Generation of single-copy transgenic mouse embryos directly from ES cells by tetraploid embryo complementation

Background Transgenic mice have been used extensively to analyze gene function. Unfortunately, traditional transgenic procedures have only limited use in analyzing alleles that cause lethality because lines of founder mice cannot be established. This is frustrating given that such alleles often reveal crucial aspects of gene function. For this reason techniques that facilitate the generation of embryos expressing such alleles would be of enormous benefit. Although the transient generation of transgenic embryos has allowed limited analysis of lethal alleles, it is expensive, time consuming and technically challenging. Moreover a fundamental limitation with this approach is that each embryo generated is unique and transgene expression is highly variable due to the integration of different transgene copy numbers at random genomic sites. Results Here we describe an alternative method that allows the generation of clonal mouse embryos harboring a single-copy transgene at a defined genomic location. This was facilitated through the production of Hprt negative embryonic stem cells that allow the derivation of embryos by tetraploid embryo complementation. We show that targeting transgenes to the hprt locus in these ES cells by homologous recombination can be efficiently selected by growth in HAT medium. Moreover, embryos derived solely from targeted ES cells containing a single copy LacZ transgene under the control of the α-myosin heavy chain promoter exhibited the expected cardiac specific expression pattern. Conclusion Our results demonstrate that tetraploid embryo complementation by F3 hprt negative ES cells facilitates the generation of transgenic mouse embryos containing a single copy gene at a defined genomic locus. This approach is simple, extremely efficient and bypasses any requirement to generate chimeric mice. Moreover embryos generated by this procedure are clonal in that they are all derived from a single ES cell lines. This facilitates the comparative analysis of lethal alleles and thereby advances our ability to analyze gene function in mammals.


Background
The analysis of gene function during mammalian embryonic development relies heavily on the use of transgenic mice. However, expression of alleles that result in embryonic lethality cannot be studied using regular transgenic procedures because lines of founder mice cannot be established. Analysis of such alleles in mammals has historically relied upon transient transgenic procedures [1]. In this approach, fertilized eggs are injected with a transgene construct and allowed to develop to a pre-determined embryonic stage in utero before being collected for phenotypic analysis. Using this method a variable number of transgenes integrate into random genomic loci. Unfortunately, the local chromosomal environment can affect expression of the integrated transgene and this can result in unpredictable transgene expression. Moreover, because no founder line is established it is not possible to repeat an analysis with a specific transgenic mouse embryo.
The problem of random transgene integration can be overcome by targeting transgenes to a specific genomic locus by homologous recombination in ES cells [2][3][4]. A method that allows the introduction of single copy transgenes into the X-linked hprt locus has been described previously by Bronson et al [2]. In this approach singlecopy transgenes are introduced 5' of the hprt locus by homologous recombination in ES cells. This is an extremely efficient procedure because targeted clones have reconstituted the hprt gene and are selected on the basis of growth in HAT medium. Following this method, upward of 90% of HAT resistant ES cell clones contain correctly targeted transgenes [2]. The hprt locus is a particularly suitable site for the integration of transgenes because it exists as an X-linked gene present as a single copy in male ES cells. Moreover, the hprt gene is ubiquitously expressed and so provides a favorable chromatin environment for transgene expression. Indeed, it has been demonstrated that the level of expression of transgenes inserted into the hprt locus is directed solely by exogenous transcriptional regulatory elements [2]. Furthermore, mice derived from independent ES cells clones containing a transgene expressed from the same promoter all exhibit comparable levels of expression [2,[4][5][6].
Although the targeting of constructs to the hprt locus facilitates consistent expression of transgenes it does not overcome the need to generate founder lines of transgenic mice. However, pioneering work by Nagy et al has shown that embryos and mice can be derived directly from R1 ES cells by aggregating them with tetraploid embryos [7,8]. This suggests that controlled expression of transgenes encoding possible gain-of-function and developmental-lethal alleles could be achieved by generating embryos and/or mice from ES cells containing transgenes inserted into the hprt locus. Unfortunately, the generation of mice and embryos by tetraploid aggregation from available Hprt negative ES cell lines has proven to be inefficient and unreliable (S.A.D. unpublished). In this report, we describe the generation of a new Hprt negative ES cell line that is compatible with the tetraploid aggregation procedure. Moreover, we demonstrate that it facilitates efficient targeting of transgenes to the hprt locus as well as the subsequent generation of clonal transgenic mouse embryos that express the transgene in a promoter-dependent manner.

Results
Nagy et al demonstrated that embryos and mice could be derived solely from R1 ES cells by aggregating them with tetraploid embryos [7,8]. Although we were able to repeat these data using R1 ES cells we were unable to generate midgestation embryos from E14TG2a-derived strains of hprt negative ES cell lines using the same approach [9][10][11]. We believed it was unlikely that Hprt itself was required to generate ES cell-derived embryos because hprt knockout mice are viable and fertile [12]. E14TG2a-derived ES cell lines contain a large spontaneous deletion encompassing 55 kb of the hprt locus [13]. Therefore, it seemed possible that sequences within this region were necessary to maintain ES cell pluripotency. However, we believed a more likely scenario was that the available lines contained unrelated genetic or epigenetic alterations that were incompatible with the aggregation procedure. Consistent with this interpretation, Eggan et al have recently shown that viable offspring are more efficiently produced through tetraploid-mediated cloning from outbred ES cell line compared to inbred ES cell lines [14]. We, therefore, believed it likely that aggregation-compatible hprt negative ES cells could be generated by targeted mutation of the hprt gene in R1 ES cells. To test this we generated a targeting vector to replace exon 1 of the mouse hprt gene with a neomycin phosphotransferase (neo) gene that conferred resistance to te pharmacological inhibitor G418 in ES cells (fig 1a). Following transfection of this plasmid into R1 cells 192 clones that were resistant to G418 were collected and grown in the presence of 6-thioguanine to select for loss of Hprt activity. Of the original 192 G418-resistant clones that were harvested only one, named F3, was also resistant to 6-thioguanine and sensitive to growth in HAT medium (fig 3b). The integrity of the targeting event was next confirmed by Southern blot analysis of F3 genomic DNA. As shown in fig 1a the correct integration of Neo at the hprt locus was predicted to delete a BamH1 restriction endonuclease cut-site from intron 1 and replace it with a BamH1 site that is unique to the mutant hprt allele. This restriction fragment polymorphism could be detected using an hprt-specific probe (fig 1a). The wild type BamH1 fragment was predicted to be 7.0 kb based For F3 cells to be a suitable vehicle for the generation of transgenic mice they should be competent to generate

Figure 2
Embryos derived from Hprt-deficient ES cells by tetraploid aggregation. A-C) Photographs of embryos at E13.0 derived from F3 cells. Pigmentation can be seen in the eyes (arrow) reflecting their ES cell origin. D) Southern blot analyses of BamH1digested genomic DNA isolated from five (#1 -#5) F3 ES cell-derived embryos that was hybridized to a radiolabeled intron 3 probe. Only the mutant allele is detected in F3 ES cell derived embryos. The wild type allele can be seen in female F3 offspring as discussed in the text.

Figure 3
A) Schematic representation of the hprt genomic locus and targeting vector with α-MHC-LacZ insert. The position of a Southern blot probe (probe b) containing LacZ specific sequences relative to Xba1 (x) and EcoRV (e) restriction enzyme cut sites is shown. B) Resistance to growth in HAT and sensitivity to G418 and 6-TG was restored in correctly targeted α MHC-LacZ ES cell lines. C) Southern blot analysis of genomic DNA from non-targeted F3 ES cells as well as from six targeted HAT-resistant ES cell lines generated from F3 cells is shown. The targeted allele is identified using probe b as a 7.0 kb EcoRV fragment. α MHC-LacZ ES cell line #4 was used to generate embryos shown in D. D) E8.5 embryos derived from α-MHC-LacZ targeted ES cells by tetraploid aggregation. β-galactosidase expression was identified by x-gal staining (blue). Expression was restricted to the developing heart (h). chimeric mice that can transmit alleles through the germline. To test the "germline potential" of F3 cells we first generated chimeric mice by aggregating F3 ES cells with CD-1 morulae and allowing them to develop to term in utero. The resulting chimeric males were bred with CD-1 females and the genotype of offspring determined by Southern blot analysis of genomic DNA from mouse tails. Fig 1c shows that agouti F1 female mice contained both the mutant (paternal) and wild type (maternal) hprt alleles. In contrast, the agouti F1 male mice contained only a wild-type hprt allele inherited from the maternal X-chromosome as expected. From this we were able to conclude that F3 hprt null ES cells are indeed competent to transmit alleles through the germline and that the resulting offspring were fertile and viable as reported previously [12].
We next addressed whether embryos could be generated solely from F3 Hprt negative ES cells by aggregating these cells with tetraploid CD-1 embryos and allowing them to develop in utero until day 13 of gestation (El 3). Figs 2a,b,c show that F3 cells could indeed give rise to normal embryos (n = 10). All embryos exhibited characteristic features of midgestation stages of development including normal rostral-caudal patterning, beating hearts, normal limb formation including the formation of fingers. F3 cells derive from the agouti (pigmented) SV129/J strain of mice. In contrast, albino CD-1 mice donated the tetraploid morulae used in the aggregation procedure. The finding, shown in figs 2a,b,c, that all the resulting embryos displayed clear retinal pigmentation strongly implied that the embryos were F3 ES cell-derived. We confirmed this by determining the genotype of the embryos by Southern blot analyses of genomic DNA. Fig 2d shows that all embryos contained a restriction fragment polymorphism that was indicative of deletion of hprt exon1. No wild type hprt allele was detected confirming that the embryos were generated entirely from F3 cells. These data demonstrate that the hprt gene is not essential for the generation of embryos from ES cells by tetraploid aggregation and that F3 cells should could provide a suitable vehicle for the direct generation of transgenic mouse embryos.
Previously it has been shown that a targeting vector, pMP8, can correct the loss of Hprt activity in E14TG2a ES cells that lack exons 1 and 2 of the hprt gene [2,15]. The design of this vector implied that it should also be capable of correcting the hprt exon 1 deficiency in F3 cells and could, therefore, be used to introduce transgenes into the hprt locus by homologous recombination. To test this we generated a modified vector based on pMP8. The practical features of this modification are that it removes the lacZ sequences in the vector backbone and reduces the overall size of the vector, both of which improve subsequent cloning steps. This vector, pMP8NEB∆LacZ, contains a 5' arm of homology consisting of 3.4 kb of mouse 5' hprt genomic DNA flanking a unique Not1 restriction endonuclease cut site. The 3' arm of homology contains 1.8 kb of the human hprt gene that includes exon1 and 2.9 kb of mouse genomic sequence that includes exon 2 (fig 3a) as described previously [15]. Homologous recombination at the modified hprt locus in F3 cells should substitute genomic DNA containing hprt exons 1 and 2 for neo thereby reconstituting a gene encoding active Hprt (fig 3a). Moreover, only cells having undergone a successful homologous recombination event should become HAT resistant and carry the targeted transgene. To assess the efficiency of this approach we attempted to target a LacZ reporter transgene whose expression is controlled by the promoter from the alpha myosin heavy chain (αMHC) gene to the modified hprt locus in F3 cells. We chose the αMHC promoter because others have demonstrated, using traditional transgenic procedures, that it directs high levels of reporter gene expression specifically in cardiac tissue of early mouse embryos [16][17][18]. Cells were selected for resistance to HAT following electroporation of the αMHC-LacZ transgenecontaining plasmid. 16 HAT resistant cells lines were collected and their genotype determined by Southern blot analysis. Fig 3c shows that, in contrast to F3 cells, a probe that specifically recognizes the LacZ transgene identified a 6.7 kb EcoRV fragment in HAT-resistant ES cell lines. Of the 16 HAT-resistant colonies retrieved 15 contained LacZ. Furthermore, fig 3b shows that these cells had become G418 and 6-thioguanine sensitive confirming that the neo gene had indeed been replaced with hprt-coding sequences in the F3 cells. These data confirm that this selection strategy is extremely efficient in identifying ES cell clones containing correctly targeted transgenes [2].
The ultimate use of this procedure is to generate clonal transgenic mouse embryos directly from the modified ES cells. Success in this would provide a system in which to express alleles with potentially lethal phenotypes in the developing embryo. We, therefore, determined whether embryos could be generated from the F3 cells that contained an αMHC-LacZ transgene and whether these embryos appropriately expressed β-galactosidase. E8.5 embryos were generated from three independent αMHC-LacZ cell lines and stained for expression of β-galactosidase. Representative embryos, shown in fig 3d all developed normally and expressed high levels of β-galactosidase in the developing heart, mimicking the expression of endogenous αMHC [16]. Importantly, none of the embryos exhibited ectopic patterns of expression lending further support to the proposal that expression of transgenes introduced to the hprt locus is primarily depend-ent on the accompanying transcriptional regulatory elements [2].

Discussion and conclusions
Here we have described the generation of an ES cell line and a procedure that will allow the phenotypic analyses of transgenic embryos expressing lethal alleles. This procedure is an important addition to the available repertoire of molecular genetic techniques available to developmental biologists because it overcomes several deficiencies associated with alternative transient transgenic approaches normally used to study gain-of-function and dominant-negative alleles. Traditional procedures rely on the generation of transgenic embryos through micro injection of fertilized mouse eggs. Using this approach an unpredictable number of transgenecopies are incorporated at a random location in the recipient genome. This can result in ectopic expression of the transgene because both copy number and the site integration can influence transgene expression.
In the approach described here, a targeting vector that contains exon 1 of the hprt gene along with a specific transgene is used to reconstitute Hprt activity in Hprt negative ES cells. Successful targeting of the transgene to the hprt locus in Hprt negative ES cells can be selected by growing transfected cells in HAT medium. The efficiency of this selection procedure is such that close to all HAT resistant colonies contain a correctly targeted transgene. Moreover, because the transgene is introduced into the hprt locus its expression is dependent upon the inclusion of specific transcriptional regulatory elements and remains relatively unaffected by the local chromatin environment [2]. A major advantage of this targeted transgenic approach compared to traditional transgenics is that transgene expression is predictable so expression at ectopic sites is largely avoided. The availability of F3 Hprt negative ES cells allows transgenic embryos to be derived directly from transgenic ES cells by aggregating the cells with tetraploid embryos and allowing them to develop in utero. Each embryo is, in effect, clonal because it is generated from a single clonal ES cell line. This allows reproducible comparison of the phenotype generated by a given transgene at different developmental stages. Moreover, because the transgene is present as a single copy at a known site of integration it circumvents the need to generate many different transgenic lines from a given construct as is the case when using a transient transgenic procedure.
There are many potential uses for this method of generating transgenic mouse embryos. The most obvious is that it provides a mechanism for generating embryos that express alleles that are predicted, or known, to be detrimental to normal embryonic development, for ex-ample dominant-negative or gain-of-function alleles. Indeed, the expression of dominant negative alleles that can potentially disrupt multiple members of a gene family has been used extensively in Xenopus laevis and has, for example, helped elucidate the action of fibroblast growth factors during mesoderm formation [19]. Such alleles are particularly useful when a null allele in the same gene causes a disruptive phenotype prior to the stage of development of interest. Choosing a promoter that expresses the dominant negative allele later in development or in a specific cell type could potentially circumvent any early phenotype associated with a null allele.
Another important use of this procedure could be to identify transcriptional regulatory regions that control expression of specific genes during development. Transgenic mice have been extensively used to identify regulatory elements that control gene expression in vivo [20,21]. This has, in turn allowed the elucidation of regulatory networks that control cell differentiation during development. In this regard, the use of F3 cells coupled with targeting transgenes to the hprt locus offers two major advantages over traditional transgenic approaches. The first is that it is extremely rapid. Once the targeting construct is in hand it takes approximately four to five weeks to generate transgenic mouse embryos. The second advantage is due to the presence of the transgene at a known and constant integration site. This potentially allows the comparison of levels of reporter gene expression driven by different promoters thereby offering a quantitative in vivo transgenic assay.
In sum, the availability of tetraploid aggregation-competent ES cells that support targeting of transgenes to the hprt locus should facilitate the rapid analysis of transcriptional regulatory elements in vivo as well as the phenotypic characterization of alleles that are incompatible with normal development. We, therefore, believe that F3 cells provide a powerful new tool for molecular geneticists.

Plasmids p∆Hprt
The targeting plasmid used to delete hprt exon1 was called p∆Hprt to indicate this deletion. It was derived from the plasmid pOS.DUP/DEL that contained the neomycin phosphotransferase (neo) gene that confers resistance to the pharmacological inhibitor G418 and also contains the Herpes Simplex Virus thymidine kinase (HSV-TK) that confers sensitivity to gancyclovir. Importantly the neo gene is flanked by loxP sequences permitting loss of G418 resistance where useful. p∆Hprt was constructed by introducing a 2.5 Kb BamHI/HindIII fragment of 5' hprt genomic DNA (E14TG2a origin) from the plasmid pMP8 into the Hpa1 site of pOS.DUP/DEL [15]. The ends of the fragments were made blunt using Klenow. The 3' arm of homology was introduced into the unique Bcl1 site of the previous plasmid as a 7 Kb BamHI fragment from pMP1(SV129 origin) that contained murine hprt exons 2 and 3.

pMP8NEB∆LacZ
The plasmid pMP8NEB∆LacZ was generated by ligating an 8 kb Xba1 fragment from pMP8 into the Xba1 site of pNEB193 (New England Biolabs) from which LacZ sequences had been removed as a NarI/EcoRI fragment [15]. This hprt replacement vector contains a 5' arm of hprt homology consisting of 3.4 kb of mouse 5' hprt genomic DNA flanking a unique Not1 cloning site. The 3' arm of homology contains 1.8 kb of the human hprt gene that includes exon1 and 2.9 kb of mouse genomic sequence that includes exon 2. A detailed description of the pMP8 targeting vector, from which these sequences were derived, has been described previously [15]. A partial nucleotide sequence of pMP8NEB∆LacZ is available from S.A.D. upon request.

pαMHC-LacZ-Hprt
This targeting vector was constructed by inserting a 9.7 kb αMHCLacZ expression cassette into the NotI site of pMP8NEB∆LacZ. The 9.7 kb αMHCLacZ expression cassette was constructed by inserting a 3.6 kb SalI/HindIII fragment from pLacF containing a lacZ gene that incorporates a nuclear localization signal into the SalI/HindI-II site of an αMHC expression module.

ES cell culture, gene targeting and selection
All ES cell lines were cultured on mitotically inactivated primary embryonic fibroblasts in ES cell medium supplemented with recombinant leukemia inhibitory factor (LIF) as described elsewhere [22]. F3 cells were generated by introducing 100 µg of Not1 digested p∆HPRT plasmid into 1.5 × 10 8 R1 ES cells by electroporation at 250 volts/500µf/resistance 8 using a BTX ECM600 electroporation system [7]. Cells containing neo were selected by supplementing the ES cell medium with 250 µg/ml Geneticin (G418, Gibco-BRL) and negative selection against Hsv-tk gene expression was achieved by including 2 µM gancyclovir (Roche). To identify cells that had lost HPRT activity G418/gancyclovir-resistant colonies were collected and expanded in 96 well plates in the presence of 250 µg/ml 6-thiogaunine. Introduction of transgenes into the hprt locus of F3 cells was achieved using the following procedure. F3 cells were grown for two passages in media containing 250 µg/ml 6-thioguanine. 100 µg of linear targeting vector plasmid was introduced into 1.5 × 10 8 F3 ES cells by electroporation. Cells were grown for 2 days before adding HAT (Sigma) to the growth medium. HAT resistant colonies were harvested after 10 days growth in HAT-containing medium.

Tetraploid and diploid embryo aggregation
ES-cell derived embryos were generated by aggregating clumps of around twenty ES cells with individual eight cell embryos or four cell embryos that had been made tetraploid by electrofusion as described previously [7,9,23]. ES cell:embryo aggregates were cultured overnight and those that formed blastocysts were transferred to the uteri of psuedo-pregnant females surrogate mothers. To generate chimeric mice the embryos were allowed to develop to term. ES cell-derived embryos were harvested at pre-determined developmental stages and DNA prepared for Southern blot analysis or fixed and processed to identify β-galactosidase expression using standard protocols.