letter
nature genetics • volume 32 • september 2002 153
High-throughput retroviral tagging to identify
components of specific signaling pathways in cancer
Harald Mikkers1, John Allen1, Puck Knipscheer1, Lieke Romeyn1, Augustinus Hart2, Edwin Vink1
& Anton Berns1
1Division of Molecular Genetics and Centre of Biomedical Genetics, 2Division of Radiotherapy, Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX
Amsterdam, The Netherlands. Correspondence should be addressed to A.B. (e-mail: a.berns@nki.nl).
Genetic screens carried out in lower organisms such as yeast1,
Drosophila melanogaster2 and Caenorhabditis elegans3 have
revealed many signaling pathways. For example, components
of the RAS signaling cascade were identified using a mutant
eye phenotype in D. melanogaster as a readout2. Screening is
usually based on enhancing or suppressing a phenotype by
way of a known mutation in a particular signaling pathway.
Such in vivo screens have been difficult to carry out in mammals,
however, owing to their relatively long generation times
and the limited number of animals that can be screened. Here
we describe an in vivo mammalian genetic screen used to identify
components of pathways contributing to oncogenic transformation.
We applied retroviral insertional mutagenesis in
Myc transgenic (EµMyc) mice lacking expression of Pim1 and
Pim2 to search for genes that can substitute for Pim1 and Pim2
in lymphomagenesis. We determined the chromosomal positions
of 477 retroviral insertion sites (RISs) derived from 38
tumors from EµMyc Pim1–/– Pim2–/– mice and 27 tumors from
EµMyc control mice using the Ensembl and Celera annotated
mouse genome databases. There were 52 sites occupied by
proviruses in more than one tumor. These common insertion
sites (CISs) are likely to contain genes contributing to tumorigenesis.
Comparison of the RISs in tumors of Pim-null mice with
the RISs in tumors of EµMyc control mice indicated that 10 of
the 52 CISs belong to the Pim complementation group. In addition,
we found that Pim3 is selectively activated in Pim-null
tumor cells, which supports the validity of our approach.
Retroviral insertions in the genome can transform host cells by
activating proto-oncogenes or inactivating tumor-suppressor
genes4. Multiple rounds of retroviral insertional mutagenesis
yield a full-blown tumor in which proviral insertions mark the
genes collaborating in stepwise tumor development. Thus,
retroviral insertions are instrumental in the clonal outgrowth
of the incipient tumor cell. In accordance with this notion, two
or three CISs within a single tumor are often occupied by
proviruses. We have previously shown co-activation of the Pim
family of serine/threonine
kinases and either Myc or
Nmyc1 in retrovirus-induced
tumors5,6. The cooperation
between the Myc and Pim
proto-oncogenes was proven
using transgenic experiments
in which EµMycEµPim1 and
EµMycEµPim2 double-transgenic
mice succumbed around
birth to pre-B cell leukemia7,8.
Although the frequent retroviral
activation of Pim1 established
the role of the Pim
genes in retrovirus-induced
lymphomagenesis, the crucial
downstream targets of the
Pim kinases are elusive. Candidate
Pim substrates such as
P100 (ref. 9), CDC25A (ref.
10), HP1γ (ref. 11),
TFAF2/SNX6 (ref. 12), SOCS1
(ref. 13) and NFATC (ref. 14)
have been described, but it is
still unclear to what extent
they contribute to Pimmediated
transformation.
Published online: 19 August 2002, doi:10.1038/ng950
Y
X
Z
Pim1 Pim2
gene X
gene Y
gene Z
tumor
Pim1 Pim2
Pim2
tumor
Pim1 Pim2
Pim2
Pim1
EµMyc
percentage
100
tumor
EµMyc Pim1–/–
EµMyc Pim1–/–
Pim2–/–
Fig. 1 Retroviral tagging in lymphoma-prone EµMyc mice that are sensitized to activation of the Pim pathway.
a, MoMuLV infection of EµMyc mice yields lymphomas, of which 40% and 15% have retrovirus-activated Pim1 and Pim2
alleles, respectively. b, In 90% of the lymphomas generated in EµMyc mice lacking expression of Pim1, the Pim pathway
has been activated through proviral insertions near Pim2. c, Retroviral insertional mutagenesis in EµMyc Pim1–/– Pim2–/–
mice is expected to yield lymphomas with activated oncogenic Pim signaling, either by mutation of a gene in a parallel
pathway (Y), a gene downstream of Pim (X) or a Pim-related gene (Z).
a b c
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
Fig. 2 Isolation of the genomic DNA sequences flanking the provirus using a
splinkerette-based PCR approach. a, Schematic representation of the amplification
procedure; provirus (green), splinkerette (gray). Because a random PCR
amplification of the sequences flanking the provirus was preferred, genomic
tumor DNA was digested with the restriction enzyme BstYI recognizing PuGATC-Py,
and not with a methylation-sensitive enzyme that selects for proviral
insertions in promoter regions of genes18,30. b, Southern-blot analysis using a
viral LTR probe shows the number of proviral insertions in a tumor. The lymphoma
analyzed carries a large number of retroviral insertions, indicating that this tumor is of oligoclonal origin. c, First radioactive splinkerette-based PCR on
the same tumor. The asterisk marks the internal MoMuLV fragment amplified. d, Nested radioactive splinkerette-based PCR on the excised fragments. e, Final
PCR amplification of the provirus flanking sequences yields ready-to-sequence DNA fragments.
letter
154 nature genetics • volume 32 • september 2002
To gain more insight into the oncogenic signaling network in
which the Pim proteins act and to identify crucial downstream
Pim targets, we established a mammalian in vivo enhancer
screen similar to the genetic screens carried out in lower organisms.
The synergism between Myc and Pim in lymphomagenesis
probably sensitizes those mice that are deficient for Pim but
express high levels of Myc to developing lymphomas in which
genes acting either downstream of or parallel to Pim have been
mutated. In fact, the percentage of retroviral activations of Pim
observed in EµMyc Pim1–/– mice was almost twice that observed
in EµMyc mice (Fig. 1a,b). We observed proviral activations of
Pim2, which encodes a protein that is 57% identical to Pim1
(ref. 15), in 90% of the lymphomas in Pim1-null mice. These
observations underscore the selective advantage of Pim activation
in the presence of high Myc levels and suggest a strategy for
identifying genes that rescue loss of Pim function in lymphomas
containing activated Myc. In the current study we implement
this strategy by infecting EµMyc newborns lacking expression of
both Pim1 and Pim2 with Moloney murine leukemia virus
(MoMuLV) (Fig. 1c), carrying out high-throughput sequence
analyses of the proviral insertion sites, mapping the insertions
and nearby candidate target genes (using the annotated mouse
Table 1 • Predicted frequencies of random proviral insertions in the mouse genome
Expected number of random CISsa Two insertions Three insertions Four insertions
Number Efr = Efr = Efr = Efr = Efr = Efr = Efr = Efr = Efr = Efr = Efr = Efr =
of tags 0.001 0.005 0.01 0.001 0.005 0.01 0.001 0.005 0.01 0.001 0.005 0.01
10,000 10 50 100 0.26 kb 1.3 kb 2.6 kb 12 kb 27 kb 39 kb 50 kb 88 kb 113 kb
5,000 5 25 50 0.5 kb 2.6 kb 5.2 kb 24 kb 54 kb 77 kb 99 kb 176 kb 227 kb
2,500 2.5 12.5 25 1.04 kb 5.2 kb 10.4 kb 47 kb 108 kb 155 kb 198 kb 351 kb 454 kb
2,000 2 10 20 1.3 kb 6.5 kb 13 kb 59 kb 135 kb 193 kb 248 kb 439 kb 567 kb
1,000 1 5 10 2.6 kb 13 kb 26 kb 118 kb 269 kb 386 kb 495 kb 878 kb 1,134 kb
500 0.5 2.5 5 5.2 kb 26 kb 52 kb 236 kb 538 kb 772 kb 991 kb 1,757 kb 2,267 kb
Ignoring end-of-chromosome effects, random proviral insertions into the mouse genome follow a Poisson distribution. The expected fraction (Efr) indicates the
fraction of the total number of proviral insertion sites expected to be random CIS clusters within the specified distance. For example, 2,500 tags will contain 2.5
CISs consisting of 2 random insertions within 1.04 kb, 2.5 CISs of 3 random insertions within 47 kb, and so on. The calculated distances are based on the available
mouse genome sequence at Celera (2.6 × 106 kb). aThe expected number of CISs is the mean number of clusters for n = ∞ experiments.
x x
xx x
tumor DNA restricted
with enzyme X is
ligated to the
splinkerette
PCR with first
primer pair
PCR with
nested
primer pair
sequence and BLAST search
against the annotated mouse
genome databases at
Ensembl and Celera
1.0 Kb
0.5 Kb
*
1.0 kb
0.5 kb
a b c d
e
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
nature genetics • volume 32 • september 2002 155
genomic sequence database at Celera Genomics and Ensembl16)
and assigning CISs to complementation groups.
The concept of insertional mutagenesis has so far been based
on the assumption that the existence of CISs is due to a selective
advantage associated with insertion in that site. Identifying a large
number of insertion sites (as in this study) may, however, increase
the likelihood of incorrectly labeling a site as a CIS. It is therefore
necessary to attach parameters of significance to the definition of
a CIS. We propose to assign to each CIS identified an estimate of
the non-randomness of its occurrence. In a set of 500 CISs, random
insertion events may account for approximately 2.5 clusters,
of 2 insertions each, within 26 kb (Table 1). Random occurrence
of larger CIS clusters (consisting of three or more retroviral
insertions) in a data set of 500 is much less likely. It should be
noted that the predicted occurrences of CISs do not objectively
equal the likelihood of insertions contributing to tumorigenesis.
This is because the measure of non-randomness does not
take into account preferential insertion due to chromatin
structure or sequence context. ‘Cold’ and ‘hot’ spots for transposon
insertions have been reported for a variety of genomes.
Table 2 • Common retroviral insertion sites in EµMyc tumors
Candidate Candidate No. No.
CIS name gene Accession ID protein family Mouse chr. Human chr. isolated tags insertions
Dkmi1 Dst ENSMUSG00000026131 actin cross-linking protein 1 6p11−p12 2 2
Dkmi2 Ly108 ENSMUSG00000015314 carcinoembryonic antigen 1 1 2 1
Cis1 Ptma ENSMUSG00000026238 nuclear protein 1 2q35−q34 1 ND
Nki3a Zfhx1b ENSMUSG00000026872 zinc-finger homeobox protein 2 2q22 1 ND
Dkmi3 Ptpn1 ENSMUSG00000027540/ TYR phosphatase 2 20q13.1−13.2 5 ND
Dkmi4 Set/ ND/ NM 023871 nucleosome assembly protein/ 2 9q34 3 ND
1190004A01Rik ENSMUSG00000026785 protein kinase/ ND
ENSMUSG00000015335
Gfi1b Gfi1b ENSMUSG00000026815 transcription factor 2 9q34.13 2 2
Notch1 Notch1 ENSMUSG00000026923 receptor 2 9q34.3 2 ND
Bmi1 Bmi1 ENSMUSG00000026739 polycomb protein 2 10p13 17 17
Evi18 RasGrp1 ENSMUSG00000027347 RAS exchange factor 2 15q15 8 3
Dkmi5 ND ND ND 2 20 2 1
Dkmi6 Pkig ENSMUSG00000035268 protein kinase inhibitor 2 20q12−q13.1 2 ND
Dkmi7 Mcl1 ENSMUSG00000038612 BCL-2−related 3 1q21 3 1
Dkmi8 Cla3 ENSMUSG00000015749 ND 3 1q21.1 2 ND
Lef1 Lef1 ENSMUSG00000027985 transcription factor 3 4q23−q25 3 2c
Evi55a Camk2d ENSMUSG00000027970 SER/THR kinase 3 4 2b ND
Cis2 ND ND ND 4 9 1 ND
Nki11a Runx3 ENSMUSG00000028814 transcription factor 4 1p36 2b ND
Evi62a E2f2/ Idb3 ENSMUSG00000007872/ HLH factor/ 4 1p36 2b ND
ENSMUSG00000018983 transcription factor
Evi143a Ak4/ Lepr NM 009647/ adenylate kinase/ 4 9p24−p13 1 ND
ENSMUSG00000028529 leptin receptor
Evi58a 5830400A04Rik ENSMUSG00000029204 RAS-related 5 4p13 1 ND
Gfi1/ Evi5 Gfi1/ Evi5 ENSMUSG00000029275/ transcription repressor/ 5 1p22 15 15
ENSMUSG00000011831 cell-cycle protein
Dkmi9 Kdr ENSMUSG00000029232 TYR kinase receptor 5 4q12 3 3
Kit Kit ENSMUSG00000005672 TYR kinase receptor 5 4q12 1 3
Dkmi10 ND ENSMUSG00000035273 heparanase 5 4q21.3 3 2c
Nki16 ND ENSMUSG00000029471 SER/THR kinase 5 12q24.31 1 ND
Evi65a Coro1c/ Selp1 ENSMUSG00000004530/ actin binding protein/ 5 12q24.1 1 ND
NM 009151 selectin
Evi78a Calm2/ ND NM 007589/ calcium binding protein/ 6 2p21 1 ND
ENSMUSG00000030349 ribosomal protein
Ccnd2 Ccnd2 ENSMUSG00000000184 cell-cycle regulator 6 12p13 3 5
Cis3 ND/Wnt5b mCG49753/ F-box protein/ 6 12p13.3 1 ND
ENSMUSG00000030170 growth factor
Evi167a Sema4b ENSMUSG00000030539 receptor 7 15q26.1 2 ND
Cis4 PD LOC243990 ND 7 ND 1 ND
Dkmi11 PD mCG60113 ND 7 10q25 2 2
Dkmi12 Rras2/ Copb1 ENSMUSG00000038142/ RAS-related/ 7 11pter−p15 3 3
ENSMUSG00000030754 beta coat protein
Evi83a Swap70 ENSMUSG00000031015 coiled-coil BCR 7 11p15 1 ND
binding protein
Dkmi13 Nttp1 ENSMUSG00000037887 TYR/THR phosphatase 7 11p15.5 3 ND
Dkmi14 PD mCG57816 Drosophila protein CG5765 8 13q34 2 2
Evi86a Irs2 ENSMUSG00000038894 docking protein 8 13q34 1 ND
Evi97a ORF23 like/ ND mCG10088/mCG57228 KIAA1865/ ND 8 14q24.3 1 ND
Dkmi15 ND ND ND 8 16p12 3 4
Evi92a Gab1/ Apm1 ENSMUSG00000003033/ clathrin coat protein/ 8 4 1 ND
ENSMUSG00000031714 growth factor receptor associated
Cis5 1100001J13Rik/ ENSMUSG00000001472/ ND/ receptor 8 16q24.3 1 ND
Mshra ENSMUSG00000041188
Cis6 Lyl1 NM008535 transcription factor 8/10 19p13.2/ 10q22 1 ND
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
Table 2 • (continued)
Fli1 Fli1 ENSMUSG00000016087 transcription factor 9 11q24.1−24.3 1 ND
Ets1 Ets1 ENSMUSG00000032035 transcription factor 9 11q23.3 2 ND
Dkmi16 Madh3 ENSMUSG00000032402 transcription factor 9 15q14−q15 4 ND
Dkmi17 Tcf12 ENSMUSG00000032228 transcription factor 9 15q21 2 2
Cis7 Kif9/ PD ENSMUSG00000032489/ kinesin-related/ 9 3p21 1 ND
ENSMUSG00000032483 KELCH-like
Evi100a 2700018N07 ENSMUSG00000041012 ND 9 ND 1 ND
Cis8 Mknk2 ENSMUSG00000020190 MAPK interacting protein 10 19p13.3 1 ND
Dkmi18d PD ENSMUSG00000020258 ND 9/11 3p21/ 5q33.1 7 ND
Myb Myb ENSMUSG00000019982 transcription factor 10 6q23.3−q24 7 ND
Dkmi19 Hbs1l/Myb NM019702/ elongation factor/ 10 6q23−q24 5 10
ENSMUSG00000019982 transcription factor
Nki28a PD/ Galgt1 ENSMUSG00000040462/ ND/ transferase 10 12q13 1 ND
ENSMUSG00000006731
Tcfe2a Tcfe2a ENSMUSG00000020167 transcription factor 10 19p13.3 2 2
Evi158a Nfic ENSMUSG00000020237 transcription factor 10 19p13.3 1 ND
Evi106a 2810013G11Rik ENSMUSG00000020280 ND 11 2p16 1 ND
Evi9 Bcl11a ENSMUSG00000000861 transcription factor 11 2 4 ND
Il9r Il9r ENSMUSG00000020279 interleukin receptor 11 5q35 1b ND
Evi159a Supt4h ENSMUSG00000020485 transcription suppressor 11 17q21−q23 1 ND
Cis9 Grb7/ Znf1a ENSMUSG00000019312/ docking protein/ 11 17q21.2 1 ND
ENSMUSG00000018168 transcription factor
Cis10 ND/ Cdc6 ENSMUSG00000038013/ WASP interacting/ 11 17q21.3 2 ND
ENSMUSG00000017499 cell-cycle regulator
Cis11 Stat5a/ Stat5b/ Stat3 ENSMUSG00000004043/ transcription factors 11 17q11.2 1 ND
ENSMUSG00000020919/
ENSMUSG00000004040
Cis12 PD ENSMUSG00000034168 transcription factor 12 14q24.3 1 ND
Cis13 Irf4 ENSMUSG00000021356 transcription factor 13 6p25−p23 1 ND
Evi112a Trim25/ Txnrd1 AL022677/ transcription factor/ 11/ 10 17q11.2/ 12q23.3 1 ND
ENSMUSG00000020250 thioredoxin reductase
Dkmi20 Cryabp1 ENSMUSG00000021366 transcription factor 13 6p24−p22.3 2 3
Nki33 PD ENSMUSG00000021755 ND 13 5p13.2 1 ND
Dkmi21 Ptp4a ENSMUSG00000022606 TYR phosphatase 15 8q24 2 ND
Pim3 Pim3 AF086624 SER/THRkinase 15 22q13.3 5 9
Evi163a PD ENSMUSG00000022462 amino acid transporter 15 12q13.11 1 ND
Dkmi22 Kcnh3/ PD ENSMUSG00000037579/ potassium channel/ 15 12q13.12 2 ND
ENSMUSG00000037570 transcription factor
Cis14 PD LOC239926 ND 15 ND 1 ND
Dkmi23 PD/ Runx1 mCG60609/ ND/transcription factor 16 21q22 3 4
ENSMUSG00000022952
Evi13 Runx1 ENSMUSG00000022952 transcription factor 16 21q22.12 4 3c
Cis15 1810055P05Rik ENSMUSG00000023883 transcription factor 17 6q27 1 ND
Pim1e Pim1 ENSMUSG00000024014 SER/THR kinase 17 6p21 9 –
Evi14 Ccnd3/ Tbn pending ENSMUSG00000034165/ cell-cycle regulator/ 17 6p21 6 ND
ENSMUSG00000023980 chromatin associated protein
Dkmi24 PD/ PD mCG55784/ENSMUSG00000041683 ND 17 6p21 2 1
Dkmi25 TsgA2 ENSMUSG00000024034 phosphatidyl inositol kinase 17 21q22.3 2 ND
Dkmi26 Fsrg1 ENSMUSG00000024335 bromodomain containing protein 17 4p16.3 2 ND
Tpl2 Tpl2 ENSMUSG00000024235 SER/THR kinase 18 10p11 7 7
Evi136a Egr1 ENSMUSG00000038418 transcription factor 18 5q31.1 1 ND
Evi153a Hmg1/ mCG9361/ HMG box protein/ ND 18 13q12 1 ND
1810041M12Rik ENSMUSG00000035765
Dkmi27 Fbxw4 ENSMUSG00000040913 F-box/WD40-repeat 19 10q24−q25 4 1
Evi17a Rasgrp2 ENSMUSG00000032946 RAS exchange factor 19 11q13 1 ND
Dkmi28 Vegfb ENSMUSG00000024962 growth factor 19 11q13 2 4
Dkmi29 Cd6 ENSMUSG00000024670 scavanger receptor 19 11q13 2 1
Pim2e Pim2 ENSMUSG00000031155 SER/THR kinase X Xp11.23 2 –
Nki37a Elf4 ENSMUSG00000031103 transcription factor X Xq26 1 ND
Dkmi30 1200013B08Rik ENSMUSG00000031101 TYR kinase X Xq25-26.3 2 –f
Where gene names are specified, RNA/protein expression altered by proviruses has been demonstrated. Candidate genes are genes adjacent to the provirus.
Gene accession number at Ensembl (ENSMUSG/LOC), Celera (mCG) or NCBI. The number of insertions is the number of retroviral insertions observed in 38 EµMyc
Pim1–/– Pim2–/– tumors as determined by Southern-blot analysis. ‘ND’ indicates that the number of insertions was not determined for the CIS. CISs in bold have
been described previously or the affected genes have been identified by altered mRNA or protein expression. Underlined gene or chromosome names are those
used by Celera. ‘PD’ indicates a predicted gene according to the Ensembl analysis pipeline, and ‘PD’ indicates a predicted gene according to the Celera Discovery
System. ‘Dkmi’ indicates a double knockout Myc insertion. aRISs overlapping with CISs identified by Suzuki et al. (Evi) or Lund et al. (Nki) based on 3 or more insertions
within 100 kb. bTwo independent retroviral insertions (distance > 26 kb). cCIS consists of two clusters, of which only one has been checked by Southern-blot
analysis. dGene-rich region of 50 kb harboring, according to Celera, four genes encoding the candidate proteins RAN-related (mCG50456), PP2C-like phosphatase
(mCG19525), ACTIN depolymerization factor (Ptk9l; mCG19506), WD-repeat-containing protein (mCG19514). ePim1 and Pim2 insertions were isolated
only from EµMyc lymphomas. fNo rearrangments were observed, indicating the subclonal nature of the retroviral insertions.
letter
156 nature genetics • volume 32 • september 2002
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
nature genetics • volume 32 • september 2002 157
Confirmation of the contribution of CISs to tumorigenesis relies
on the identification of the affected gene and evidence that aberrant
expression of that gene reproduces specific aspects of the
tumor phenotype.
To identify the genomic sequences flanking most proviruses,
we designed an efficient, PCR-based splinkerette amplification
procedure (Fig. 2). A splinkerette is an adaptor molecule containing
a hairpin loop that prevents nonspecific PCR amplification17.
The applied splinkerette amplification is preferred over
the previously described inverse PCR (IPCR) method18 for two
reasons. First, this technique is not based on a long-range PCR
amplification, which may increase the size of the amplified fragments
and limit the recovery of provirus flanking sequences.
Because the complete annotated mouse genomic sequence is
available at Ensembl/Celera, the size of the amplified sequences
flanking proviruses no longer has an advantage for identifying
CISs. Second, this method does not require cloning in bacterial
hosts, which increases the speed of the isolation of proviralflanking
sequences. We used this splinkerette procedure to analyze
the sequences of 477 RISs from 38 lymphomas from
EµMycPim1–/–Pim2–/– mice and 27 lymphomas from control
EµMyc mice. This group represents approximately 60% of all
retroviral insertions present in the tumors, a substantial fraction
of which were of oligoclonal origin. The sequence-analyzed
fraction of RISs corresponded to an average of approximately
seven insertions per tumor (Table 4). We then compared the 477
RIS sequences against the Celera annotated mouse genomic
database and found that 176 of the RISs represented 52 CISs
(Table 2; for a complete overview, see Web Table A online).
Comparison of the RISs with previously identified CISs, and the
RISs and CISs characterized by Suzuki et al. (ref. 19; this issue)
and Lund et al. (ref. 25; this issue), yielded a total of 91 independent
CISs in this tumor panel (Table 2 and Table 4).
In EµMyc mice lacking expression of Pim1, the pressure to
activate the Pim pathway by means of proviral insertions in
Pim2 is very high. In EµMyc
mice nullizygous with respect
to both Pim1 and Pim2, the
selective advantage conferred
by retroviral activation of the
Pim pathway is likely to remain
unchanged. Therefore, genes
that can substitute for Pim1
and Pim2 in lymphomagenesis
can be expected to fall into one
of the following categories
(Fig. 1c): genes encoding proteins
that directly substitute for
the function of Pim1 or Pim2; genes that encode targets or other
downstream components of the Pim1/Pim2 signaling pathway;
or genes whose proteins function in pathways parallel to Pim1 or
Pim2 that independently activate a similar crucial oncogene target.
One common retroviral insertion that was identified in five
independent lymphomas from EµMyc Pim1–/– Pim2–/– mice
affected Pim3, the third member of the Pim family and thus a
prime candidate for the first category of genes that might substitute
for Pim1 or Pim2. At the amino-acid level, Pim3 is 71% and
61% identical to Pim1 and Pim2, respectively. Knockout experiments
have shown a high degree of redundancy between Pim1
and Pim3, suggesting a similar function for the encoded proteins
(H.M., unpublished data). Southern-blot analysis showed insertions
near Pim3 in 9 of 38 lymphomas from EµMyc Pim1–/–
Pim2–/– mice. The discovery that Pim3 is preferentially activated
in tumors lacking expression of Pim1 and Pim2 underscores the
pathway-specificity of this screen.
To assign a gene to the Pim complementation group (which
consists of genes encoding proteins that can either fully or partially
substitute for Pim in lymphomagenesis), retroviral insertions
near the corresponding genes should be preferentially
absent in tumors of mice that express Pim1 and/or Pim2 or, if present,
should be mutually exclusive with insertions near one of the
Pim genes. To test the validity of this hypothesis, we carried out
Southern-blot analysis for insertions near Pim3 in 89 lymphomas
from EµMyc, EµMyc Pim1–/– and EµMyc Pim2–/– control mice, of
which 61% showed retroviral activation of either Pim1 or Pim2.
Retroviral insertions near Pim3 were observed in only one tumor,
and this tumor did not carry insertions near Pim1 or Pim2. We
subsequently analyzed the whole tumor panel of 89 lymphomas
from EµMyc control and 38 EµMyc Pim1–/– Pim2–/– mice by
Southern blotting, using CISs depicted in Table 2 as probes. Nine
CISs, identified as Kit, Ccnd2, Tpl2, Dkmi1, Dkmi9, Dkmi11,
Dkmi15, Dkmi20 and Dkmi28, were found to be mutually exclusive
with Pim1, Pim2 and Pim3 (Table 3). Within this group of
Table 3 • Common retroviral insertion sites substituting for Pim1 and Pim2 in lymphomagenesis
CIS name Gene Protein family No. insertionsa Insertion type P valueb
Pim3 Pim3 SER/THR kinase 9 5′ promoter, 5′ or 3′ enhancer 0.000
Kit Kit TYR kinase receptor 3 5′ enhancer 0.025
Tpl2 Tpl2 SER/THR kinase 7 activating truncation 0.000
Ccnd2 Ccnd2 cell-cycle regulator 5 5′ promoter and 5′ enhancer 0.002
Dkmi1 ND actin cross-linking 2 0.088
Dkmi9 ND TYR kinase receptor 3 0.025
Dkmi11 ND ND 2 0.088
Dkmi15c ND ND 4 0.007
Dkmi20 ND transcription factor 3 0.025
Dkmi28 ND growth factor 4 0.007
aNumber of proviral insertions in 38 EµMyc Pim1–/– Pim2–/– tumors detected by Southern-blot analysis. bP value of 38 EµMyc Pim1–/– Pim2–/– tumors compared with
89 control tumors using Fisher’s exact test. cCIS consists of two loci 40 kb or 125 kb apart according to Celera and Ensembl, respectively. Protein descriptions in
italics represent the proteins encoded by candidate genes ‘ND’ means that the genes affected by the retroviral insertions have not been determined.
Table 4 • Cancer loci are efficiently identified by retroviral tagging
No. tumors No. tags No. tags per tumor No. tags CISsa % tags CISs CISs per tumor
65 477 7.40 230 48 3.53
known CISs 86 18 1.32
new CISs 90 19 1.38
RIS/CISsb 39 8 0.6
new CISs (RIS/RIS)c 15 3 0.23
PIM-substituting CISs 25d 0.68e
aNumber of proviral tags identifying a CIS. bSingle RISs from this study belong to CISs identified by Suzuki et al.19 or
Lund et al.20. cComparison of the RISs from this study with the RISs isolated by Suzuki et al.19 and Lund et al.20 revealed
additional CISs. dSouthern-blot analysis showed 34 RISs substituting for Pim. eCalculations are based on 38 EµMyc
Pim1–/–Pim2–/– tumors.
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
158 nature genetics • volume 32 • september 2002
Pim-complementing loci, four of the affected genes (Pim3, Kit,
Ccnd2 and Tpl2) were identified by altered expression (data not
shown). The observation that these loci belong to the Pim complementation
group suggests that the proteins encoded by the
affected genes act either downstream of or parallel to Pim.
Despite the unknown position of the proteins encoded by the
Pim-complementing genes relative to Pim signaling, the varying
nature of these proteins argues that Pim proteins, like members of
the Myc family, have a central role in a complex signaling network.
In a pathway model that fits this hypothesis, Pim acts as a modulator
of cross-talk between stem-cell factor−induced Kit signaling
and cytokine signaling pathways (see Web Fig. A online). To induce
a maximum proliferative effect, cytokines require the synergistic
action of stem-cell factor21–23. Genes induced by interleukins but
not stem-cell factor, such as Pim, are prime candidates for involvement
in the cross-talk mediating the synergistic proliferative
effect8,24. The modulating role for Pim is supported by the observations
that enforced Pim1 expression reconstitutes the number of
lymphocytes in Rag-deficient and common-γ−deficient mice25,
and that Pim1 is recruited to the receptor complexes where it associates
with the suppressor of cytokine signaling13.
The use of genetically modified mice in combination with
high-throughput analyses of retroviral insertions and the availability
of the complete mouse genomic sequence have permitted
us to focus on specific oncogenic signaling pathways by
means of an in vivo mammalian genetic screen. The strategy
described here can indicate whether the protein encoded by a
candidate gene belongs to a particular signaling network, and
permits a more focused approach in subsequent biochemical
analyses. Thus, it represents a mammalian equivalent of the
powerful D. melanogaster and C. elegans genetic screens. In addition,
the methodology we used allows combination of data from
independent panels, as illustrated by the additional CISs that were
identified upon comparison of the RISs from different panels.
Methods
Mice and MoMuLV infection. The EµMyc mice26 were bred with Pim1–/–
Pim 1neo59 mice24 and Pim2–/– Pim2 K180 mice (J.A., unpublished data)
to generate EµMyc Pim1–/–, EµMyc Pim2–/– and EµMyc Pim1–/– Pim2–/–
mice. We infected newborns with 1 × 105 infectious units of MoMuLV. We
killed moribund mice and isolated lymphomas. All animal experiments
were approved by the Dutch Animal Research Committee.
Statistical analysis. Because exact calculations and simulations show that
end-of-chromosome effects can be ignored for a set of several hundred
proviral insertions, random insertions in the genome follow a Poisson distribution
in which the distance between two adjacent insertions is exponentially
distributed. This means that the probability that the distance is at
most x equals 1 – e−x/µ, where µ equals the mean distance G/(b + 1), where
G is the sequenced genome size (2.6 × 109) and b is the total number of
insertions. For small values of x/µ (<0.05), the probability can be approximated
by x/µ.
The number of insertions to the right of a selected insertion in a fixed
window W also follows a Poisson distribution. This means that the probability
of at least m such extra insertions equals 1 – exp(−λ)(1 + λ + λ2/2 +
… + λm−1/(m − 1)!), where λ equals the mean number of insertions in window
W: W × b/G and (m – 1)! = 1 × 2 × 3 × … × (m – 1). For m = 1 (a cluster
of two insertions), this equals 1 – exp(–λ), or approximately λ. For m =
2 (a cluster of three insertions), this equals 1 – exp(–λ)(1 + λ), or approximately
1 – (1 – λ + λ2/2) × (1 + λ) = λ2/2. For a cluster of 4 insertions, this
equals 1 – exp(–λ)(1 + λ + λ2/2), or approximately 1 – (1 – λ + λ2/2–λ3/6 +
λ4/24) × (1 + λ + λ2/2) = λ3/6 – λ4/8. These approximations only hold for a
small mean number of insertion clusters in the window (<0.05).
Southern-blot analysis of CISs. Genomic tumor DNA (10 µg) was digested
with the appropriate restriction enzyme, separated on a 0.7% agarose gel and
transferred to Hybond-N membranes (Amersham). We analyzed the number
of proviral insertions and the number of insertions into the known CISs
Pim1, Pim2, Bmi1 and Gfi1 using the probes and restriction enzymes as
described previously5,15,28,29. Genomic fragments, free of repetitive
sequences, that flanked the proviruses and hybridized to a CIS were used as
probes to analyze the frequency at which a provirus inserted into these loci.
Isolation of the proviral insertion sites. Tumor DNA (3 µg) was digested
with BstYI (New England Biolabs) and the enzyme was subsequently inactivated.
We generated the splinkerette adaptor by annealing the splinkerette
oligonucleotides HMSpAA and HMSpBB (primer sequences are
available on request). Both oligonucleotides contain modifications of a
splinkerette described previously17. The oligonucleotides (150 pmol each)
were denatured at 95 °C for 3 min and subsequently cooled to room temperature
at a rate of 1 °C per 15 s using a thermocycler (PTC100, Perkin
Elmer). We ligated 600 ng of genomic tumor DNA digested with BstYI to
the splinkerette oligonucleotide (molar ratio 1:10) with 4 U T4 DNA ligase
(Roche Diagnostics) in a final volume of 40 µl. To avoid amplification of
the internal 3′ MoMuLV fragment, we digested the ligated fragments with
10 U of EcoRV in a total volume of 100 µl. Ligation mixtures were desalted
in a Microcon YM-30 (Amicon BioSeparations).
We amplified MoMuLV-flanking sequences with a radioactive long terminal
repeat (LTR)–specific primer, AB949, and a splinkerette primer,
HMSp1 (primer sequences are available upon request). Primer AB949 (10
pmol) was radioactively labeled with [γ-32P]ATP (3 µCu) using T4 polynucleotide
kinase (PNK) (0.2 U; Roche Diagnostics).
The 50 µl PCR mixture contained 150 ng ligated tumor DNA, 10 pmol
each primer, 300 nmol dNTPs, 1 U PfuITurbo and 1 × PfuITurbo buffer
(Stratagene). The hot-start PCR conditions were 3 min at 94 °C (2 cycles);
15 s at 94 °C, 30 s at 68 °C, 3.5 min at 72 °C (27 cycles); 15 s at 94 °C, 30 s at
66 °C, 3.5 min at 72 °C, and 5 min at 72 °C. We concentrated radioactive
PCR fragments using a Microcon-YM30 (Amicon BioSeparations) and
then separated them on a 3.5% denaturing polyacrylamide gel. The gels
were dried onto 3-mm Wattman paper and exposed overnight to X-Omat
AR films (Kodak). We excised amplified fragments from the gel and boiled
them for 30 min in 100 µl TE. We used 1 µl of the DNA solution for a
nested amplification with a 32P-labeled virus-specific primer, HM001, and
a non-radioactive splinkerette-specific primer, HMSp2 (primer sequences
are available upon request). We carried out nested PCR with 5 pmol of
each primer, 200 nM of each dNTP, 1.75 mM Mg, 1 U Taq polymerase
(Gibco BRL) and 1 × PCR buffer (Gibco BRL) in a final volume of 20 µl.
The PCR conditions were 15 s at 94 °C, 30 s at 60 °C, 3 min at 72 °C for
either 25 (for fragments < 400 bp) or 28 cycles (for fragments > 400 bp).
We separated the re-amplified fragments on a 3.5% denaturing polyacrylamide
gel and isolated them as described above. We then re-amplified 1 µl
of the amplified fragments in a non-radioactive PCR of 25 cycles under the
conditions as described for the radioactive nested PCR.
We treated the nested PCR mixture with 0.5 U exonuclease and 0.5 U
shrimp alkaline phoshatase, according to the manufacturer’s (Amersham)
instructions. We used about 25 ng of the PCR product in the sequence reaction
containing BigDye terminator mix (Perkin Elmer) and primer HM001.
In addition, we used HMSp2 as primer for sequencing of amplified fragments
larger than 500 bp. We carried out automated sequence analysis on an ABI
377 (Perkin Elmer). The sequences were processed with Sequencher 3.1.1 and
subjected to BLAST analysis against the annotated mouse genome databases
at Celera (release 1.2) and Ensembl (version 6.3a.1).
GenBank accession numbers. The accession numbers for the 477 flanking
sequences of the retroviral insertions in the EµMyc tumors (Dkm) are
AY127080 through AY127557). Further information is available at
http://protagdb.nki.nl.
Note: Supplementary information is available on the Nature
Genetics website.
Acknowledgments
We wish to thank R. Regnerus for assistance in genotyping the mice;
N. Bosnie, L. Rijswijk, A. Zwerver, T. Maidment, C. Spaans and F. van der
Ahé for animal care; and J. Jonkers and R. van Amerongen for critical
reading of the manuscript. This work was supported by the Dutch Cancer
Society (H.M.) and the Leukemia Society of America (J.A.).
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
letter
nature genetics • volume 32 • september 2002 159
Competing interests statement
The authors declare that they have no competing financial interests.
Received 23 April; accepted 8 July 2002.
1. Nasmyth, K. At the heart of the budding yeast cell cycle. Trends Genet. 12, 405−
412 (1996).
2. Wassarman, D.A., Therrien, M. & Rubin, G.M. The Ras signaling pathway in
Drosophila. Curr. Opin. Genet. Dev. 5, 44−50 (1995).
3. Sternberg, P.W. & Han, M. Genetics of RAS signaling in C. elegans. Trends Genet.
14, 466−472 (1998).
4. Jonkers, J. & Berns, A. Retroviral insertional mutagenesis as a strategy to identify
cancer genes. Biochim. Biophys. Acta 1287, 29−57 (1996).
5. van Lohuizen, M. et al. Predisposition to lymphomagenesis in Pim1 transgenic
mice: cooperation with c-Myc and N-Myc in murine leukemia virus-induced
tumors. Cell 56, 673−682 (1989).
6. Selten, G., Cuypers, H.T., Zijlstra, M., Melief, C. & Berns, A. Involvement of c-Myc
in MuLV-induced T cell lymphomas in mice: frequency and mechanisms of
activation. EMBO J. 3, 3215−3222 (1984).
7. Verbeek, S. et al. Mice bearing the EµMyc and EµPim1 transgenes develop pre-Bcell
leukemia prenatally. Mol. Cell. Biol. 11, 1176−1179 (1991).
8. Allen, J.D., Verhoeven, E., Domen, J., van der Valk, M. & Berns, A. Pim2 transgene
induces lymphoid tumors, exhibiting potent synergy with c-Myc. Oncogene 15,
1133−1141 (1997).
9. Leverson, J.D. et al. PIM-1 kinase and P100 cooperate to enhance c-Myb activity.
Mol. Cell 2, 417−425 (1998).
10. Mochizuki, T. et al. Physical and functional interactions between PIM-1 kinase
and CDC25A phosphatase. Implications for the PIM-1-mediated activation of the
c-MYC signaling pathway. J. Biol. Chem. 274, 18659−18666 (1999).
11. Koike, N., Maita, H., Taira, T., Ariga, H. & Iguchi-Ariga, S.M. Identification of
heterochromatin protein 1 (HP1) as a phosphorylation target by PIM-1 kinase and
the effect of phosphorylation on the transcriptional repression function of
HP1(γ). FEBS Lett. 467, 17−21 (2000).
12. Ishibashi, Y. et al. PIM-1 translocates sorting NEXIN6/TRAF4-associated factor 2
from cytoplasm to nucleus. FEBS Lett. 506, 33−38 (2001).
13 Rainio, E.M., Sandholm, J. & Koskinen P.J. Cutting edge: transcriptional activity of
NFATc1 is enhanced by the PIM-1 kinase. J. Immunol. 168, 1524−1527 (2002).
14. van der Lugt, N.M. et al. Proviral tagging in EµMyc transgenic mice lacking the
Pim1 proto-oncogene leads to compensatory activation of Pim2. EMBO J. 14,
2536−2544 (1995).
15. Allen, J.D. & Berns, A. Complementation tagging of cooperating oncogenes in
knockout mice. Semin. Cancer Biol. 7, 299−306 (1996).
16. Craig, N.L. Target site selection in transposition. Annu. Rev. Biochem. 66, 437−474
(1997).
17. Devon, R.S., Porteous, D.J. & Brookes, A.J. Splinkerettes—improved vectorettes
for greater efficiency in PCR walking. Nucleic Acids Res. 23, 1644−1645 (1995).
18. Li, J. et al. Leukaemia disease genes: large-scale cloning and pathway predictions.
Nature Genet. 23, 348−353 (1999).
19. Suzuki, T. et al. Retroviral tagging in the post-genome era identifies new genes
involved in cancer. Nature Genet. 32, 86–94 (2002).
20. Lund, A.H. et al. Genome-wide retroviral insertional tagging of cancer genes in
Cdkn2a-deficient mice. Nature Genet. 32, 80–85 (2002).
21. Miyazawa, K. et al. Recombinant human interleukin-9 induces protein tyrosine
phosphorylation and synergizes with steel factor to stimulate proliferation of the
human factor-dependent cell line, M07e. Blood 80, 1685−1692 (1992).
22. Tsuji, K., Lyman, S.D., Sudo, T., Clark, S.C. & Ogawa, M. Enhancement of murine
hematopoiesis by synergistic interactions between steel factor (ligand for c-kit),
interleukin-11, and other early acting factors in culture. Blood 79, 2855−2860
(1992).
23. Laird, P.W. et al. In vivo analysis of Pim-1 deficiency. Nucleic Acids Res. 21, 4750−
4755 (1993).
24. Jacobs, H. et al. PIM1 reconstitutes thymus cellularity in interleukin 7- and
common γ chain-mutant mice and permits thymocyte maturation in Rag- but not
CD3γ-deficient mice. J. Exp. Med. 190, 1059−1068 (1999).
25. Losman, J. et al. IL-4 signaling is regulated through the recruitment of
phosphatases, kinases, and SOCS proteins to the receptor complex. Cold Spring
Harb. Symp. Quant. Biol. 64, 405−416 (1999).
26. van Lohuizen, M. et al. Identification of cooperating oncogenes in EµMyc
transgenic mice by provirus tagging. Cell 65, 737−752 (1991).
27. te Riele, H., Maandag, E.R., Clarke, A., Hooper, M. & Berns, A. Consecutive
inactivation of both alleles of the Pim1 proto-oncogene by homologous
recombination in embryonic stem cells. Nature 348, 649−651 (1990).
28. Cuypers, H.T. et al. Murine leukemia virus-induced T-cell lymphomagenesis:
integration of proviruses in a distinct chromosomal region. Cell 37, 141−150
(1984).
29. Scheijen, B., Jonkers, J., Acton, D. & Berns, A. Characterization of Pal1, a common
proviral insertion site in murine leukemia virus-induced lymphomas of c-Myc and
Pim1 transgenic mice. J. Virol. 71, 9−16 (1997).
30. Nakamura, T., Largaespada, D.A., Shaughnessy, J.D. Jr, Jenkins, N.A. & Copeland,
N.G. Cooperative activation of Hoxa and Pbx1-related genes in murine myeloid
leukaemias. Nature Genet. 12, 149−153 (1996).
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics
erratum
nature genetics
High-throughput retroviral tagging to identify components of specific signaling
pathways in cancer
H Mikkers, J Allen, P Knipscheer, L Romeyn, A Hart, E Vink & A Berns
Nature Genet. 32, 153–159 (2002).
By error, several corrections were not made to proofs while preparing the manuscript for the press.
In the reference list, reference 25 (Losman et al.) should be inserted as reference 13. As a consequence, references 13–24 should be
renumbered as 14–25.
In the text, the following changes should be made:
On page 154, in the second column, reference 16 should be placed at the end of the sentence “These observations underscore the selective
advantage…” Reference 16 should also be removed from the first line on page 155.
On page 155, in the second column, reference 17 should be placed at the end of the sentence “‘Cold’ and ‘hot’ spots for transposon
insertions…”
On page 157, in the first full paragraph, reference 17 should be reference 18, reference 18 should be reference 19, reference 19 should be
reference 20, and reference 25 should be reference 21.
On page 158, in the first full paragraph, references 21–23 should be references 22,23.
errata
nature genetics • volume 32 • october 2002 331
Genetics, cytokines and human infectious disease: lessons from weakly pathogenic
mycobacteria and salmonellae
T H M Ottenhoff, F A W Verreck, E G R Lichtenauer-Kaligis, M A Hoeve, O Sanal & J T van Dissel
Nature Genet. 32, 97–105 (2002).
On page 97, paragraph 2, line 13, ‘IL-29’ should be ‘IL-27’.
On page 98, line 11, ‘seemed to be’ should be ‘was’, as this has been demonstrated to be the case.
On page 98, Fig. 1, a mistake in the color coding occurred, and the affected genes that appear in purple should be colored red.
On page 100, line 1, ‘IL12Rb1’ should read ‘IL12Rβ1’.
Distal ureter morphogenesis depends on epithelial cell remodeling mediated by
vitamin A and Ret
E Batourina, C Choi, N Paragas, N Bello, T Hensle, F D Constantini, A Schuchardt, R L Bacallao & C L Mendelsohn
Nature Genet. 32, 109–115 (2002).
doi:10.1038/ng952
The article was missing a reference to Web Movie A, which should have appeared on the last line of page 110 together with the reference
to Web Fig. A.
Targeted mutation of Cyln2 in the Williams syndrome critical region links CLIP-115
haploinsufficiency to neurodevelopmental abnormalities in mice
C C Hoogenraad, B Koekkoek, A Akhmanova, H Krugers, B Dortland, M Miedema, A van Alphen, W M Kistler, M Jaegle, M Koutsourakis,
N Van Camp, M Verhoye, A van der Linden, I Kaverina, F Grosveld, C I De Zeeuw & N Galjart
Nature Genet. 32, 116–127 (2002).
doi:10.1038/ng954
The article mistakenly contained a note stating that supplementary information was available on the Nature Genetics website. Instead, it
should have contained a brief paragraph in the Methods section listing a separate website where additional information can be found.
High-throughput retroviral tagging to identify components of specific signaling
pathways in cancer
H Mikkers, J Allen, P Knipscheer, L Romeyn, A Hart, E Vink & A Berns
Nature Genet. 32, 153–159 (2002).
doi:10.1038/ng950
By error, several corrections were not made to proofs while preparing the manuscript for the press.
In the reference list, reference 25 (Losman et al.) should be inserted as reference 13. As a consequence, references 13–24 should be
renumbered as 14–25.
In the text, the following changes should be made:
On page 154, in the second column, reference 16 should be placed at the end of the sentence “These observations underscore the selective
advantage…”. Reference 16 should also be removed from the first line on page 155.
On page 155, in the second column, reference 17 should be placed at the end of the sentence “‘Cold’ and ‘hot’ spots for transposon
insertions…”.
On page 157, in the first full paragraph, reference 17 should be reference 18, reference 18 should be reference 19, reference 19 should be
reference 20, and reference 25 should be reference 21.
On page 158, in the first full paragraph, references 21–23 should be references 22,23.
New genes involved in cancer identified by retroviral tagging
T Suzuki, H Shen, K Akagi, H C Morse III, J D Malley, D Q Naiman, N A Jenkins & N G Copeland
Nature Genet. 32, 166–174 (2002).
doi:10.1038/ng949
By error, the subpanel labels for Fig. 1 on page 172 were offset to the right of the corresponding subpanels.
©2002NaturePublishingGrouphttp://www.nature.com/naturegenetics