mismatch repair mechanism

Proc. Nati. Acad. Sci. USA Vol. 87, pp. 4680-4684, June 1990 Genetics

Introduction of specific point mutations into RNA polymerase II by gene targeting in mouse embryonic stem cells: Evidence for a DNA mismatch repair mechanism (site-specific mutagenesis in mammalian cells/homologous recombination/a-amanitin resistance)

CAROL MIERNICKI STEEG*, JAMES ELLIS*t, AND ALAN BERNSTEIN*tt *Division of Molecular and Developmental Biology, Mount Sinai Hospital Research Institute, 600 University Avenue, Toronto, Ontario M5G 1X5, Canada; and tDepartment of Medical Genetics, University of Toronto, Toronto, Ontario M5S 1A8, Canada

Communicated by Paul Berg, April 6, 1990

experiments. However, contrary to our expectations, the two closely linked point mutations frequently segregated in the targeted clones, suggesting that the targeting mechanism was not double crossover or gene conversion. Instead, we propose that DNA mismatch repair was responsible for the introduction of a single point mutation into the RPII215 locus. The efficiency and directionality ofthis repair pathway in mammalian cells vary, depending upon the particular nucleotide pair mismatch (21). Thus, if DNA mismatch repair plays a role in the targeted site-specific mutagenesis of other genes, these parameters may determine the efficiency with which specific point mutations can be introduced into the genes of mammalian cells.

We have introduced two specific point mutaABSTRACT tions, located 20 base pairs apart, into the endogenous murine gene that encodes the largest subunit of RNA polymerase II (RPII21S). The first mutation conferred resistance to the mushroom toxin a-amanitin (amar), and the second mutation generated a restriction fragment length polymorphism without altering the protein sequence. Targeted amar clones were generated at a frequency of 1 in 30 totipotent embryonic stem cells that expressed stably integrated DNA vectors after electroporation. Thirty to 40% of these clones had acquired both mutations, whereas, surprisingly, the remaining clones had acquired the specific amar point mutation but lacked the restriction fragment length polymorphism. We suggest that the latter clones were generated by independent DNA mismatch repair rather than by double crossover or gene conversion. These results demonstrate that it is possible to introduce specific point mutations into an endogenous gene in embryonic stem cells. Thus it should be possible to introduce single base substitutions into other cellular genes, including nonselectable genes, by optimizing the efficiency of gene transfer and/or the sensitivity of screening for targeted clones.

MATERIALS AND METHODS Plasmids and Construction of the Targeting Vector. Plasmid

pMClNeo-PolyA (9) was obtained from Stratagene. Plasmids pE26-7 and pRPMG were gifts of J. Corden. pE26-7 contains a 13-kb genomic fragment of mouse RPII215 with a single point mutation at position 6819 that confers a-amanitin resistance (22). pRPMG ("RNA polymerase minigene") contains a 19-kb insert composed of the entire mouse RPII215 genomic sequence (23) minus the first intron; the minigene

The creation of specific alterations in mammalian genes at their native chromosomal loci has been made possible by gene targeting, or recombination between introduced DNA and a homologous region in the host cell chromosome (1-3). By using gene targeting in totipotent embryonic stem (ES) cells, predetermined modifications have been introduced into the mouse germ line (4-7). Gene targeting in ES cells thus provides a general approach to assess the role of a particular gene in the whole animal. In addition, it permits the generation of mouse models of human genetic disease (8). To date, gene targeting has been used to effect the specific disruption (5, 6, 9-16) or correction (3, 4, 17-20) of a number of chromosomal loci in mammalian cells. However, the utility and versatility of this technology would be greatly increased if specific point mutations could be efficiently introduced into the genes of ES cells. Such subtle alterations in coding and regulatory regions would allow detailed study of gene structure and function. Here, we report the introduction by gene targeting of two specific point mutations, located 20 base pairs (bp) apart, into the gene that encodes the largest subunit of murine RNA polymerase II (RPII215). In contrast to earlier targeting strategies, in which the introduced DNA generally contained a large internal region of nonhomology (such as the bacterial neo gene), our vector DNA contained only 2 bp of nonhomology in several kilobases (kb) of DNA homologous to the chromosomal target locus. ES cell clones containing both point mutations were generated in three

also carries the a-amanitin resistance (ama9 point mutation. Plasmid pE26-7 was digested with EcoRI and EcoRV, and the 5.8-kb fragment containing the point mutation was subcloned into the EcoRI site of pUC19. The Sph I and HindIII sites in the pUC polylinker were subsequently destroyed, generating plasmid pamaHS. The 0.6-kb Sph I-HindIII fragment containing the mutation was subcloned into M13mpl9. Sitedirected mutagenesis was carried out using the "pulse" modification (24) to the procedure described by Zoller and Smith (25) and employing a uracil-substituted DNA template (26). The Sph I-HindIII insert was isolated from the mutant phage produced and used to replace the wild-type sequence in pamaHS, thereby creating pamaHSB. Culturing, Electroporation, and Selection of ES Cells. ESD3 cells (27) were cultured on a feeder layer of mitomycin C-inactivated primary mouse embryo fibroblasts (27) in Dulbecco's modified Eagle's medium/15% fetal calf serum/0.1 mM 2-mercaptoethanol/1 mM sodium pyruvate. Prior to electroporation, ES-D3 cells were resuspended in phosphatebuffered saline at a concentration of 6.25 x 106 cells per ml, and vector DNA was added. The cells were aliquoted into electroporation cuvettes (0.8 ml each) and subjected to a single 250-V pulse at 500 ,uF in a Bio-Rad Gene Pulser. They were subsequently plated onto gelatinized tissue culture Abbreviations: ES, embryonic stem; RFLP, restriction fragment length polymorphism; amar, a-amanitin resistance; PCR, polymerase chain reaction. tTo whom reprint requests should be addressed.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. 4680

Genetics:

Steeg et al.

dishes in Buffalo rat liver cell-conditioned (BRL) medium (28) at a density of 2.5 x 106 cells per 90-mm dish. Small aliquots were removed and plated separately to determine the percentage of cells surviving electroporation and plating. BRL medium was changed 24 hr after electroporation, and selection was begun after an additional 24 hr; thereafter, daily medium changes were made for 10 days. Selection medium for cells electroporated with pRPMG or with pamaHSB initially contained 1 jtg of a-amanitin per ml (Calbiochem); the concentration was increased to 2 ttg/ml after 2 days of selection. Cells electroporated with pMClNeo-PolyA were selected in 100 ,ug of G418 per ml (GIBCO). Resistant colonies were picked with finely drawn-out Pasteur pipettes and cultured on feeder layers. Southern Blot Analysis. Genomic DNA (5 ,ug per reaction) was digested with appropriate restriction enzymes, separated on a 0.7% agarose gel, transferred to nitrocellulose, and hybridized to a random-primed probe (Pharmacia oligolabeling kit) by standard methods (29). Blots were washed at a final concentration of 0.1 x SSC (0.015 M NaCI/0.0015 M sodium citrate)/0.1% SDS at 650C. DNA Amplification and Sequencing. Two-hundred nanograms of genomic DNA was amplified by the polymerase chain reaction (PCR) (30) using Taq DNA polymerase (Thermus aquaticus DNA polymerase) (Perkin-Elmer/Cetus). The recommended reaction buffer was adjusted to an optimal concentration of 1.2 mM MgCl2. Primers for PCR were RP1 (5'-GCTGTGTAGCCCAGGTTGAG-3') and RP3 (5'-GGGTGGAGAAACGATGGTGC-3'). Reactions were performed on an Ericomp thermal cycler set for an initial incubation at 94°C (31/2 min) followed by 30 cycles of 94°C (1 min), 57°C (1 min), 72°C (2 min), and a final incubation at 72°C (7 min). The amplified product was digested with Sph I and HindIII and separated on a 0.7% agarose gel, and the 600-bp fragment was purified using Geneclean (Bio 101). This fragment was ligated into pUC18, and recombinant plasmids were sequenced using Sequenase (USBC) and the RP2 primer (5'-AGTGGGTGTGAGACCAGCCA-3').

RESULTS Targeting Scheme. The pamaHSB targeting vector used for the site-specific mutagenesis of the RPII215 chromosomal locus of ES cells is illustrated in Fig. 1. The vector contains a 5.8-kb EcoRI fragment, extending from intron 7 to exon 21 of a doubly mutant RPII215 allele. The first mutation is a single base-pair change (A to G) at position 6819 in exon 15, which confers dominant resistance to the mushroom toxin a-amanitin (22). Since the 5.8-kb genomic fragment lacks the promoter and 5' and 3' coding regions of RPII215, integration of pamaHSB into random sites in the host cell chromosome does not confer a-amanitin resistance. Thus, only recombination events that introduce the amar point mutation into the native RPII215 locus would be selected by growing cells electroporated with pamaHSB in medium containing aamanitin. Using site-directed mutagenesis in M13, we introduced a second point mutation into the targeting vector 20 bp downstream from the first. The replacement of C with T at this position destroyed a BamHI site, while the amino acid encoded (isoleucine) remained the same. The silent BamHI mutation was created to provide an independent means of screening amar colonies for targeted clones. Since the distance between them was very short, we expected the two point mutations to cosegregate in all targeting events. Generation of Targeted amar ES Cell Clones. Plasmid pamaHSB was digested with EcoRI prior to electroporation into ES-D3 cells. ES-D3 cells were also electroporated in the absence of DNA to determine the spontaneous mutation rate to a-amanitin resistance, and with pMClNeo-PolyA (9) or an

Proc. Natl. Acad. Sci. USA 87 (1990)

8

Xb

15

I

E

II

BH

4681

21

I

H Xb

amaHSB

An

BarnH

VVT

CAG AAT GTA GAG GSGC AAG CGG ATC CCA GTC hTA CAT CTC CCCG TTC GCC TAG GGT

AMA

CAG GAT GTA GAG GGC AAG CGG ATT CCA GTC CTA CAT CTC CCG TTC GCC TAA GGT

Asp FIG. 1. Targeting vector for the site-specific mutagenesis of the mouse RPI1215 gene. The top panel depicts a portion of the RPII215 genomic sequence. Exons are shown in black, with exon number above; introns are in white. Xb, Xba I; E, EcoRI; B, BamHI; H, HindIII. The solid black bar depicts the 5.8-kb genomic fragment of the amar mutant allele of RPJI215 that was used as the insert in the pamaHSB targeting vector. The sequence of the wild-type RPII215 gene beginning at nucleotide 6816 in exon 15 and the same region of the pamaHSB vector are shown in the bottom panel. The point mutation that confers a-amanitin resistance and the silent point mutation that creates a BamHI restriction fragment length polymorphism (RFLP) are indicated by arrows.

amar RPII215 minigene (pRPMG) to assess the frequency of random integration and expression of introduced DNA. After electroporation, the cells were plated in BRL medium and allowed to recover for 48 hr. Selection with a-amanitin or with G418 was then begun and continued for 10 days. Resistant clones were picked, expanded, and analyzed. Table 1 gives the results of three independent gene targeting experiments. Pooling the data from experiments 2 and 3, a single spontaneous amar clone arose from 3 x 106 ES-D3 cells that survived electroporation in the absence of DNA. In parallel, 107 ES-D3 cells surviving electroporation with the pamaHSB targeting vector yielded a total of 180 amar clones. The elevated frequency at which the latter clones were generated, in contrast to the spontaneous mutation rate, strongly suggests that they were derived from gene targeting. To determine the frequency of random integration events and the efficiency of electroporation, we used a plasmid that could be selected with the same drug as the targeted clones. Thus, the positive control plasmid pRPMG, which contains a functional amar RPII215 minigene, was electroporated and selected with a-amanitin in parallel to the pamaHSB targeting vector (see Table 1). In both experiments 2 and 3, targeted amar clones were generated at a frequency of 1 in 30 cells that expressed stably integrated pRPMG DNA (correcting for cell numbers and for the molarity of the introduced DNA). The overall efficiency of electroporation in experiment 3 (see Table 1, electroporation of pRPMG and pMClNeo-PolyA) was high relative to the other two experiments, and the absolute number of targeted amar clones generated in this experiment was 5-fold greater than in experiment 2.

4682

Genetics:

Steeg et al.

Proc. Natl. Acad. Sci. USA 87 (1990)

Table 1. Gene targeting by electroporation of pamaHSB to generate amar ES cells Drug-resistant ES cells ES cells DNA electroporated nM ,ug electroporated, no. surviving, no. colonies, no. Plasmid Exp. 71 G418 3.9 x 105 5.0 x 106 20 1 pMClNeo-PolyA 10.5 9.4 400 5.0 x 10 3.9 x 106 t pamaHSB 5.2 x 105 5.0 x 106 Oama 2 No DNA 30 ama 5.2 x 105 5.0 x 106 35 3.3 pRPMG 30 ama 5.2 x 106 5.0 X 107 9.4 400 pamaHSB

BamHI RFLP

Frequency* pRPMG pMClNeo-PolyA

1/3 7/22

t 1/29

0/1 1 ama 2.5 x 106 2.5 x 107 No DNA 180 ama 5.0 x 105 5.0 x 106 40 3.8 pRPMG 289 G418 5.0 x 105 5.0 x 106 40 pMClNeo-PolyA 21 1/9 1/30 12/29 150 ama 5.0 x 106 5.0 x 107 9.4 400 pamaHSB *Frequency of amar colonies from pamaHSB electroporation as compared to number of random integrations of pRPMG or pMClNeo-PolyA (corrected to equivalent DNA molarity and surviving cell numbers). tSelection in experiment 1 was with a-amanitin (Boehringer Mannheim), which was past its expiration date, at a concentration of 1 ,ug/ml. This produced many false positive clones. It was necessary to pool the clones on individual plates and reselect using fresh a-amanitin (Calbiochem) at 2 ,g/ml. An accurate targeting frequency cannot be calculated, but the three twice-selected clones that were analyzed did arise from different original plates. Subsequent experiments used fresh a-amanitin (Calbiochem) at 2 ug/ml. 3

Characterization of amar ES Cell Clones by Southern Blot Hybridizations. DNA from the amar clones generated by electroporation of ES-D3 cells with pamaHSB was subjected to Southern blot analysis to screen for the BamHI RFLP. The probe used was a 1.8-kb Bgl II-HindIII restriction fragment of RPII215 that spans the BamHI site in exon 15 (see Fig. 2). As shown in Fig. 2, digestion of wild-type ES-D3 DNA with BamHI followed by hybridization to this probe yielded two

-amaHSB XE H X

H

E

II

I

i

I

B

B

B

5.8

5.2 -

B

X

E

probe H

0IW0

-ps 1

1-

_.5.85.2--s

FIG. 2. Southern blot analysis of DNA from the parent ES-D3 cell line and a typical amar targeted ES cell clone (9-3). A map of the native RPJJ215 chromosomal locus is given, with the solid black bar representing the 1.8-kb Bgl II-HindIII fragment used as a probe, and the open bar representing the 5.8-kb insert of the pamaHSB targeting vector. Lanes of the autoradiogram are arranged in pairs, ES-D3 on the left and clone 9-3 on the right. Fragment sizes are given in kb. B, BamHI; X, Xba I; E, EcoRI; H, HindIII.

bands of 5.8 and 5.2 kb. In contrast, 30-40% of the amar clones generated in each of three independent experiments possessed the introduced BamHI RFLP (see Table 1). Southern blot analysis of one such clone (9-3), generated in experiment 1, is shown in Fig. 2. Digestion of DNA from this clone with BamHI gave rise to a new band of 11 kb, whereas the intensities of the 5.8- and 5.2-kb bands were halved as predicted. When subjected to additional restriction digests with enzymes that cleave outside the region homologous to the introduced DNA, the restriction patterns of wild-type ES-D3 and clone 9-3 were identical across a region spanning 27.5 kb and encompassing essentially the whole RPII215 locus (Fig. 2). Identical results for 20 of 54 clones analyzed strongly suggest that the simultaneous acquisition of aamanitin resistance and the BamHI polymorphism by these clones was the result of a gene targeting event. None of the 54 amar clones contained either plasmid or RPII215 sequences integrated at random sites elsewhere in the genome (data not shown). Amplification and Sequence Analysis of Target Loci of ama' ES Cell Clones. Surprisingly, 60-70% of the amar clones generated in each experiment did not possess the BamHI RFLP. However, these clones arose at a much higher rate than the spontaneous mutation frequency (see Table 1, experiments 2 and 3), suggesting that they too resulted from gene targeting events. To determine whether these clones possessed the specific amar point mutation (A to G at position 6819 in exon 15) and to confirm that the clones possessing the BamHI polymorphism also had the specific amar mutation, we amplified the targeted region using PCR (30), subcloned the amplified fragment, and sequenced the DNA from individual subclones derived from six amar colonies. Two amar colonies that possessed the BamHI polymorphism by Southern analysis were found to have the two expected point mutations in one allele (see Fig. 3 for sequence of antisense strand). Four amar colonies that had not demonstrated the BamHI RFLP were confirmed to have intact BamHI sites in both alleles but had acquired the amar point mutation (T to C in the antisense strand) at position 6819 in one allele (see Fig. 3). No other mutations were detected within a 140-bp region centered around position 6819 in any of the targeted amar clones analyzed. The single spontaneous amar mutant that arose after electroporation of ES-D3 cells in the absence of DNA also possessed an intact BamHI site but, in contrast, was wild-type at position 6819 (data not shown). This latter observation supports the conclusion that the clones possessing the specific point mutation at position 6819 resulted from gene targeting events.

Genetics: RPI S

Steeg et al. B

Proc. Natl. Acad. Sci. USA 87 (1990) Amnar

H

G A T C .,

_...l f

,,W.

G A T C

-.

RP11215

.

(WT)

s

IAsp

-*#. O"

aw*

T = A

RP3

RP2

G A T C

Bamr

p pamnaHSB aG a-

-

4683

a

GG

T

Bam+

Arras

-oft

db

Heteroduplex formation

.F

_&

I

1.1

a"

::B:_.

Z lam Am --f

*

Bam Hi =G T

FIG. 3. DNA sequence analysis of amar targeted ES cell clones. (Upper) Region of the RPIJ215 gene that was amplified from the amar clones by PCR and subsequently sequenced. RP1, RP2, and RP3 were the 20-bp primers used for PCR or for sequencing. B, the BamHI site in exon 15; S, Sph I; H, HindIII. (Lower) Sequence ladders generated from the antisense strand of a particular allele (position 6816 at top). Right: wild-type allele of amar clone SA. Center: targeted allele of amar clone 5A, which has a single point mutation as indicated by the arrow. Left: targeted allele of clone 18B, which has two specific point mutations as indicated by the arrows. Clones SA and 18B were generated in experiment 2.

Independent mismatch repair

-G

-c-

-C -GBam+ Amar

DISCUSSION The experiments presented here demonstrate the feasibility of introducing specific point mutations into an endogenous gene in ES cells. We have introduced two point mutations into the RPII215 chromosomal locus, at a frequency of 1 targeted cell per 30 that expressed stably integrated DNA after electroporation. This relative efficiency is higher than is generally achieved by electroporation and direct positive selection and is comparable to that achieved by a positivenegative selection strategy (11, 15). One explanation for the high relative efficiency that we observed may be that, by using a vector containing only two single base-pair substitutions within 5.8 kb of homology, we avoided large internal regions of nonhomology that may hinder recombination by gene conversion (31). A similar relative frequency was observed for the disruption of the Hox).1 gene by microinjection of a vector containing only 20 bp of nonhomology (13). Implicit in our targeting scheme was the assumption that the short distance (20 bp) between the two point mutations in the introduced DNA would result in their cosegregation in the majority of targeted recombination events. In fact, 30-40% of the selected amar events contained both mutations and are most easily explained by gene conversion or double crossover. In contrast, the remaining 60-70%o of the amar colonies had the specific amar point mutation but lacked the BamHI polymorphism. It is very unlikely that these clones arose from frequent crossovers or resolution of gene conversion within the 20-nucleotide region. Rather, we suggest that each point mutation was independently corrected by a DNA mismatch repair pathway. The specific mismatches in our experiment would be two GT mispairs if the sense strand of the introduced DNA invaded and formed a heteroduplex with the endogenous gene (see Fig. 4) or two AC mispairs if the antisense strand were involved. Previous experiments have shown that =95% of GT mismatches and 75% of AC mismatches in mammalian cells undergo correction (21, 32). Over 90% of GT mismatches are corrected to GC (21, 32, 33), whereas there is no directional bias for AC repair (21). The directionality of GT repair would account for the high proportion of targeted clones with only one point mutation, by favoring the amar phenotype in the case of the first mismatch but restoring the BamHI site in the second case (see Fig. 4). The targeted clones containing both point mutations could have arisen from independent mismatch repair of the AC

T= G

Amas

-C-A Amar BamII

(selected against)

III,

IV

FIG. 4. Proposed mechanism for the targeted introduction of point mutations into RPII215: Heteroduplex formation and independent DNA mismatch repair. The top panel depicts the endogenous RPII21S locus (solid lines) and the homologous region of the pamaHSB targeting vector (open lines). WT, wild type; s, sense strand; a, antisense strand. The specific base pairs shown confer a-amanitinresistance or sensitivity and the presence or absence of a BamHI restriction site as indicated. The center panel depicts the heteroduplex formed if the sense strand of the pamaHSB vector is the invading strand. The possible products resulting from independent DNA mismatch repair and subsequent resolution of the heteroduplex are shown below (roman numerals). The predominance of GT mismatch repair to GC is illustrated by the thick arrow. In three targeting experiments, 60-70% of the amar clones analyzed were from class I. The remaining 30-40o were from class II. These could have arisen from AC mismatch repair following invasion of the antisense strand of pamaHSB or, alternatively, from double crossover or gene conversion. Classes III and IV, in which the first mismatch is repaired to AT, would not survive selection with a-amanitin.

mismatches of a heteroduplex formed by the invasion of the pamaHSB antisense strand. Efficient DNA mismatch repair enzymes might also account for the high relative targeting frequency. Segregation of point mutations was also observed in experiments in which heteroduplex DNAs containing several mismatches were injected into mammalian cells (34) and in extrachromosomal recombination experiments performed by the same group in which homologous DNAs were coinjected into mammalian cells (35). The authors postulated that independent mismatch repair was responsible for the segregation in the former case and gene conversion was responsible in the latter. Given their results, our observations on RPII215 are most likely not locus-specific. If DNA mismatch repair of a heteroduplex is involved in gene targeting, then some point mutations may be easier to introduce than others, depending on the efficiency and directionality of repair of the specific mismatch. Further experiments will be necessary to confirm this hypothesis, but it should be considered in devising future strategies for the site-specific mutagenesis of mammalian genes.

It has been suggested that repair of a heteroduplex intermediate was involved in the introduction of de novo mutations in the target locus in two other gene targeting studies

4684

Genetics: Steeg et al.

(20, 36). However, in each case, the target locus or the vector DNA contained a large deletion as well as single base-pair mismatches in the region of homology. The absence of de novo mutations in the targeted RPII215 genes described here suggests that heteroduplex-induced mutagenesis (36) may require a large internal region of nonhomology near single base-pair mismatches. Previous studies have shown that resistance to a-amanitin inhibits the in vitro differentiation of certain cells (37). The amar ES cell clones we have generated will be useful for further investigation of this phenomenon, both in vitro and in ES cell-derived chimeric and transgenic mice. In addition, the strategy for site-specific mutagenesis described here could easily be extended to those mammalian genes for which point mutations can be directly selected. The direct correlation between the absolute number of targeted clones and the efficiency of electroporation suggests that point mutations could also be introduced into nonselectable genes. For example, in our most efficient experiment, one amar clone was generated per 3.3 x 104 ES cells surviving electroporation. Although detection and isolation of these rare targeted events would be laborious, it should be possible to screen pools of electroporated ES cells by PCR of the target locus and hybridization to mutation-specific oligonucleotides. Further optimization of gene transfer efficiency, perhaps by microinjecting DNA, would increase the absolute number of targeting events and thereby ease the screening process. The introduction of specific point mutations into the coding or regulatory regions of mammalian genes will allow finestructure analysis of the functions of protein domains and of cis-acting DNA elements. When combined with the ability of totipotent ES cells to contribute to the mouse germ line, this approach will make it possible to generate mice carrying specific point mutations, hence providing a powerful means to address a wide variety of biological questions. We thank J. Corden for the generous gift of the pE26-7 and pRPMG plasmids. We are grateful to A. Joyner, A. Gossler, A. Auerbach, and C. Moens for advice on culturing ES cells. We also thank J. Rossant and P. Dubreuil for critical reading of the manuscript. This work was supported by grants from the Medical Research Council and the National Cancer Institute of Canada. J.E. was supported by a studentship from the Natural Sciences and Engineering Research Council of Canada. 1. Lin, F.-L., Sperle, K. & Sternberg, N. (1985) Proc. Natl. Acad. Sci. USA 82, 1391-1395. 2. Smithies, O., Gregg, R. G., Boggs, S. S., Koralewski, M. A. & Kucherlapati, R. S. (1985) Nature (London) 317, 230-234. 3. Thomas, K. R., Folger, K. R. & Capecchi, M. R. (1986) Cell 44, 419-428. 4. Thompson, S., Clarke, A. R., Pow, A. M., Hooper, M. L. & Melton, D. W. (1989) Cell 56, 313-321. 5. Schwartzberg, P. L., Goff, S. P. & Robertson, E. J. (1989) Science 246, 799-803.

Proc. Natl. Acad. Sci. USA 87 (1990) 6. Zijlstra, M., Li, E., Sajjadi, F., Subramani, S. & Jaenisch, R. (1989) Nature (London) 342, 435-438. 7. Koller, B. H., Hagemann, L. J., Doetschman, T., Hagaman, J. R., Huang, S., Williams, P. J., First, N. L., Maeda, N. & Smithies, 0. (1989) Proc. Natl. Acad. Sci. USA 86,8927-8931. 8. Capecchi, M. R. (1989) Trends Genet. 5, 70-76. 9. Thomas, K. R. & Capecchi, M. R. (1987) Cell 51, 503-512. 10. Doetschman, T., Maeda, N. & Smithies, 0. (1988) Proc. Natl. Acad. Sci. USA 85, 8583-8587. 11. Mansour, S. L., Thomas, K. R. & Capecchi, M. R. (1988) Nature (London) 336, 348-352. 12. Sedivy, J. M. & Sharp, P. A. (1989) Proc. Natl. Acad. Sci. USA 86, 227-231. 13. Zimmer, A. & Gruss, P. (1989) Nature (London) 338, 150-153. 14. Joyner, A. L., Skarnes, W. C. & Rossant, J. (1989) Nature (London) 338, 153-156. 15. Johnson, R. S., Sheng, M., Greenberg, M. E., Kolodner, R. D., Papaioannou, V. E. & Spiegelman, B. M. (1989) Science 245, 1234-1236. 16. Koller, B. H. & Smithies, 0. (1989) Proc. Natl. Acad. Sci. USA 86, 8932-8935. 17. Doetschman, T., Gregg, R. G., Maeda, N., Hooper, M. L., Melton, D. W., Thompson, S. & Smithies, 0. (1987) Nature (London) 330, 576-578. 18. Baker, M. D., Pennell, N., Bosnoyan, L. & Shulman, M. J. (1988) Proc. Natl. Acad. Sci. USA 85, 6432-6436. 19. Jasin, M. & Berg, P. (1988) Genes Dev. 2, 1353-1363. 20. Brinster, R. L., Braun, R. E., Lo, D., Avarbock, M. R., Oram, F. & Palmiter, R. D. (1989) Proc. Natl. Acad. Sci. USA 86, 7087-7091. 21. Brown, T. C. & Jiricny, J. (1988) Cell 54, 705-711. 22. Bartolomei, M. S. & Corden, J. L. (1987) Mol. Cell. Biol. 7, 586-594. 23. Ahearn, J. M., Jr., Bartolomei, M. S., West, M. L., Cisek, L. J. & Corden, J. L. (1987) J. Biol. Chem. 262, 10695-10705. 24. Nisbet, I. T. & Beilharz, M. W. (1985) Gene Anal. Tech. 2, 23-29. 25. Zoller, M. J. & Smith, M. (1982) Nucleic Acids Res. 10, 6487-6500. 26. Kunkel, T. A., Roberts, J. D. & Zakour, R. A. (1987) Methods Enzymol. 154, 367-382. 27. Doetschman, T. C., Eistetter, H., Katz, M., Schmidt, W. & Kemler, R. (1985) J. Embryol. Exp. Morphol. 87, 27-45. 28. Smith, A. G. & Hooper, M. L. (1987) Dev. Biol. 121, 1-9. 29. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Lab., Cold Spring Harbor, NY). 30. Saiki, R. K., Scharf, S., Faloona, F., Mullis, K. B., Horn, G. T., Erlich, H. A. & Arnheim, N. (1985) Science 230, 1350-

1354.

31. Ellis, J. & Bernstein, A. (1989) Mol. Cell. Biol. 9, 1621-1627. 32. Brown, T. C. & Jiricny, J. (1987) Cell 50, 945-950. 33. Wiebauer, K. & Jiricny, J. (1989) Nature (London) 339, 234236. 34. Folger, K. R., Thomas, K. & Capecchi, M. R. (1985) Mol. Cell. Biol. 5, 70-74. 35. Folger, K. R., Thomas, K. & Capecchi, M. R. (1985) Mol. Cell. Biol. 5, 59-69. 36. Thomas, K. R. & Capecchi, M. R. (1986) Nature (London) 324, 34-38. 37. Crerar, M. M., Leather, R., David, E. & Pearson, M. L. (1983) Mol. Cell. Biol. 3, 946-955.