Long targeting arms do not increase the efficiency of homologous recombination in the β-globin locus of murine embryonic stem cells

Conventional gene therapy strategies generally involve the use of retroviral or lentiviral vectors to introduce therapeutic genes and linked regulatory sequences into random sites within the target cell genome. Random integration approaches to gene therapy are associated with a number of problems, including integration site–dependent expression, viral genome silencing, and the risk of insertional mutagenesis.^1,2 

An alternative strategy is the use of homologous recombination (HR) to “correct” a mutant gene. The major challenge of this strategy is the low efficiency of HR in mammalian cells. The factors that are most crucial for determining the gene targeting efficiency in mammalian cells are not yet well understood. One parameter that has clearly been shown to improve HR efficiency is the use of long regions of isogenic DNA in the targeting vector.^3-7  Thomas and Capecchi first described an exponential relationship between length of homology and targeting frequency in the HPRT locus.³  A subsequent study by Hasty et al⁵  also revealed an increase in targeting rate with an increase in targeting arm size from 1.3 to 6.8 kb. In a systematic study by Deng and Capecchi, the targeting frequency was highly dependent on homology length up to 14 kb DNA, and then reached a plateau at 18.4 kb.⁷  To our knowledge, no systematic studies have yet been reported comparing vectors with “standard” targeting arms (ie, total arm length of 6-8 kb) to vectors with arms longer than 18.4 kb.

To explore this issue for the β-globin locus, we tested the ability of large targeting vectors to correct an engineered mutation of the murine β-globin gene. Our results showed that these vectors did not improve targeting efficiency in this locus, which suggests that the relationship between targeting arm size and HR efficiency is more complicated than previously thought.

The generation of mouse embryonic stem (ES) cell lines bearing the mutation changing GCT (encoding Ala) to ATC (encoding Ile) at position 6 of β-major globin gene (β^6I) has been described previously.⁸  The loxP-flanked PGK-neo cassette located in the targeted locus was subsequently removed via Cre-mediated recombination to create ES lines heterozygous for the β^6IΔPGK-neo allele (Figure 1A).

The targeting vectors used in this study (described in Figure 1) all contained targeting DNA sequences that were derived from a P1/PAC clone no. 1935, which contains a 110-kb insert isogenic to the β-globin locus in β^6I mutant ES cells (129/SvJ background). The ES cell culture, electroporation of linearized targeting vector DNA, G418 selection, and DNA preparation from G418-resistant clones were performed as previously described.⁸ 

We previously created a mouse ES cell line that contains an engineered mutation changing GCT (encoding Ala) to ATC (encoding Ile) in the position 6 of the mouse β-major globin gene (β^6I).⁸  The loxP-flanked PGK-neo selectable marker cassette was subsequently removed from the mutant locus via Cre/LoxP-mediated recombination. This model was designed to facilitate the testing of very large targeting vectors, so that analysis of homologous recombination could use probes that are contained within the targeting sequences. The β^6I mutation creates a novel EcoRV site in the β-major gene (Figure 1A; β^6IΔPGK-neo allele; E*, novel EcoRV site), which allows the mutant allele to be distinguished from the “wild-type,” corrected allele with Southern analysis. Because the probe is internal to the targeting sequences, it also detects randomly integrated targeting vectors.

We created replacement-type vectors that contained a total of 8, 16, 24, or 110 kb of wild-type isogenic DNA from the mouse β-globin gene locus in the targeting arms (Figure 1B). We compared the abilities of these vectors to correct the β^6I mutation in RW-4 (129/SvJ) ES cells bearing the β^6IΔPGK-neo allele. Positive selection was provided by a PGK-neo selectable marker gene (Figure 1B hatched box, PGK-neo) positioned downstream from the wild-type β-major gene (Figure 1B open box; β^WT) at the same position as a residual 42-bp lox-P site in the β^6IΔPGK-neo mutant allele. The placement of PGK-neo at this site allowed the targeting DNA used in these vectors to be uninterrupted by the heterologous lox-P site retained in the target locus. All vectors were linearized by restriction enzyme digestion at a unique site within the vector backbone that was close (< 26 bp) to one edge of the targeting fragment. Targeting vector DNA (40 μg) was transfected for each experiment. Therefore, fewer molar equivalents of the larger vectors were transfected, which was reflected in proportionally reduced numbers of G418-resistant colonies per microgram transfected DNA (data not shown). The ratio of homologous to random integration events (ie, the “efficiency” of HR) is still comparable for vectors of all sizes, because the total integration efficiency is “normalized” by the same PGK-neo cassette present in each vector.

To measure HR frequency, each targeting vector was transfected into β^6I mutant ES cells, and G418-resistant clones were screened by Southern analysis (Figure 2). A targeting vector with the same arms was previously used to create the β^6I mutation in RW-4 cells; 4 of 220 G418-resistant ES clones were shown to be correctly targeted.⁸  A representative Southern blot of BglII/EcoRV-digested ES cell DNA hybridized with the internal probe is shown in Figure 2. The pβC8 vector yielded one ES clone with a corrected allele of 207 tested (Figure 2, Table 1; pβC8). Three of 146 clones transfected with pβC24 contained a corrected β-globin gene (Table 1). We did not detect any targeted clones with either the 16-kb or 110-kb vectors (Table 1). The overall frequency of HR is underestimated by a factor of 2 because we detected HR of only the mutant β-globin allele. The 95% confidence intervals for HR frequencies for each of these vectors are overlapping and not statistically different from the targeting frequency of the vector used to make the original mutation.

Our results extend several earlier studies and provide a crude estimate of the critical length of homology that gives a “saturating” gene targeting efficiency in the β-globin locus. The absolute targeting frequency attained in our system was in the range of 10^–5 with targeting frequencies in the range of 1% to 2% of G418-resistant clones (similar to that previously described).^3,7,9  The HR frequencies obtained in this study are statistically identical to our overall experience using traditionally sized targeting vectors (Figure 2, Table 1).

Why did long regions of homology not improve targeting efficiency? One possibility is that large DNA fragments are unstable in host cells during integration. However, studies from our group and others have shown that purified human gDNA fragments of 100 kb are usually unrearranged after electroporation and integration into murine ES cell genomes^10,11  (R.M.K., T.J.L., unpublished observation, January 1999). It is also possible that HR is less efficient when great distances separate the ends of the targeting arms.

Our results suggest that the relationship between targeting arm size and HR efficiency is more complicated than previously appreciated,¹²  because targeting efficiency for the β-globin locus is not improved by very large vectors. The exact relationship between vector size and targeting efficiency is probably influenced by other variables, such as the targeted locus itself, vector design, and the status of cellular HR machinery,¹³  all of which are under investigation.

We thank Rick Goforth and the Siteman Cancer Center Embryonic Stem Cell Core for technical assistance. Kim Trinkaus provided excellent statistical support. Nancy Reidelberger provided expert assistance with manuscript preparation.