The use of gene targeting to develop animal ... - Semantic Scholar

The use of gene targeting to develop animal models for human genetic diseases DAVID W.MELTON Depurtrnetit of Molecrrlur Riology, University o f Ediriburgh, Mayfield Houd, Edinburgh Ef I0 .VR, U.K .

Iritrodirctiori

In the early 1980s it became possible to produce animals carrying experimentally introduced foreign genes in their genome. T h e foreign gene is referred t o as a transgene and the animals as being transgenic. This technology, applied initially to mice and subsequently to other mammals, has been o f considerable benefit to basic biological and medical research (for a review. see [ 11). and promises to have very important practical and commercial applications. Transgenic mice are routinely used in the study o f mammalian gene expression, in particular to identify DNA sequences which control the tissue specific variations in the expression of a given gene. O u r knowledge of the actions and interactions o f oncogenes - genes that are instrumental in the development of tumours - is largely derived from cxperiments involving transgenic mice. T h e usual method for generating transgenic animals involves the microinjection of the chosen gene into the nucleus of a fertilized egg. However. the use o f transgenic animals is complicated and limited by two properties associated with the conventional methods for introducing D N A into the genome. First, the introduced gene can potentially integrate at any site in the genomc. Consequently, undesirable mutations may arise by insertion into an essential gene. Alternatively. a transgene may integrate into a region of the genome which modifies o r even abolishes its expression. Secondly. the number o f copies o f the transgene integrated can vary several hundredfold. T h e level o f expression o f an introduced transgene may also be affected by the number o f copies present. Thus, useful as the approach is, the interprctation o f results is often subject to complications. If more precise changes could be made t o the mammalian germ line. it would be possible to modify and study the expression of a gene in its natural chromosomal position in a whole animal. Many human genetic diseases and some c;incers result from mutation of a single gene. By inactivating such genes in mice it might prove possible t o produce mouse models for human diseases and permit evaluation of novel forms o f treatment, such as somatic gene replacement therapy, where functional genes would be introduced in an attempt t o complement the genetic deficiency in an individual [ 2 I. Et?ibnorricstem c d l s

l ' h c alternative systcm for making transgenic animals that makes such gene manipulation possible involves the use o f embryonic stem ( E S ) cells. This systcm has been developed in the mouse 131. T h e properties of these cells which make them extremely valuable as a routc into the mouse germ line are illustrated in Fig. I . ES cells are isolated from the inner cell mass of early embryos, called blastocysts. They arc the cells which normally differentiate to form all the various cell types in the mouse. Undcr appropriate. very stringent conditions these cells can be cultured iri v i m without losing this ability. Hence, if they are reintroduced by microinjection into the inner cell mass of a different embryo, which is subsequently reimplanted into a foster mothcr and allowed to VOl. 18

Twenty-seventh Colworth Medal Lecture Delivered on 19 July 1990 at the University of Aberdeen -r

v-

-

.

~

s-

-.

.

v

I

-

a

'

F

c

. * .' L '

...

j.

..

.

.

D R DAVID W.MELTON develop, thc tissues o f the resultant mouse will have contributions from both the host embryo and the injected ES cells. Such a mouse which contains two genetically distinct cell types is known as a chimaera. If the ES cells and the host embryo cells carry genes for different coat colours. then both colours will be evident in the coat o f the chimaera. If ES cells are to be used as a route into the germ line. it is essential that they contributc t o the germ cells o f thc chimaera and produce gametes. s o that ES-cell-derived genes can be passed on t o offspring. Inheritance of ES-cell-derived genes is indicated by the presence o f offspring that have the ES-cellderived coat colour. Thus, by breeding it is possible t o obtain mice which are genetically identical t o cultured ES cells. Consequently, if the genome of cultured ES cells is manipulated, using the gene targeting procedures to be described. mice can be derived which carry the same modification in every cell of their body and transmit it to future generations.

I035

I036

L

\

Isolate blastocyst

-

BIOCHEMICAL SOCIETY TRANSACTIONS

Ribose-5-P +ATP

1-1

I piFiFJ

PRPP

\

jnReirnDlant / into foster

5-Phosphoribosyl-l - amine

1 I

cell mass host embryo

- I --

Adenine -AMPPiPP

IMP

GMP

1

Gldnine

Gene targeting

Adenosine

Chimaera

Hypoxanthine

Xanthine oxidase

Uric Acid Urate axidase Not humans

1

Allantoin Fig. 2. I’itritie metuholi.sm Host-embryo-derived progeny

Targeted ES-cellderived progeny

Fig. 1. llhtstrutioti of’ the itse of’ ernhryotiic stem cells as a roiite into the mummuliuti germ line

Abbreviations: ADA, adenosine deaminasc; AMP, adcnosine monophosphate; AMPRT, amidophosphoribosyltransferase; ATP, adenosine triphosphate; GMP, guanosine monophosphate; HPRT, hypoxanthine-guanine phosphoribosyltransferase; IMP, inosine monophosphate; PNP, purine nucleotide phosphorylasc; PRPP, 5’-phosphoribosyl1-pyrophosphate.

Gene turgetitig

Gene targeting procedures rely on the ability of DNA, introduced into a cell, to locate and recombine with a homologous chromosomal region. This was first demonstrated by Smithies et ul. 141 who succeeded in making targeted alterations to the human B-globin gene in cultured cells. Inconveniently, the frequency of targeting is low relative to the frequency of random integration events, typically only one targeting event for every thousand random integrations. If the targeting procedure is to be generally useful, it is clearly necessary to be able to identify efficiently the rare targeting events in a cell population. 7urgeting the III’K 7‘getie

For this reason studies on the hypoxanthine-guanine phosphoribosyltransferase (HPRT) gene have proved useful in developing gene targeting strategies. HPRT is involved in the purine salvage pathway in mammalian cells. HPRT offers two major advantages for this work. First, the gene is X-chromosome linked, so thcre is one copy in males and only one functional copy in females, due to X-chromosome inactivation. Secondly, it is possible to select both for and against HPRT expression in cultured cells. HPRT catalyses the transfer of a phosphoribosyl group from 5’-phosphoribosyl- 1-pyrophosphate (PRPP) to either of the purine bases, hypoxanthine or guanine, to form inosine or guanosine monophosphate, respectively (see Fig. 2). Cells can also make IMP and GMP by de iiovo synthesis. The first step in the de tiovo pathway is catalysed by PRPP synthetase. If the de tiovo pathway is blocked by the antifolate drug amino-

pterin, only cells with HPRT activity will survive. This is the basis of the HAT (hypoxanthine, aminopterin, thymidine) selection technique. Cells lacking HPRT may be selected using a base analogue, such as 6-thioguaninc, which is phosphoribosylated by HPRT with fatal consequences for the cell. HPRT is a housekeeping gene which is expressed at low levels in most tissues. However, a higher lcvel o f expression has been reported in the brain. In humans, HPRT deficiency results in the severe neurological disorder Lesch-Nyhan syndrome. We are interested in the mechanism for the elevated expression in the brain and the reason for its peculiar sensitivity to HPRT deficiency. HPRT-deficicnt micc have been obtained from HPRT-deficient ES cells isolated in culture with the aim of studying the disease 15, 61. Although the HPRT-deficient mice have slightly reduccd brain dopamine levels compared with normal animals, no spontaneous behavioural abnormalities characteristic o f Lesch-Nyhan syndrome have been reported s o far. However, a rccent report suggests that behavioural abnormalitics can bc induced in these animals by administration o f amphetamines, which are known to affect brain dopamine levels 171. Thc reason for the different response o f mice and humans to HPRT deficiency is unclear. The purine biochemical pathways operating in man and rodents are slightly different. Rodents may be less dependent o n the purinc salvage pathway Alternatively, rodents have an additional enzyme activity, urate oxidase, which prevents accumulation o f the insoluble purine waste product uric acid (sec Fig. 2). The HPRT deficiency has been corrected by genc targeting in ES

1037

27th COLWORTH M E D A L LECTURE cells 1x1 and the corrected gene has been introduced into the mouse germ line 191. Thus, we have used gene targeting in E S cells t o produce mice in which an HPRT deficiency has been corrected in every cell of their body. Although we d o not at present have a good disease phenotype that we can study and attempt t o correct, this system is proving very useful for the study of H P R T gene expression in mice. T h e structure o f the normal HPRT gene is shown in Fig. 3. T h e protein coding sequences of the gene are divided into nine distinct cxon blocks, indicated by the numbered open boxes. T h e promoter region that controls transcription of the gene is represented by the striped box. T h e fully shaded box at the beginning of the gene and the stippled box at its end. represent parts o f the gene which are transcribcd into the HPRT messenger RNA, but d o not code for the HPRT protein. T h e structure of the HPRT mutation isolated by tlooper et ul. [ 51 is shown below the normal gene. T h e mutation consists o f a deletion of the first two cxons and the promoter rcgion of the H P R T gene. T h e Figure also shows the structure of the targeting vector which was constructed to correct the mutation. T h e correcting vector contains the two exons that arc deleted in the mutant ES cell line and the transcription control region. Although, in this case, the correcting vector contains the normal HPRT promoter, the box representing this region is striped in the opposite direction to its counterpart in the HPRT gene itself. This serves t o indicate that different control sequences could also be used to manipulate the expression of the corrected gene. To locate the target gene the correcting vector needs t o contain a stretch o f D N A sequence identical t o it. In this case this is provided by the exon 3 region. A restriction enzyme was used t o cut the vector D N A within exon 3. t o produce linear molecules, before introduction into the HPRT-deficient ES cells, a s this is found t o promote recombination. T h e standard method for introducing D N A into ES cells is by elcctroporation. This involves suspending the cells in a solution containing the vector D N A and applying a short pulse of electricity, which causes the cell membranes to become temporarily permeable. If, as shown in Fig. 3, the targeting vector inserts into the genome by recombination with the mutant HPRT gene, the mutation will be corrected and gene activity will be restored. However, cells in which the vector integrates randomly will remain HPRT deficient because the correcting vcctor itself does not contain the full HPRTcoding region. T h e outcome o f the gene targeting event is illustrated at the bottom o f Fig. 3. T h e corrected gcne has the same structure as the normal gene. cxccpt that the intron between exons 1 and 2 is reduced in size. Another consequence o f targeted correction of the gene is that exon 3 becomes duplicated. T h e second copy is upstream of the gene and does not interfere with its exprcssion. To isolate cells in which the targeted correction had taken place. the cell population was subjected t o selection for

1

2n

-m

U

r

3.

F-ITiX 1 W

5

6

789

Y

3

3 n "

4

2

3 "

Y

Y

4

5

6

789

? 6, Y

7&

HPRTgene HPRT deletion

Corrected gene

Fig. 3. C i)rrc'ctiori of' uri HI'K7' deJicieiiiy by gcwe targetiiig irsirig urz irisertiori vector VOl. 18

E'eutirres of'thr IIl'U 7'gerie targetirig systern

T h e HPRT gene targeting experiments described involve insertion events, with around 3 kb of homology to the target locus. When expressed relative to the number o f cells treated, the correction frequency seems low, around 5 x l o - ' . This is because ES cells d o not take u p and intcgrate D N A readily. By carrying out positive control experimcnts using a fully functional, but truncated version o f the HPRT gene itself, an H P R T minigene, which docs not require homologous recombination for expression, we believe that around 1 % o f the integrations which result in expression are homologous events. We have now looked at a large number of independent targeted cloncs and only rarely d o we see any evidence for non-homologous integrations in addition t o the homologous, correcting, event. This suggests that it is perfectly feasible to make precise genome alterations without introducing undesirable changes elsewhere. We have also studied the stability of the corrected gene structures. We were concerned that the corrected gene, with its upstream exon 3 duplication, might be prone to a reversal of the original recombination event. We have no evidence for any corrected gene instability. Again this seems encouraging for the prospects of mammalian germ line manipulation. Expressiori of'the corrected HPR Tgerie iri r7iic.e

Correcting vector

4,

HPRT enzyme activity. Only corrected cells will have HPRT activity and survive in these conditions. O u r first vcctor contained 6 5 0 bp o f the natural HPRT promoter region. By substituting other promoters for the natural HPRT promoter in the correcting vector, this approach may be used to alter the transcriptional control region of the endogenous H P R T gene itself. With this aim we have constructed two other correcting vectors - one containing just a 35 bp core region from the H P R T promoter [ 101and the second containing the promoter from a gene with a different pattern of expression to HPRT - the metallothionein gene promoter. All three vectors worked and gave a similar correction frequency around five correction events per population o f 10 million HPRT-deficient ES cells electroporated. T h e level o f H P R T mRNA in individual corrected clones was determined by extracting R N A and doing Northern hybridizations. After probing for H P R T mRNA, the filters were reprobed for an unrelated mRNA, actin mRNA, to allow comparisons between tracks. T h e HPRT/actin signal ratios for the different corrected clones were obtained by densitometry and expressed relative to the value obtained for wild-type ES cells. T h e corrected gene with the full HPRT promoter region functioned less well than the normal gene. In two independent clones we obtained a value of 0.3. We obtained the same value in a clone with the metallothionein promoter and a slightly higher level of HPRT mRNA (0.5) in a clonc with the minimal HPRT promoter region. T h e reason for the lower level o f expression from the corrected gene will be considered after its expression in mice has been described.

Mice containing the corrected gene. under thc control o f the full HPRT promoter region, were produced by injecting corrected ES cells into host blastocysts. T h e ES cell line is homozygous for two different recessive alleles which both lighten coat colour, while host embryos are wild-type at both these loci. Thus, resulting chimaeras are readily idcntified by the presence o f light patches in an otherwise dark coat. O n e of the eight malc chimaeras tested showed germ line transmission of the ES-cell-dcrived coat-colour markers t o some of its progeny. T h e expression of the corrected gene in a number of tissues from the ES-cell-derived mice has been compared with the expression of the wild-type gene. In the wild-type, HPRT m R N A levels are highest in brain, inter-

1038 mediate in heart, lung and kidney, and lowest in liver. To be able to compare HPRT mRNA levels between mice. the Northern filters were reprobed with an actin probe and HPRT/actin ratios were obtained. With one exception, the pattern o f expression of the corrected gene was similar t o the wild-type, with the highest HPRT mRNA level in brain. The HPRT/actin ratios for each tissue in the corrected mouse were calcdated as a percentage of the value in the same tissue in the wild-type. The values in liver. brain, heart and kidney agreed reasonably well with the 30% value obtained from ES cells. The exception was lung, where the level of HPRT mRNA was several fold over the expected value and approximated to the level of expression of the wild-type gene in lung. S o , with this exception, correction has restored the appropriate pattern o f expression to the gene, but at a reduced level compared to the wild-type. Keusori ji)r poor corrected gene expressiori Several reasons for the low level of corrected gene cxpression were considered (see Fig. 3). First, that the correcting vector was not providing a sufficient amount o f 5' flanking information. Only 650 bp of this region is being introduced. while the deletion has removed over 1 0 kb of 5' flanking DNA. Secondly, that the reduced size of the first intron in the corrected gene might have a deleterious effect o n expression. Thirdly, that the upstream copy of the plasmid vector might be interfering with expression. This last possibility was the easiest to investigate. To d o this, a modified version of the correcting vector was constructed. This vector was identical to the original vector except that the plasmid sequence is n o w present at the restriction site within exon 3. This results in the release of plasmid sequence when the vector is cut. s o that the corrected gene does not contain an upstream copy o f the plasmid. Both vectors gave the same frequency of correction. but a difference emerged when the level of HPRT mRNA in the corrected clones was examined. The HPRT/ actin ratios, normalized to the wild-type value. for three independent clones made with the original correcting vector were : 0.3.0.7 and 1.3, while the values for three clones made with the modified vector were: 2.0, 2.4 and 2.5. So the presence of the upstream plasmid sequence appears to have a deleterious, but rather variable. effect on expression in different clones. We have now corrected a gene deficiency and restored activity to greater than the wild-type level. Targeting other genes A variety of strategies have been developed for targeting other genes, where the facility of direct selection is not available. Most have involved gene inactivation with the target locus being disrupted by a dominantly selectable marker gene. The marker used is normally the bacterial neomycin phosphotransferase (tieo)gene. Cells expressing this gene are resistant to the synthetic antibiotic G-4 18. Typically, the targeting vector will contain a fragment of the gene to be inactivated, with a functional neo gene inserted into one of its exon elements. An homologous recombination with the target locus will result in the replacement of part of the endogenous gene with vector sequence and the inactivation of the gene. Typically, one in a thousand of the clones surviving G-4 18 selection will be targeted, the remainder arising from non-homologous integrations of the targeting vector. Targeted clones have been identified by screening individual clones, or pools of clones, in one of two ways. In Southern blot hybridization the cell DNA is cut with restriction enzymes, electrophoresed to separate the resulting restriction fragments by size, and the fragments are then bound to a membrane, where a radioactive hybridization probe is used to detect the specific alteration to the structure of the target

BIOCHEMICAL SOCIETY TRANSACTIONS locus 11 I]. Alternatively, the polymerase chain reaction technique may be used. Here cell DNA is used as the tcmplatc in repeated cycles of DNA synthesis. The sequence of one of the primers for the synthesis is derived from the rreo gene and the sequence o f the second primer is from a region o f cellular DNA flanking the target locus. A specific DNA fragment of defined length is only generated in the synthesis reaction if the cell DNA used as template contains a targeted gene [ 121. Although the development o f the polymerase chain reaction method has facilitated scrcening. it still remains a laborious process. The cell population may be further enriched for targctcd cells in one of two ways t o simplify their isolation. In the first method. the expression of the dominantly selectable marker is made dependent on its correct integration into the target locus. by requiring that it use transcriptional or translational control signals from the target locus 11 3, 141. Random integrations of the targeting vector will be unlikely to result in expression o f the marker gene. This method is only appropriate if the target locus is expressed at a sufficiently high level in the target cells t o drive expression o f the sclectable marker. If this is not the casc, a second stratcgy. termed positive-negative selection. is used (see Fig. 4). Here the marker (rieo) gene used t o disrupt the target locus contains all necessary control signals for efficient expression. Random integrations of the targeting vector are sclccted against by the inclusion in the vector o f a second marker gene. In the procedure as originally described by Mansour et 111. I I S ] . the second marker was the thymidine kinase gene from Herpes simplex virus (HSVTK). Thc viral enzyme has ii different substrate specificity to the mammalian enzyme, s o that thymidine analogues, such as gancyclovir, can be used sclectively to kill cells expressing the viral enzyme. In an homologous recombination event the HSVTK gene will be lost in the replacement o f part o f the targct locus with the tieo-gene-containing vector sequence. Thus. targeted cells will be neo+ HSVTK-. Cells with random integrations o f the vector, which usually occur via the ends o f the linear DNA molecule, will be neo+ HSVTK' and can be sclectcd against. These procedures have been used t o target a variety o f genes which are expressed in ES cells, such as proto-oncogenes I 1 1 , 151 and genes with a role in development 11 21. In addition, procedures using a functional dominantly selectable gene have also been used t o target genes which arc not expressed in ES cells [ 16-18]. This is an important finding Gene

HSVTK

Inactivating vector

Inactivated gene (Neo' HSVTK

)

Fig. 4. Illustration of the positive-riegutive selection procedure to inactivate u typical gene (corituiriitig five exoris) by gene targeting usirig a replacement vector 1990

'

27th COLWORTH MEDAL LECTURE because it shows that at least some non-expressed genes are accessible t o the homologous recombination mechanism. Just as with targeted alterations to the HPRT gene 19, 191. some of these other targeted genes have now been transmitted through the mouse germ line [ 1 I . 181. and animals with interesting phenotypes are beginning t o be described [ 20. 2 1 I.

I030 I am grateful t o the members o f my research group in t’dinhurgh for their invaluable contributions over the past six years: Angela Pow. Carolanne McEwan. Simon Thompson. Nik Somia. Jim Selfridge. Patrick Costello. Angus King. Arthur McKie. Marion Milne and Chao-Liang Wu. I am indehted to all the mcmhers o f the Department o f Molecular Biology. particulerly to Kichard Ambler, Ken Murray and David Finncgan. for their help ;ind support since I moved t o Edinburgh. Work in my laboratory is supported hy the A.F.R.C., the C.R.C. and the M.K.C.

I’ropcrties of’ihc tcirgetitig p r o c w s I . Jaenisch. R. ( I9XX) .Scic,tice 240. I 4hX- I474 It is important to realize that the targeting procedures 2. Friedmann, T. ( 1989) Scic~ticz244. 1275- I 2X I currently in use are inefficient, typically resulting in less than 3. Bradley. A,. Evans. M.. Kaufman. M. H. & Kohertson. E. one targeting event per million ES cells electroporated. This ( 1984) iV(i11ir~(Li>tidi>ti)309. 255-256 is in part due to the reduced ability of ES cells, compared 4. Smithies, 0..Gregg, R. G.. Boggh. S. W., Kor;tlewski. M. A. & with many other cell types, t o take up and integrate exoKucherlapati. R. S. ( 1985) ~Vuriirc,/ / 2 0 t i d o t i ) 317. 230-234 genous DNA. One solution has been t o introduce targeting 5. Hooper, M. L., Hardy. K.. Handyside. A,. Hunter. S.!L Monk. vector DNA by direct microinjection. resulting, in one M. ( 1087) Ncrrirre (Lotidoti) 326. 2 0 2 - 2 0 5 instance, in a reported targeting frequency of almost 1% [ 2 2 ] . 6 . Kuehn, M. R.. Bradley. A,. Kohertson. E. J. &i Evans. M. J . ( 10x7) Nuriirc, ( L o t i d o t i ) 326. 205-2OX The two basic types of targeting vector described (insertion, 7. Jinnah, H. A,, Gage. F. H. & Friedmann. I.. ( IOVO) J. C’?II see Fig. 3 and replacement, see Fig. 4)appear t o work with Hiochem. (Suppl. 14A j. 369 similar efficiency 123I. although replacement vectors do, in X. Doetschman. T., Gregg, K. G.. Maeda. N.. Hooper. M. I-., general. produce less disturbance t o the structure o f the Melton. D. W.. Thompson. S. & Smithies. 0. 119x7) iViilitrc target locus. Reported targeting frequencies vary widely (l.i>tidi>ti)330. 576-57X between genes. but there is general agreement that the fre9. Thompson. S.. Clarke, A. K.. Pow. A . M.. Hooper. M. I-. & quency increases with the extent ot homology between Melton. D. W. ( 19x9) C d / ((irmhridge. .MN.Y.s.) 56. 3 13-32 I vector and target locus. Although it is likely that the frc- 10. Melton. 0. W. ( 19x7) in 0,rfi)rd . S i r r ~ r yo t i fiikcrryiric. ( k t i c s (Maclean, N., ed.), vol. 4, pp. 33-70. Oxford Univcr\ity Press. quency of homologous recombination will vary widely in Oxford different regions of the genome. much of the variation in I , Schwartzberg, P. L., Goff. S. 1’. &i Kohertson. E. J . ( I O X V ) targeting frequency reported t o date is equally likely to result Science 246. 790-XO3 from differences in vector design and experimental tech2. Joyner, A. L.. Skarnes. W. C , “4 Ko\sant. J. ( I V X V i Ntirrirc nique. Through use o f the indirect selection procedures (Lorir/oti) 338. 153- I 5 6 dcscribcd it has been possible t o enrich the cell population 3. Dorin. J. R., Inglis, J. I>. & Portcous. i I>. J. [ I 9 X O ] .S(~ic,tic.o 243. for targeted clones t o a frequency o f greatcr than 10% in 1357- I360 some cases. 4. Jasin. M. & Berg, P. [ I 9XX) C k t i c s /)(,I.. 2. 1353- I 3 0 3

I’rospc~c~tsji)r Retie tcirpJtitig

Although gene targeting has been used to alter the transcriptional control o f the HPRT gene, most of the changes made t o other loci have involved insertional gene inactivation. Some o f the common human genetic diseases. such as cystic fibrosis, result from a single amino acid change t o the gene product. Targeting procedures for making subtler alterations are currently being developed. Their use is likely t o lead to more useful animal models o f human disease. In addition, selection procedures are heing modified to utilize marker genes which are expressed on the cell surface. This permits screening o f populations f o r targeted cells using automatic cell sorters 1241. This approach could prove particularly important for targeting cells of the blood and immune system. It is hoped that these developments will facilitate the study and treatment o f many human diseases. Currently, we arc trying to produce mice with DNA repair gene deficiencies, which we hope might prove useful models for human diseases such as xeroderma pigmentosum, where there is a reduced ability t o repair DNA damage and a high incidence o f tumours.

Vol. I x

5. Mansour. S. L.. Thomas. K. I. H. & Jaenisch. R. ( 1990) Nktirrc, ( L o t i d o t i ) 344. 742-746 Z I . DeChiara. 1’. M., Efstratiadis. A. & Kohertson. E. J. (1090J N(irirre (Lotidoti) 345. 7X-XO 22. Zimmer. A. & Gruss. P. ( 1989) ,V(iriirc, (l.otidoti) 338. 150- I53 23. Thomas. K. R. & Capecchi. M. K. (IVX7J C’c// (C’ordwidgc, ‘ M N S S . ) 51. 503-512 21. Jasin, M.. Davis. K. W. & I3erg. P. ( 1090) ( ; c t i c \ I h . . 4. 157-166

Received 10 July I990