A large-scale, gene-driven mutagenesis approach for
the functional analysis of the mouse genome
Jens Hansen*, Thomas Floss*, Petra Van Sloun†
, Ernst-Martin Fu¨ chtbauer‡§
, Franz Vauti¶
, Hans-Hennig Arnold¶
,
Frank Schnu¨ tgen†
, Wolfgang Wurst*ʈ
**, Harald von Melchner†
**, and Patricia Ruiz**††‡‡
*Institute of Developmental Genetics, GSF-National Research Center for Environment and Health, D-85764 Neuherberg, Germany; †Laboratory for Molecular
Hematology, University of Frankfurt Medical School, D-60590 Frankfurt am Main, Germany; ‡Department of Developmental Biology, Max Planck Institute of
Immunobiology, D-79108 Freiburg, Germany; ¶Department of Cell and Molecular Biology, Institute of Biochemistry and Biotechnology, TU Braunschweig,
D-38106 Braunschweig, Germany; ʈDepartment for Molecular Neurogenetics, Max Planck Institute of Psychiatry, D-80804 Munich, Germany; and
††Department of Vertebrate Genomics, Max Planck Institute for Molecular Genetics, D-14195 Berlin, Germany
Communicated by Sherman M. Weissman, Yale University School of Medicine, New Haven, CT, May 30, 2003 (received for review November 5, 2002)
A major challenge of the postgenomic era is the functional characterization
of every single gene within the mammalian genome.
In an effort to address this challenge, we assembled a collection of
mutations in mouse embryonic stem (ES) cells, which is the largest
publicly accessible collection of such mutations to date. Using four
different gene-trap vectors, we generated 5,142 sequences adjacent
to the gene-trap integration sites (gene-trap sequence tags;
http:͞͞genetrap.de) from >11,000 ES cell clones. Although most of
the gene-trap vector insertions occurred randomly throughout the
genome, we found both vector-independent and vector-specific
integration ‘‘hot spots.’’ Because >50% of the hot spots were
vector-specific, we conclude that the most effective way to saturate
the mouse genome with gene-trap insertions is by using a
combination of gene-trap vectors. When a random sample of
gene-trap integrations was passaged to the germ line, 59% (17 of
29) produced an observable phenotype in transgenic mice, a
frequency similar to that achieved by conventional gene targeting.
Thus, gene trapping allows a large-scale and cost-effective production
of ES cell clones with mutations distributed throughout the
genome, a resource likely to accelerate genome annotation and the
in vivo modeling of human disease.
With the completion of sequencing of the human and mouse
genomes, the interest in tools suitable for performing
genome-wide mutagenesis has increased significantly. Two major
mouse-mutagenesis programs have evolved: one is phenotypedriven
and based on chemical (ethyl-nitroso-urea) mutagenesis (1,
2), and the other is gene-driven and based on insertional mutagenesis
(3, 4).
Large-scale insertional mutations in mammalian cells are
induced most effectively with gene traps, a class of DNA or
retroviral vectors that insert a promoterless reporter gene into a
large collection of chromosomal sites. By selecting for gene
expression, recombinants are obtained in which the reporter
gene is fused to the regulatory elements of an endogenous gene.
Transcripts generated by these fusions faithfully reflect the
activity of individual cellular genes and serve as molecular tags
to identify and͞or clone any genes linked to specific functions
(3–5). Application of this technique in a genome-wide manner
should allow the identification of most, if not all, active transcripts
in the genome and thus is an important tool for genome
annotation. More importantly, gene trapping in mouse embryonic
stem (ES) cells enables the establishment of ES cell libraries
with mutations in most genes, which then can be used to make
mice. This opens the possibility to assign a function to each gene
in the context of an entire organism.
Several smaller-sized mutagenesis screens with gene-trap vectors
have been reported (4, 6–9). However, the use of single
gene-trap vectors in each screen, the unavailability of a complete
mouse genome sequence, and a comparatively low number of
analyzed insertions precluded a systematic assessment of the
technology.
Based on the analysis of 5,142 sequence tags obtained from
gene-trap insertions across the mouse genome, we show here
that gene-trap vectors can disrupt all functional classes of genes,
including disease genes, and are highly mutagenic in transgenic
mice. We also show that individual gene-trap vectors complement
each other in gene targeting, suggesting that the most
effective way of saturating the mouse genome with mutations is
by using a combination of different gene-trap vectors.
Materials and Methods
ES Cell Cultures and Gene-Trap Vectors. The E14.1 (129͞Ola), CJ7
(129͞Sv), R1 (129͞Sv ϫ 129͞Sv-CP), and TBV-2 (129͞SvP) ES
cell lines were grown on irradiated (x-rays, 32 Gy) or mitomycin
C (Sigma)-treated (10 ␮g͞ml for 2.5 h) mouse embryonic
fibroblast feeder layers in DMEM (GIBCO͞BRL) supplemented
with 10–20% (vol͞vol) preselected and heat-inactivated
FCS (Invitrogen), 2 mM glutamine, 1ϫ nonessential amino
acids, 1 mM sodium pyruvate, 0.1 mM 2-mercaptoethanol (all
from Invitrogen), 1,000 units͞ml leukemia inhibitory factor
(Esgro, Chemicon), and optionally 5 ␮g͞ml penicillin and streptomycin
(GIBCO͞BRL). ES cells were either electroporated
with pT1␤geo and pT1ATG␤geo plasmid vectors or infected
with U3␤geo and ROSA␤geo retroviruses as described (3, 10).
Gene-trap-expressing ES cell clones were selected in 200 ␮g͞ml
G418 (GIBCO͞BRL), manually picked, expanded, and stored
frozen in liquid nitrogen. For gene-trap sequence tag (GTST)
recovery all clones were arrayed into 48-well plates, lysed, and
subjected to 5Ј rapid amplification of cDNA ends.
5؅ Rapid Amplification of cDNA Ends and Sequencing. cDNAs were
prepared from the polyadenylated RNA by using a RoboAmp
robotic device (MWG Biotec, Ebersberg, Germany) with a processing
capacity of 96 samples per day. Samples of 2 ϫ 105
cells were
lysed in 1 ml of lysis buffer containing 100 mM Tris⅐HCl, pH
8.0͞500 mM LiCl͞10 mM EDTA͞1% lithium-dodecyl sulfate
(LiDS)͞5 mM DTT. Polyadenylated RNA was captured from the
lysates by biotin-labeled oligo(dT) primers according to manufacturer
instructions (Roche Diagnostics, Indianapolis) and placed on
streptavidin-coated 96-well plates (AB Gene, Surrey, U.K.). After
washing, solid-phase cDNA synthesis was performed in situ by using
random hexamers and SuperScript II reverse transcriptase (Invitrogen).
To remove excess primers the cDNAs were filtered through
multiscreen PCR plates (Millipore). The 5Ј ends of the purified
cDNAs were tailed with dCTPs by using terminal transferase,
terminal deoxynucleotidyl transferase (Invitrogen), following manufacturer
instructions.
Abbreviations: ES, embryonic stem; GTST, gene-trap sequence tag.
§Present address: Institute of Molecular and Structural Biology, Aarhus University, C. F.
Mollers Alle, 8000 Aarhus C, Denmark.
**W.W., H.v.M., and P.R. contributed equally to this work.
‡‡To whom correspondence should be addressed. E-mail: ruiz@molgen.mpg.de.
9918–9922 ͉ PNAS ͉ August 19, 2003 ͉ vol. 100 ͉ no. 17 www.pnas.org͞cgi͞doi͞10.1073͞pnas.1633296100
For PCR amplification of GTSTs, the following vector-specific
primers were used: (i) pT1␤geo and pT1ATG␤geo: 5Ј-CTA
CTA CTA CTA GGC CAC GCG TCG ACT AGT ACG GGI
IGG GII GGG IIG-3Ј and 5Ј-GCC AGG GTT TTC CCA GTC
ACG A-3Ј; and 5Ј-CTA CTA CTA CTA GGC CAC GCG TCG
ACT AGT AC-3Ј and 5Ј-TGT AAA ACG ACG GCC AGT
GTG AAG GCT GTG CGA GGC CG-3Ј (nested); and (ii)
U3␤geo and the ROSA␤geo: 5Ј-GCC ATT CAG GCT GCG
CAA-3Ј; and 5Ј-CAA GGC GAT TAA GTT GGG TAA TG-3Ј
(nested). Amplification products were directly sequenced
by using AB377 or ABI3700 sequencing machines (Applied
Biosystems).
GTST Analysis. After filtering sequences against repeats and
removing all vector sequences from the GTSTs, a PHRED score
was assigned to each individual nucleotide. GTSTs qualified as
informative if they were at least 50 nt long and exhibited a
minimum mean PHRED score of 20 (Fig. 4, which is published
as supporting information on the PNAS web site, www.pnas.org).
Homology searches were performed by using the publicly available
sequence databases and the BLASTN algorithm. Databases
included GenBank, UniGene, Online Mendelian Inheritance in
Man (OMIM) (all at www.ncbi.nlm.nih.gov), ENSEMBL (www.
ensembl.org), RIKEN (www.rarf.riken.go.jp), and GeneOntology
(www.geneontology.org).
ES Cell Injections, Breeding, and Genotyping. 129Sv͞J (TBV-2, R1,
and E14.1) ES cell-derived chimeras were generated by injecting
C57BL͞6 blastocysts. The resulting male chimeras were bred to
C57BL͞6 females, and agouti offspring were tested for transgene
transmission by tail blotting. Animals heterozygous for gene-trap
insertions were backcrossed to C57BL͞6 mice, and phenotypes
were assessed in homozygous F2 offspring.
Results and Discussion
We used the gene-trap vectors pT1␤geo, pT1ATG␤geo,
ROSA␤geo, and U3␤geo to transduce a promoterless ␤-galactosidase-neomycin
phosphotransferase (␤geo) reporter gene
into mouse ES cells. In pT1␤geo, pT1ATG␤geo, and
ROSA␤geo, ␤geo is flanked by an upstream 3Ј splice consensus
sequence (splice acceptor) and a downstream polyadenylation
site to ensure its activation from integrations into introns
(‘‘intron trap’’) (11–13). U3␤geo lacks a splice acceptor sequence
and therefore is activated mostly from integrations into exons
(‘‘exon trap’’) (10, 14). Because all these gene-trap vectors
require a cellular promoter for activation, the maximum number
of genomic targets equals the number of expressed genes. The
vectors pT1␤geo and pT1ATG␤geo were transduced as DNA
into ES cells by electroporation. The vectors U3␤geo and
ROSA␤geo were transduced as retroviruses into ES cells by
infection.
From 11,266 ES cell clones containing gene-trap insertions in
expressed genes, we isolated 8,423 sequences adjacent to the
gene-trap integration sites (GTSTs). As summarized in Table 1,
5,142 of these sequences provided useful GTSTs. The other
sequences were either of low quality or were too short (Ͻ50
nucleotides) to be informative (see Materials and Methods and
Fig. 4).
GenBank (NCBI) homology analysis revealed that 3,750
(72.9%) of the GTSTs belonged to known genes, 623 (12.1%)
were ESTs, and 769 (15%) had no match in the database (Table
1). In comparison to our previous analysis (7), the number of
matches to known genes increased by 26%, clearly reflecting the
sustained progress in sequencing of the human and mouse
genomes. Moreover, when nonmatching ‘‘novel’’ (previously
uncharacterized) sequences (769) were aligned to the ENSEMBL
database, 41% (389) produced a match (Table 2). However,
despite the availability of a nearly complete mouse genome
sequence, 7.4% (380 of 5,142; Tables 1 and 2) failed to produce
a match in any database. Although this could be the result of
some strain-specific variations between mouse genomes, it may
also reflect the fact that some sequences are not yet available
from the genome sequence, which still contains gaps.
Fifty-five percent of the genome-matching GTSTs were in
annotated genes. Interestingly, the frequency of U3␤geo insertions
into predicted introns was almost twice as high as that
obtained with all the other vectors (Table 2), confirming previous
studies showing that the U3-type exon-trap vectors can be
activated also from integrations into the introns of expressed
genes (9, 15). Unexpectedly, 50 of 110 GTSTs obtained with the
other vectors were also part of predicted introns (Table 2),
although intronic sequences should have been removed by
splicing (3, 4). Although in nine instances the intron-matching
GTSTs resulted from aberrant splicing, we assumed that the
other 41 GTSTs are actually part of exons annotated incorrectly
by the current gene-prediction programs. To substantiate this
Table 1. Summary of GTST results and homology analysis in GenBank (release 133)
Gene-trap vector pT1␤geo pT1ATG␤geo U3␤geo ROSA␤geo Total
No. of insertions sequenced 3,866 1,581 2,100 876 8,423
No. of GTSTs 2,526 771 1,111 734 5,142†
NR* homology 2,093 (82.9%) 579 (75.1%) 627 (56.4%) 451 (61.4%) 3,750 (72.9%)
EST homology 190 (7.5%) 103 (13.4%) 192 (17.3%) 138 (18.8%) 623 (12.1%)
No homology 243 (9.6%) 89 (11.5%) 292 (26.3%) 145 (19.8%) 769 (15.0%)
Cut off e value Յ 10Ϫ20.
*NR, nonredundant.
†The accession numbers for GTSTs also present in GenBank are BZ689860–BZ691019.
Table 2. Genome matches of ‘‘novel’’ GTSTs
Gene-trap vector
Novel
GTSTs
Genome
matches
In annotated
genes
In predicted
exons
In predicted
introns
All 769 389 214 84 130
Without 3Ј splice site (U3␤geo) 292 189 104 24 80
With 3Ј splice site (pT1␤geo,
pT1ATG␤geo, and ROSA␤geo)
477 200 110 60 50
According to ENSEMBL, version 13.30.1.
Hansen et al. PNAS ͉ August 19, 2003 ͉ vol. 100 ͉ no. 17 ͉ 9919
GENETICS
hypothesis, we selected 10 annotated genes for additional expression
studies. By using RT-PCR and primers complementary
to the intron-annotated GTSTs and to the corresponding downstream
exons (Fig. 5A, which is published as supporting information
on the PNAS web site), we obtained amplification
products in five instances. Direct sequencing of these products
revealed splicing of the GTSTs to the downstream exons (Fig.
5B), indicating that a significant proportion of intron-matching
GTSTs indeed are part of mispredicted exons.
To localize the GTSTs cytogenetically, we screened the UniGene
database using the GenBank accession number as an
identifier. Allowing for an e value Յ10Ϫ20
, we identified 1,349
GTSTs in mapped UniGene clusters that were distributed
among all chromosomes except the Y chromosome (Table 3).
There was a direct correlation between the number of GTSTs on
a given chromosome and the number of UniGene clusters on
that chromosome, indicating that gene-trap insertions are dispersed
throughout the genome and occur more frequently in
chromosomes with a high density of genes (Fig. 1).
Several preferred integration sites or ‘‘hot spots’’ were observed,
some of which were hit Ͼ20 times. Examples include the
UniGene clusters 38,186 and 36,541, the growth-arrest gene
Gas5, the C-terminal-binding protein 2, and the Jumonji
(mouse) homolog (Table 8, which is published as supporting
information on the PNAS web site). We identified a total of 441
UniGene clusters containing two or more gene-trap insertions,
which corresponds to 25% of the recovered UniGene clusters
and suggests that 75% of all genes are randomly accessible for
gene-trap insertions. Forty-five percent of the hot spots contained
multiple (more than two) insertions of more than one of
the vectors and thus were vector-independent. Of the remaining
vector-specific hot spots, 12% were recognized only by pT1␤geo,
10% by pT1ATG␤geo, 16% by U3␤geo, and 17% by ROSA␤geo
vectors. Moreover, the gene-trap hot spots were not sequencespecific
and were not related to gene size (Fig. 2), suggesting that
they are most likely defined by secondary chromatin structure.
Considering that over half of all the hot spots are vector-specific,
we believe that the most effective way to saturate the genome
with gene-trap insertions is with gene-trap vector combinations.
To estimate how effectively the various vectors trap genes that
had not been trapped before, we determined the number of
insertions required by each vector to trap a novel UniGene
cluster. Fig. 3 shows that the vectors with a splice acceptor site
(pT1␤geo, pT1ATG␤geo, and ROSA␤geo) trapped a different
gene with almost every insertion. However, results from
pT1␤geo, for which more insertions are available, suggest that
the trapping efficiency decreases with an increasing number of
insertions, presumably because of a gradual reduction of the pool
of trappable genes (Fig. 3 Insert). In contrast, U3␤geo, which
does not contain a splice acceptor, consistently required two or
more insertions to hit a novel UniGene cluster (Fig. 3). The
inferior gene-trapping efficiency of U3␤geo reflects its comparatively
small pool of genomic integration targets, consisting
mainly of the exons of expressed genes. As a result, U3␤geo
integrated more frequently into a given genomic hot spot than
any of the other vectors. With an average insertion frequency of
4.1 insertions per hot spot, U3␤geo exceeded the average
hot-spot insertion frequency of the other vectors by almost 2-fold
(Table 8).
Because gene inactivations induced by gene-trap vectors with
a splice acceptor sequence partly depend on effective splicing,
the frequency of aberrant splicing events was determined by
analyzing the splice junctions induced by each individual vector.
Because the frequency of aberrant splicing was essentially similar
for all gene-trap vectors (pT1␤geo ϭ 3.5%; pT1ATG␤geo ϭ
5.5%; ROSA␤geo ϭ 4.0%), we conclude that the splice acceptor
sequences used in this analysis are equally efficient [i.e., engrailed
splice acceptor sequence for pT1␤geo and pT1ATG␤geo
(11, 12) and adenovirus major late transcript splice acceptor
sequence for ROSA␤geo (13)]. Interestingly, Ͼ80% of the
aberrantly spliced integrations into annotated genes were atypically
in exons, suggesting that ectopic splice sites inside exons are
recognized ineffectively by the splicing enzymes.
Table 3. Distribution of gene-trap insertions among chromosomes
Chromosome
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 X Y Total
No. of gene-trap
insertions
85 115 58 96 105 66 80 71 73 74 112 47 44 37 64 40 58 43 48 33 0 1,349
The UniGene database (release 120) was screened by using GenBank accession numbers as identifiers and an e value Յ10Ϫ20.
Fig. 1. Correlation between gene-trap insertions and the number of UniGene
clusters per chromosome.
Fig. 2. Number of gene-trap (GT) insertions into annotated hot spots. All
genes with two or more insertions were classified as hot spots. Gene lengths
were derived from GenBank.
9920 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.1633296100 Hansen et al.
Because the relative mutagenicity of the gene-trap vectors
likely depends on their position within a gene, we looked at the
insertion site of each gene-trap vector with regard to its location
within the full-length cDNA. Table 4 shows that the vast majority
of retroviral gene-trap insertions involved the 5Ј half of genes,
confirming a reported preference of retroviral integrations (9,
16). Interestingly, Ͼ50% of the U3␤geo insertions were in 5Ј
untranslated regions (Table 4), presumably due to a relatively
high stringency of selection that requires gene-trap vectors
without a splice acceptor to insert close to an active cellular
promoter. Although plasmid vectors also exhibited a slight
preference for the 5Ј ends of genes, insertions were distributed
more evenly over the coding region of a gene, indicating that
even longer fusion proteins are stable (Table 4). Finally, one
U3␤geo integration was recovered from an intronless gene
(glutathione peroxidase 4͞ENSMUSG00000038809). Although
this was a unique event, it demonstrates that U3␤geo vectors can
also disrupt single exon genes.
To analyze the functional spectrum of the genes represented
in the GTST library, we classified the trapped UniGene clusters
based on their known or putative function by using the GeneOntology
database. Table 5 shows that the vectors used in this
study inserted into all functional classes of mammalian genes,
although with different frequencies, which suggests that the
effective trapping of some specific classes of genes may require
more specialized gene-trap vectors (17, 18).
Because the development of mouse models for human disease is
a major goal of the human genome project, we also searched our
library for integrations into genes involved in human disease. Using
the Online Mendelian Inheritance in Man (OMIM) database, we
found 204 GTSTs that corresponded to 90 previously characterized
disease genes (Table 9, which is published as supporting information
on the PNAS web site). ES cell clones with these insertions can
be used to produce mouse mutant strains that may replicate the
genetic defects and the symptomology of specific human disorders,
and that may be useful for testing therapeutic methods. For
example, we recently characterized a mouse strain with a phenotype
closely resembling congenital nephrotic syndrome (19, 20).
To analyze the frequency of obvious phenotypes developing
after gene-trap insertions, we injected 29 randomly selected ES
cell clones into blastocysts and produced mutant mice from
them. As shown in Table 6, 59% of the mice developed an
obvious phenotype when bred to homozygosity, a frequency
comparable with conventional gene targeting and to reported
gene-trap screens (6, 13). Interestingly, over half of the observed
phenotypes were embryonic or perinatal-lethal (Table 7), suggesting
that a significant proportion of the genes expressed in ES
cells are required for embryonic development.
We conclude that gene-trap mutagenesis is an efficient approach
for annotating and dissecting the function of mammalian
genes. Its large-scale implementation has already enabled the
worldwide establishment of several databases containing GTSTs
from hundreds of mouse genes (4, 6–9). Collectively, these
databases provide an unprecedented resource for the scientific
community in the postgenomic era, because clones from the
corresponding ES cell libraries can be used immediately to
cost-effectively generate mouse models of human disease.
Clearly, the goal of understanding the function of every gene in
the genome could be attained more quickly with the establishFig.
3. Frequency of gene-trap (GT) insertions into unique UniGene clusters.
Data points represent the number of novel UniGene clusters accumulating
with every 50 insertions. For further explanation see text.
Table 4. Gene-trap vector insertion site preference in full-length cDNAs (according to RefSeq)
Gene trap
vector
Total no. of
insertions
Insertions in
5Ј UTR (%)
Insertions in
5Ј CDS (%)
Insertions in
3Ј CDS (%)
Insertions in
3Ј UTR (%)
pT1␤geo 1,385 222 (16.0) 589 (42.5) 511 (36.9) 63 (4.6)
pT1ATG␤geo 395 97 (24.6) 160 (40.5) 125 (31.6) 13 (3.3)
U3␤geo 324 176 (54.3) 101 (31.2) 31 (9.6) 16 (4.9)
ROSA␤geo 302 100 (33.1) 152 (50.3) 37 (12.3) 13 (4.3)
Total 2,406 595 (24.7) 1,002 (41.6) 704 (29.3) 105 (4.4)
5Ј CDS, first half of coding sequence; 3Ј CDS, second half of coding sequence.
Table 5. Functional gene classes targeted by gene-trap insertions
Class* Annotated genes Trapped genes (%)
Ligand binding or carrier 2,002 240 (12.0)
Enzyme͞enzyme regulator 1,491 138 (9.3)
Transcription factors͞cofactors 416 45 (10.8)
Transporter 437 25 (5.7)
Signal transducer 779 24 (3.1)
Structural protein 193 11 (5.7)
Chaperone 49 11 (22.5)
Translation regulator 27 8 (30.0)
Motor 39 7 (17.9)
Defense͞immunity 44 1 (2.3)
Cell adhesion molecule 78 1 (1.3)
Apoptosis regulator 30 1 (0.3)
*According to GeneOntology.
Table 6. Frequency of phenotypes obtained with gene-trap
vectors
Gene-trap
vector
Phenotypes͞mutant
strains Frequency, %
pT1␤geo 6͞11 55
pT1ATG␤geo 5͞9 56
U3␤geo 7͞9 78
All 17͞29 59
Hansen et al. PNAS ͉ August 19, 2003 ͉ vol. 100 ͉ no. 17 ͉ 9921
GENETICS
ment of ES cell libraries with mutations in every single gene.
Because each gene-trap vector seems to have its own set of
specific hot spots, we conclude that the most effective generation
of an ES cell library saturated with mutations should involve a
collection of different gene-trap vectors. The ongoing collaboration
within the international mouse-mutagenesis consortium
(22) is likely to achieve complete saturation of the mouse
genome within the next few years.
We thank Drs. William. C. Skarnes for the pT1␤geo vectors, H. Earl
Ruley for the U3␤geo, and Philippe Soriano for the ROSA␤geo vectors.
We acknowledge Susanne Bourier, Franziska Ko¨hler, Katharina
Kuhlmeier, Sava Michailidou, Ines Peiser, Armin Reffelmann, Irina
Rodionova, Cordula Schulz, Beata Thalke, Sandra Schwarzmeier, Beate
Walther, and Carsta Werner for excellent technical assistance. This work
was supported by grants from the Bundesministerium fu¨r Bildung und
Forschung to the German Gene Trap Consortium.
1. Brown, S. D. & Balling, R. (2001) Curr. Opin. Genet. Dev. 11, 268–273.
2. Beier, D. R. (2000) Mamm. Genome 11, 594–597.
3. Floss, T. & Wurst, W. (2002) Methods Mol. Biol. 185, 347–379.
4. Stanford, W. L., Cohn, J. B. & Cordes, S. P. (2001) Nat. Rev. Genet. 2, 756–768.
5. Hicks, G. G., Shi, E. G., Chen, J., Roshon, M., Williamson, D., Scherer, C. &
Ruley, H. E. (1995) Methods Enzymol. 254, 263–275.
6. Mitchell, K. J., Pinson, K. I., Kelly, O. G., Brennan, J., Zupicich, J., Scherz, P.,
Leighton, P. A., Goodrich, L. V., Lu, X., Avery, B. J., et al. (2001) Nat. Genet.
28, 241–249.
7. Wiles, M. V., Vauti, F., Otte, J., Fuchtbauer, E. M., Ruiz, P., Fuchtbauer, A.,
Arnold, H. H., Lehrach, H., Metz, T., von Melchner, H. & Wurst, W. (2000)
Nat. Genet. 24, 13–14.
8. Zambrowicz, B., Friedrich, G., Buxton, E., Lilleberg, S., Person, C. & Sands,
A. (1998) Nature 392, 608–611.
9. Hicks, G. G., Shi, E., Li, X. M., Li, C. H., Pawlak, M. & Ruley, H. E. (1997)
Nat. Genet. 16, 338–344.
10. von Melchner, H., DeGregori, J. V., Rayburn, H., Reddy, S., Friedel, C. &
Ruley, H. E. (1992) Genes Dev. 6, 919–927.
11. Gossler, A., Joyner, A. L., Rossant, J. & Skarnes, W. C. (1989) Science 244,
463–465.
12. Skarnes, W. C., Auerbach, B. A. & Joyner, A. L. (1992) Genes Dev. 6, 903–918.
13. Friedrich, G. & Soriano, P. (1991) Genes Dev. 5, 1513–1523.
14. Scherer, C. A., Chen, J., Nachabeh, A., Hopkins, N. & Ruley, H. E. (1996) Cell
Growth Differ. 7, 1393–1401.
15. Wempe, F., Yang, J. Y., Hammann, J. & von Melchner, H. (2001) Genome Biol.
2, research0023.
16. Rohdewohld, H., Weiher, H., Reik, W., Jaenisch, R. & Breindl, M. (1987)
J. Virol. 61, 336–343.
17. Skarnes, W., Moss, J., Hurtley, S. & Beddington, R. (1995) Proc. Natl. Acad.
Sci. USA 92, 6592–6596.
18. Gebauer, M., von Melchner, H. & Beckers, T. (2001) Genome Res. 11,
1871–1877.
19. Patrakka, J., Martin, P., Salonen, R., Kestila, M., Ruotsalainen, V., Mannikko,
M., Ryynanen, M., Rapola, J., Holmberg, C., Tryggvason, K. & Jalanko, H.
(2002) Lancet 359, 1575–1577.
20. Rantanen, M., Palmen, T., Patari, A., Ahola, H., Lehtonen, S., Astrom, E.,
Floss, T., Vauti, F., Wurst, W., Ruiz, P., et al. (2002) J. Am. Soc. Nephrol. 13,
1586–1594.
21. Sterner-Kock, A., Thorey, I. S., Koli, K., Wempe, F., Otte, J., Bangsow, T.,
Kuhlmeier, K., Kirchner, T., Jin, S., Keski-Oja, J. & von Melchner, H. (2002)
Genes Dev. 16, 2264–2273.
22. Nadeau, J. H., Balling, R., Barsh, G., Beier, D., Brown, S. D., Bucan, M.,
Camper, S., Carlson, G., Copeland, N., Eppig, J., et al. (2001) Science 291,
1251–1255.
Table 7. Mutant phenotypes induced by gene-trap insertions
Line Gene name Symbol Phenotype
A006B04 Sprouty homolog 4 (Drosophila) Spry4 Limb deformation
A20010 Novel Pigmentation defects
M004D05 Selenocysteine tRNA gene transcription-activating factor Staf Sterility
M016A06 PHD finger protein 2 Phf2 Dwarfism
W027B02 (20) Nephrin 1 Nphs1 Nephrotic syndrome
W036C08 Baculoviral IAP repeat-containing 6 Birc6 Placenta defects
W044B06 Neurochondrin Ncdn-pending Lacrimal gland hypertrophy
W052E02 Synaptojanin 2 Synj2 Perinatal-lethal
3C7 (21) Latent transforming growth factor ␤-binding protein 4 Ltbp4 Pulmonary emphysema, colorectal cancer
F053A01 ITSN Itsn Embryonic-lethal
F045D05 Ect2 oncogene Ect2 Embryonic-lethal
M016E07 Splicing factor (CC1.3) CC1.3 Embryonic-lethal
M017A08 Heterogeneous nuclear ribonucleoprotein C (C1͞C2) Hnrpc Embryonic-lethal
M019E03 KIAA0240 KIAA0240 Embryonic-lethal
M020A01 Plectin 1, intermediate filament-binding protein, 500 kDa Plec1 Embryonic-lethal
W023D11 Novel Embryonic-lethal
W078F01 Peroxisomal biogenesis factor 14 Pex14 Embryonic-lethal
F035B07 Nuclear factor of ␬ light chain gene enhancer in B cells 1,
p105
Nfkb1 Not obvious
F041B05 HSPC063 protein ESTs, weakly similar to I53869 zinc finger
protein–mouse (Mus musculus)
EST Not obvious
M017C03 EST EST Not obvious
M019D01 Chromobox homolog 1 (Drosophila HP1 ␤) Cbx1 Not obvious
W008G09 ESTs, highly similar to T34020 zinc finger protein–rat
(Rattus norvegicus)
EST Not obvious
W024F10 Homo sapiens cDNA FLJ30453 fis, clone BRACE2009307,
weakly similar to P120 PROTEIN
Pkp4 Not obvious
W027F01 Msx-interacting zinc finger Miz1 Not obvious
W047A01 Baculoviral IAP repeat-containing 6 Birc6 Not obvious
W056E05 Dystrophin, muscular dystrophy Dmd Not obvious
W063E06: Fibroblast growth factor-inducible 13 Fin13 Not obvious
W073D02 Dentin matrix protein 1 Dmp1 Not obvious
Aquarius Aquarius Aqr Not obvious
9922 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.1633296100 Hansen et al.