For several years now it has been recognised that the expression of some genes in the mammalian genome is dependent upon their parental origin. Some genes are only expressed following paternal transmission, while some others are only expressed following maternal transmission. Gcncs are differentially repressed by an epigenetic marking process called genomic imprinting that acts during gameto- genesis when the two parental alleles are segregated in the male and female germlines. Although thc developmental consequenccs of genomic imprinting have been studied extensively by analysing the phenotypes of diploid partheno- genetic and androgenetic embryos, and of chimeras inade with thosc embryos, only three of the imprinted genes con- tributing to these phenotypes have so far bccn identified For reviews see r e f s 1 and 2). T~~ genes, H I 9 (located on chromosome 7) and IRf2r (chromosome 17), are paternally repressed and one genc, I d 2 (chromosome 7) is maternally repressed“, 4. s).

Considerable attention to these genes is now being paid by several labs 10 try and find the molecular basis of the imprints; for example. extcnsive searches for allelic differ- ences in DNA methylation and DNase I hypersensitivity are underway(6). Other questions that can now be addressed con- cern the mechanisms by which imprints are first established in the germline and are subsequently maintained in the soma. The imprinting process musl involve the action of ‘imprint- ing genes’. One way in which these genes can be studied is to search for genetic variation in imprinting and to identify, by genetic means, the genes responsible for any variation.

This strategy has recently been used by Forejt and Gre- gorova to study imprinting of the mouse mutation T-assori- afed maternal effect (Tme)(7). The Tme locus resides within the T-complex of mouse chromosornc 17 and it is delcted in two t haplotypes, pfI) and PL7,uh2(8). In both cases, a dominant lethal effect is observed in late gestation embryos when the mutant chromosomes are inherited from their mothers but not from their fathers. This suggests that the Tme gene is imprinted and expressed from the maternal but repressed in the paternal chromosome 17. An extensive molecular analy- sis by Barlow and colleagues had previously physically mapped Tine within a 8OOkb stretch of DNA with the genes for plasminogen (PZg) and superoxide dismutase 2 (Sod-2) at its flanks and containing the genes for the. insulin-like growth factor type-2 receptor (Igf2r) and T-complex peptide I (Tcp-

Each of these four genes was assayed for expression in embryos following either maternal or paternal transmission of the deletion chromosome and it was found that Z@r was only expressed when it was inherited from the mother. The conclusion was therefore drawn that Igf2r is imprinted and is closely linked or identical to Tmd3) (Fig. 1). Thus, together with Igf2 and H l Y , I@r was one of the first imprinted genes to be discovered(4, s).

Forejt and Grcgorova analysed the imprinting of Tme using mice from a number of different genetic backgrounds drawn from Mus m. musculus and Miis m. domesticus sub- species. The authors crossed Thr)l+ fcmales (Mus m. domesti- ous background) with males from a number of different strains of inbred and wild mice from Mus m. mitsculus and Mus In. domesticus sub-species and looked for rescue of the Tme cmbryonic lethal phenotype by scoring the viable prog- eny for the dominant short-tail phenotype (the dominant short-tail phenotype is associated with the deletion of Bra- ~ l i i ~ ~ y or T that maps within P I P ) . Intriguingly, they found that fetuses with a maternally inheritcd dcletion of Tme (Thp) could be rescued when the father came from the M M S m. mus- culus sub-species, but there was no rescue with all strains of Mus m. domesticus mice tested (Table I). The short tail phenotype was observed at a lower rrequency than expected for complete rescue (35% instead of 50%); however, whether this was because of incomplete rescue or because of incom- plete penetrance of the marker (short tail) is not entirely clear.

To analyse the genetic control of Tnie imprinting further, a cross was madc between ?“P/+ females and rescuing (PWD (musculris)) x non-rescuing (C3H (domesticus)) F1 males. This produced 14% short-tailed progeny, which is close to half the frequency seen in the Thpl+ x PWD cross (36.8%) (Table 1) and could suggest the action in the offspring of a single modifying allele segregating from the F1 father. Importantly in this experiment, it was found that both the PWD and C3H Tme alleles were present in thc offspring, suggesting that rescue is not effected by allelic differences at the Tme locus itsclf but rather by an unlinked genotype- specific modifier.

Next, the mode of action of the modifier was addressed in more detail. When the Tme gene and the musculus modifier allele were both inherited from the male (and the maternal Tme was deleted), rescue of the Tme phenotype was observed as expected (P (pp /+ x C57BIl10) x 6 PWD), however when the inusculus modifier was inherited from the mother and the Tme gene was inherited from the father no rescue of the Tme phenotype was seen (O(Tkf[ll+ x PWD) x 8 C57BIl10). The authors interpret these findings as revealing the existence of an imprinting gene (designated Imprintor- I, or I r n p l ) , whose domesticus allele (Imp-I(]) acts to repress Tme on the paternal chromosome, but whose rn.u.scuZus allele (fmp-ln2) allows Tme to escape from impiinting (which leads to rescue of fetuses with the maternally inherited deletion). Furthermore, the observation that the rescue only works if finp-Im is present in the father could argue for a gametic action of the modifier on Tme rather than acting post-fertili- sation. 1rr1p-l could therefore be a gene that acts in the male germline to imprint Tnze. However, an alternative explana-

Imprinting of Tme /Igf2r

Chr 17

? $ no Igf2r/Tine Tgf2r/Tme expression expression

Chr 17

? ( f Thl' Dclction

Fetuses die Ol'fspring survive

Fig. 1. Fetuses inheriting thc r"p deletion of niouse chromosome 17 from their mother die because imprinting represses the paternal alleles of IKfz and Trne that arc present within the deletion (I'@ and Tn7r may be identical). Thus no product of the imprinted genes is available lo con- tribute to normal development. Paternal inheritance of the ?zp deletion does not compromise development since the maternal alleles of I&? and Trnu are not imprinted.

tion for these results would be that Imp-I is itself imprinted and only expressed when paternally inherited. Indeed, there is a precedent for this in the modification of a transgene locus TKZ75 I , in which a genotype-specific modifier present in BALB/c mice represscs the transgene locus post-zygotically, but only after maternal transmission of the modifier(9)). Rcs- cue by an imprinted gene that acts after fertilisation in the offspring would also explain the reduced rescue frcquency that is observed when F1 males (Inip-InYimp-Id heterozy- gotes) are used instead of inurculi4s males ( Imp- 1"' homozy- gotes).

The greatest surprise came when the expression of lg/2r was analysed in rescued embryos and newborn mice and it was found still to be imprinted. This observation has dra- matic conscquences for the interpretation of the genetic data. Firstly, the continued paternal repression of /@r could mean that l@r and Ttire are in fact two separate but tightly linked genes in the PLub2 deletion, and furthermore, imprinting in this 8001ch region would be regulatcd on a gene by gene basis rather than the two genes being co-regulated within a com- mon domain. The imprinting of Tme but not Zgj2r would be dependent upon the prezygotic action of a genotype-specific modifier or imprinting gene.

Although the involvement of species-specific imprinting genes acting in the germline is very attractive, this explana-

tion of the results does not hold if Tine and /$?r are indeed the same gene. In thc interspecific hybrids, thc lack of I@r gene product could be somchow compensated for by the functions of thc miisculus allele of another gene. Igf2r encodes a protein with at least two distinct functions. Its best defined role is as the cation-independcnt mannose-6-phos- phate receptor, in which it is involved in trafficking lysoso- ma1 enzymes into the lysosomes by binding their mannose-6- phosphate residues. The second role of this molecule is to bind IGF-I1 through a distinct binding Currcnl evi- dence suggests that this function acts to reduce the available IGF-I1 ligand for mitogenic activation which occurs through binding the IGF-I receptor. The sophisticated regulation o f fetal growth involving the reciprocal imprinting of the IGF-I1 ligand and the ICF-I1 type 2 receptor has been elegantly dis- cussed elsewhere(' 12).

To address the possibility of a compensating gene function in M u s in. musculus mice, Forejt and Gregorova identified the cation-dependent mannose-&phosphate receptor (which maps to Chromosome 6) as their best candidate for such a gene, however, in a test cross it was found not to segregate with ZinpfJ1. While this addresses the mannose-6-phosphate binding function of the receptor, the IGF-11 binding function was not addressed. In the absence of a functional receptor, there will be significantly more IGF-I1 ligand available.

Table 1. Summary ofcrosses shobving rescue Of'Thp by Mus. m. mu.vcu1u.s Crosses Observations

Paternal Tnzc deleted, maternal Tmr prescnt: Viable short-tailed young present Maternal ' / m e deleted, Paternal Trne present: Short tailed fetuses die late in gestation 35% short-tailed offspring 14% short-tailcd offspring, random segregation of T ~ P alleles and of cation-dependent mannose-6-phosphate reccptor (Chr 6) No rescue of short-tailed offspring

P +'l'i+ x d M . rn. tni6.r.

P +/+ x 6 FI(M. m. nius. x M . in. ciom.)

? F1(9'p/+ x M . m. mus.) x 6 M. m. dnni.

7'11J/+ mice are M. tn. dom. background

Given the possible critical relationship between the levels of the ICF-I1 ligand and the IGF-I1 type 2 receptor that have been proposed by Haig and Graham(”), the increased amount of IGF-I1 ligand could have detrimental conse- quences. Is it then possible that the musculus species used in this study have a lowered level of IGF-I1 expression and that this could explain the rescue? If so, one would expect to see the musculus Zgf2 allele in the rescued fetuses that result froin the cross with F1 males. Since the lethal effect of Thl’ is seen late in gestation, and Igf2 expression is turned off in the immediate postnatal period, it may be that only a small reduction in the expression of I g t 2 is required to effect a res- cue. Since imprinting can be seen as an elaborate mechanism for regulating gene dosage any interference in the status quo could be detrimental. The altered equilibrium between IGF- 11 and its type-2 receptor is evident in the experiments of Forejt and Gregorova since the rescued Tme mice are about 16% larger at birth. A further increase in fetal growth due to higher available IGF-I1 levels may well be detrimental, to the fetus itself and indirectly through exhausting available maternal resources. Mice that lack IGF-I1 entirely are severely growth retarded(s) and it is noteworthy that various attempts to generate transgenic mice carrying extra copies of the Igf2 gene have failed (Sasaki, pers comm). A role for IGF-II in the rescue of Tme would also be supported by the genetic data obtained by Forejt and Gregorova if the interpre- tation is taken that Imp-1 is a maternally imprinted gene act- ing postzygotically rather than a non-imprinted modifier that acts at the gametic stage.

Despite some of the unanswered questions raised by Forejt and Gregorova’s paper, it is important because it focuses attention on the issue of how imprinting may be influenced by changes in genetic background and whether some or the genes involved in the imprinting process can be identified by genetic means. Although good precedent for such an approach towards understanding imprinting comes from the analysis of some transgene loci, surprisingly few background effects have been associated with the imprinting of endoge- nous genes(”). Thus, the question of whethcr polymorphic

alleles of genes that control imprinting exist in interbreeding populations, must for the present remain open. Rather than talking about ‘Zmprintors ’, the experiments of Forejt and Gregorova may tell us more about the intricate struggle that goes on between imprinted genes.

References 1 Monk M. and Surani A., eds. (1990). Gmomir. hiprinting. L)eve/opp,nenf 1990 Supplement. The Coinpany or Biulogisl\ Ltd, Cambridge, LJK 2 Surani A. and Reik W., eds. (1992). Genomic imprinting in mouse and man. Srm. DPY. Bioi. 3, 73-160 3 Barlow D.P., Stoger R., Hermann B.G., Saito K. and Schweifer N. (199 I ). The mouse insulin-like growth factor type-2 rcccptor is imprinted and clo\ely linked t o the Trne locus. Nature 349,8447. 4 Bartolomei M.S., Zemel S., and Tilghman S.M. (199 I ) . Parental imprinting of the tnouseH19gene. Nufure351, 153-155. 5 De Chiara T M, Robertson E J, Efstratiadis A (199 I ) . Parental imprinting of the nioubr insulin-like growth factor I1 gene. Celi 64.849-860 6 Sasaki H., Jones P.A., Chaillet J.R., Ferguson-Smith A.C., Rartun S.C., Reik W. and Surani M.A. (1992). Parental tinprinting. potentially active chromatin of the teprcued iniateriial allelc of Ihe muuse insulin-like growth factor I1 (Ic@) genc. Genes Dei,. 6, 1843.1 R56. 7 Porejt J. and Gregorova S. (1992). Genetic analysis of genomic imprinting: An imprintor-l gene controls inactivation of the pitternid copy of the mouse Tnze locus. Cell70,443-450. 8 Winking H. and Silver L.M. (1984). A rcconihina~it mouse f haplotype that expressed a dominant lethal maternal effcet. Ganeli(.\ 108, 101 3- 1020. 9 Allen N.D. and Mooslehner K.A. ( IYY2) . Imprinting. transgene tncthylation and, genotypespecific modilication. Sern Dw. Biol. 3,87-98 10 Roth R.A., Knizacina K.S., Stcele-Perkins G. and Purchio A.F. (1990). lnsulin- like growth factor-Ill mannose-6-phosphatc receptor: structure and [unction. In Growth Factors: From G e m to Clinical Application, (ed. V. R. Sara). (New York: Raven Press) pp. 73-83. 11 Haig D. and Graham C. (1991). Cienomic imprinting and the strange case of the insulin-like growth faclor I1 receptor. Cell64. 1045-1046. 12 Moore T. and Haig D. (1991). Genoinic imprinting in inanimaltan development: a parental tug-of-war. Trerids Genet. 7,45-49.

Nick Allen and Wolf Reik are in the Department of Molecular Embryology, AFRC Institute of animal physiology and genetics research. Bahrahatn, Cambridge CB2 4AT, England.