letter
160 nature genetics • volume 32 • september 2002
We have used large-scale insertional mutagenesis to identify
functional landmarks relevant to cancer in the recently completed
mouse genome sequence. We infected Cdkn2a−/− mice
with Moloney murine leukemia virus (MoMuLV) to screen for
loci that can participate in tumorigenesis in collaboration with
loss of the Cdkn2a-encoded tumor suppressors p16INK4a and
p19ARF. Insertional mutagenesis by the latent retrovirus was
synergistic with loss of Cdkn2a expression, as indicated by a
marked acceleration in the development of both myeloid and
lymphoid tumors. We isolated 747 unique sequences flanking
retroviral integration sites and mapped them against the mouse
genome sequence databases from Celera and Ensembl. In addition
to 17 insertions targeting gene loci known to be cancerrelated,
we identified a total of 37 new common insertion sites
(CISs), of which 8 encode components of signaling pathways
that are involved in cancer. The effectiveness of large-scale
insertional mutagenesis in a sensitized genetic background is
demonstrated by the preference for activation of MAP kinase
signaling, collaborating with Cdkn2a loss in generating the lymphoid
and myeloid tumors. Collectively, our results show that
large-scale retroviral insertional mutagenesis in genetically predisposed
mice is useful both as a system for identifying genes
underlying cancer and as a genetic framework for the assignment
of such genes to specific oncogenic pathways.
The Trp53 and Rb pathways are two of the principal pathways
controlling cell proliferation that have been identified in
human and mouse cells. The Cdkn2a locus is involved in both
the p53 and Rb pathways by virtue of encoding p16INK4a, a
regulator of CDK4/6-mediated Rb1 phosphorylation, and
p19ARF, a modulator of Mdm2-mediated degradation of p53.
Loss of Cdkn2a expression enhances the survival and proliferative
potential of cells in the face of aberrant oncogenic signaling
and cell-cycle entry1. The importance of Cdkn2a in tumor suppression
is underscored by its frequent inactivation in human
tumors by deletion, mutation or epigenetic silencing2. Mice
deficient for both p16INK4a and p19ARF are viable but highly
Fig. 1 MoMuLV-induced tumorigenesis in Cdkn2a–/– mice. a, Disease-free survival of MoMuLV-infected Cdkn2a–/– mice after neonatal infection. Diamonds represent
MoMuLV-infected Cdkn2a–/– mice (n = 115); triangles represent MoMuLV-infected Cdkn2a–/+ mice (n = 37); squares represent MoMuLV-infected Cdkn2a+/+
mice (n = 27); and ‘x’ marks represent uninfected Cdkn2a–/– mice (n = 6). We did not detect any evidence of a tumor in four of the infected Cdkn2a–/– mice.
b−e, Tumors in MoMuLV-infected Cdkn2a–/– mice. b, Spleen of a MoMuLV-infected Cdkn2a–/– mouse with a nodular tumor at the lower left. c, Hematoxylin and
eosin staining of a spleen section of a MoMuLV-infected Cdkn2a–/– mouse with a histiocytic sarcoma, shown on the right at low magnification. d, The same section
as c at higher magnification, stained with anti-F4/80 and horseradish peroxidase. Tumor cells uniformly stained with the antibody, whereas the staining in
the normal portion of the liver is typical of that of hepatic mononuclear phagocytes29. e, Serial section from the same mouse as in d, stained with hematoxylin
and eosin. This section shows the tumor at the lower left and a relatively normal hepatic architecture at the upper right.
Genome-wide retroviral insertional tagging of genes
involved in cancer in Cdkn2a-deficient mice
Anders H. Lund1*, Geoffrey Turner2*, Alla Trubetskoy2, Els Verhoeven1, Ellen Wientjens1, Danielle Hulsman1,
Robert Russell3, Ronald A. DePinho4, Jack Lenz2 & Maarten van Lohuizen1
*These authors contributed equally to this work.
1Division of Molecular Genetics, The Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX, Amsterdam, The Netherlands. Departments of 2Molecular
Genetics and 3Pathology, Albert Einstein College of Medicine, Bronx, New York, USA. 4Department of Adult Oncology, Genetics and Medicine, Dana-Farber
Cancer Institute, Harvard Medical School, Boston, Massachusetts, USA. Correspondence should be addressed to J.L. (e-mail: lenz@aecom.yu.edu) or M.v.L.
(e-mail: m.v.lohuizen@nki.nl).
Published online: 19 August 2002, doi:10.1038/ng956
0
25
50
75
100
0 100 200 300 400
age (d)
MoMuLV
–/–
uninfected
–/–
MoMuLV
+/–
percentsurviving
wild type
a
b c
d e
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
nature genetics • volume 32 • september 2002 161
prone to tumors, succumbing to lymphomas and fibrosarcomas
early in life3. Transcriptional regulation of Cdkn2a is complex,
as evidenced by the growing number of proteins known to
regulate its expression. These proteins include Bmi1, Dmp1,
E2f family members, Ets, JunB, Myc and TBX2 (refs 4−10).
Aberrant upstream signaling through Abl, β-catenin and Src
and through the Ras–Raf–MAP kinase pathway can also effect
activation of Cdkn2a8,11−15.
To identify genes that act in collaboration with the combined
loss of p16INK4a and p19ARF in tumorigenesis, we infected
neonatal Cdkn2a–/– mice3 with MoMuLV. This retrovirus induces
tumors by insertional activation of proto-oncogenes or, more
rarely, by insertional inactivation of tumor-suppressor genes.
Although infection by MoMuLV elicits T-cell lymphomas, tumor
susceptibility and phenotype is strain-dependent16. Cdkn2a–/–
mice infected with MoMuLV developed tumors with a significantly
shorter mean latent period than that of Cdkn2a+/– or
Cdkn2a+/+ littermate controls that were infected with MoMuLV
(P < 0.0001, log-rank test) and that of uninfected Cdkn2a–/– siblings
(Fig. 1a). This finding demonstrates a synergy between viral
infection and loss of Cdkn2a expression in tumorigenesis.
We observed that 80% of Cdkn2a–/– mice infected with MoMuLV
developed lymphomas. Approximately 60% of these mice had Tcell,
12% B-cell, and 8% mixed B- and T-cell lymphomas. In addition,
approximately 55% had histiocytic sarcomas, which were
distinctive nodular tumors (Fig. 1b−c), usually in the liver or spleen.
These sarcomas stained positive for the macrophage marker F4/80
Table 1 • Integration frequencies into known common
insertion loci
Cdkn2a–/– tumorsa wt and +/– tumors
Myc 10/51 20% 8/46 17%
Nmyc1 1/16 6% 2/23 9%
Pim1 5/51 10% 0/45 0%
Gfi1 14/51 27% 13/47 28%
Bmi1 0/51 0% 0/47 0%
aFrequencies of genomic rearrangements observed per number of tumors
analyzed.
Fig. 2 Mapping of 565 unique retroviral insertion sites onto the mouse genome assembly from Celera Genomics. Blue lines indicate single MoMuLV insertions,
green lines loci previously identified by retroviral insertional mutagenesis, and red lines newly discovered cancer-related CIS loci, with candidate gene names in
red. Chromosome ideograms were modified with permission from C.V. Beechey and E.P. Evans (see Methods). Chromosome Y has not been assembled by Celera
Genomics, and no tags were assigned to it.
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
162 nature genetics • volume 32 • september 2002
(Fig. 1d) and were transplantable into immunodeficient SCID mice
(data not shown). Moreover, 40% of MoMuLV-infected Cdkn2a–/–
mice sustained both lymphomas and histiocytic sarcomas. Histiocytic
sarcomas occurred in four uninfected Cdkn2a–/– mice, albeit
at later times than in MoMuLV-infected mice. We observed histiocytic
sarcomas in approximately 10% of Cdkn2a+/– mice infected
with MoMuLV, but found none in infected wildtype controls, indicating
that Cdkn2a deficiency or reduction predisposes to histiocytic
sarcomas. All of the Cdkn2a+/– and Cdkn2a+/+ mice infected
with MoMuLV developed lymphomas, with no significant difference
in the rate of tumorigenesis.
On the basis of the demonstrated synergy between MoMuLV
insertional mutagenesis and loss of Cdkn2a function, we began a
large-scale analysis of retroviral insertion site sequences to identify
new oncogenes and tumor-suppressor genes. First, Southern-blot
analysis of known CISs in MoMuLV-induced tumors revealed that
genetic ablation of Cdkn2a did not significantly alter the insertion
frequencies into these common loci (Table 1). Although these
observations are based on a limited set of genes, it is likely that
many specific tumorigenic events may benefit from loss of Cdkn2a
expression, thus sensitizing the screen towards the identification of
a broader spectrum of loci associated with cancer.
From a panel of 104 tumors—57 lymphoid (55%) and 47
myeloid (45%)—we isolated a total of 747 unique MoMuLV integration
sites and directly sequenced them using a combination of
inverse PCR and splinkerette-aided insertion site amplification.
Homology searches in GenBank and in the nearly complete
mouse genome databases from Celera Genomics and Ensembl
allowed us to unambiguously map 565 viral insertions (Fig. 2).
Several viral integrations were found to target the same locus in
independent tumors, thus defining a CIS and indicating the presence
of a cancer-associated gene16. The mapping data revealed
that 163 (28.8%) of the sequences were clustered in 45 CISs; 37 of
these have not been described previously (Table 2). In addition,
we identified single viral integrations into many loci with known
links to cell-cycle regulation or cancer pathogenesis, such as Cdk2,
Cdk8, Cdc25c, Mapk1, Fra1 and E2f2. Of the new CISs, 14 were
specific to lymphomas, 6 were specific to the histiocytic sarcomas,
Table 2 • CISs identified in tumors of Cdkn2a–/– mice
Number GenBank Ensembl mouse Celera mouse Human
of Candidate accession Mouse chromosome chromosome chromosome
Locusa Celera IDb tags geneb number chromosome positionc positionc position Cell typed
NKI-1 No ID 2 Rgs1 NM_015811 1 145163982−145186084 135762279−no hit 1q31 L
NKI-2 mCG21261 2 Ggta1 NM_010283 2 35932137−35932421 31638654−31639099 9q33-q34 L
NKI-3 mCG8151 3 Zfhx1b NM_015753 2 45582724−45586359 41327870−41331182 2q22 L
NKI-4 mCG64382 8 Ptpn1 NM_011201 2 168789627−168845131 165422527−165471534 20q13.1−q13.2 L, HS
NKI-5 mCG8829 2 Mef2d S68893 3 88925116−88929012 82005550−82008904 1q12−q23 L, HS
mCG8841 Iqgap1 NM_016721 15q26
NKI-6 mCG16756 3 RIKEN cDNA NM_029789 3 95571261−95602603 87303050−87427968 1p21 L, HS
NKI-7 mCG22374 2 Nfkb1 NM_008689 3 106285108−no hit 127772992−127857753 4q24 L, HS
NKI-8 mCG7369 2 Rabggtb NM_011231 3 154489997−154490618 146325593−146326261 1p31 L
NKI-9 mCG2332 2 putative AK013175 4 43733132−43783247 41397143−41447334 ND L, HS
NKI-10 mCG65233 3 RIKEN cDNA XM_13153610 4 104505611−104665343 101393203−101543172 ND L
NKI-11 mCG17461 3 Runx3 NM_019732 4 132639134−132789991 131734840−131779243 1p36 L, HS
NKI-12 mCG7764 2 RIKEN cDNA NM_026383 4 133460472−133463348 132194843−132197692 ND L, HS
NKI-13 mCG23363 2 Gnb1 NM_008142 4 150934253−151010713 152078608−152155328 1pter−p31.2 L
NKI-14 mCG10252 2 Klf3 NM_008453 5 63873382−63995956 59883104−60002945 4p16.1−p15.2 L, HS
NKI-15 mCG7209 2 Kdr NM_010612 5 75154747−75155055 70951478−70952215 4q12 L, HS
NKI-16 mCG11116 3 Camkk2, mo NM_031338 5 120807339−120868275 116383995−116512644 12q24.2 L, HS
NKI-17 mCG16286 2 KIAA0053, mo NM_014882 6 88356371−88401020 86035121−86077929 ND L, HS
NKI-18 mCG19794 2 Hcph NM_013545 6 125626101−125626125 122564894−122564979 12p13 HS
NKI-19 mCG62286 2 EST AV083176 6 130155495−130155708 127047268−127047481 ND L
NKI-20 mCG13312 3 Kras2 NM_031515 6 145911824−145916127 140215453−140219542 12p12.1 L, HS
NKI-21 mCG7264 2 Plaur NM_011113 7 16611803−16692064 9871320−9950546 19q13 L
NKI-22 mCG51745 2 Rras2 NM_025846 7 103758158−103810818 94601976− 94604902 11pter−p15.5 L
NKI-23 mCG22407 3 PSK, mo NM_016151 7 116737933−116883522 112639555−112918319 16 L, HS
mCG22413 Maz AA915408 16p11.2
NKI-24 mCG13090 2 ZNF220 NM_006766 8 21537403−21537452 18309141−18309166 8p11 HS
mouse ortholog
NKI-25 mCG57225 3 EST BB653745 8 126716162−126866935 120231285−120365173 ND L, HS
NKI-26 mCG21612 2 NHE7, AF298591 9 95206300−95210881 90165786−90170216 X L
mouse ortholog
NKI-27 mCG2827 6 Hbs1l NM_019702 10 21080626−21195919 17946731−18022547 6q23−q24 L
NKI-28 mCG5175 2 Galgt1 NM_008080 10 127813456−127815198 109653266−109655579 12q13.3 L
NKI-29 mCG7666 6 Rbp1B, mo AF199338 11 88413609−88417214 87602127−87605591 17q25 L, HS
NKI-30 mCG1480 2 Dgke NM_019505 11 89769880−89818410 90822843−90872179 17q22 HS
NKI-31 mCG20534, 2 Fkbp6 NM_033571 11 101188993−101220645 102752458−102783766 7q11.23 L
NKI-32 mCG20453 2 Kif13a AB037923 13 46365600−46409349 41811568−41852808 6p23 HS
NKI-33 mCG55520 2 EST BF549796 13 110028533−110034445 101660853−101666913 ND HS
NKI-34 mCG62490 2 EST AV287890 17 3423327−3423365 891751−892712 ND HS
NKI-35 mCG3890 4 Cabp2 NM_013878 19 6700612−6910493 2625769−2835944 11q13.1 L, HS
NKI-36 mCG20866 2 DUSP5, mo NM_004419 19 53530620−5354765- 52164129−52180624 10q25 L
NKI-37 mCG5050 2 Elf4 NM_062654 X 34081336−34081401 33637118−33637542 Xq26 L, HS
aNewly discovered common insertion loci. Loci specific for the Cdkn2a–/– background, relative to previous screens27,28, are indicated in bold. NKI, Netherlands
Cancer Institute. bCandidate genes are based on their proximity to integrations. mo, mouse ortholog. cThe position of both proximal and distal integration is
indicated. dL, lymphoma; HS, histiocytic sarcoma; ND, not determined.
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
nature genetics • volume 32 • september 2002 163
and the remaining 17 consisted of sequences from both lymphomas
and histiocytic sarcomas. As many mice had both histologically
evident lymphoma and histiocytic sarcoma, it is likely
that at least a fraction of the DNA samples were derived from tissue
that was a mixture of both types of tumors.
Many of the newly discovered CISs encode proteins that can be
assigned to functional pathways previously associated with mitogenic
processes or transformation. The identification of two Ras
homologs, Kras2 and Rras2, is consistent with a large body of data
establishing cooperative interactions between activated H-Ras and
loss of Cdkn2a function15. Viral insertions near both Gnb1 (at
locus NKI-13) and Gnb2 (single insertion) underline the importance
of G-protein βγ-subunits in the regulation of signaling pathways
involved in cell proliferation and survival17. Moreover,
NKI-1, NKI-5 and NKI-17 also encode G-protein regulatory factors,
and single viral insertions were mapped near Rap1a, Rap1b
and an uncharacterized member of the Ras gene family. These
findings further highlight the importance of deregulated Ras-like
signaling in tumorigenesis. Three CISs, NKI-4, NKI-18 and NKI-
36, target phosphatase genes. Both Ptpn1 and Hcph have been
proposed to activate Src through removal of an inhibitory tyrosine
phosphate within the Src-homology 2 (SH2) domain18. The CIS
NKI-36 targets the mouse ortholog of the gene (DUSP5) encoding
dual-specificity protein phosphatase, which is thought to negatively
regulate the activity of the ERK1 MAP kinase19. Notably, the
insertions in NKI-36 are found within the gene, indicating a
potential tumor-suppressor function. The urokinase plasminogen
activator receptor (NKI-21) is central to several pathways implicated
in tumor progression. Aside from functions in angiogenesis
and invasion, the urokinase plasminogen activator receptor activates
the ERK1/2 MAP kinases, probably through interactions
with β-integrins and receptor tyrosine kinases20.
Transcription factors constituted a second large group of candidate
oncogenes. The factor NFκB, identified in NKI-7, has
important functions in cell-cycle progression, transformation
and survival21. Myeloid elf-1 like factor (Elf4, NKI-37), a member
of the Ets family, is a potent transcriptional activator in both
lymphoid and myeloid cells. Notably, Elf4 recognizes consensus
DNA-binding sequences similar to those recognized by other Ets
proteins, which directly upregulate the CDKN2A promoter7. The
protein ZNF220 is implicated in subtypes of acute myeloid
leukemia (AML) carrying a t(8;16)(p11;p13) translocation22.
The identification of the mouse ZNF220 ortholog as a CIS in histiocytic
sarcomas (NKI-24) points to the use of this model to
identify genes specifically involved in myeloid tumorigenesis.
The targeting of a member of the AML gene family, Runx3
(Aml2, NKI-11), is notable. Human RUNX3 is located at 1p36, a
region associated with structural aberrations in a number of neoplasias,
including acute and chronic leukemia. In addition,
RUNX3 was recently proposed to be an important tumor suppressor
in human gastric cancers23, raising the question of how
Runx family members may exert both oncogenic and tumorsuppressor
functions. In NKI-3, the integration hot spot targets
Zfhx1b, encoding a transcriptional repressor recently shown to
repress E-cadherin and promote tumor-cell migration24.
The validity and sensitivity of the screen is evidenced by the
identification of many genes previously implicated in MoMuLVinduced
tumorigenesis (Table 3). Specifically, 22 integrations targeted
Myc, a major collaborator in tumorigenesis with loss of
Cdkn2a25. Reciprocally, no insertions among the 104 tumors
analyzed targeted Bmi1, a known CIS previously shown to
encode a repressor of Cdkn2a4.
We confirmed the ability of the screen to identify collaborators
of Cdkn2a loss using three independent approaches. First,
Southern-blot analysis of a selected subset of the newly discovered
CISs in tumors from Cdkn2a–/– and wildtype littermate
controls revealed Cdkn2a–/–-specific MoMuLV insertions in 2 of
45, 4 of 45 and 2 of 45 tumors analyzed for Zfhx1b, Ptpn1 and
Rbp1B, respectively. We found no viral insertions at these loci in
a panel of 38 Cdkn2a+/– and Cdkn2a+/+ tumors.
Table 3 • Viral insertions into known and new CISs
Number Mouse Human
Locus of tags chromosome chromosome
Known CISs
Myc 22 15 8q24.12−q24.13
Gfi1 16 5 1p22
Tpl2 8 18 10p11.2
Pim1 7 17 6p21.2
Myb 4 10 6q22−q23
Ccnd1 3 7 11q13
Ccnd3 3 17 6p21
MycN1 3 12 2p24.1
Ets1 1 9 11q23.3
Evi26 1 14 10
Fgf8 1 19 10q24
Fos 1 12 14q24.3
His1 1 2 2q14−q21
Notch1 1 2 9q34.3
Sint1 1 11 17q25
Sox4 1 13 6p22.3
Zfp36 1 7 19q13.1
Newly discovered CISsa
Dkmi5 1 3 1q21
Dkmi17 1 15 8q24
Dkmi22 1 19 11q13
Evi38 1 1 1q32
Evi42 1 2 9q34
Evi50 1 2 NDb
Evi62 1 4 1p36.13−p36.12
Evi73 1 5 7p22−p21
Evi78 1 6 2p21
Evi82 1 7 19
Evi92 1 8 4
Evi102 1 10 ND
Evi105 1 10 ND
Evi112 1 11 17q11.2
Evi130 1 17 6p21
Evi134 1 17 2p22.3−p21
Evi146 1 6 12pter−p13.31
Evi148 1 11 5q31.1
aSingle insertions from Cdkn2a–/– tumors identified as CISs27,28. bNot determined.
0
0
2
4
6
8
10
12
14
16
18
1 2 3 4 5
time (d)
relativegrowth
wt MEFs + vector
wt MEFs + Tpl2
wt MEFs + RAS
Cdkn2a–/– MEFs + vector
Cdkn2a–/– MEFs + Tpl2
Cdkn2a–/– MEFs + RAS
Fig. 3 Ectopic expression of Tpl2 collaborates with loss of Cdkn2a. The graph
shows proliferation curves for wildtype (wt) and Cdkn2a–/– primary mouse
embryo fibroblasts infected with pBabepuro-Tpl2, pBabepuro-RASV12 or
empty vector. Error bars indicate standard deviations for triplicate samples.
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
164 nature genetics • volume 32 • september 2002
Second, we obtained experimental validation for a notable
CIS within the screen: Tpl2, encoding a previously identified
MAP kinase kinase kinase known to activate several MAP-kinase
pathways26. Extending on the data generated from insertion site
retrieval, Southern-blot analyses revealed MoMuLV insertions
into Tpl2 in 16 of 24 myeloid tumors analyzed, consistent with
an important role of Tpl2 in myeloid tumorigenesis. As myeloid
tumors arose exclusively in the Cdkn2a–/– background, no control
myeloid tumors were available for analysis. Whereas five
Tpl2 insertions were found among tumors from mice sustaining
mixed lymphoid and myeloid tumors or harboring sporadically
appearing Cdkn2a–/– tumor types, no MoMuLV
integrations were found at Tpl2 among 47 Cdkn2a–/– and 30
Cdkn2a+/+ or Cdkn2a+/– lymphomas. Ectopic expression of
Tpl2 in mouse embryo fibroblasts (MEFs) resulted in diminished
growth, resembling that seen in wildtype MEFs expressing
oncogenic Ras. In contrast, the proliferation rate of
Cdkn2a–/– MEFs infected with a Tpl2-expressing retroviral vector
was similar to that of mock-infected cells (Fig. 3), and the
cells eventually reached a higher cell density than did mockinfected
Cdkn2a–/– MEFs (data not shown). These findings
demonstrate a collaboration between Tpl2 and loss of Cdkn2a
and further support MAP-kinase activation as an important
feature in Cdkn2a activation.
Third, to analyze the extent to which we have identified genes
that act in collaboration with loss of the tumor suppressor
Cdkn2a, we carried out extensive comparative data analyses
against the retroviral tagging screens described in the accompanying
papers by Suzuki et al.27 and Mikkers et al.28. Of the 37
new CISs we identified, 18 loci were also identified as single or
multiple hits in the screens described therein27,28. Thus, 19 new
loci associated with cancer were identified on the Cdkn2a–/–
background; these loci are targets for future investigations
(Table 2). In addition, at several loci where we found overlaps
among the three screens, we saw a significant skewing toward
specific targeting of Cdkn2a–/–. For example, we found nine
insertions in Ptpn1 on the Cdkn2a–/– background, whereas this
locus is hit only once in each of the screens by Mikkers et al.28
and Suzuki et al.27 (P < 0.001, Fisher exact test). Similarly, targeting
of the Rbp1B and Zfhx1b loci is significantly biased
towards the Cdkn2a–/– background (P < 0.02 and P < 0.05,
respectively). Although genetic predispositions, such as loss of
Cdkn2a, favor activation of a specific subset of oncogenic pathways,
the observed overlap between the three screens (including
Myc, Myb and Gfi1) probably reflects common events in cancer
that fall outside the functional groups identified in each of the
systems used.
Single insertions from the Cdkn2a–/– screen overlapped 18 of
the CISs identified by Mikkers et al.28 or Suzuki et al.27, raising
the proportion of insertions within CISs to 36.4%. Combined
CIS analysis of the three screens (where a CIS is defined as three
insertions within a window of 100 kb) revealed 15 additional CIS
loci28. This finding underscores the cumulative nature of largescale
insertional mutagenesis screens, the data from which have
been assembled into a publicly available database.
Many of the candidate genes identified in this mouse model
have been linked to cancer in humans or belong to gene families
associated with human neoplastic diseases. Our identification
of 37 new CISs, along with the large number identified in the
accompanying papers27,28, indicates that the ability to contribute
to tumorigenesis is a potential property of many genes,
not just a crucial few. Apart from its use in discovering new
cancer genes, the power of the approach lies in the ability to
focus on, and place these genes in, a limited set of signaling
pathways associated with cancer. Owing to the specific genetic
predisposition used in this screen, the majority of the loci can
probably be functionally linked to the Cdkn2a locus, when the
relevant candidate cancer genes have been verified, as is
demonstrated by the preference for MAP-kinase activation via
Tpl2. Evidence exists from both human tumors and rodent
model systems showing that inactivation of Cdkn2a or downstream
pathways is crucial in tumorigenesis. Thus, identification
and functional characterization of genetic components
acting on or in collaboration with these pathways is central to
the understanding of many cancers.
Methods
MoMuLV tumorigenesis. Cdkn2a–/– and Cdkn2a+/– mice and wildtype
littermates3 were injected intraperitoneally with 105 infectious units of
MoMuLV within 72 h of birth. We killed diseased mice, necropsied them
and then carried out Cdkn2a genotyping by PCR. We froze most of each
tumor for genetic studies. Fragments of tumors were fixed in either 10%
formalin or Bouin’s fixative, embedded in paraffin, sectioned and
stained with hematoxylin and eosin. T-cell lymphomas appeared as lymphoblastic-like,
diffuse tumors with pale-staining nuclei that were vesicular,
pleiomorphic and frequently large. B-cell lymphomas appeared as
lymphoblastic-like, diffuse tumors with moderately basophilic nuclei
and prominent nucleoli. In some B-cell lymphomas, a fraction of cells
showed immunoblastic morphology. We analyzed lymphomas by Southern
blotting to assess T-cell receptor and immunoglobulin gene
rearrangements (experimental details are available upon request). Cellsurface
markers for hematopoietic lineages were determined by
immunohistochemistry. We carried out histiocytic sarcoma transplantation
into immunodeficient mice by subcutaneous injection, near the
scapulae of BALB/c SCID mice, of approximately 4.5 × 105 disaggregated
cells from macroscopically visible histiocytic liver nodules in virally
infected Cdkn2a–/– mice. Tumors occurring in the livers of recipient mice
histologically resembled those of donors. We confirmed that these
tumors were of donor origin by PCR genotyping for the Cdkn2a allele
with exon 2 deleted.
Retrieval and analysis of MoMuLV integration site sequences. We
amplified 463 viral insertion sites using splinkerette-aided amplification
essentially as described in Mikkers et al.28. We retrieved 284 insertionsite
sequences using inverse PCR on SacII-digested tumor DNA as
described by Suzuki et al.27. All sequences have been deposited with
GenBank (for exact insertion site positions, see Web Tables A–C online
and the website listed below). Sequences of the MoMuLV−specific
primers used for the nested inverse PCR are available upon request. We
purified amplification products from agarose gels and subjected them to
direct automated sequencing. To identify candidate genes, we filtered
the insertion-site sequences for the presence of repetitive DNA and carried
out homology searches using BLASTn in GenBank databases and in
the mouse genome sequence databases from Celera Genomics (version
12) and Ensembl (Mouse Genome Sequencing Consortium, Mouse
Assembly 3 February 2002 freeze).
Whereas CISs traditionally have been demonstrated using Southern
blotting, the presence of the complete mouse genome allows for in silico
CIS determination. Following a previously described statistical analysis28,
we defined retroviral common insertion sites as two or more integrations
within 26 kb, or three insertions within 300 kb. For a set of 500
insertions, these windows give a tolerable statistically calculated background
of approximately 2.5 CISs occurring at random. For insertions
flanking a gene, the accepted distance between insertions was set to
100 kb. Although the functional distance between viral insertions and
candidate oncogenes is known to differ between loci, the statistical
threshold set here is in accordance with viral integration patterns surrounding
previously characterized common insertion sites. In some
loci, no candidate gene could be found in the immediate vicinity of an
insertion. Moreover, a current limitation to the insertional mutagenesis
approach is the limited extent and fidelity of the mouse genome annotation.
This will improve over time. When analyzing overlaps between the
Cdkn2a–/– screen and the screens carried out by Mikkers et al.28 and
Suzuki et al.27, sequence tags falling within 100 kb on either side of a CIS
were defined as overlapping.
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
nature genetics • volume 32 • september 2002 165
Cell-culture assays. We cloned the Tpl2 cDNA26 into pBabe-puro. The
retroviral vector pBabe-RASV12, procedures for growth, retroviral infection
and proliferation analyses of primary mouse embryo fibroblasts have been
described25. Briefly, MEFs were infected at passage 2, selected with 2 µg ml−1
of puromycin at passage 3 and seeded in triplicate with 25.000 cells per well
in six-well plates at passage 4 for proliferation analysis. We measured relative
proliferation using crystal violet staining. We repeated the experiment three
times with independent batches of MEFs.
URLs. Data from the insertional mutagenesis screens are available at
http://protagdb.nki.nl; Mo-MLV viral insertion site positions can be found
at http://protagdb.nki.nl; the reference Standard Ideogram/Anomaly Breakpoints
of the Mouse (Beechey, C.V. & Evans, E.P.) can be found at
http://www.mgu.har.mrc.ac.uk/anomaly/anomaly-intro.html.
GenBank accession numbers. The retroviral flanking sequences have been
deposited to GenBank, submission numbers AF524262–AF524826.
Note: Supplementary information is available on the Nature
Genetics website.
Acknowledgments
We thank L. Johnson, R. Stanley, D. Polsky, C. Cordon-Cardo, G. Hart,
L. Spunk Jacobsen and R. Westrop for assistance. This work was supported by
a grant from the National Institutes of Health (to J.L.) and a US National
Institutes of Health training grant. A.H.L. was supported by a grant from the
Danish Medical Research Council.
Competing interests statement
The authors declare that they have no competing financial interests.
Received 23 April; accepted 8 July 2002.
1. Sherr, C.J. The Cdkn2a network in tumour suppression. Nature Rev. Mol. Cell Biol.
2, 731−737 (2001).
2. Ruas, M. & Peters, G. The p16INK4a/CDKN2A tumor suppressor and its relatives.
Biochim. Biophys. Acta 1378, F115−177 (1998).
3. Serrano, M. et al. Role of the INK4a locus in tumor suppression and cell mortality.
Cell 85, 27−37 (1996).
4. Jacobs, J.J., Kieboom, K., Marino, S., DePinho, R.A. & van Lohuizen, M. The
oncogene and Polycomb-group gene bmi-1 regulates cell proliferation and
senescence through the ink4a locus. Nature 397, 164−168 (1999).
5. Inoue, K., Roussel, M.F. & Sherr, C.J. Induction of ARF tumor suppressor gene
expression and cell cycle arrest by transcription factor DMP1. Proc. Natl Acad. Sci.
USA 96, 3993−3998 (1999).
6. Bates, S. et al. p14ARF links the tumour suppressors RB and p53. Nature 395, 124−
125 (1998).
7. Ohtani, N. et al. Opposing effects of Ets and Id proteins on p16INK4a expression
during cellular senescence. Nature 409, 1067−1070 (2001).
8. Passegue, E. & Wagner, E.F. JunB suppresses cell proliferation by transcriptional
activation of p16(INK4a) expression. EMBO J. 19, 2969−2979 (2000).
9. Zindy, F. et al. Myc signaling via the ARF tumor suppressor regulates p53dependent
apoptosis and immortalization. Genes Dev. 12, 2424−2433 (1998).
10. Jacobs, J.J. et al. Senescence bypass screen identifies TBX2, which represses
Cdkn2a (p19(ARF)) and is amplified in a subset of human breast cancers. Nature
Genet. 26, 291−299 (2000).
11. Damalas, A., Kahan, S., Shtutma -dependent growth arrest and cooperates with
Ras in transformation. EMBO J. 20, 4912−4922 (2001).
12. Zhu, J., Woods, D., McMahon, M. & Bishop, J.M. Senescence of human fibroblasts
induced by oncogenic Raf. Genes Dev. 12, 2997−3007 (1998).
13. Radfar, A., Unnikrishnan, I., Lee, H.W., DePinho, R.A. & Rosenberg, N. p19(Arf)
induces p53-dependent apoptosis during abelson virus-mediated pre-B cell
transformation. Proc. Natl Acad. Sci. USA 95, 13194−13199 (1998).
14. Lin, A.W. et al. Premature senescence involving p53 and p16 is activated in
response to constitutive MEK/MAPK mitogenic signaling. Genes Dev. 12, 3008−
3019 (1998).
15. Serrano, M., Lin, A.W., McCurrach, M.E., Beach, D. & Lowe, S.W. Oncogenic ras
provokes premature cell senescence associated with accumulation of p53 and
p16INK4a. Cell 88, 593−602 (1997).
16. Jonkers, J. & Berns, A. Retroviral insertional mutagenesis as a strategy to identify
cancer genes. Biochim. Biophys. Acta. 1287, 29−57 (1996).
17. Radhika, V. & Dhanasekaran, N. Transforming G proteins. Oncogene 20, 1607−
1614 (2001).
18. Bjorge, J.D., Jakymiw, A. & Fujita, D.J. Selected glimpses into the activation and
function of Src kinase. Oncogene 19, 5620−5035 (2000).
19. Ishibashi, T., Bottaro, D.P., Michieli, P., Kelley, C.A. & Aaronson, S.A. A novel dual
specificity phosphatase induced by serum stimulation and heat shock. J. Biol.
Chem. 269, 29897−29902 (1994).
20. Aguirre Ghiso, J.A., Kovalski, K. & Ossowski, L. Tumor dormancy induced by
downregulation of urokinase receptor in human carcinoma involves integrin and
MAPK signaling. J. Cell Biol. 147, 89−104 (1999).
21. Joyce, D. et al. NF-κB and cell-cycle regulation: the cyclin connection. Cytokine
Growth Factor Rev. 12, 73−90 (2001).
22. Borrow, J. et al. The translocation t(8;16)(p11;p13) of acute myeloid leukaemia
fuses a putative acetyltransferase to the CREB-binding protein. Nature Genet. 14,
33−41 (1996).
23. Li, Q.-L. et al. Causal relationship between the loss of RUNX3 expression and
gastric cancer. Cell 109, 113−124 (2002).
24. Comijn, J. et al. The two-handed E box binding zinc finger protein SIP1
downregulates E-cadherin and induces invasion. Mol. Cell 7, 1267−1278 (2001).
25. Jacobs, J.J. et al. Bmi-1 collaborates with c-Myc in tumorigenesis by inhibiting cMyc-induced
apoptosis via Cdkn2a. Genes Dev. 13, 2678–2690 (1999).
26. Patriotis, C., Makris, A., Chernoff, J. & Tsichlis, P.N. Tpl-2 acts in concert with Ras
and Raf-1 to activate mitogen-activated protein kinase. Proc. Natl Acad. Sci. USA
91, 9755−9759 (1994).
27. Suzuki, T. et al. New genes involved in cancer identified by retroviral tagging.
Nature Genet. 32 86–94 (2002).
28. Mikkers, H. et al. High-throughput retroviral tagging to identify components of
specific signaling pathways in cancer. Nature Genet. 32 73–79 (2002).
29. Roth, P., Dominguez, M.G. & Stanley, E.R. The effects of colony-stimulating factor-
1 on the distribution of mononuclear phagocytes in the developing osteopetrotic
mouse. Blood 91, 3773−3783 (1998).
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics