohistochemical data support a distinct
pathway of genotoxin-induced sarcomagenesis that involves
loss of Nf1, normal p53 activity, and a lack of EGFR amplifica-
CANCER CELL : OCTOBER 2005
tion. The recurring copy number changes detected by CGH
analysis of these tumors imply that limited numbers of cooper-
ating events promote clonal outgrowth in vivo.
The observation that female Nf1+/−mice developed breast
cancer after exposure to RAD or RAD in combination with CY
was unexpected and intriguing. The incidence of SMNs has
been investigated over many decades in patients who were
treated for Hodgkin’s disease as adolescents or young adults.
Among these individuals, breast cancer is the most common
solid SMN in women, with a cumulative probability of up to
42% after 30 years (Bhatia et al., 1996). Although RAD dose,
combined treatment with chemotherapy, and age at the time
of treatment strongly influence the risk of subsequent develop-
ment of secondary breast cancer, the genetic pathways that
underlie malignant transformation are poorly understood. CGH
analysis of these tumors uncovered a striking pattern of recurr-
ing copy number changes involving chromosome 12 and other
DNA segments (Figure 3 and Supplemental Data). Together,
our data imply that loss of Trp53 and hyperactive Ras cooper-
ate with a common series of secondary genetic changes in
mammary tumorigenesis after genotoxin exposure.
Hyperactivation of the PI3 kinase/Akt cascade underlies
growth factor-independent survival in immortalized Nf1-defi-
cient hematopoietic cells, whereas aberrant MAP kinase sig-
naling drives proliferation and autocrine production of granulo-
cyte-macrophage colony stimulating factor (GM-CSF) (Donovan
et al., 2002a). In addition, recent studies in Nf1-deficient as-
trocytes and MEFs have uncovered deregulated activation of
S6 kinase, a downstream effector of mTOR (Dasgupta et al.,
2005; Johannessen et al., 2005). Dephosphorylation of tuberin
(the protein encoded by the TSC2 tumor suppressor gene) by
activated Akt is an important mechanism that contributes to
hyperactive mTOR/S6 kinase in Nf1-deficient cells (Johannes-
sen et al., 2005). We found that breast cancer cell lines from
Nf1+/−mice responded to EGF with prolonged activation of
Ras, ERK, and Akt, and it will be of interest to determine how
this modulates the activation status of downstream effectors
and contributes to specific cellular phenotypes. Based on the
data of Li and colleagues (Li et al., 2002) showing elevated
levels of EGFR expression in most sarcoma cell lines from cis
Nf1+/−;Trp53+/−mice, we investigated signaling in response to
EGF in the S4CL cell line. The modest increase in ERK phos-
phorylation that we observed in response to EGF is consistent
with our data showing that sarcomas that arise in Nf1+/−mice
that are exposed to genotoxins do not express detectable
levels of EGFR.
Nf1+/−mice are a robust and tractable in vivo system that
can be harnessed to address the mutagenic potential of cancer
treatments and to test preventive strategies. This “first genera-
tion” model can also be refined further; for example, it is fea-
sible to selectively administer a range of RAD doses to the
mammary glands of mice, and to test how age, previous preg-
nancies, and hormonal manipulations modulate tumorigenesis.
A more challenging experimental question involves investigat-
ing if the therapy-induced cancers that develop in this model
are largely refractory to conventional chemotherapeutic agents,
and if this is true, this model may prove useful for testing novel
therapeutic strategies. Unlike hematopoietic cancers, there are
no simple techniques for transplanting primary solid tumors
into immunocompetent recipients, which would greatly facili-
tate studies of this nature. Our data also have implications for
A R T I C L E
treating tumors that arise in the patients with NF1. In particular,
we found that a single low dose of RAD cooperated strongly
with heterozygous Nf1 inactivation in tumorigenesis. Children
with NF1 are predisposed to optic track gliomas and low-grade
astrocytomas, which may be difficult to manage. Our data sug-
gest that the potential risk of SMN should be considered when
deciding whether to irradiate these and other NF1-associated
tumors and that patients who require RAD should be followed
carefully for subsequent treatment-induced cancers.
Mouse stains, breeding, and treatment
All experimental procedures involving mice were reviewed and approved by
the UCSF Committee on Animal Research. Heterozygous Nf1 mutant mice
that were maintained in the 129/Sv strain background (Jacks et al., 1994b)
were mated with wild-type C57Bl/6 mice (Jackson Labs, Bar Harbor, ME)
to generate wild-type and Nf1+/−mice. These mice were genotyped and
assigned to one of the four experimental groups. Study mice were housed
in a pathogen-free environment at the UCSF Animal Care Facility. Mice ex-
posed to CY (Mead Johnson Oncology Products) received intraperitoneal
injections of CY (200 mg/ml in sterile water) medial to the proximal aspect
of the femur. RAD-treated mice received a single dose of 3 Gy, which was
administered using a cesium-137 source (JL Shepherd & Associates, San
Fernando, CA) at a rate of w250 cGy/min.
Complete blood counts (CBCs) were measured on blood samples taken
premortem from tail vein bleeding and postmortem by intracardiac puncture
in Hema-Vet 850 hematology analyzer (CDC Technologies, Inc., Oxford,
CT). Blood and bone marrow smears were made on glass slides, and cyto-
spin slides were prepared by suspending 100,000 viable cells in 200 ?l of
PBS and centrifuging them for 10 min at 400 rpm. The slides were stained
with Wright-Giemsa (Fisher) and examined using a Nikon Eclipse E400
microscope. Tumor tissues were fixed in 10% formalin (Fisher) and embed-
ded in paraffin, cut, and stained with hematoxylin and eosin. Slides of tumor
tissues were stained with antibodies against CK and S100 (Dako, Carpin-
teria, CA) according to the manufacturer’s instructions. EGFR expression,
p16 expression, and p21 expression were assessed on 10 ?m thick cryostat
sections cut from frozen tumor specimens. Briefly, the slides were fixed in
cold acetone or 4% paraformaldehyde and then blocked with serum fol-
lowed by incubation with avidin then biotin blocking reagents (Vector
Laboratories). Sections were immunostained using standard procedures
with rabbit antibodies against EGFR (SC-03), p16 (SC-1207), and p21 (SC-
397) (Santa Cruz Biotechnology) and enhanced with ABC Elite Vector Stain
Substrate Kit (Vector Laboratories) using the manufacturer’s protocol. Stain-
ing was visualized with 3,3#-diaminobenzidine (DAB Substrate Kit, Vector
Laboratories) and counterstained with Gill’s hematoxylin (Santa Cruz Bio-
technology). Blood, bone marrow, and spleen sections from mice with
hematopoietic malignancies were reviewed by a hematopathologist with ex-
pertise in murine myeloid malignancies (S.C.K.). A veterinary pathologist
with experience in evaluating mouse cancer models (A.B.) and a pathologist
with expertise in human sarcomas (A.H.) independently reviewed the solid
Genotyping and mutation analysis
Mice were genotyped at the Nf1 locus, and LOH was performed on solid
tumor specimens using Southern blot analysis as described elsewhere
(Jacks et al., 1994b). A PCR-based assay was employed to assess myeloid
malignancies for LOH at Nf1. LOH analysis was performed at the D11Mit29
locus amplifying tumor DNA samples with 3# and 5# γ-33P-labeled primers
(Research Genetics, Carlsbad, CA). The amplification products were re-
solved on a polyacrylamide gel and visualized by autoradiography.
Cells were maintained at 37°C in humidified incubator with 5% CO2. To
generate tumor cell lines, a 10–100 mm3piece tissue was first washed with
Dulbecco’s modified essential medium (DMEM) containing penicillin and
streptomycin and minced with a sterile scalpel. After adding 2–3 ml of
CANCER CELL : OCTOBER 2005 345
DMEM with 10% FBS (Hyclone, Logan, UT), the tissue was passed through
a 16G needle (Becton-Dickinson, Franklin Lakes, NJ), plated in two wells of
a six-well plate (Corning, Corning, NY), and cultured at 37°C. The medium
was changed every 1–3 days, and cells that proliferated in culture were
trypsinized, passaged, and frozen. MEFs were generated from E13.6 F1
embryos by first dissecting the fetal liver and head, and then mincing and
trypsinizing the remaining tissues, and culturing the cells in DMEM supple-
mented with 10% FBS, glutamine, and β-mercaptoethanol (Sigma, St.
Acquisition and analysis of CGH data
Array CGH was carried out on arrays of 2500 mouse BACs as described
previously (Snijders et al., 2005). Briefly, test and reference DNA samples
from 129/Sv × C57Bl/6 female F1 mice were labeled by random priming to
incorporate Cy3- and Cy5-dUTP, respectively. The labeled DNAs together
with Cot-1 DNA (Invitrogen) were hybridized to the arrays for 48 hr. Follow-
ing washing 16 bit 1024 × 1024 pixel DAPI, Cy3 and Cy5 images were
collected with a custom-built CCD camera system as described previously
(Pinkel et al., 1998). “UCSF SPOT” software (Jain et al., 2002) was used to
automatically segment the spots based on the DAPI images, perform local
background correction, and calculate various measurement parameters, in-
cluding log2 ratios of the total integrated Cy3 and Cy5 intensities for each
spot. A second custom program, SPROC, was used to obtain averaged
ratios of the triplicate spots for each clone, standard deviations of the tripli-
cates, and plotting position for the BACs on the May 2004 freeze of the
mouse genome sequence (http://genome.ucsc.edu/). Data files were edited
to remove ratios on clones for which only one of the triplicates remained
after SPROC analysis and/or the standard deviation of the log2 ratios of the
triplicates was >0.2.
To address the distinction between types of tumors based on their geno-
mic profiles, we applied clustering approaches (Fridlyand et al., 2004). The
clones missing in more than 20% of the samples were filtered out, and each
sample was segmented using DNA copy (Olshen et al., 2004). The tumor-
specific experimental error was estimated as median absolute deviation
(MAD) of the differences between the observed log2 ratios and the means
of their corresponding segments. The clones with log2 ratios further away
from their segment mean than 4 × MAD retained their original log2 ratios,
whereas the rest of the clones were assigned the value of their segment
mean. The distance between a pair of samples was computed as 1 − Pear-
son correlation calculated over the autosomal clones. The resulting distance
matrix was used as an input to hierarchical (see Figure 3) and k-means
clustering with two groups. To evaluate the relative similarity of the two
clusters we compared the within sum of squares of each cluster, which is
an output of the k-means procedure in the R statistical package.
Cell lines or MEFs were plated a density of 1.5–2 × 106cells per 10 cm
plate in DMEM with 10% FBS. After 16–20 hr, the cells were washed and
transferred to DMEM without serum. After 24 hr, the cells were stimulated
with EGF (Peprotech, Rocky Hill, NJ) at 10 ng/ml for 5–60 min. The cells
were lysed in Holstrom buffer, protein levels were quantified using the modi-
fied Bradford assay (Bio-Rad, Hercules, CA), and the samples were re-
solved using SDS-PAGE in 10%–12.5% precast gels (Bio-Rad). The pro-
teins were transferred to an Immobilon filter (Amersham, Piscataway, NJ)
and probed with antibodies labeled with horseradish peroxidase. The filters
were incubated with ECL (Amersham) reagent and then exposed to film.
We used the following primary antibodies to detect specific proteins: anti-
pan-Ras antibody (Oncogene Research Products, San Diego, CA), anti-
pERK (Santa Cruz Biotechnology, Santa Cruz, CA), anti-ERK antibody (Cell
Signaling, Beverly, MA), anti-pAkt antibody (a kind gift from David Stokoe),
anti-Akt antibody (Cell Signaling). Ras-GTP levels were quantified using an
excess of Raf1 RBD-GST conjugated agarose beads (Upstate, Charlottes-
ville, VA) according to the manufacturer’s instructions. For studies of p53
function, cells that were irradiated to a dose of 5 Gy were incubated at 37°C
for 3 hr or incubated with doxorubicin for approximately 16 hr. The cells
were lysed, and Western blotting was performed as described above with
antibodies that recognize p53 (Santa Cruz Biotechnology) and p21 (Santa
A R T I C L E
Sample size and statistical analysis
The primary endpoint for statistical analysis was the effect of RAD, alone
or in combination with CY, on the incidence of cancer in Nf1+/−mice. Based
on previous data, we estimated the risk of tumor development in untreated
F1 Nf1+/−mice to be 10% over 16 months. Since the primary goal was to
examine if RAD cooperates with heterozygous Nf1 inactivation in tumori-
genesis, we allocated twice as many mice to the RAD cohorts in order to
The Supplemental Data include two figures and one table and can be found
with this article online at http://www.cancercell.org/cgi/content/full/8/4/337/
We are grateful to Abigail Aiyagari, Charles Fezzie, Ben Yen, Robert Cardiff,
Angell Shieh, and Doan Le for advice and for assisting with various aspects
of this study. We are also indebted to David Stokoe, who provided the anti-
pAkt antibody that we used in these studies. This work was supported by
US Army Neurofibromatosis Research Program projects DAMD 17-02-1-
0638 and DAMD17-98-1-8608, NIH grants R01 CA72614 and U01 CA84221,
and the Jeffrey and Karen Peterson Family Foundation (all to K.M.S.); by
NIH training grant T32ES07106 (R.C.C.); and by NIH grants U01 CA84118
and R01 CA101359 (D.G.A.). S.C.K. is a Scholar of the Leukemia and Lym-
phoma Society of America, and T.E.J. is an Investigator of the Howard
Hughes Medical Institute.
Received: February 9, 2005
Revised: June 24, 2005
Accepted: August 26, 2005
Published: October 17, 2005
Armitage, J.O., Carbone, P.P., Connors, J.M., Levine, A., Bennett, J.M., and
Kroll, S. (2003). Treatment-related myelodysplasia and acute leukemia in
non-Hodgkin’s lymphoma patients. J. Clin. Oncol. 21, 897–906.
Beck, A. (1922). Zur Frage des Röntgensarkoms, Zugleich ein Beitrag zur
Pathogenese des Sarkoms. Munch. Med. Wochenschr. 69, 623–624.
Beech, D., Pollock, R.E., Tsan, R., and Radinsky, R. (1998). Epidermal
growth factor receptor and insulin-like growth factor-I receptor expression
and function in human soft-tissue sarcoma cells. Int. J. Oncol. 12, 329–336.
Bhatia, S., and Sklar, C. (2002). Second cancers in survivors of childhood
cancer. Nat. Rev. Cancer 2, 124–132.
Bhatia, S., Robison, L.L., Oberlin, O., Greenberg, M., Bunin, G., Fossati-
Bellani, F., and Meadows, A.T. (1996). Breast cancer and other second neo-
plasms after childhood Hodgkin’s disease. N. Engl. J. Med. 334, 745–751.
Birindelli, S., Perrone, F., Oggionni, M., Lavarino, C., Pasini, B., Vergani, B.,
Ranzani, G.N., Pierotti, M.A., and Pilotti, S. (2001). Rb and TP53 pathway
alterations in sporadic and NF1-related malignant peripheral nerve sheath
tumors. Lab. Invest. 81, 833–844.
Boguski, M., and McCormick, F. (1993). Proteins regulating Ras and its rela-
tives. Nature 366, 643–653.
Boice, J.D., Jr., Engholm, G., Kleinerman, R.A., Blettner, M., Stovall, M.,
Lisco, H., Moloney, W.C., Austin, D.F., Bosch, A., Cookfair, D.L., et al. (1988).
Radiation dose and second cancer risk in patients treated for cancer of the
cervix. Radiat. Res. 116, 3–55.
Brachman, D.G., Hallahan, D.E., Beckett, M.A., Yandell, D.W., and Weich-
selbaum, R.R. (1991). p53 gene mutations and abnormal retinoblastoma
protein in radiation-induced human sarcomas. Cancer Res. 51, 6393–6396.
Bridge, R.S., Jr., Bridge, J.A., Neff, J.R., Naumann, S., Althof, P., and Bruch,
L.A. (2004). Recurrent chromosomal imbalances and structurally abnormal
breakpoints within complex karyotypes of malignant peripheral nerve
CANCER CELL : OCTOBER 2005
sheath tumour and malignant triton tumour: a cytogenetic and molecular
cytogenetic study. J. Clin. Pathol. 57, 1172–1178.
Cichowski, K., and Jacks, T. (2001). NF1 tumor suppressor gene function:
Narrowing the GAP. Cell 104, 593–604.
Cichowski, K., Shih, T.S., Schmitt, E., Santiago, S., Reilly, K., McLaughlin,
M.E., Bronson, R.T., and Jacks, T. (1999). Mouse models of tumor develop-
ment in neurofibromatosis type 1. Science 286, 2172–2176.
Cichowski, K., Santiago, S., Jardim, M., Johnson, B.W., and Jacks, T.
(2003). Dynamic regulation of the Ras pathway via proteolysis of the NF1
tumor suppressor. Genes Dev. 17, 449–454.
Cowan, J.M., Beckett, M.A., Tarbell, N., and Weichselbaum, R.R. (1990).
Symmetrical chromosome rearrangements in cell lines established from hu-
man radiation-induced sarcomas. Cancer Genet. Cytogenet. 50, 125–137.
Curtis, R.E., Boice, J.D., Jr., Stovall, M., Bernstein, L., Greenberg, R.S.,
Flannery, J.T., Schwartz, A.G., Weyer, P., Moloney, W.C., and Hoover, R.N.
(1992). Risk of leukemia after chemotherapy and radiation treatment for
breast cancer. N. Engl. J. Med. 326, 1745–1751.
Darby, S.C., Doll, R., Gill, S.K., and Smith, P.G. (1987). Long term mortality
after a single treatment course with X-rays in patients treated for ankylosing
spondylitis. Br. J. Cancer 55, 179–190.
Dasgupta, B., and Gutmann, D.H. (2003). Neurofibromatosis 1: Closing the
GAP between mice and men. Curr. Opin. Genet. Dev. 13, 20–27.
Dasgupta, B., Yi, Y., Chen, D.Y., Weber, J.D., and Gutmann, D.H. (2005).
Proteomic analysis reveals hyperactivation of the mammalian target of rapa-
mycin pathway in neurofibromatosis 1-associated human and mouse brain
tumors. Cancer Res. 65, 2755–2760.
DeClue, J.E., Heffelfinger, S., Benvenuto, G., Ling, B., Li, S., Rui, W., Vass,
W.C., Viskochil, D., and Ratner, N. (2000). Epidermal growth factor receptor
expression in neurofibromatosis type 1-related tumors and NF1 animal
models. J. Clin. Invest. 105, 1233–1241.
de Vathaire, F., Francois, P., Hill, C., Schweisguth, O., Rodary, C., Sarrazin,
D., Oberlin, O., Beurtheret, C., Dutreix, A., and Flamant, R. (1989a). Role of
radiotherapy and chemotherapy in the risk of second malignant neoplasms
after cancer in childhood. Br. J. Cancer 59, 792–796.
de Vathaire, F., Schweisguth, O., Rodary, C., Francois, P., Sarrazin, D., Ober-
lin, O., Hill, C., Raquin, M.A., Dutreix, A., and Flamant, R. (1989b). Long-
term risk of second malignant neoplasm after a cancer in childhood. Br. J.
Cancer 59, 448–452.
Dixon, D., and Maronpot, R.R. (1991). Histomorphologic features of sponta-
neous and chemically-induced pulmonary neoplasms in B6C3F1 mice and
Fischer 344 rats. Toxicol. Pathol. 19, 540–556.
Dobson, C.L., Warren, A.J., Pannell, R., Forster, A., Lavenir, I., Corral, J.,
Smith, A.J., and Rabbitts, T.H. (1999). The mll-AF9 gene fusion in mice
controls myeloproliferation and specifies acute myeloid leukaemogenesis.
EMBO J. 18, 3564–3574.
Doherty, M.A., Rodger, A., and Langlands, A.O. (1986). Sarcoma of bone
following therapeutic irradiation for breast carcinoma. Int. J. Radiat. Oncol.
Biol. Phys. 12, 103–106.
Donehower, L., Harvey, M., Slagle, B., McArthur, M., Montgomery, C., Butel,
J., and Bradley, A. (1992). Mice deficient for p53 are developmentally normal
but susceptible to spontaneous tumours. Nature 356, 215–221.
Donovan, S., See, W., Bonifas, J., Stokoe, D., and Shannon, K.M. (2002a).
Hyperactivation of protein kinase B and ERK have discrete effects on sur-
vival, proliferation, and cytokine expression in Nf1-deficient myeloid cells.
Cancer Cell 2, 507–514.
Donovan, S., Shannon, K.M., and Bollag, G. (2002b). GTPase activating
proteins: critical regulators of intracellular signaling. Biochim. Biophys. Acta
Fridlyand, J., Snijders, A.M., Pinkel, D., Albertson, D.G., and Jain, A.N.
(2004). Hidden Markov models approach to the analysis of array CGH data.
J. Multivariate Anal., in press.
Halling, K.C., Scheithauer, B.W., Halling, A.C., Nascimento, A.G., Ziesmer,
S.C., Roche, P.C., and Wollan, P.C. (1996). p53 expression in neurofibroma
A R T I C L E
and malignant peripheral nerve sheath tumor. An immunohistochemical
study of sporadic and NF1-associated tumors. Am. J. Clin. Pathol. 106,
Halperin, E.C., Greenberg, M.S., and Suit, H.D. (1984). Sarcoma of bone
and soft tissue following treatment of Hodgkin’s disease. Cancer 53, 232–
Hawkins, M.M., Draper, G.J., and Kingston, J.E. (1987). Incidence of second
primary tumours among childhood cancer survivors. Br. J. Cancer 56,
Huvos, A.G., Woodard, H.Q., Cahan, W.G., Higinbotham, N.L., Stewart,
F.W., Butler, A., and Bretsky, S.S. (1985). Postradiation osteogenic sarcoma
of bone and soft tissues. A clinicopathologic study of 66 patients. Cancer
Imamura, N., Abe, K., and Oguma, N. (2002). High incidence of point muta-
tions of p53 suppressor oncogene in patients with myelodysplastic syn-
drome among atomic-bomb survivors: a 10-year follow-up. Leukemia 16,
Ingram, D.A., Yang, F.C., Travers, J.B., Wenning, M.J., Hiatt, K., New, S.,
Hood, A., Shannon, K., Williams, D.A., and Clapp, D.W. (2000). Genetic and
biochemical evidence that haploinsufficiency of the Nf1 tumor suppressor
gene modulates melanocyte and mast cell fates in vivo. J. Exp. Med. 191,
Inwards, C.Y., and Unni, K.K. (1995). Classification and grading of bone
sarcomas. Hematol. Oncol. Clin. North Am. 9, 545–569.
Jacks, T., Remington, L., Williams, B.O., Schmitt, E.M., Halachmi, S., Bron-
son, R.T., and Weinberg, R.A. (1994a). Tumor spectrum analysis in p53-
mutant mice. Curr. Biol. 4, 1–7.
Jacks, T., Shih, S., Schmitt, E.M., Bronson, R.T., Bernards, A., and Wein-
berg, R.A. (1994b). Tumorigenic and developmental consequences of a tar-
geted Nf1 mutation in the mouse. Nat. Genet. 7, 353–361.
Jain, A.N., Tokuyasu, T.A., Snijders, A.M., Segraves, R., Albertson, D.G.,
and Pinkel, D. (2002). Fully automatic quantification of microarray image
data. Genome Res. 12, 325–332.
Johannessen, C.M., Reczek, E.E., James, M.F., Brems, H., Legius, E., and
Cichowski, K. (2005). The NF1 tumor suppressor critically regulates TSC2
and mTOR. Proc. Natl. Acad. Sci. USA 102, 8573–8578.
Kogan, S.C., Ward, J.M., Anver, M.R., Berman, J.J., Brayton, C., Cardiff,
R.D., Carter, J.S., de Coronado, S., Downing, J.R., Fredrickson, T.N., et al.
(2002). Bethesda proposals for classification of nonlymphoid hematopoietic
neoplasms in mice. Blood 100, 238–245.
Kushner, B.H., Heller, G., Cheung, N.K., Wollner, N., Kramer, K., Bajorin, D.,
Polyak, T., and Meyers, P.A. (1998). High risk of leukemia after short-term
dose-intensive chemotherapy in young patients with solid tumors. J. Clin.
Oncol. 16, 3016–3020.
Laskin, W.B., Silverman, T.A., and Enzinger, F.M. (1988). Postradiation soft
tissue sarcomas. An analysis of 53 cases. Cancer 62, 2330–2340.
Le Beau, M.M., Albain, K.S., Larson, R.A., Vardiman, J.W., Davis, E.M.,
Blough, R.R., Golomb, H.M., and Rowley, J.D. (1986). Clinical and cytoge-
netic correlations in 63 patients with therapy-related myelodysplastic syn-
dromes and acute nonlymphocytic leukemia: further evidence for character-
istic abnormalities of chromosomes no. 5 and 7. J. Clin. Oncol. 4, 325–345.
Li, F.P., and Fraumeni, J.F., Jr. (1982). Prospective study of a family cancer
syndrome. JAMA 247, 2692–2694.
Li, H., Velasco-Miguel, S., Vass, W.C., Parada, L.F., and DeClue, J.E. (2002).
Epidermal growth factor receptor signaling pathways are associated with
tumorigenesis in the Nf1:p53 mouse tumor model. Cancer Res. 62, 4507–
Liapis, H., Marley, E.F., Lin, Y., and Dehner, L.P. (1999). p53 and Ki-67 prolif-
erating cell nuclear antigen in benign and malignant peripheral nerve sheath
tumors in children. Pediatr. Dev. Pathol. 2, 377–384.
Ling, B.C., Wu, J., Miller, S.J., Monk, K.R., Shamekh, R., Rizvi, T.A., Decour-
ten-Myers, G., Vogel, K.S., DeClue, J.E., and Ratner, N. (2005). Role for the
CANCER CELL : OCTOBER 2005 347
epidermal growth factor receptor in neurofibromatosis-related peripheral
nerve tumorigenesis. Cancer Cell 7, 65–75.
Lothe, R.A., Smith-Sorensen, B., Hektoen, M., Stenwig, A.E., Mandahl, N.,
Saeter, G., and Mertens, F. (2001). Biallelic inactivation of TP53 rarely con-
tributes to the development of malignant peripheral nerve sheath tumors.
Genes Chromosomes Cancer 30, 202–206.
Mahgoub, N., Taylor, B., Le Beau, M., Gratiot, M., Carlson, K., Jacks, T., and
Shannon, K.M. (1999). Myeloid malignancies induced by alkylating agents in
Nf1 mice. Blood 93, 3617–3623.
Major, I.R. (1979). Induction of myeloid leukaemia by whole-body single
exposure of CBA male mice to x-rays. Br. J. Cancer 40, 903–913.
Major, I.R., and Mole, R.H. (1978). Myeloid leukaemia in x-ray irradiated
CBA mice. Nature 272, 455–456.
Malkin, D., Jolly, K.W., Barbier, N., Look, A.T., Friend, S.H., Gebhardt, M.C.,
Andersen, T.I., Borresen, A.L., Li, F.P., Garber, J., et al. (1992). Germline
mutations of the p53 tumor-suppressor gene in children and young adults
with second malignant neoplasms. N. Engl. J. Med. 326, 1309–1315.
Maris, J.M., Wiersma, S.R., Mahgoub, N., Thompson, P., Geyer, R.J., Hur-
witz, C.G., Lange, B.J., and Shannon, K.M. (1997). Monosomy 7 myelodys-
plastic syndrome and other second malignant neoplasms in children with
neurofibromatosis type 1. Cancer 79, 1438–1446.
Marsche, E. (1922). Tuberkulose und Sarkom. Zentralbl. Chir. 49, 1057–
Matesich, S.M., and Shapiro, C.L. (2003). Second cancers after breast can-
cer treatment. Semin. Oncol. 30, 740–748.
Menon, A.G., Anderson, K.M., Riccardi, V.M., Chung, R.Y., Whaley, J.M.,
Yandell, D.W., Farmer, G.E., Freiman, R.N., Lee, J.K., Li, F.P., et al. (1990).
Chromosome 17p deletions and p53 gene mutations associated with the
formation of malignant neurofibrosarcomas in von Recklinghausen neurofi-
bromatosis. Proc. Natl. Acad. Sci. USA 87, 5435–5439.
Mertens, F., Rydholm, A., Bauer, H.F., Limon, J., Nedoszytko, B., Szadow-
ska, A., Willen, H., Heim, S., Mitelman, F., and Mandahl, N. (1995). Cytoge-
netic findings in malignant peripheral nerve sheath tumors. Int. J. Cancer
Mole, R.H., Papworth, D.G., and Corp, M.J. (1983). The dose-response for
x-ray induction of myeloid leukaemia in male CBA/H mice. Br. J. Cancer
Murray, E.M., Werner, D., Greeff, E.A., and Taylor, D.A. (1999). Postradiation
sarcomas: 20 cases and a literature review. Int. J. Radiat. Oncol. Biol. Phys.
Nakanishi, H., Tomita, Y., Myoui, A., Yoshikawa, H., Sakai, K., Kato, Y., Ochi,
T., and Aozasa, K. (1998). Mutation of the p53 gene in postradiation sar-
coma. Lab. Invest. 78, 727–733.
Neglia, J.P., Meadows, A.T., Robison, L.L., Kim, T.H., Newton, W.A., Ruy-
mann, F.B., Sather, H.N., and Hammond, G.D. (1991). Second neoplasms
after acute lymphoblastic leukemia in childhood. N. Engl. J. Med. 325,
Nichols, K.E., Malkin, D., Garber, J.E., Fraumeni, J.F., Jr., and Li, F.P. (2001).
Germ-line p53 mutations predispose to a wide spectrum of early-onset can-
cers. Cancer Epidemiol. Biomarkers Prev. 10, 83–87.
Olshen, A.B., Venkatraman, E.S., Lucito, R., and Wigler, M. (2004). Circular
binary segmentation for the analysis of array-based DNA copy number
data. Biostatistics 5, 557–572.
Papageorgio, C., Seiter, K., and Feldman, E.J. (1999). Therapy-related mye-
lodysplastic syndrome in adults with neurofibromatosis. Leuk. Lymphoma
Pendlebury, S.C., Bilous, M., and Langlands, A.O. (1995). Sarcomas follow-
ing radiation therapy for breast cancer: a report of three cases and a review
of the literature. Int. J. Radiat. Oncol. Biol. Phys. 31, 405–410.
Pinkel, D., Segraves, R., Sudar, D., Clark, S., Poole, I., Kowbel, D., Collins,
C., Kuo, W.L., Chen, C., Zhai, Y., et al. (1998). High resolution analysis of
DNA copy number variation using comparative genomic hybridization to
microarrays. Nat. Genet. 20, 207–211.
A R T I C L E
Plaat, B.E., Molenaar, W.M., Mastik, M.F., Hoekstra, H.J., te Meerman, G.J.,
and van den Berg, E. (1999). Computer-assisted cytogenetic analysis of 51
malignant peripheral-nerve-sheath tumors: sporadic vs. neurofibromatosis-
type-1-associated malignant schwannomas. Int. J. Cancer 83, 171–178.
Preston, D.L., Kusumi, S., Tomonaga, M., Izumi, S., Ron, E., Kuramoto, A.,
Kamada, N., Dohy, H., Matsuo, T., Matsui, T., et al. (1994). Cancer incidence
in atomic bomb survivors. Part III. Leukemia, lymphoma and multiple my-
eloma, 1950–1987. Radiat. Res. 137, S68–S97.
Remmelink, M., Decaestecker, C., Darro, F., Goldschmidt, D., Gebhart, M.,
Pasteels, J.L., Kiss, R., and Salmon, I. (1998). The in vitro influence of eight
hormones and growth factors on the proliferation of eight sarcoma cell lines.
J. Cancer Res. Clin. Oncol. 124, 155–164.
Rowley, J.D., Golomb, H.M., and Vardiman, J. (1977). Nonrandom chromo-
somal abnormalities in acute nonlymphocytic leukemia in patients treated
for Hodgkin disease and non-Hodgkin lymphoma. Blood 50, 759.
Sekyi-Otu, A., Bell, R.S., Ohashi, C., Pollak, M., and Andrulis, I.L. (1995).
Insulin-like growth factor 1 (IGF-1) receptors, IGF-1, and IGF-2 are ex-
pressed in primary human sarcomas. Cancer Res. 55, 129–134.
Side, L.E., and Shannon, K.M. (1998). The NF1 gene as a tumor suppressor.
In Neurofibromatosis Type 1, M. Upashyaya and D.N. Cooper, eds. (Oxford:
Bios Scientific Publishers), pp. 133–152.
Smith, P.G., and Doll, R. (1982). Mortality among patients with ankylosing
spondylitis after a single treatment course with x rays. Br. Med. J. (Clin.
Res. Ed.) 284, 449–460.
Smith, S.M., Le Beau, M.M., Huo, D., Karrison, T., Sobecks, R.M., Anastasi,
J., Vardiman, J.W., Rowley, J.D., and Larson, R.A. (2003). Clinical-cytoge-
netic associations in 306 patients with therapy-related myelodysplasia and
myeloid leukemia: the University of Chicago series. Blood 102, 43–52.
Snijders, A.M., Nowak, N.J., Huey, B., Fridlyand, J., Law, S., Conroy, J.,
Tokuyasu, T., Demir, K., Chiu, R., Mao, J.H., et al. (2005). Mapping segmen-
tal and sequence variations among laboratory mice using BAC array CGH.
Genome Res. 15, 302–311.
Stiller, C.A., Chessells, J.M., and Fitchett, M. (1994). Neurofibromatosis and
childhood leukemia/lymphoma: A population-based UKCCSG study. Br. J.
Cancer 70, 969–972.
CANCER CELL : OCTOBER 2005
Thirman, M.J., and Larson, R.A. (1996). Therapy-related myeloid leukemia.
Hematol. Oncol. Clin. North Am. 10, 293–320.
Tucker, M.A., Coleman, C.N., Cox, R.S., Varghese, A., and Rosenberg, S.A.
(1988). Risk of second cancers after treatment for Hodgkin’s disease. N.
Engl. J. Med. 318, 76–81.
van Leeuwen, F.E., Klokman, W.J., Hagenbeek, A., Noyon, R., van den Belt-
Dusebout, A.W., van Kerkhoff, E.H., van Heerde, P., and Somers, R. (1994).
Second cancer risk following Hodgkin’s disease: a 20-year follow-up study.
J. Clin. Oncol. 12, 312–325.
Vogel, K.S., Klesse, L.J., Velasco-Miguel, S., Meyers, K., Rushing, E.J., and
Parada, L.F. (1999). Mouse tumor model for neurofibromatosis type 1. Sci-
ence 286, 2176–2179.
Wang, J., Coltrera, M.D., and Gown, A.M. (1994). Cell proliferation in human
soft tissue tumors correlates with platelet-derived growth factor B chain
expression: an immunohistochemical and in situ hybridization study. Cancer
Res. 54, 560–564.
Watson, M.A., Perry, A., Tihan, T., Prayson, R.A., Guha, A., Bridge, J.,
Ferner, R., and Gutmann, D.H. (2004). Gene expression profiling reveals
unique molecular subtypes of Neurofibromatosis Type I-associated and
sporadic malignant peripheral nerve sheath tumors. Brain Pathol. 14, 297–
Weiss, S.W., Langloss, J.M., and Enzinger, F.M. (1983). Value of S-100 pro-
tein in the diagnosis of soft tissue tumors with particular reference to benign
and malignant Schwann cell tumors. Lab. Invest. 49, 299–308.
Wick, M.R., Swanson, P.E., Scheithauer, B.W., and Manivel, J.C. (1987).
Malignant peripheral nerve sheath tumor. An immunohistochemical study of
62 cases. Am. J. Clin. Pathol. 87, 425–433.
Wong, F.L., Boice, J.D., Jr., Abramson, D.H., Tarone, R.E., Kleinerman, R.A.,
Stovall, M., Goldman, M.B., Seddon, J.M., Tarbell, N., Fraumeni, J.F., Jr.,
and Li, F.P. (1997). Cancer incidence after retinoblastoma. Radiation dose
and sarcoma risk. JAMA 278, 1262–1267.
Yasui, K., Arii, S., Zhao, C., Imoto, I., Ueda, M., Nagai, H., Emi, M., and
Inazawa, J. (2002). TFDP1, CUL4A, and CDC16 identified as targets for
amplification at 13q34 in hepatocellular carcinomas. Hepatology 35,