It is over a decade since the first demonstration that mouse embryonic stem cells could be used to transfer a predetermined genetic modification to a whole animal1. The extension of this technique to other mammalian species, particularly livestock, might bring numerous biomedical benefits, for example, ablation of xenoreactive transplantation antigens, inactivation of genes responsible for neuropathogenic disease and precise placement of transgenes designed to produce proteins for human therapy. Gene targeting has not yet been achieved in mammals other than mice, however, because functional embryonic stem cells have not been derived. Nuclear transfer from cultured somatic cells provides an alternative means of cell-mediated transgenesis2, 3. Here we describe efficient and reproducible gene targeting in fetal fibroblasts to place a therapeutic transgene at the ovine 1(I) procollagen (COL1A1) locus and the production of live sheep by nuclear transfer. We previously showed that transfection of fetal fibroblasts with nuclear transfer offers an efficient and practical method of producing sheep carrying randomly integrated transgenes3, and similar work has been reported in cattle4. Gene targeting is a more powerful method of genetic manipulation and requires essentially the same procedures of transfection and drug selection of cultured cells. Although experimental gene targeting was first carried out in somatic cell lines5, 6, the use of embryonic stem (ES) cells now predominates; however, there has been no definitive comparison of gene targeting efficiency in primary somatic cells and ES cells (but see ref. 7 for review). We wished to determine whether practically useful gene targeting could be achieved in primary ovine fetal fibroblasts, and whether these cells could produce viable animals by nuclear transfer. The ovine COL1A1 gene represented a suitable target with which to establish gene targeting in fetal fibroblasts for three reasons. First, we expected that gene-targeting events would be very rare compared with random integrations. COL1A1 is highly expressed in fibroblasts, allowing promoter-trap enrichment of gene-targeting events. Second, few ovine genes have been cloned and characterized. COL1A1 is well studied and highly conserved in several species, facilitating molecular cloning and construction of ovine gene-targeting vectors. Third, mutations in COL1A1 can cause connective tissue disorders in humans, for example, osteogenesis imperfecta 8, 9. The ability to generate models of human genetic disorders by gene targeting in animals other than mice might be valuable for clinical research; however, we chose to target a site that would not significantly affect type I collagen protein function or expression to avoid affecting fetal development.

We used two gene-targeting vectors to target ovine COL1A1 (Fig. 1). Each vector incorporated two regions of COL1A1 homology, derived from a contiguous fragment of Poll Dorset fetal fibroblast3 (PDFF2) genomic DNA. COLT-1 was designed to insert a promoterless neo selectable marker between the COL1A1 translational stop and polyadenylation signal, such that transcription of the targeted locus resulted in a bicistronic messenger RNA. An internal ribosomal entry site (IRES)10 immediately 5' of neo facilitated translation. COLT-2 had the same structure, but included a transgene as a separate transcription unit located 3' of neo. This transgene, termed AATC2, comprised human 1-antitrypsin (AAT) complementary DNA within an ovine -lactoglobulin (BLG) expression vector designed to direct expression in the lactating mammary gland11.

We transfected COLT-1 DNA into early passage PDFF2 female and PDFF5 male ovine primary fetal fibroblasts; we transfected COLT-2 DNA into PDFF2 cells. Stable G418 resistant clones were derived. About 30 days of culture elapsed between fetal disaggregation and cryopreservation of gene-targeted cell clones. DNA samples of each cell clone were initially screened by polymerase chain reaction (PCR) using primers designed to amplify a 3.4-kilobase (kb) fragment across the 5' junction of the targeted locus (Fig. 1 ). In each case, a high proportion of G418 resistant cell clones were found to have undergone gene targeting (5 out of 36 PDFF2 COLT-1, 4 out of 56 PDFF5 COLT-1, and 46 out of 70 PDFF2 COLT-2 cell clones analysed). We called COLT-1 cell clones 'PDCOL' and COLT-2 cell clones 'PDCAAT'. The DNA sequence of PCR products amplified from three PDCOL and two PDCAAT cell clones was determined across the 5' junction; each was consistent with gene targeting (data not shown).

Figure 2a shows Southern analysis of the 5' junctions of a series of PDCAAT cell clones. All samples showed the presence of a 7-kb BamHI fragment from the normal COL1A1 locus. Each clone identified as positive by PCR also showed a diagnostic 4.7-kb Bam HI fragment spanning the 5' junction of the COLT-2 targeted locus. This is consistent with the presence of one targeted and one normal COL1A1 allele. PDCAAT cell clones 87 and 99 also showed the presence of additional bands, indicating additional integrations of COLT-2.

PDFF5 cells were also targeted at high frequency with COLT-1. PDFF2 and PDFF5 cells have different parentage from within the PPL outbred flock of Poll Dorset sheep. This indicates that it may not always be essential to isolate DNA from the same individual, or an animal of the same inbred strain, to achieve efficient gene targeting12; however, the degree of sequence divergence, if any, between the targeted COL1A1 alleles in these cells has not yet been determined.

Northern analysis of cell clones PDCAAT 81 and 90 is shown in Fig. 2b. Hybridization with human 1(I) procollagen cDNA detected a 4.8-kb mRNA species in both non-transfected cells and PDCAAT cell clones, consistent with expression of normal COL1A1. A larger species of about 6.8 kb was present only in the PDCAAT cell clones, consistent with a bicistronic COL1A1­IRES­neo fusion mRNA. Hybridization of the same RNA samples with a neo probe also detected a 6.8-kb mRNA in the targeted clones, again consistent with a bicistronic mRNA. These results confirmed that gene targeting had occurred. This analysis also showed that, unlike mouse, rat and human, which express two endogenous 1(I) procollagen mRNA species from different polyadenylation sites13, 14, sheep express a single mRNA species.

Although these experiments were designed to avoid disruption of COL1A1 gene expression, the mRNA from the targeted locus is less abundant than the wild type (Fig. 2b). Whether this reflects different mRNA stability or transcriptional activity has yet to be determined. However, elements which affect transcription have been identified at the 3' end of COL1A1 in other species15, 16, and targeted DNA insertion may affect their function.

We carried out northern analysis to determine whether placement of the AATC2 transgene adjacent to the highly expressed COL1A1 gene resulted in aberrant expression of the BLG promoter in fibroblasts. Hybridization with human AAT cDNA failed to detect AAT mRNA expression in either PDCAAT81 or 90 cells (data not shown), indicating no apparent loss of BLG promoter specificity.

Four targeted cell clones, all derived from PDFF2 female cells (PDCOL6, PDCOL13, PDCAAT81 and PDCAAT90), were selected for nuclear transfer on the basis of their vigour and normal metaphase chromosome number. Nuclear transfer was carried out on twelve occasions and the results are summarized in Table 1. Fourteen lambs were live-born: seven died within 30 hours of birth, and one each after 3 days, 8 days, 7.5 weeks and 12 weeks. Three lambs are currently alive and thriving at almost one year of age. The first two live-born lambs are shown in Fig. 3.

Post mortem examination of lambs that died in utero or after birth revealed a range of abnormalities. Although there was no consistent pattern, we observed a high incidence of kidney defects (frequently renal pelvis dilation), liver and brain pathology. These findings are similar to a previous nuclear transfer study using the same cells3, and are therefore probably due to some aspect of the cell treatment or nuclear transfer procedure and not a consequence of gene targeting per se. Several researchers have reported developmental abnormalities associated with somatic cell nuclear transfer4, 17, 18. Although there is some indication that inappropriate expression of imprinted genes may be involved19, definitive investigations have yet to be carried out. Understanding and rectifying this problem is a continuing priority.

Figure 4 shows Southern analysis of 5' and 3' junctions of the COL1A1 targeted locus in representative nuclear transfer lambs. Hybridization with the same 5' and 3' probes used to analyse PDCAAT cell clones revealed fragments consistent with the presence of one targeted and one normal COL1A1 allele in the two lambs shown. This was confirmed using a 5' probe external to the vector.

Sixteen lambs and fetuses were analysed by Southern blot, of which fifteen confirmed the presence of a targeted allele. One lamb (990504), derived from PDCOL13, showed only the normal COL1A1 gene, which probably indicates that the cell isolate was oligoclonal. Non-transfected cells have been detected within some G418 selected cell isolates in previous experiments (unpublished data).

Lamb 990507 derived from clone PDCAAT90 was hormonally induced to lactate and milk samples analysed by western blotting (data not shown). AAT was detected at a concentration of 650 µg ml-1, which compares favourably with the highest level previously reported for an AAT cDNA transgene in sheep carrying multiple random gene inserts (18 µg ml -1)20. This indicates that the COL1A1 locus supports transgene expression even though it is not actively expressed in mammary epithelium21.

We have shown that gene targeting can be carried out efficiently in somatic cells and that viable animals can be produced by nuclear transfer. We have also obtained preliminary data (that is, PCR fragment size and sequence) indicating similarly efficient targeting at the -1,3-galactosyl-transferase locus in porcine fibroblasts (unpublished data). Notably, the use of nuclear transfer does not require embryonic stem or embryonic germ cells, and circumvents the generation of chimaeric animals, which would be costly and time consuming in livestock. Fibroblasts are also being used in clinical trials to provide a protein production system after ex vivo gene therapy in human patients22, and it has been suggested that the introduction of therapeutic transgenes by homologous recombination could avoid undesirable effects arising from random integration23. Nuclear transfer in animals such as sheep provides a rigorous means of testing the suitability of specific loci for transgene placement. If the COL1A1-targeted sheep continue to show no locus-related deleterious effects, this would indicate that this target locus may provide a permissive and benign environment for the insertion of therapeutically useful genes.

Methods

Gene-targeting vectors The promoter trap vector COLT-1 comprised a 3-kb region of the 3' end of the ovine COL1A1 gene from a point roughly 2.9-kb 5' of the translation stop site to an SspI site 131-bp 3' of the stop site; a 0.6-kb IRES region10 corresponding to bases 1,247­1,856 of the pIREShyg vector (Clontech); a 1.7-kb region containing the bacterial neomycin gene and a portion of the 3' end of the human growth hormone gene containing the polyadenylation site, essentially as described24; an 8.3-kb region of the 3' end and flanking region of the ovine COL1A1 gene from an SspI site 131-bp 3' of the translational stop site to a XhoI site roughly 8.4-kb 3' of the stop site; and the bacterial cloning vector pSL1180 (Pharmacia). DNA fragments homologous to the ovine COL1A1 gene were derived from a single genomic clone isolated from a library of genomic fragments of PDFF2 in bacteriophage . The promoter trap transgene placement vector COLT-2 was constructed by inserting an MluI fragment containing the AATC2 transgene into COLT-1 at a unique EcoRV site at the 3' end of the IRES­neo region.

Preparation, culture and transfection of primary fibroblasts Derivation of ovine PDFF2 and PDFF5 cells has been described3. PDFF cells were grown throughout in BHK 21 (Glasgow MEM) medium supplemented with 2 mM glutamine, 1 mM sodium pyruvate, 1X non-essential amino acids and 10% fetal calf serum in standard tissue-culture vessels in a humidified atmosphere composed of 2% CO2, 5% O2 and 93% N2, and were passaged by standard trypsinization. PDFF cells were used at passage three, and had undergone 5­6 days of culture following fetal disaggregation. Cells were plated at 5 105 cells in a 25-cm2 flask and transfected the next day with eit

NOTICE: In accordance with Title 17 U.S.C. Section 107, this material is distributed for research and educational purposes only. **