UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Imprinting and gene silencing are in the Air

Sleutels, F.J.G.T.

Link to publication

Citation for published version (APA):Sleutels, F. J. G. T. (2001). Imprinting and gene silencing are in the Air.

General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.

Download date: 24 Apr 2020

Chapterr 1

Thee origins of genomic imprintin g in mammals

F.. Sleutels and D.P. Barlow

Advancess in Genetics, Volume 46, Homology Effects, Editors:: C. -ting Wu, Jay C. Dunlap, Academic Press, USA,

Inn Press.

Thee origins of genomic imprintin g in mammals Frankk Sleutels* and Denise P. Barlow

ÖAWÖAW Institute of Molecular Biology, Billrothstrasse 11, A502ÖA502Ö Salzburg, Austria *The*The Netherlands Cancer Institute, PlesmanlaanPlesmanlaan 121, 1066CX, Amsterdam, The Netherlands

Mammalss are diploid organisms that inherit a complete chromosome set from each par-ent.. The vast majorit y of genes are equally expressed from both parental chromosomes. However,, in a small subset of genes, a process known as genomic imprintin g results in parental-specificc gene expression. The imprinted gene is expressed on one parental chro-mosome,, but silent on the other. Parental specific gene expression does not result from geneticc changes, but instead results from modifications to DNA or chromatin that are describedd as "epigenetic". Genomic imprintin g is thus an exceptionally good tool to study epigeneticc gene regulation because both the active and silent allele are retained in the samee nucleus. Historically , imprintin g has only been viewed as a gene-silencing mecha-nism.. However, the actual data obtained from studies of imprinte d genes, are not com-patiblee with this view and, instead, indicates that imprint s were in many cases, acquired ass a gene-activating mechanism. To accommodate these findings, a model is proposed wherebyy imprint s can activate, or de-repress, genes previously silenced by an epimuta-tion. .

11 1

Contentss Chapter 1

1.. Abstract

2.. Definition of genomic imprinting

3.. Evolution and function of imprinting

i.. Sexual reproduction ii .. Placental development iii .. Intra-uterine embryonic growth iv.. Avoidance of fetal immune rejection v.. Post-natal nurturing

4.. Origins of genomic imprinting i.. Model A ii .. Model B

5.. Predictions for an imprinting mechanism

6.. Organization and epigenetic modification of imprinted genes i.. Clusters ii .. Alleli c methylation iii .. CpG islands iv.. Direct repeats v.. Asynchrony vi.. Non-coding RNAs vii .. Homology

7.. DNA methylation and genomic imprinting

8.. The function of methylation imprints: Model C

9.. Innocent bystanders, future challenges

10.. Concluding remarks

11.. Acknowledgements

12.. References

13.. Appendix 1: Abbreviations

12 2

Chapterr 1

2.. Definition of genomic imprinting

Mammalss are diploid organisms arising from the fusion of two parental gametes that each donatee one set of autosomal chromosomes (22 autosomes in humans, 19 in mice) plus one sex chromosomee (X or Y) to an offspring. Diploid cells thus contain two parental copies (or parentall alleles) of each autosomal gene that are predicted to show the same transcription state sincee loss of one copy can be easily tolerated for the majority of genes. Genomic imprinting iss a rare and unusual process that results in parental specific gene expression of autosomal loci (inn males and females) and of loci on the diploid X chromosome (in females). (See also reviewss [1-5]). The transcription of imprinted genes is determined by their parental origin and thee two parental alleles can be either maternally expressed and paternally silent or vice versa (maternallyy silent and paternally expressed). Thus, imprinted genes show haploid expression despitee the cell being diploid for the locus. About 50 mammalian imprinted genes have now beenn identified in humans and mice with approximately equal numbers of paternal and mater-nal-expressedd examples (see [6, 7], for a full listing and chromosome maps of all known imprintedd mouse and human genes and other imprinted phenotypes). The total number of imprintedd genes is likely to be relatively small compared to the total number of genes in the genome.. Estimates based on genetic experiments in mice and on all the known genetic muta-tionss in mice and humans that, in contrast to imprinted loci, obey Mendel's First law and are functionallyy equivalent on maternal and paternal inheritance, provide an estimate of 100-200 [6].. However, this number may now be reduced by 60% following the reappraisal of the total numberr of genes in the mammalian genome [8].

3.. Evolution and function of imprinting

Genomicc imprinting appears to be conserved in all placental mammals (with the caveat that onlyy a limited number of species have yet been tested). It is also present in marsupial mam-malss where two autosomal genes plus X chromosome inactivation are known to be imprinted [9-11].. Monotreme or oviparous mammals lack imprinting of two tested autosomal genes, and althoughh they show a partial and tissue-specific X inactivation, it is not yet known if this is randomm or imprinted [11]. Genomic imprinting is thought to be absent from non-mammalian vertebratess but, except for one gene shown not to be imprinted in the chicken [9], this hypoth-esiss has not been generally tested amongst all vertebrates. The absence of genomic imprinting inn non-mammalian vertebrates is, instead, based on the argument that genomic imprinting imposess a requirement for sexual reproduction in mammals (see below point 3i), and asexual parthenogeneticc reproduction has been observed to occur in all non-mammalian vertebrates [12-15].. Genomic imprinting thus likely arose in the vertebrate lineage, after the formation of mammalss but prior to the divergence of the marsupial and placental subclasses. However, the presencee of parental specific gene expression in the endosperm of angiosperm plants (e.g., Arabodopsis)Arabodopsis) and parental specific chromosome inactivation linked to sex determination in scalee insects (Coccidea) indicates not only that genomic imprinting evolved several times, but thatt it serves more than one biological function [16,17]. (See also [18,19] and Grossniklaus, thiss volume).

13 3

Genomicc imprinting in mammals today is already known to serve the function of chromosomee dosage compensation by inactivating one X chromosome in female mammals, thuss equalizing the dosage of X-linked genes between males (XY) and females (XX). X chro-mosomee inactivation is imprinted so that only the paternal X is inactivated in all tissues of marsupials.. In placental mammals, however, imprinted X inactivation is restricted to the pla-centaa and extraembryonic tissues, whilst somatic tissues show random X inactivation (reviewedd in [11, 20], see also Lee, this volume). Genomic imprinting of autosomal genes in mammalss is unlikely to serve the same function of dosage compensation, since males and femaless have the same number of autosomes. An indication of the other possible function(s) off imprinting in mammals comes from the finding that imprinted genes are present in mam-malss but absent from all other vertebrate classes. Mammals are very similar to other verte-bratess in terms of organism physiology and genetic developmental programs [21]. However, inn contrast to oviparous vertebrates, mammals depend solely on sexual reproduction and fur-thermore,, in contrast to oviparous animals that deposit a fixed amount of nutrient per embryo inn an egg, mammals do not strictly limit the extent of nutrient transfer between the maternal parentt and embryo. All three classes of mammals (monotreme, marsupial and placental), trans-ferr nutrients and other resources from the maternal parent to the embryo, but only in placen-tall mammals does an open circulatory connection exist between the embryo and the maternal parentt in utero. Al l three subclasses of mammals, in addition, have a postnatal growth period wheree offspring are fully dependent on maternal milk production. In both pre- and postnatal growth,, the mammalian embryo or offspring can reprogram the maternal environment to influ-encee the extent of the nutrient transfer. These and other unique features of mammalian repro-ductionn are listed below with indications of possible role(s) for imprinted genes.

3i.3i. Sexual Reproduction

Mammals,, alone of all the vertebrates, do not reproduce asexually. The development of imprintedd genes that are essential for embryonic survival, but show parental-specific silenc-ing,, wil l make sexual reproduction essential. Many of the imprinted genes studied by targeted inactivationn show an essential developmental function (Table 1).

3ii.3ii. Placental Development

Thee placenta is an embryonic tissue necessary for embryonic growth and development. It has beenn argued that spontaneous activation of eggs leading to parthenogenesis could be danger-ouss for the development of intra-uterine embryonic growth strategies. The use of imprinting too silence maternal genes that regulate placental development and invasion of maternal tissue, wouldd prevent the implantation of spontaneously activated non-fertilized eggs. Active placen-taltal development and invasion genes, introduced at fertilization by the paternal gamete would thenn control pregnancy (Trophoblast defense hypothesis, [22, 23]. At this time, none of the paternallyy expressed imprinted genes studied by targeted inactivation show an essential implantationn function, although some such as Igf2 and Peg3 play a modest role in placental growthh (Table 1). N.B. abbreviations used in the text, and full names of genes are listed in Appendixx 1.

3Hi.3Hi. Intra-Uterine Embryonic Growth

Thee placental mammalian embryo grows inside the uterus attached via the placenta to the

14 4

Chapte rr 1

maternall blood supply where it is nourished and protected. Although the final size of any organismm is genetically determined [24], there is a wide variation in birthweight that is influ-encedd both by the uterine environment (large uteri will produce large offspring, [25]), and the embryoo itself (through production of embryonic hormones that manipulate the maternal envi-ronmentt and the extent of maternal nutrient transfer). Birthweight is a crucial parameter for postnatall survival both for polytocous and monotocous (animals with multiple or single off-springg in one litter) species. Both underweight and overweight offspring have increased mor-talityy rates [26]; however, overweight offspring may present an increased threat to the evolu-tionn of intra-uterine growth because they endanger the mother and thus also older siblings. Becausee of the necessity to achieve optimum birthweight, strict dosage control of growth reg-ulatoryy genes that may be achieved by silencing one parental allele may have been a necessary requirementt for mammalian reproduction [27]. In support of this function, many of the imprintedd genes studied by targeted inactivation, as well as many of the as yet uncloned but mappedd imprinted loci, show growth regulatory functions (Table 1 and [6]). There is little evi-dence,, however, to indicate that silencing one locus is necessary for tight regulation of gene expression. .

Thee ability of the embryo to influence its own growth at maternal expense is the basiss of one of the best known ideas to explain the evolution of imprinting, known as the Parentall Conflict Hypothesis [28]. This hypothesis proposes that mammalian breeding strate-giess in which females breed with multiple males will produce a conflict of interest between the twoo parental genomes. The interests of the maternal genome are best served by equal distri-butionn of her resources amongst all her offspring; however, the interests of the paternal genomee are best served by ensuring that offspring with his genome receive the largest share of maternall nutrients at the expense of offspring with different paternal genomes. This theory pre-dictss that growth-promoting genes will be paternally expressed while growth-suppressing geness will be maternally expressed. As can be seen from Table 1, there is very strong evidence fromfrom the function of imprinted genes to support this hypothesis as at least one of the biologi-call functions of imprinting. However, one prediction of the theory, that the coding sequence off imprinted genes will evolve rapidly (to allow competition between different paternal alle-les),, is not supported, since the imprinted genes so far examined show a low level of genetic variationn [18].

3iv.3iv. Avoidance Of Fetal Immune Rejection Thee direct connection between the maternal and fetal circulation exposes the maternal immune systemm to fetal antigens. Paternal fetal antigens will be recognized as foreign and stimulate a maternall immune response. Silencing fetal expression of paternal antigens may have been necessaryy to subvert the maternal immune response during the evolution of intra-uterine embryonicc growth. However, there is as yet no evidence for this view and none of the imprint-edd genes studied by targeted inactivation show an immune function (Table 1). Moreover, it has recentlyy been shown that the fetus actively defends itself against immune attack using non-imprintedd genes to suppress maternal immune responses e.g. by complement inhibition [29] andd by inhibition of T cell activation by tryptophan catabolism [30].

3v.3v. Postnatal Nurturing Alll newborn mammals are fully dependent on maternal milk for nourishment and maternal

15 5

behaviorr for heat and protection in a post-natal period that is of variable length. The neonate iss able to stimulate milk production by suckling and other behavior patterns, as well as stimu-latingg appropriate maternal responses by many means, including vocalization [31]. Maternal ovulation,, and thus the production of new embryos, is also suppressed by suckling in some species.. Using the same arguments as in point (iii ) above, it could be claimed that strict con-troll of genes regulating these behaviors is necessary to limit maternal contribution to the neonate.. This would maintain the mother's fitness to care for the present offspring that con-tinuess well beyond the weaning period, allow her to care for previous offspring, as well as ensuree her ability to produce new offspring in the future. Genes affecting nurturing or suck-lingg behavior in the postnatal period would also be selected under the Parental Conflict theo-ry.. Two of the imprinted genes studied by targeted inactivation show a maternal behavioral functionn (Table 1).

Thee identification of the in vivo function of 17 of the 50 known imprinted genes in mice (Table 1)) shows that a large majority have growth regulatory functions in late embryonic and post-natall stages. As shown in Table 1, growth promoting effects are only shown by paternally expressedd genes (6 cases indicated by +). No paternal-expressed imprinted gene has been shownn to have a growth suppressor effect. In contrast, growth suppressing effects have only

Imprintedd geae LossOfFanctloaraeaotrpc References ad Note* Inn Mice

1 1 2 2

3 3 4 4

5 5

6 6 7 7 8 8

9 9 10 0

11 1 12 2 13 3

14 4 15 5

16 6

IgC C Ins2 2

Snrpn/Snurf f Smpn+IC C 3'Snrpn-Ube3a a Grfl/Ras-Grfl l

Mest/Pegl l Peg3 3 Neee din

Masl l GnasXl l

Gaas s M6P-lrf2 r r «B «B

Masa2 2 Oftfl l l

Ubc3a a

Growthh reduction Viablee - no growth effect

Viablee - no growth affect Growthh reduction - lethal Growthh reduction - lethal Longg term memory Postnatall growth Growthh reduction, nurturing Growthh reduction, nurturing Lethall - no growth effect

Behaviorr phenotype Growthh reduction

Growt hh increase Growt hh increase - lethal N oo phototype

Placenta!! defect- lethal LetlaLL proliferatio n defect-worgmnt mm growth defect Motorr dysfunction and teaming deficit t

+ +

--

--+ +

+ + + + + +

--

--+ +

+ + + +

--+ + + +

--

T371 1

Micee lacking Ins2 and Insl have non-imprinted growthh retardation,Insl is not imprinted [165] T166.. 1671 Thesee deletions may affect the same gene(s) [166,167]. . [168] ] T1691 1

rnoi i [171] ]

Nulll mice are viable in some genetic backgroundss [172]

Noww known not to be imprinted [112,173] Bothh genes are transcribed in the same directionn from different promoters and splice too exon 2, but from different parental chromosomess f 174,175]

wa a ExnlHyomcjgrowthisttKTeasedonlyifirace e s hee over-express Ig£2 [32,33] flTS l l

P?7,, 178!

um um 177 Gtl2 Growth decrease by transgene + Transgene insertion may affect the upstream

insertionn upstream to GÜ2 paternally expressed DBcl gene [159,180]

Tablee 1. The function of imprinted genes as determined by loss of function alleles. Paternallyy expressed imprinted genes are in white boxes at the top of the Table, maternally expressed imprintedd genes are in the shaded boxes at the bottom of the Table. The column indicates if the pheno-typee could have relevance for an imprinting function in any aspect of mammalian development as dis-cussedd in the text. See also a similar Table in Hurst and McVean [18].

16 6

Chapterr 1

beenn shown by maternally expressed genes (2 cases, H19 is excluded since absence of the HI 9 RNAA alone does not generate a growth phenotype [32]). Because of the similarity between the pairedd Gtl2 and Dlkl loci and the paired Igf2 and HI 9 loci, it is likely, but not yet proven, that thee growth decrease associated with a transgene insertion upstream to Gtl2 reflects a change inn Dlkl expression (Table 1 and [159], discussed in more detail in the section below on mater-nall and paternal imprints). With this caveat in mind, it can also be stated that no maternal-expressedd imprinted gene has been shown to have a growth promoting effect. This finding pro-videss very strong support for the Parental Conflict theory. However, 8 of the studied genes do nott affect growth, which may indicate the Parental Conflict theory is false. Conversely, it may alsoo indicate that imprinting serves several different biological function in mammals today, as arguedd by Hurst and McVean [18], or even a function not specific to mammals, as argued by Pardo-Manuell de VHlena et al. [19]. Any one of the above five listed features of mammalian reproductionn may have reinforced selection of genomic imprinting in mammals.

Thee gene inactivation experiments have provided much information. However, studyingg a loss of function allele is actually the wrong experiment, since there is a difference betweenn the function of a gene and the significance of imprinting it. A moree informative way too examine the biological function of imprinting in mammals would be to change mono-allel-icc imprinted expression to biallelic expression. If no phenotypic consequences result from increasingg the gene dosage two fold, then imprinting this gene has no significance for the bio-logicall function of imprinting. This is independent of the fact that a loss of function allele showss that the gene has an essential function. Such a gene must be regarded as an "innocent bystander"" of the imprinting mechanism [22]. "Loss of imprinting" alleles have only been gen-eratedd for two genes (Igf2 and M6P-Igf2r) by targeted deletion of an Imprint Control Element (ICE).. These experiments generated a growth phenotype with a reciprocal effect to that shown byy a loss of function mutant [33-35]. However, deletion of an ICE can change expression of moree than one gene, and although the phenotype is likely to be due to the two primary imprint-edd genes regulated by the ICE (respectively, Ig/2 and M6P-Igf2r), the effects of the other deregulatedd genes may influence the phenotype.

4.. Origins of genomic imprinting

Thee biological function of genomic imprinting is, however, only one half of the puzzle. The otherr half is an explanation of the molecular mechanism underlying parental-specific gene silencing.. Before we examine the current features of imprinted genes, it may be informative to considerr how imprinted genes first arose in die ancestral mammalian population. In consider-ingg this, the assumption is made that an ancestral mammalian population lacked the imprint-ingg mechanism but did not lack the genes that are imprinted in the population today. This is likelyy since genes imprinted in mammals today have homologues that are not imprinted in non-mammaliann vertebrates. Thus, imprinted gene expression is likely to have evolved in a populationn that initially showed biallelic expression of these genes. In addition, since the majorityy of genes in the modern day mammalian genome are non-imprinted and show biallel-icc expression, it is logical to initially consider the imprinting mechanism as one that acts to silencee one of the two parental alleles. Models A and B outline below two possible scenarios

17 7

forr acquisition of imprinting by parental-specific silencing of genes that show biallelic expres-sionn in an ancestral population.

4141 MODEL A AA genetic 'Gain of Imprinting' mutation arose in a single animal that introduced a parental-specific,, epigenetic gene-silencing mechanism. An example could be the gain of expression in thee gametes of one parent only of a gene silencing protein such as a tfe novo methyltransferase. However,, to allow for the spread of the Gain of Imprinting (GOI) mutation in the ancestral population,, there must have existed one susceptible gene that could be "imprinted" i.e., a gene whosee methylation by the de novo methyltransferase reduced expression from one parental allelee and so conferred an immediate advantage to the organism. It is important to appreciate thatt the new GOI allele would not increase in frequency without phenotypic change and hem-izygouss RNA expression by itself is unlikely to have been advantageous for the organism (but seee Pardo-Manuel de Villena et al. [19], for a different point of view). An example of an appro-priatee susceptible gene could be a dosage-sensitive growth suppressor or growth promoter gene.. Animals that were effectively hemiyzgous for a growth suppressor gene because of the GOIGOI mutation would be larger than wildtype littermates and, as a consequence, may be more effectivee competitors for food and mates, thus allowing the spread of the GOI allele through-outt the ancestral population. Animals that were effectively hemiyzgous for a growth promot-err gene would be smaller than littermates, and although it is less easy to suggest an immedi-atee advantage for size reduction, an explanation is needed since many of the imprinted genes listedd in Table 1 have a growth promoting effect. There are a few possibilities: first, produc-tionn of smaller-sized offspring may have been related to the development of a polytocous (multiplee offspring in one litter) breeding strategy in mammals. A "small but many" breeding strategyy may have had advantages over a "large but few" strategy. Second, small sized off-springg may have provided immediate benefit to the mother by conserving her resources, this mayy have allowed increased nurturing behavior mat would benefit the survival of present and pastt offspring. A third possibility is that imprinted expression of growth suppressor genes occurredd early in evolution, and imprinted expression of growth promoting genes occurred in responsee to reduced levels of suppressor genes. This latter suggestion would require that growthh promoter genes were involved in the same genetic pathways as growth suppressor genes,, as is shown by some of the genes listed in Table 1 (e.g., M6P-Igf2r, Igf2, HI9 and pos-siblyy GrblO, see Table 1 for references). A final suggestion is that the imprint may have altered thee expression pattern or level of expression on the active allele of a growth promoting gene, therebyy allowing increased growth despite the complete silence of one allele.

Thee initial GOI mutation and the nature of the susceptibility of the growth suppres-sorr gene most likely lacked the sophistication of the present day imprinting mechanism. However,, removal of the imprint from the germline of the next generation and re-acquisition accordingg to the sex of the offspring, would be an essential requirement to allow parental-specificity.. In this model, a different GOI mutation, that would not imprint the same set of genes,, would have been acquired later by the reciprocal parental gamete. The acquisition of a GOIGOI mutation suggests the possibility that a large number of existing susceptible genes were simultaneouslyy imprinted and showed hemizygous expression in diploid cells of the ancestral mammal.. Although such radical change seems improbable, selection for the GOI mutation by imprintingg one key gene that benefited organism fitness, would also allow selection for

18 8

Chapterr 1

imprintedd expression of other genes that have no obvious benefit to the organism. This may explainn why some genes imprinted in mammals today appear to have no essential function (Tablee 1, see also discussion below on innocent bystander genes).

4ii.4ii. MODEL B AA parental-specific, epigenetic gene-silencing mechanism already existed in ancestral mam-mals,, but served a different purpose. An example could be gamete-specific silencing by de novonovo methyltransferase of infectious exogenous retroviruses and endogenous transposable ele-ments,, which would serve as a genome defense function. The maternal gamete would be able too methylate and silence some forms of foreign invasive DNA, whilst the paternal gamete wouldd be able to methylate and silence complementing forms [36]. With this gamete-specific mechanismm in place in the ancestral population, a genetic mutation could arise in a single ani-mal,, in a gene or eft-linked regulatory element, that rendered a gene susceptible to this pre-existingg silencing mechanism. A possible mutation strategy that would serve this purpose wouldd be the integration of a virus or transposon into a dose-sensitive growth suppressor or growthh promoter gene. The result would be the same as outlined in model A: the mutated allele iss silenced by transmission through one gamete only and heterozygous advantage from hap-loidd expression of the mutated gene in the organism allows spread of the mutated "imprinted allele"" throughout the population. An increase in the number of genes subject to imprinting withinn the framework of model B is likely to have involved acquisition of the susceptible trait byy individual genes that were individually tested for function by selective forces.

Inn both model A and model B, the increased population frequency of the mutated allelee resulting from heterozygous advantage cannot lead to homozygous silencing of the gene, sincee the silencing mechanism is parental-specific. With both models, it is also clear that spreadd of the first imprinted allele in the ancestral population must have conferred an advan-tagee on an individual organism. However, it is not clear that imprinted genes in modern day mammalss are maintained by the same selective force that drove their initial establishment in ancestrall populations.

5.. Predictions for an imprinting mechanism

Imprintedd genes, in many cases, show parental allele-specific gene silencing that is stable in alll cells of the organism. This shows that the imprinting mechanism must be maintained on the samee parental allele during cell division. Both the active and silent parental alleles are retained inn the nucleus and potentially exposed to the same trans-acting transcription regulators but onlyy one parental allele is affected, thus the imprinting mechanism must act in cis to the chro-mosomee that inherits it. Imprinted genes show the same parental allele-specific expression in succeedingg generations (e.g., males containing an imprinted gene that has an active paternal allelee and a silent maternal allele, will transmit both parental alleles in an active form to their ownn offspring); this shows that imprints are erased in the germline and then reformed accord-ingg to the sex of each new offspring. Imprinted genes can be found in animals with identical maternall and paternal genomes (e.g., inbred mice) although this is not a requirement since out-bredd populations such as humans, also have imprinted genes. This finding shows two further

19 9

importantt features of the imprinting mechanism. First: since imprinting can operate with geneticallyy identical parental chromosomes, the mechanism is not genetic and must be epige-neticc (i.e., a modification of DNA or chromatin able to regulate gene expression). Second: sincee the imprinting mechanism can distinguish between genetically identical parental chro-mosomes,, imprints cannot be acquired after the cell becomes diploid and must therefore be acquiredd when the parental chromosomes are in separate compartments such as during game-togenesiss or just after fertilization prior to pronuclei fusion. Thus, imprints are predicted to be epigeneticc modifications that are acquired during gametogenesis by one gamete only, herita-blee by the same parental allele during mitosis, associated with a ci's-acting mechanism able to silencee expression, and lastly, reversible in the germline of the next generation.

Thee degree of silencing of the repressed allele of an imprinted gene can show con-siderablee variation. Some imprinted genes show tissue-specific loss of silencing, e.g., the paternall allele of the mouse Igf2 gene is not silenced in the choroid plexus, the human IGF2 genee shows biallelic expression in juvenile liver [37, 38] and the paternal allele of the mouse M6P-Igf2rM6P-Igf2r gene is partially activated in adult brain [39, 40]. Imprinted genes can also show developmentall control of imprinted expression. The mouse M6P-Igf2r gene shows low-level biallelicc expression in preimplantation embryos, while imprinted expression is only seen after embryonicc implantation, in parallel with X chromosome inactivation [20, 41,42], The Mash2 genee initially also shows biallelic expression before the onset of imprinted expression [34]. In contrast,, imprinted expression of Kcnql is restricted to the early embryo and placental tissues andd the gene is expressed from both alleles in later tissues [34]. Several imprinted genes e.g. mousee M6P-Igf2r and human CDKN1, also show a low level of expression of the silent allele inn all tissues when samples are analyzed by sensitive means [43, 44]. Finally, although imprintedd expression is generally highly conserved between mice and humans, several imprintedd genes show species variation in imprinted expression. The most dramatic example iss M6P-Igf2x, which shows full imprinted expression in the mouse embryo and adult, while imprintedd expression in humans is limited to post-implantation embryos (20-24 weeks) and, furthermore,, is limited to a small proportion of individuals. All tissues at birth and all adult tis-suess in humans show full biallelic expression of M6P-Igf2r [45-47]. Polymorphic imprinted expressionn has also been observed for the human Serotonin-2A receptor [48] human Wilm's tumortumor gene [49] and human IGF2 [50]. This variation in the degree of silencing of the repressedd allele indicates that the imprinting is not an "all or nothing" mechanism and that the repressedd allele may be reactivated under some conditions.

6.. Organization and epigenetic modification of imprinted genes s

Accordingg to the models A and B, outlined above, imprinting is considered to arise as a par-ent-specificc gene silencing mechanism. However, the first results describing the modifications off imprinted genes in the mouse yielded surprising results. The paternally expressed Igf2 pro-moterr was shown to have potentially active chromatin as well as a somatic methylation imprint upstreamm of the active paternal allele [51], while the maternally expressed M6P-Igf2r gene was shownn to carry a gametic methylation imprint only on the expressed allele [52]. If the imprint-

20 0

Chapterr 1

ingg mechanism acted by parental allele-specific silencing, the prediction would have been that thee silent maternal Igf2 allele and the silent paternal M6P-Igf2r allele should have been mod-ified.. The next gene to be studied, the maternally-expressed H19 gene did, however, support aa parental allele-specific silencing mechanism as the basis of imprinting, since it showed spe-cificc methylation of the silent paternal promoter [53]. As further imprinted genes were isolat-edd and their epigenetic modifications studied, several features common to imprinted genes becamee apparent:

6161 CLUSTERS Imprintedd genes are clustered but do not occupy a special chromosome position. Imprinting of thee mouse H19 gene was initially tested because it mapped close to Igf2 on chromosome 7, the positivee result forming the basis of the hypothesis that imprinted genes are clustered and reg-ulatedd in concert by long range control elements [54]. This hypothesis has been spectacularly successfull and has allowed the identification of more than half of the known imprinted genes. Thee extent of clustering in the mouse can be seen in the imprinting maps maintained by Cattanachh and Beechey [6] and also from sequence analysis of large regions containing homol-ogouss imprinted clusters in human and mice. Mouse chromosome 7 is an extreme example and containss three clusters of imprinted genes mapping to the proximal, central and distal regions whosee homologous regions are separated in the human respectively, on to chromosome 19,15 andd 11. An approximately one megabasepair region from the distal mouse chromosome 7 clus-terr and the same from the homologous human chromosome 1 lpl5 region, were sequenced and shownn to contain 17 genes, of which 8 showed imprinted expression [55, 56]. However, it is noww known that these 8 imprinted genes are regulated by at least two, and maybe more, dis-tinctt imprint control elements [57-59]. Transgenic experiments indicate that clustering may havee more relevance for some genes than others. For example, many genes do not maintain theirr imprinted status as transgenes in ectopic sites (Igf2, [60], Cdknh [61]), while others requiree large tracts of flanking DNA or need to be present in multiple copies (M6P-Igfir: [62], HI9:HI9: [63], Snrpn: [64, 65], Xist: [20]).

6ii.6ii. ALLELIC METHYLATION DNAA methylation fits all the predicted requirements for an imprinting mechanism as outlined above,, and the majority of imprinted genes do have parental allele-specific DNA methylation. DNAA methylation in mammals modifies cytosine to 5-methyl cytosine in any CpG dinu-cleotidee pair [66]. Other cytosine dinucleotide pairs can be methylated but this only occurs transientlyy in very early embryos [67]. Methylation patterns are dynamic during mouse devel-opmentt with two cycles of methylation loss and gain [68]. First, methylation is lost during germm cell formation and gained as the gametes mature. Second, after fertilization, methylation iss lost by the preimplantation embryo then regained after embryonic implantation, coincident withh gastrulation in the mouse. The allele-specific methylation of imprinted genes has two forms.. It can be acquired in the gamete and thus be a candidate for the "imprint" or, it can be acquiredd in the soma and thus be a candidate for maintenance of imprinted expression. For methylationn to be a candidate for the "imprint" it must be acquired by one gamete only, sur-vivee thee preimplantation demethylation wave and be maintained on the same parental allele in thee diploid cell after DNA replication. Gametic methylation imprints with this behavior have beenn identified on several genes; M6P-Igf2r [52], H19 [69], RasGrfl [70], U2afl-rsl [71],

21 1

SnrpnSnrpn [72]. Somatic methylation imprints that are acquired after the cell is diploid and thus cannott be the mark that distinguishes between the parental alleles, have also been shown to occurr on many imprinted genes and can also be found on several genes that also carry gamet-icc methylation imprints, e.g., M6P-Igf2r [52]. An analysis of mice deficient for the mainte-nancee methyltransferase Dnmtl gene, and thus deficient for global genomic methylation, has confirmedd that DNA methylation plays a major role regulating imprinted expression [73] (see discussionn below).

6iii.6iii. CpG ISLANDS AA comparative analysis of a one megabasepair imprinted region from mouse chromosome 7 andd human chromosome 11, prompted the observation that imprinted genes tend to be associ-atedd with one or even two CpG islands [55, 74], although a recent review observed that only 88%% of imprinted genes have islands, mouse Ins2 being one exception [2]. This is in contrast too the rest of the genes in the mammalian genome of which approximately 50% have CpG islandd promoters. CpG islands are defined as stretches of DNA 500-1500bp long with a CG:GCC ratio of more than 0.6 and they are normally found at promoters and contain the 5'end off the transcript (reviewed in [75]). Transcription of CpG promoters appears to be regulated byy the same panoply of transcription factors as the set of CpG poor promoters in mammals, butt there is little information about their direct regulation. However, it is known that CpG islandd promoters normally lack TATA boxes, often have multiple transcription starts, tend to bee associated with genes with "housekeeping" functions, and are normally unmethylated irre-spectivee of expression status. There are also indications that CpG islands act as replication foci inn the mammalian genome (reviewed in [76]). The absence of DNA methylation on CpG din-ucleotidess located inside islands is in contrast to the majority of CpG dinucleotides that lie out-sidee islands and are invariably methylated. Since methyl-CpG can undergo a spontaneous tran-sitionn to TpG that is not efficiently repaired, this leaves the mammalian genome with a deficit off CpG dinucleotides compared with GpC dinucleotides, outside islands [77]. If methylation iss used as a gene regulatory signal, as is the case for imprinted genes (described in point ii) it makess evolutionary sense to place the important CpGs within islands. For a large number of imprintedd genes (but not all) the CpG island promoters are methylated either in the gamete or inn the soma (see references in point ii). Methylated CpG islands pose two problems. First, it is nott clear how CpG dinucleotides are retained in methylated islands, since methyl-CpG/TpG transitionn will reduce the CpG content. The explanation may lie in the fact that imprinted CpG islandss are only methylated for a relatively short time in the germ line [78], thus they are not exposedd for a long time to the hazards of deamination. It is also possible that deamination is reducedd following methylation of CpG dinucleotides in CpG islands, or that repair is more efficient.. The second problem, which poses one of the major challenges in the imprinting field, iss to understand why imprinted CpG islands are methylated when the majority of CpG islands thatt are linked to non-imprinted genes, appear to be actively protected against methylation [79, 80]] (see comments below).

6iv.6iv. DIRECT REPEATS Severall but not all imprinted genes contain or are closely linked to a region rich in direct repeats,, that shows no homology to the high copy interspersed repeat families present in mam-maliann DNA. The direct repeats are generally found in or close to CpG islands and their pres-

22 2

Chapterr 1

encee but not their sequence, can be conserved between mice and humans [81]. Imprinted genes containingg direct repeats include: Xist [82], M6P-Igf2r [52, 83], PW71 [84], Cdknl [85], H19 [81,, 86], Igf2 [87, 88], U2afl-rsl [71], IPW[%% Snrpn [90], RasGrfl [70, 91], Magel2 [92], lambdaPenlllambdaPenll [93] and Peg3 [94] The presence of direct repeats has been suggested to attract gameticc methylation in a sequence-independent manner by inducing DNA to form unusual secondaryy structure during replication [81]. The large number of imprinted genes containing repeatss supports this proposal. New results from a comparative sequence analysis of the humann and mouse Impact genes also support a role for repeats in attracting methylation. The mousee Impact gene contains direct repeats in its CpG island promoter, and is maternally methylatedd and silenced. In contrast, the human IMPACT homologous gene, lacks the direct repeatss and lacks imprinted methylation and expression [95]. But, there are also indications thatt repeats play no role in attracting the methylation imprint since it has also been shown that aa G-rich repeat element close to the H19 imprint control element, and a central core from the U2afl-rslU2afl-rsl CpG island repeat, are not relevant for imprinting [96,97]. An alternative viewpoint iss that differential methylation is achieved by specific sequences that attract or repel methyla-tionn and in support of this, a sequence-specific 113 bp element from the M6P-Igf2r intron 2 methylatedd island has been demonstrated to attract methylation in a gamete specific manner [98]. .

6v.6v. ASYNCHRONY Fluorescentt in situ hybridization used to label chromosomes during the cell cycle at S phase, hass identified an apparent asynchrony of replication between the parental alleles of an imprint-edd gene [99]. The region showing asynchrony can extend beyond the cluster of imprinted geness to include non-imprinted genes. The asynchrony can go in both directions (maternal earlyy or late) but in contrast to the behavior of non-imprinted genes, there is no correlation betweenn late replication and silencing [100]. In contrast to the fluorescent in situ hybridization technique,, bromodeoxyuridine incorporation only detects replication asynchrony in highly expressingg cells, and shows that the extent of asynchrony between imprinted parental alleles cann be less than between loci along a single chromosome. It has been suggested that the asyn-chronyy observed with both these techniques may indicate allelic differences in chromatin structuree rather then DNA replication [101,102]. In support of this, increased chromatin pack-agingg on non-transcribed alleles for Snrpn and M6P~Igf2r has been described [103]. In addi-tion,, several imprinted genes have now been shown to have differential chromatin organiza-tionn over the maternal and paternal promoter of an imprinted gene [104, 105] (reviewed in [106]).. However, it is not yet known if chromatin plays a part in acquisition of the imprint, or inn the maintenance of imprinted expression, or perhaps has no significance for imprinting.

6vi.6vi. NON-CODING RNAs Thee identification of the human XIST gene, which is essential for imprinted and random X chromosomee inactivation, as a non-coding RNA that lacks a structural function was the first indicationn that RNAs may possess cw-regulatory properties in mammals [82, 107] (see also Lee,, this volume). The first identified imprinted non-coding RNA gene, named HI9, was noted too have organizational features in common v/i\hXist, such as large exons and small introns and associationn with direct repeats [54,108]. Of the 50 known imprinted genes so far identified in mammals,, a surprising 27% are non-coding RNAs (see Table 2 for references). These 14

23 3

imprintedd non-coding RNAs can be roughly divided into three categories: (i) "antisense recip-rocall allele"; these are mostly paternally expressed non-coding RNAs that overlap the silenced allelee of a maternally expressed imprinted gene. Examples are Air/M6P-Ig/2r, UBE3Aas/UBE3A,UBE3Aas/UBE3A, Kcnqlotl/Kcnql, Nespas/Nesp, Cop2as/Cop2. The TsiX/Xist is an exam-plee of a maternally-expressed antisense non-coding RNA overlapping the silenced maternal Xistt allele (that is itself a non-coding RNA), but this parental-specific expression pattern is onlyy found in placental tissues, (ii) "antisense same allele" a non-coding RNA overlaps a pro-tein-codingg gene expressed from the same parental allele. The two known examples are also paternallyy expressed; Igf2as/Igf2t ZFP127as/ZFP127. (iii ) "no overlap detected"; these includee non-coding RNAs that lie close to imprinted protein-coding gene(s) but do not over-lap.. The examples include two maternally expressed non-coding RNAs linked to the silent

1 1

2 2

3 3

4 4

5 5

6 6

7 7

8 8

9 9

M. M.

i i i

':li. . .

13 3

14 4

Imprinted d non-codmg g

RNA A Nespas s (2D) )

€opg2as/Mitl l RNA A «SP) )

Pwcrl l (TO O Zfpl27as s (7C) ) Ipw w (7C) ) PARI,, PARS 05q) )

UBE3A-AS S (15q> >

HI* *

Igjla»» '::

(7D) )

Gtl2 2 (12D) )

: t f&: . . . . Xistt 'placenta (X) )

TsiX X (X) )

Expressed d Allele e

PP M ovetiap P P

P P

P P

P P

P P

P P

P P

P P

P P

F F

P P

P P

;¥ ¥

M M

M M

Antisense e reciprocal reciprocal allele e Antisense e reciprocal reciprocal allele e

Noo overlap detected d Antisense e samee allele Noo overlap detected d Noo overlap detected d

Antisense e reciprocal l allele e Noo overlap

reciprocalreciprocal :< Antisense e samee allele Antisense e reciprocal l

Ütófe e Noo overlap detected d reciprocal l Aartsffts* * reciprocal l allele e sequence e specific c function n Antisense e reciprocal l allele e

Name,, Linked Coding Gene, and Genetically Determined Function (if tested)

Nespp (Neuroendocrine Secretory Protein ) antisense RNA [181]

Copg22 (COatomerComPjes subunit Gamma 2J antisense RNA overlap* the3' endd of the upstream Mest I gene. MitI{Mest-1inked Jnprinted Irwtscrïpt 1 in intronn 20of Copg2). Copg2as and Mil l may belong to die same transcript. It is alsoo not excluded ém they could derive from Mest i read through transcripts 11821. . Prader-Willii Chromosome Region 1/Probable sequence specific function as a SnoRNAfllOl l Zfpp 12 7 (Zinc-Finger Protein 127) antisense RNA [ 183]

Imprintedd in Prader-Willi syndrome [89]

Fivee PAR (Prader-Willi Angelmans Region) RNAs have been isolated only from humans.. PARI, 5 are paternally expressed. However, it is unclear if the PAR RNAss are one or many transcripts 1184] UBE3AA (Ubiquitin E6-AP protein ligase 3A) antisense RNAs have only been foundd in humans in the PWS-AS region [185]

HI99 fetal BvermRNA(cioM!y linked to Igf2) [54]

lgi2(insulir^likee grow* fiictor 2) antisense RNA [S8]

KCï^ ll (FfeStiiö̂ k^-hTtea4iar«ly,tne«BBerl) «wMÉJpn^ftiwenptt t [Ï36J

Genee Trap Locus 2 (closely linked to Dlk) [ 159,186]

M6P-Igf2rr (msulm-like growth factor 2 receptor) antisense RNA known as Air (Antisensee Igf2rRNA> [62,! 12]

XX chromosome Inactive Specific Transcript, 'imprinted only in placenta and extra embryonicc tissues. N.B. marsupial Xist is not yet identified. Reviewed in [ 11,20] andd Lee, this volume Overlapss and regulates Xist expression (and wins the best name contest) [113. 187] ]

Tablee 2. 27% of known human and mouse imprinted 'genes' are non-coding RNAs. Thee 14 known imprinted non-coding RNAs are listed and grouped by linkage groups indicated by the dark andd light shading. Chromosome numbers and regions (P, proximal, C, central, D, distal) are listed in col-umnn 2. The lane headed 'overlap' indicates three groups, i) antisense reciprocal allele: non-coding RNAs thatt are expressed from one parental allele and overlap a silenced protein coding gene expressed on the otherr parental allele, ii ) antisense same allele: a non-coding RNA that overlaps an active protein coding genee expressed on the same parental allele, iii ) No overlap detected: non-coding RNAs that lie close to an imprintedimprinted coding gene but do not overlap, and are expressed from the opposite parental allele in some cases.. See text for further details. Antisense transcripts are indicated in some cases by 'as' and in other casess by an independent name as indicated in the last column.

24 4

Chapterr 1

allelee of a paternally expressed protein-coding gene; H19/Igf2, Gti2/Dlkl. In addition, sever-all paternal-expressed non-coding RNAs have been isolated from the human Prader-Willi // Angelmans syndromic regions: PWCR1, IPW, and two incompletely characterized tran-scripts,, PARI and PARS. Of die 14 non-coding RNAs listed in Table 2, only two may have a sequencee dependant function. The Xist transcript has been shown to be necessary and suffi-cientt for X chromosome and also autosome inactivation [109]. The PWCR1 RNA is homolo-gouss to SNO-guide RNAs that act to modify other RNAs, however, a function has not been geneticallyy demonstrated [110].

Thesee non-coding RNAs, particularly the overlapping antisense transcripts, are unusuall in the mammalian genome and most share common features. For example, non-cod-ingg RNAs mostly show strong expression, many are known to be very long transcripts, sever-all contain interspersed repeats, they also show large exons, and contain few or no introns. The lackk of introns in imprinted genes has been noted previously [111]. An extreme example is the AirAir non-coding RNA that overlaps the M6P-Igf2r gene; it is 108 kb long, has no identified intronss and appears to be co-linear with the genomic sequence [112]. Thus, in comparison to mammaliann coding genes, non-coding RNAs appear to show reduced ability to splice and reducedd ability to terminate. The presence of families of interspersed repeats in non-coding RNAss argues against a sequence specific or structural function. In several cases, a deletion of thee promoter of the non-coding RNA has been shown to cause re-expression of the overlapped silentt gene. This has been shown for the Air promoter [62], TsiX [113] and KCNQ10T1 [58]. However,, precise deletion of the HI 9 non-coding gene and promoter has no effect on imprint-edd expression of the linked protein-coding Igf2 gene in endoderm tissues, although de-repres-sionn was seen in mesoderm tissue [114]. Instead, imprinted expression of both HI 9 and Igf2 iss regulated by an element lying upstream to HI9 that appears to function as an insulator that bindss die CTCF insulator protein in a methylation dependant manner [115], (reviewed in [116]).. This finding raises the possibility that non-coding RNAs are not the cause of allele spe-cificc silencing of the linked imprinted gene, but are a consequence of this silencing and may reflectt a local alteration in chromatin structure that is permissive for spurious transcription. Thuss it is by no means certain that imprinted non-coding RNAs participate directly in allele-silencing.. The experiments cited above that demonstrate that deletion of a non-coding RNA promoterr causes de-repression of a linked protein-coding imprinted gene, do not distinguish betweenn a role for the RNA itself and a role for the fragment that was deleted.

6vii.6vii. HOMOLOGY Thee discovery of the widespread existence of RNAi (RNA interference [117], see also Mello, thiss volume) as a means to silence genes with a similar but altered sequence, has raised the questionn of whether RNAi could be involved in gene silencing in the case of imprinted genes withh overlapping non-coding transcripts. RNAi operates through a sense/antisense pairing mechanismm and has been shown to exist in a large range of invertebrates and also in two ver-tebrates,, the pregastrulation Xenopus embryo [118] and the preimplantation mammalian embryoo [119]. There are several good arguments against RNAi action in imprinted allele-spe-cificc silencing: the imprinting mechanism is cw-acting, restricted to the nucleus and can be ini-tiatedd if a single locus is present in the nucleus. In contrast, RNAi is /ra/is-acting and cyto-plasmic.. However, it hass recently been suggested that RNA silencing can act in cis, to repress transgeness in plants [120] (see also Matzke, this volume). Thus, in view of the large number

25 5

off non-coding RNAs linked to imprinted loci, further experiments will be needed to fully test iff RNAi or any form of RNA silencing can be excluded.

Thee imprinting mechanism does not require the presence of both parental chromo-somess since it occurs in parthenogenetic and androgenetic embryos (i.e., diploid embryos con-tainingg only maternal or paternal germline derived chromosomes, [121,122] and is maintained inn uniparental disomic conditions (both copies or part of, one chromosome are derived from onee parent). Imprinting also remains parental specific when the copy number of an imprinted locuss is altered to one copy, e.g., in deletion syndromes in mice and humans [6] or increased too multiple copies, e.g., in imprinted transgenes [63]. This indicates that counting or homolo-gyy interactions between the two parental alleles are unlikely to play a role in imprinting. A chromosomall analysis of the imprinted human Prader-Willi Syndrome region and a genetic analysiss of the imprinted mouse Insulin (Ins2) gene did, however, indicate that homologous interactionss occurred at these loci [123, 124]. However, it has also been shown that imprint-ingg in the Prader-Willi region and also the M6P-Igf2r locus occurs in the absence of obvious homologouss interactions [125, 126]. Thus, in view of the known features of the imprinting mechanism,, especially its m-acting mechanism, it is more likely that while homologous inter-actionss are more common than expected in mammalian genes, they are not relevant for imprinting. .

Nott all of the epigenetic and organizational features associated with imprinted genes are unusual.. Many features, such as clustering and CpG island promoters, are found in non-imprintedd genes. DNA methylation and replication delay are also typical of non-imprinted geness that show tissue-specific silencing. Non-coding RNAs are more unusual, but are increasinglyy being found in association with biallelically expressed genes (e.g., Beta-globin: [127],, Hox 11: [128], RPS14: [129], bFGF: [130], Hsp70: [131], Nmyc: [132]). Direct repeats havee not yet been linked to biallelically expressed genes but it will not be clear if this repre-sentss a unique feature of imprinted genes until more of the completed human sequence is avail-ablee for analysis.

7.. DNA methylation and genomic imprints

Thee study of allele specific methylation of imprinted genes has in many ways, laid the foun-dationn of our current understanding of the imprinting mechanism. In addition to its use to iso-latee novel imprinted genes [133], the identification of regions within imprinted genes that showw allele-specific methylation (now referred to as DMRs, Differentially Methylated Region) hass led directly to the isolation of critical imprint control elements and of elements that mod-ifyy imprinted expression. Examples are the M6P-Igf2r intron2 DMR [62], HI9 upstream DMR [57],, Snrpn promoter DMR [134], Kcnqlotl intron 10 DMR [58] and Igf2 DMR [135]. Furthermore,, DMRs have led to the identification of unsuspected imprinted non-coding RNAs transcribedd within protein-coding genes (examples are Air [62], Kcnqlotl [136]).

AA genetic test of the involvement of DNA methylation in imprinting was performed inn mice carrying defective alleles of the Dnmtl maintenance methyltransferase gene [73],

26 6

Chapterr 1

Imprinte dd locus Igf2 Igf2

M6P~Igf2r M6P~Igf2r Kcnql Kcnql Dlkl Dlkl Peg3 Peg3 GrblO GrblO Cdknl Cdknl Snrpn Snrpn Mash2 Mash2 GrblO GrblO Rasgrfl Rasgrfl Sgce Sgce ZacI ZacI U2afl U2afl

Pegl/MestI Pegl/MestI H19 H19

Meg3/Gt!2% Meg3/Gt!2% Afcr-imprinted d Xist-iandom Xist-iandom

PC/NC C PC C PC C PC C PC C PC C PC C PC C PC C PC C PC C PC C PC C PC C PC C PC C NC C NC C NC C

M+ + P P M M M M P P P P M M M M P P M M M M P P P P P P P P P P M M M M P P

MAE-R R

M-(methodd 1) Repressed d Repressed d Repressed d Repressed d

nd d nd nd

P+M M Increased* *

M M nd d nd nd nd nd nd d nd d nd nd

P+M M Increased* * P+M/P& &

P+M M

MM - (method 2) Repressed d

nd d nd d nd d

Repressed d Repressed d Increased* *

nd d nd d nd d nd d nd d nd d nd d nd d

Repressed d nd d nd d

Increased d

M -- (method 3) unchangedd in PG cells

nd nd nd nd nd nd

MM in PG cells nd nd

Pinn AG cells unchangedd in AG cells

ne e unchangedd in AG cells unchangedd in AG cells unchangedd in AG cells unchangedd in PG celts unchangedd in AG cells unchangedd in PG cells

PP in AG cells nd d nd d

Ret t a,e,g g

a a b b c c

e,g g e e

b,e,g g d* * b b g g e e R R e e g g E E

a,e,g g c c

a,f,e e

Tablee 3. DNA Methylation And Expression Of Imprinted Genes. Thee effect of DNA methylation has been analyzed using two methods based on removal of the Dnmtl enzyme.. These two methods are generally in agreement with their results regarding the H19 gene. Method 11 uses allele specific RNA analysis of Dnmtl null embryos at E8.5 days post coitum. Method 2 uses GeneCHEPP analysis of E13.S virus transformed or TPS3 null MEFs following a conditional deletion of DnmtlDnmtl in vitro. Method 3 uses demethylating agents azacytidine or deoxyazacytidine applied to E 13.5 androgeneticc (AG, diploid for paternal genome) or parthenogenetic (PG, diploid for maternal genome) MEFs.. PC, protein coding; NC, non-coding; P, paternal expression; M, maternal expression; M+, expressedd allele in the presence of wildtype genomic methylation; M -, expressed allele in the presence of 90%% reduced genomic methylation. MAE-R, random monoallelic expression; nd, not determined; ne, not expressedd in this cell type, a [73], b [34], c [159], d [72], e [137], f [138], g [139]. N.B., TdagSl, described ass imprinted in reference (e), is not an imprinted gene [164]. P&, Xist expression becomes biallelic in extraa embryonic tissue of in E8.5 embryos (e) but, imprinted X inactivation is unchanged (f). % the parentall alleles were not distinguished. # increased expression may result from increased transcription of thee expressed allele, and may not represent activation of the repressed allele.

Thesee mice showed a 70-90% reduction in global genomic DNA methylation and died in the earlyy post implantation stage (between day 8.5 - 9), however the expression of several imprint-edd genes could be analyzed before embryonic death ensued when embryos appeared to display aa normal morphology. These experiments have been recently complemented by a conditional genee targeting strategy that removes Dnmtl in vitro in differentiated mouse fibroblasts (MEFs) preparedd from E13.5 day post-implantation embryos. GeneChip analysis was used to identify aa large number of genes regulated by methylation [137]. In these reduced methylation condi-tionss (methods 1 and 2 in Table 3), three different responses were observed for imprinted pro-tein-codingg genes: i) genes such as Igf2, M6P-Igf2r, Kcnql, Dlkl, Peg3 and GrblO were repressed,, thus both parental alleles are now silent, ii ) two genes, Cdknl and Snrpn, were activee on both parental alleles, and iii ) one gene, Mash2, showed no change. One type of responsee was shown by the three non-coding RNAs so far examined: HI 9, Gtl2 and Xist all showedd increased expression, and although this is assumed to result from activation of the silentt allele, this has only been shown for HI 9 and Xist. Xist is included in this pattern because thee imprinted expression of Xist in placenta switches to biallelic expression. A recent report interestingly,, shows that imprinted X inactivation in placental tissues is maintained in the

27 7

absencee of DNA methylation despite the expression of Xist from both parental alleles [138]. Tablee 3 also shows a conflicting result for the H19 gene between these two assay systems. H19 hass repeatedly been shown to be activated on both alleles in the Dnmtl null embryo in differ-entt crosses [34, 73], but is repressed in the GeneChip analysis of El 3.5 fibroblasts. The rea-sonn is not clear but may relate to the fact that these conditional Dnmtl fibroblasts are trans-formedd by SV40 or have a TP53 null phenotype [137]. A third method (Table 3) using demethylatingg agents has generated results that are broadly similar to those obtained with removall of the Dnmtl enzyme [139]. However this method also identified a larger number of imprintedd genes whose expression was unchanged following treatment with the demethylating agent,, including two (Igf2 and Snrpn) that had showed changes in the absence of Dnmtl, The reasonss for these differences are not yet clear.

Overall,, the results in Table 3 show very clearly that DNA methylation is involved in the imprintingg mechanism at some level. However, the pattern of results is a surprise. Referring backk to the two models, A and B, on the origins of imprinting as an allele-specific gene silenc-ingg mechanism, it would be predicted that removal of methylation imprints (or removal of methylationn modifications mat maintain allele silencing) would lead to expression from both parentall alleles. Five of the genes studied (two coding genes, Cdknl and Snrpn, and three non-codingg RNAs, HI 9, Gtl2, Xist) do show this behavior indicating that methylation suppresses thee silent allele of these imprinted genes. However, the behavior of six protein-coding genes Igf2,Igf2, M6P-Igf2r, Kcnql, Dlkl, Peg3 and GrblO indicates that methylation is needed to main-tainn expression of the active allele of these imprinted genes. This manner of action of DNA methylationn was unexpected, since methylation is normally regarded as a suppressive modifi-cationn for gene expression [140,141].

Ass described above in section 5, it is predicted that genomic imprints are acquired duringg gametogenesis. Experiments have been performed that transfer early oocyte nuclei and malee primordial germ cell nuclei (that are predicted to be "imprint-free") into oocytes, that are thenn activated and will undergo embryonic development to an early post-implantation stage [142,143].. The resultant embryos are predicted to contain maternal or paternal chromosomes lackingg imprints. Analysis of these embryos containing maternal chromosomes (Table 4) indi-catess that some genes are activated by maternal imprints (M6P-Igf2r and Cdknl) while some (Peg3,(Peg3, Pegl/Mest and Snrpn) are silenced by maternal imprints that are acquired in mature full-grownn oocytes. The analysis of paternal chromosomes in androgenetic embryos, supports thee interpretation that (Igf2, M6p-Igf2r and Cdknl) are normally activated by an imprint while otherr genes (Peg3, Pegl/Mest, Nnat and Snrpn) are predicted to be silenced by an imprint.

Inn comparing the models A and B, it is expected that genes shown to be activated by imprint acquisitionn as measured in Table 4 should also be shown to be activated by DNA methylation ass shown in Table 3. Igf2 and M6P-Igf2r behave in both test systems as epigenetically acti-vatedd genes, but Cdknl does not. Similarly, Snrpn behaves as an epigenetically repressed geness in both systems but Peg3 does not. The reason for the differences between these two test systemss is not fully clear at this moment. However, the important message is that both systems havee clearly shown that imprinted genes can be both activated or repressed by epigenetic imprints. .

28 8

Chapterr 1

1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 8 9 9

gtt t t

1x12 1x12 M6P-Isj2r M6P-Isj2r

Pert Pert Cdhtl Cdhtl Snrpn Snrpn Mash2 Mash2 Nnat Nnat

Pegl/Mesl Pegl/Mesl H19 H19

r r NC C

PC C PC C PC C PC C PC C PC C PC C PC C NC C

MEfl P P

P P M M P P M M P P M M P P P P M M

« . l i p r i r t i i h d f r h M M

i i

-fe e nott tested

-ft t nott tested

-fe e nott tested nott tested

-fe e nott tested

«»EtpWiMte4feW 4 4

4vftra44 few # M taffy

aa4«iJ*JUIp*w*(fe ) )

- f e / - ng g ++ fe/-ni! -- f B / + ng ++ fg/ -ng - fg // + n« nott tested nott tested - f e// + nu ++ fg/ + ng

----+ +

--+ +

-+ + + + + +

Tablee 4. Imprinted gene expression from "imprint-free" germ cell genomes. Thiss table lists the expression of imprinted genes in post-implantation embryos manipulated by nuclear transplantt techniques to contain only the maternal genome (parthenogenote embryos) or only the paternal genomee (androgenote embryos). A full grown oocyte is predicted to represent a stage after imprint acqui-sition,, while the non-growing oocyte and male primordial germ cell nuclei are predicted to represent stagess before imprint acquisition. Three paternally-expressed imprinted genes (Pegl/Mestl, Peg3, Snrpn) thatt are expressed in parthenogenote embryos from the imprint-free 'ng' genome are assumed to be repressedd by an imprint on the maternal chromosome. Two maternally-expressed imprinted genes (M6P-Igftr,Igftr, Cdhtl) that are expressed in parthenogenote embryos from the fg, but not the ng, genome are assumedd to be activated by an imprint on the maternal chromosome. The androgenote embryos contains onlyy a paternal genome lacking imprints. Expression of maternally-expressed imprinted genes such as H19,H19, is assumed to indicate a gene normally repressed by an imprint on the paternal chromosome. Expressionn of paternally-expressed imprinted genes (Peg3, Pegl/Mestl, Nnat, Snrpn) is assumed to indi-catee genes normally repressed by an imprint on the maternal chromosome. Absence of expression (Ig/2, M6P-Igflr,M6P-Igflr, Cdknl) is assumed to indicate genes normally activated by an imprint on paternal or maternal chromosome.. PC, protein coding; NC, non-coding RNA; M/P, expressed allele in normal differentiated cells;; +, expression detected; +fg, expression from the full-grown oocyte genome; + ng, expression from thee non-growing oocyte genome; -fg or -ng, expression not detected; * the cell type expressing Mash2 mayy be absent from this embryo type, a [142], b [143]. See text for further explanation of these results.

8.. The function of methylation imprints: model C

Imprintedd genes have drawn attention to an unexpected role for methylation in regulating gene

expression.. Methylation imprints as studied in the Dnmtl -null model are thus shown to be nec-

essaryy for expression of the majority of protein-coding genes studied and necessary for repres-

sionn of all the non-coding RNAs studied so far. In two cases, Igf2 + H19, and DM + Gtl2,

thee genes show a reciprocal response to the absence of methylation: the protein-coding gene

iss silenced while the linked non-coding RNA is activated. These results were initially inter-

pretedd as indicating a form of expression-competition between protein-coding genes and non-

codingg RNAs [144], and while this may still be a possibility for some coding/non-coding pairs,

itt has been excluded for Igf2 and H19 [114]. However, whatever the mechanism that regulates

neighboringg or overlapping pairs of imprinted protein-coding and non-coding RNAs, the abil-

ityy of DNA methylation to maintain gene expression requires an explanation that would inte-

gratee into a model that explains the origin of imprinting.

Modelss A and B, described above, outlined a scenario whereby imprinting arose as

ann allele-specific silencing mechanism, acting on genes that were biallelically expressed in the

29 9

ancestrall population. The demonstration that demethylation silences many imprinted genes, indicatess this cannot be the correct interpretation at least for these genes. Since methylation imprintss in modern mammals are needed for expression of the active parental allele, it is log-icall to assume that this parental allele was silent before acquiring a methylation imprint. Thus imprinting,, at least of these genes, must have been acquired in two steps: first the ancestral biallelicallyy expressed gene acquired a silencing mutation, second the silencing mutation acquiredd a methylation imprint that suppressed its action. Model C (Figure 1) suggests a sce-narioo for these two steps that proposes that imprints can act to de-repress epimutations. In modell C, it is proposed that an insertion mutation arose in a single gene in a member of the ancestrall mammalian population that exerted a cw-acting silencing effect. An example could bee an insertion of a retrovirus or retrotransposon or DNA transposon, into a non-exon gene regulatoryy sequence (step 2, Figure 1). The sequence of the target gene mRNA is thus unchangedd but the gene is silenced as a result of the insertion. As with model A and B describedd above, hemizygous expression of the mutated gene is proposed to confer an imme-diatee advantage to the organism, such as would occur if the mutated gene was a dose-sensitive

1.. Ancestral mammal

Biallelic Biallelic -<£---£--£-

2.2. Epigenetic mutation

Onee allele silenced heterozygouss advantage

3.. Epimutated allele frequency increasess in the population

homozygotee disadvantage

4.. Methylation imprints

Epimutatedd allele de-repressed

fixationn of epimutated allele Heterozygotee advantage regained

S* S*

5.. Loss of methylation

exposess eptmutation bothboth alleles silenced £ £

Figuree 1. Model C: Imprints de-repress epimutations. Thee steps leading from biallelic expression to parental-specific gene activation are displayed. The two parentall chromosomes, depicted as black lines, are not distinguished in steps 1, 2 and 3. The white ellipse indicatess a gene promoter, the black ellipse indicates the insertion. The direction of transcription is indi-catedd by arrows, a black circle indicates repression. The asterisks indicate the imprint modification (for examplee a methylation imprint). In steps 4 and 5, the parental chromosomes are distinguished by the imprintt on one chromosome and are depicted with a black or gray line.

30 0

Chapterr 1

growthh suppressor or growth promoter gene. Animals with hemizygous expression of such a genee could be more competitive in terms of food and mates because they are either larger and moree competitive, or smaller but more numerous, thus allowing the spread of the mutated allelee through the ancestral population. Homozygosity for the mutated silent allele is proposed too be disadvantageous but difficult to avoid as the frequency of the mutated allele increased in thee population (step 3). The acquisition of an imprint, such as DNA methylation, that acts to suppresss the effect of the insertion, would restore expression to the mutated allele. Thus the originall insertion mutation can be referred to as an "epimutation" since its action would have beenn to induce epigenetic gene silencing. For imprinted gene expression to occur it is neces-saryy that the imprint is imposed only on one mutated parental allele. Thus model C requires thee pre-existence of a parental-specific, epigenetic modification mechanism. As described for modelmodel B, this could be gamete-specific silencing of retroviruses or transposable elements, that servess a genome defense function. However, in model C (in contrast to model B), gamete-spe-cificc modification acts to nullify the effects of the insertion (step 4) and so activate the mutat-edd allele in cis. With model C, loss of the imprint can be predicted to result in silencing of both alleless (step 5).

Supportt for model C, that imprints can suppress the effects of epimutations, comes primarily fromm the finding that demethylation silences a large number of imprinted protein-coding genes.. However, there are also other arguments that can be used to support this model.

a)) A prediction of model C is that removal of the insertion mutation that exerts the cis-actingg silencing effect (and is proposed to be marked by the DMR), would activate the silencedd allele, rendering it equivalent to the active methylated allele. This has been most clearlyy demonstrated in vivo for the M6P-Igf2r gene where it has been shown that deletion of thee intron 2 located DMR, restores full expression to the silent paternal allele [35, 62]. Reactivationn has also been shown for the silent paternal KCNQ1 «Uele after deletion of the intronn 10 DMR in a somatic cell hybrid system [58]. Deletion of the H19 upstream DMR, however,, only restores full expression to Ig/2 when H19 is also deleted [33], deletion of the DMRR leaving HI 9 intact, results only in partial Ig/2 activation [57]. Finally, mesoderm-spe-cificc reactivation of Ig/2 has also been shown following deletion of DMR1 [135].

b)) Indirect support for model C, comes from the identification of new insertions that cann induce epimutations in populations today. A good example is that of spontaneous IAP (Intracisternall A-Particle) retroposon insertions into the mouse agouti locus. These insertions havee occurred multiple times in the upstream region and in 5'introns in both sense and anti-sensee orientations, and have induced agouti overexpression by acting as enhancers or promis-cuouss promoters. Interestingly, the extent of epigenetic regulation is affected by parental trans-mission.. Paternal transmission correlates with increased methylation and reduced activity of thee mutated agouti locus [145, 146] A second example of an insertional epimutation is the Movl3Movl3 mutation caused by an intronic retrovirus insertion that interferes with tissue-specific regulationn of the mouse collagen Collal gene, however, in this case parental effects have not beenn noted [147].

c)) In plants, there has long been abundant evidence that transposon insertions can exertt regulatory effects on neighboring genes, as originally described by McClintock [148]. Moree interesting with regards to supporting the proposal outlined in model C, that imprints can actt to de-repress epimutations, is the ability of methylation to mask the effects of transposon

31 1

insertionn in plants ([149] and references therein). Transposons and retrotransposons can con-stitutee from 10% (e.g., Arabodopsis) to 80% (e.g., in maize) of a plant's genome and there are manyy descriptions of transposon insertions into essential genes. Some Robertson's mutator or En/SpmEn/Spm transposon insertions in maize have been described that silence target genes by sup-pressingg transcription or elongation. However, this silencing effect can be suppressed by methylationn of the transposon and the target gene is then expressed normally (reviewed in [149]).. This action of methylation, to restore expression to a gene following transposon inser-tion,, has clear parallels with model C. In mammals, the genome is also replete with trans-posons,, retrotransposons, retroviruses and their degenerate copies (for example, 40% of humann chromosome 21 and 42% of chromosome 22 is composed of interspersed repeat fam-iliess [150, 151]). It has also been shown that absence of methylation leads to a very large increasee in transcription of IAP retroposons in the Dnmtl null mouse [152]. Despite this, sim-ilarr methylation-suppression of transposon mutation effects have not yet been described in mammals. .

d)) As a final argument in support of model C, there is emerging evidence, particu-larlyy with the increasing availability of the genome sequence, that the mammalian genome has evolvedd a symbiotic relationship with the large numbers of transposons, retrotransposons and retrovirusess that form the repeated DNA comprising a large proportion of the genome (reviewedd in [153-155]). There is a new appreciation that host genomes can make use of trans-posablee elements e.g., to generate new patterns of tissue specific gene expression (as described forr the Movl3 mutation above), to insert new introns, for telomere maintenance (mammalian telomerasee is similar to the reverse transcriptase of non-LTR retrotransposons), to mediate V(D)JJ recombination in B and T cell receptors in mammals (the human Recombinase Activatingg Qenes, Rag 1 and 2, are structurally and functionally similar to a DNA transposon), too induce syncitium formation in placenta (an HERV virus envelope gene is specifically expressedd in placenta during syncitium formation) ([156] and see references in [155]). The DMRss identified so far in imprinted genes in mammals do not have any sequence similarity to thee interspersed repeat families now found in mammals. However, in view of the ability of the genomee to co-opt transposable elements for its own benefit, the possibility exists that DMRs representt ancient insertions of mobile elements that have lost their original identity during genomee evolution, but retained their ability to exert a cis-silencing effect. And, it is this unusu-all form of gene regulation that has been co-opted for the benefit of the host genome.

9.. Innocent bystanders, future challenges

InnocentInnocent bystanders Sincee the discovery of the first mammalian imprinted genes in 1991 a considerable amount of informationn has been generated concerning the function of imprinted genes and the mecha-nismm imprinting them. This information has been obtained by treating all imprinted genes as thee same entity (as in Table 1). However, it is possible that while many genes are imprinted, notnot all are equal. In order to collect relevant information about the function of imprinting and aboutt the imprinting mechanism, we need to discriminate between genes whose imprinted functionn was selected during mammalian evolution, and those "innocent bystander" genes

32 2

Chapterr 1

whoo were carried along for the ride. Innocent bystander genes [22] are genes that show imprintedd expression because they lie close to other imprinted genes and are in some manner, susceptible.. The important but unappreciated test, is to examine the effect on the organism of biallelicc expression of an imprinted gene. If there is no phenotypic effect, then such a gene wouldd fit the classification of an innocent bystander. The function of innocent bystander genes, theirr epigenetic modification, structural organization and features, wil l not be informative in generatingg a true picture of the function and mechanism of imprinting. The future challenge heree is to devise a test that wil l distinguish the innocent bystanders from the true participants thatt played a role in the evolution of imprinting in mammals. As mentioned above, this type off test has only been performed for two imprinted genes, Igft [33, 157] and M6P-Igf2r [35] andd both these genes show a phenotypic effect that is reciprocal to the loss of function pheno-type. .

Non-codingNon-coding RNAs Thee results we have so far on the imprinting mechanism, that imprints activate the expressed allelee for many imprinted protein-coding genes studied (Table 3), forces us to pay closer atten-tionn to the silent and non-methylated allele of these imprinted genes. The silent, non-methy-latedd alleles of Ïgf2, M6P-Igf2r, Kcnql, Dlkl, Peg3 and GrblO are silenced by an epigenetic mechanismm that does not use DNA methylation. In four cases, silencing correlates with expres-sionn of a non-coding RNA in cis; for two of these genes, M6P-Igftr and Kcnql, the non-cod-ingg RNA is antisense and arises from an internal promoter, while in the two other genes, Igft andd DM, the non-coding RNA is sense and approximately 100 kb downstream (Figure 2). The futuree challenge here is to test if the non-coding RNA is the cause of silencing, or the result. Thesee experiments are technically difficult because we do not know at this moment, how to silencee a non-coding RNA without changing the sequence of its promoter. And this is the experimentt that we would all like to do. Thus, experiments that have deleted the promoter of aa non-coding RNA and restored expression of the cw-linked protein-coding gene fail to dis-tinguishh between a role for the non-coding RNA and a role for the deleted DNA element (e.g., deletionn of the Air promoter [62] and of the Kcnqortl promoter [58]).

AA large amount of information has been obtained on the function of the HI 9 non-codingg RNA in regulating the imprinted expression of the upstream Ig/2 gene (reviewed in [116,, 158]). One key experiment, that deleted the HI 9 promoter and gene body but left imprintingg of Igf2 intact in endoderm tissue, provides a strong argument against non-coding RNAss having a role in imprinting. However, this experiment did not fully remove transcrip-tionn through the deleted HI9 locus and did cause loss of Igft imprinting in mesoderm [114], Inn view of these results, it was a surprise to find a second imprinted gene pair with very sim-ilarr characteristics. The organization of Dlkl and Gtl2 resembles that of Igft and HI 9. Both protein-codingg genes (Igft and Dlkl) are paternally expressed, while both non-coding RNAs (HI(HI 9 and Gtl2) are maternally expressed (see references in Tables 1 and 2). Igft and Dlkl are alsoo both separated from their respective non-coding RNAs by approximately 100 kb. And, lastly,, in each case the protein-coding and non-coding RNAs are regulated in the same man-nerr by DNA methylation (Table 3). This does not represent a locus duplication since Igft and HIHI 9 have no sequence similarity to Dlkl and Gtl2, and further work wil l be needed to deter-minee if the similarities between these two gene pairs indicates any function for the non-cod-ingg Gtl2 and H19 RNAs.

33 3

Thee large number of imprinted non-coding RNAs (Table 2) have been described as "genes"" and mostly treated as entities distinct from their linked protein-coding genes. At this pointt we do not know if this is a valid distinction, and possibly we should be hesitant to name themm as genes. If these non-coding RNAs are cis regulatory RNAs, there may be some justi-ficationfication in defining them as genes, however, if they are merely the result of silencing a cis-linkedd protein-coding gene and reflect only promiscuous expression from altered chromatin, theyy would not fit a definition of a gene.

MaternalMaternal and Paternal imprints AWAW the models described above (A, B or C) require the parental gametes to respond different-lyy to an imprinted gene. This indicates that the recognition signal for maternally imprinted geness will be different from that of paternally imprinted genes. The recent identification of the Dlkl/Gtl2Dlkl/Gtl2 imprinted pair of protein-coding/non-coding genes and their similarity to the Igf2/H19Igf2/H19 pair contrasts to the 5 known gene pairs that show the arrangement of an imprinted protein-codingg gene overlapped by an antisense non-coding RNA. These two systems may representt paternally and maternally imprinted genes. Figure 2A shows the key features of paternally-imprintedd genes represented by the /g/2///79 gene pair. Figure 2B shows the key

--A.. Paternal imprinted genes

e.g.,, IgG, Dlkl Mat. .

B.. Maternally imprinted genes e.g.,, M6P-Igf2r, Kcnql,

UBE3A,, Nesp, Copg2

<T T

—— f T imprint,, protein coding gene, non-coding RNA, no transript

Figuree 2. A generalized scheme for Maternal and Paternal imprinted genes. (A)) indicates a paternal imprinted gene (the two known examples of a paternally-expressed protein-cod-ingg gene linked to a maternally-expressed non-coding RNA are given). (B) indicates a maternal imprint-edd gene (the five known examples of maternally expressed protein-coding genes overlapped by paternal-ly-expressedd non-coding RNAs are given). The gray ellipse indicates expressed promoters, the black ellipsee indicates repressed promoters. Solid lines indicate transcription of protein-coding genes, dotted liness indicate transcription of non-coding RNAs. The black lollipop indicates repression. Asterisks indi-catee a methylation (or other epigenetic) imprint, Mat, maternal chromosome (gray line), Pat, paternal chro-mosomemosome (black line). N.B., three other protein-coding genes UBE3A, Nesp, Copg2, and their respective antisensee RNAs, that have been not yet been fully characterised may also fit into this category.

34 4

Chapterr 1

featuress of maternally-imprinted genes represented by the M6P~Igf2r/Air transcripts. Igf2 is a paternally-expressedd protein-coding gene that lies approximately 100 kb upstream to the HI 9 locuss (Figure 2A). HÏ9 is a maternally expressed non-coding RNA and has the same tran-scriptionn orientation. A methylation imprint inherited from the paternal gamete is present on a largee CpG rich element 2 kb upstream to the defined HI 9 transcription start and paternal-spe-cificc methylation of this DMR is thought to spread onto the flanking H19 promoter during preimplantationn embryonic development. The methylated paternal HI 9 promoter is silent, and thee unmethylated HI9 promoter on the maternal chromosome is active (reviewed in [116]). Thee Dlkl/Gtl2 gene pair has the same relative organization and Gtl2, similarly, has a paternal-specificc methylation mark but the details have not yet been clarified [159,160]. In contrast to thesee gene pairs, M6P-Igf2r is a maternally-expressed protein-coding gene that contains the promoterr for the Air non-coding RNA in intron 2 (Figure 2.2). The paternally expressed Air RNAA overlaps the upstream silent M6P-Igf2r promoter that is 30 kb distant. A methylation imprintt inherited from the maternal gamete is present on a large CpG island in intron 2 that containss the Air promoter. The methylated Air promoter is silent on the maternal chromosome andd active on the paternal chromosome [112] (reviewed in [161]). The Kcnql/Kcnqotl, UBE3A/AS-UBE3AUBE3A/AS-UBE3A and the Nesp/as-Nesp transcripts have a similar expression pattern, and KcnqlKcnql also has a similar methylation pattern and regulation ([58] and references in Table 2). Thus,, although the data is incomplete, diere is a possibility that Figures 2A and 2B represent twoo distinct parental-specific imprinting systems. Paternal imprints, as depicted in Figure 2A, correlatee with a silent paternal non-coding RNA and an active paternal protein-coding mRNA. Whilee in contrast, maternal imprints, as depicted in Figure 2B, correlate with a silent maternal non-codingg RNA and an active maternal protein-coding mRNA. As yet, despite the availabil-ityy of sequence for the DMRs of most of these genes, no sequence homology can be found, evenn amongst the separate groups of maternally and paternally methylated DMRs. The future challengee here will be to identify the features common to maternally imprinted genes and thosee common to paternally imprinted genes. Another major challenge in this area is to iden-tifyy the molecular pathway leading to de novo methylation in the gametes. The methylation enzymess so far identified, Dnmtl, Dnmt3A, Dnmt3B (reviewed in [66,162]) have not yet been directlyy tested for their involvement in the acquisition of gamete specific methylation imprints, althoughh Dnmtl is known to be involved in the maintenance of the imprint [73].

Non-methylationNon-methylation imprinting Thee action of DNA methylation on imprinted genes shows that it is used to generate the expressionn difference between parental alleles, albeit in an unexpected manner. There are how-ever,, two known cases where DNA methylation may not be responsible for the parental-spe-cificc difference. Imprinted expression of the autosomal Mash2 gene in the embryonic placen-taa is not altered in Dnmtl null mice deficient in global genomic methylation [34, 163]. Similarly,, Sado et al. [138] have shown that imprinted X inactivation does not change in the placentaa of Dnmtl null mice and these authors have suggested that a distinct methylation-inde-pendentt imprinting mechanism may exist in extra-embryonic tissues. It is worthwhile to note thatt imprinted X inactivation, which is the only form of chromosome dosage regulation in marsupials,, also does not appear to depend on DNA methylation (reviewed in [11]). Furthermore,, it has also recently been shown in marsupials, that imprinted expression of the

35 5

M6P-Igf2rM6P-Igf2r gene is not linked to a differentially methylated intronic CpG island. All these resultss may indicate a distinct imprinting mechanism in marsupials that may be conserved in placentall mammals in the placenta for X inactivation and also for some autosomal imprinted genes.. The future challenge here is to isolate more marsupial homologues of imprinted genes andd to validate the existence of non-coding RNAs in marsupials. The latter is of particular interestt since all attempts to isolate a marsupial Xist homologue have not yet been successful (J.. Graves, personal communication).

10.. Concluding remarks

Inn this review, we have drawn attention to the behavior of the large number of imprinted genes inn the absence of DNA methylation, in order to consider how imprinting evolved in mammals. Inn doing this, we have tried to make two clear distinctions. Firstly, we have distinguished betweenn how imprints work to bring about parental-specific expression and why imprinted monoallelicc expression was selected in the ancestral population. Although model C explains howw the imprint works in terms of activating a silenced allele, this should not be confused with thee function of imprinting and we are not arguing that the function of imprinting is to activate silencedd genes. The function of imprinting can only be explained by considering why parental allele-specificc expression would have been advantageous to the ancestral population, whenn the alternativee would have been biallelic expression. The second distinction is in considering exactlyy how imprints work to bring about imprinted expression. In this case we have attempt-edd a distinction between the function of the methylation imprint and what the methylation markk actually does. The function of the imprint, is in many cases (as shown in Table 3) to acti-vatee protein-coding genes, while the methylation mark is proposed to be silencing a linked non-codingg RNA. It should be noted that model C does not exclude epimutations targeting positivee transcription regulatory sequences. Insertions into promoters or enhancers or bound-ariess could attract methylation and lead to gene silencing, but in the absence of methylation no effectt on expression is seen. In this way, methylation acquired in one gamete would lead to imprintedd silencing of a protein-coding gene. It should be noted that for the majority of pro-tein-codingg genes, methylation imprints are activating as defined in the Dnmtl null model (Tablee 3). However, it should also be noted that the assay in Table 4, using "imprint-free' genomes,, identified a larger number of protein-coding genes that are silenced by genomic imprints. .

Att the beginning of this review, models A and B were suggested as ways in which imprintingg could be acquired as a gene silencing mechanism. The actual data obtained from studiess of imprinted genes indicates that, for a large number of genes, imprints were acquired ass a gene activating mechanism (as proposed in Model C). Our attention is now focussed on thee silent but unmethylated parental allele of an imprinted gene. This makes imprinting stud-iess much more exciting because, now, hitherto hidden means of epigenetic gene silencing, that doo not use DNA methylation, can be revealed. A vast array of epigenetic gene silencing mech-anismss must exist in the mammalian genome to regulate gene expression, and although a lot iss known about some mechanisms, they likely represent only the tip of the iceberg. Imprinting willl have a lot to teach us about how epigenetic gene regulatory mechanisms function in the

36 6

Chapterr 1

genome. . Ass a parting comment, if model C is the correct interpretation of how imprints func-

tion,, it will also throw down a challenge to the proponents/opponents of the popular Parental Conflictt theory [28]. This theory proposes that imprinting evolved in mammals because of the conflictingg interests of maternal and paternal genes in relation to the transfer of nutrients from thee mother to her offspring. The imprinting mechanism is proposed to manipulate embryonic growthh by silencing growth-suppressing genes in the paternal germline and by silencing growth-promotingg genes in the maternal germline. If imprints are used to derepress epimuta-tions,, as outlined in model C, then the parental genomes act in the opposite manner; the mater-nall genome activates growth-suppressing genes while the paternal genome activates growth-promotingg genes. Although the end result would be the same; maternally expressed growth-suppressingg genes and paternally expressed growth-promoting genes, the biological signifi-cancee of this may need reappraisal.

11.. Acknowledgements

Wee thank Andras Paldi, Jenny Graves, Wolf Reik, and Carmen Sapienza for answering ques-tionss during the writing of this review, Karl Pfeifer and Hamish Spencer for sending me copies off their reviews, and Wolf Reik, Marjori Matzke, Günther Kreil and Vic Small, for comments onn the manuscript. DPB is supported by the Austrian Academy of Sciences, FS by the Dutch Cancerr Society.

37 7

12.. References

11 Suiani, M.A. (1998) Cell 93, 309-12. 22 Reik, W. and Walter, J. (2001) Nat Rev Genet 2, 21 -32. 33 Tilghman, S.M. (1999) Cell 96, 185-93. 44 Bartolomei, M.S. and Tilghman, S.M. (1997) Annu Rev Genet 31,493-525. 55 Solter, D. (1998) Int J Dev Biol 42, 951-4. 66 Beechey, C.V., Cattanach, B.M. and Selley, R.L. (2000). 77 Morison, I.M., Paton, CJ. and Cleverley, S.D. (2001) Nucleic Acids Res 29, 275-6. 88 Claverie, J.M. (2001) Science 291, 1255-7. 99 O'Neill, M.J., Ingram, R.S., Vrana, P.B. and Tilghman, S.M. (2000) Dev Genes Evol 210, 18-

20. . 100 Killian, J.K., Byrd, J.C., Jirtle, J.V., Munday, B.L., Stoskopf, M.K., MacDonald, R.G. and

Jirtle,, R.L. (2000) Mol Cell 5, 707-16. 111 Graves, J.A. (1996) Annu Rev Genet 30, 233-60. 122 Sarvella, P. (1973) Nature 243, 171. 133 Cuellar, O. (1977) Science 197, 837-43. 144 Cole, C.J. and Townsend, C.R. (1990) J Exp Zool Suppl 4, 174-6. 155 Corley-Smith, G.E., Lim, CJ. and Brandhorst, B.P. (1996) Genetics 142, 1265-76. 166 Messing, J. and Grossniklaus, U. (1999) Results Probl Cell Differ 25, 23-40. 177 Herrick, G. and Seger, J. (1999) Results Probl Cell Differ 25, 41 -71. 188 Hurst, L.D. and McVean, G.T. (1998) Curr Opin Genet Dev 8, 701-8. 199 Pardo-Manuel de Villena, R, de la Casa-Esperon, E. and Sapienza, C. (2000) Trends Genet 16,

573-9. . 200 Avner, P. and Heard, E. (2001) Nat Rev Genet 2, 59-67. 211 Ferrier, D.E., Minguillon, C, Holland, P.W. and Garcia-Fernandez, J. (2000) Evol Dev 2, 284-

93. . 222 Varmuza, S. and Mann, M. (1994) Trends Genet 10, 118-23. 233 Haig, D. (1994) Trends Genet 10, 346-9. 244 Conlon, I. and Raff, M. (1999) Cell 96, 235-44. 255 Snow, M.H. (1981) Br Med Bull 37, 221-6. 266 Wilcox, A.J. and Russell, I.T. (1983) Int J Epidemiol 12, 319-25. 277 Solter, D. (1988) Annu Rev Genet 22, 127-46. 288 Moore, T. and Haig, D. (1991) Trends Genet 7, 45-9. 299 Xu, C, Mao, D., Holers, V.M., Palanca, B., Cheng, A.M. and Molina, H. (2000) Science 287,

498-501. . 300 Munn, D.H., Zhou, M., Attwood, J.T., Bondarev, L, Conway, S.J., Marshall, B., Brown, C. and

Mellor,, A.L. (1998) Science 281, 1191-3. 311 Nowak, R., Porter, R.H., Levy, R, Orgeur, P. and Schaal, B. (2000) Rev Reprod 5, 153-63. 322 Jones, B.K., Levorse, J.M. and Tilghman, S.M. (1998) Genes Dev 12, 2200-7. 333 Leighton, P.A., Ingram, R.S., Eggenschwiler, J., Efstratiadis, A. and Tilghman, S.M. (1995)

Naturee 375, 34-9. 344 Caspary, T., Cleary, M.A., Baker, C C, Guan, X.J. and Tilghman, S.M. (1998) Mol Cell Biol

18,3466-74. . 355 Wutz, A., Theussl, H.C., Dausman, J., Jaenisch, R., Barlow, D.P. and Wagner, E.F. (2001)

Developmentt 128, 1881-7. 366 Barlow, D.P. (1993) Science 260, 309-10. 377 DeChiara, T.M., Efstratiadis, A. and Robertson, E.J. (1990) Nature 345, 78-80. 388 Vu, T.H. and Hoffman, A.R. (1994) Nature 371, 714-7. 399 Schweifer, N., Valk, P.J., Delwel, R., Cox, R., Francis, R, Meier-Ewert, S., Lehrach, H. and

Barlow,, D.P. (1997) Genomics 43, 285-97. 400 Hu, J.F., Oruganti, H., Vu, T.H. and Hoffman, A.R. (1998) Mol Endocrinol 12, 220-32. 411 Szabo, P.E. and Mann, J.R. (1995) Genes Dev 9, 3097-108.

38 8

Chapterr 1

422 Lerchner, W. and Barlow, D.P. (1997) Mech Dev 61,141-9. 433 Wang, Z.Q., Fun, MR., Barlow, D.P. and Wagner, E.F. (1994) Nature 372,464-7. 444 Matsuoka, S., Thompson, J.S., Edwards, M.C., Bartletta, J.M., Grundy, P., Kalikin, L.M.,

Harper,, J.W., Elledge, S.J. and Feinberg, A.P. (1996) Proc Natl Acad Sci USA 93, 3026-30. 455 Xu, Y., Goodyer, C.G., Deal, C. and Polychronakos, C. (1993) Biochem Biophys Res Commun

197,, 747-54. 466 Kalscheuer, V.M., Mariman, E.C., Schepens, M.T., Rehder, H. and Ropers, H.H. (1993) Nat

Genett 5, 74-8. 477 Ogawa, O., McNoe, L.A., Eccles, M.R., Morison, I.M. and Reeve, A.E. (1993) Hum Mol

Genett 2,2163-5. 488 Bunzel, R., Blumcke, I., Cichon, S., Normann, S., Schramm, J., Propping, P. and Nothen, M.M.

(1998)) Brain Res Mol Brain Res 59, 90-2. 499 Nishiwaki, K„ Niikawa, N. and Ishikawa, M. (1997) Jpn J Hum Genet 42, 205-11. 500 Giannoukakis, N., Deal, C, Paquette, J., Kukuvitis, A. and Polychronakos, C. (1996) Biochem

Biophyss Res Commun 220, 1014-9. 511 Sasaki, H., Jones, P.A., Chaillet, J.R., Ferguson-Smith, A.C., Barton, S.C., Reik, W. and Surani,

M.A.. (1992) Genes Dev 6, 1843-56. 522 Stöger, R., Kubicka, P., Liu, C.G., Karri, T., Razin, A., Cedar, H. and Barlow, D.P. (1993) Cell

73,61-71. . 533 Ferguson-Smith, A.C., Sasaki, H., Cattanach, B.M. and Surani, M.A. (1993) Nature 362, 751-

5. . 544 Bartolomei, M.S., Zemel, S. and Tilghman, S.M. (1991) Nature 351, 153-5. 555 Onyango, P., Miller, W., Lehoczky, J., Leung, C.T., Birren, B„ Wheelan, S., Dewar, K. and

Feinberg,Feinberg, A.P. (2000) Genome Res 10, 1697-710. 566 Engemann, S., Strodicke, M., Paulsen, M., Franck, O., Reinhardt, R., Lane, N., Reik, W. and

Walter,, J. (2000) Hum Mol Genet 9,2691-706. 577 Thorvaldsen, J.L., Duran, K.L. and Bartolomei, M.S. (1998) Genes Dev 12, 3693-702. 588 Horike, S., Mitsuya, K., Meguro, M., Kotobuki, N., Kashiwagi, A., Notsu, T., Schulz, T.C.,

Shirayoshi,, Y. and Oshimura, M. (2000) Hum Mol Genet 9, 2075-83. 599 Mannens, M. and Alders, M. (1999) Ann Med 31,4-11. 600 Lee, J.E., Tantravahi, U., Boyle, A.L. and Efstratiadis, A. (1993) Mol Reprod Dev 35, 382-90. 611 John, R.M., Hodges, M., Little, P., Barton, S.C. and Surani, M.A. (1999) Hum Mol Genet 8,

2211-9. . 622 Wutz, A., Smrzka, O.W., Schweifer, N., Schellander, K., Wagner, E.F. and Barlow, D.P. (1997)

Naturee 389, 745-9. 633 Pfeifer, K., Leighton, PA. and Tilghman, S.M. (1996) Proc Natl Acad Sci USA 93, 13876-83. 644 Blaydes, S.M., Elmore, M., Yang, T. and Brannan, C.I. (1999) Mamm Genome 10, 549-55. 655 Shemer, R., Hershko, A.Y., Perk, J., Mostoslavsky, R„ Tsuberi, B., Cedar, H„ Buiting, K. and

Razin,, A. (2000) Nat Gen 26, 440-3. 666 Bestor, T.H. (2000) Hum Mol Genet 9, 2395-402. 677 Ramsahoye, B.H., Biniszkiewicz, D., Lyko, F., Clark, V, Bird, A.P. and Jaenisch, R. (2000)

Procc Natl Acad Sci USA 97, 5237-42. 688 Oswald, J., Engemann, S., Lane, N., Mayer, W., Olek, A., Fundele, R., Dean, W, Reik, W. and

Walter,, J. (2000) Curt Biol 10,475-8. 699 Tremblay, K.D., Saam, J.R., Ingram, R.S., Tilghman, S.M. and Bartolomei, M.S. (1995) Nat

Genn 9,407-13. 700 Shibata, H., Yoda, Y, Kato, R., Ueda, T., Kamiya, M., Hiraiwa, N., Yoshiki, A., Plass, C,

Pearsall,, R.S., Held, W.A., Muramatsu, M., Sasaki, H., Kusakabe, M. and Hayashizaki, Y. (1998)) Genomics 49, 30-7.

711 Shibata, H., Ueda, T, Kamiya, M., Yoshiki, A., Kusakabe, M., Plass, C., Held, W.A., Sunahara, S.,, Katsuki, M., Muramatsu, M. and Hayashizaki, Y. (1997) Genomics 44, 171-8.

722 Shemer, R., Birger, Y, Riggs, A.D. and Razin, A. (1997) Proc Natl Acad Sci USA 94, 10267-72. .

733 Li , E., Beard, C. and Jaenisch, R. (1993) Nature 366, 362-5. 744 Paulsen, M., El-Maarri, O., Engemann, S., Strodicke, M., Franck, O., Davies, K., Reinhardt,

R.,, Reik, W. and Walter, J. (2000) Hum Mol Genet 9, 1829-41.

39 9

755 Cross, S.H. and Bird, A.P. (1995) Curr Opin Genet Dev 5, 309-14. 766 Antequera, F. and Bird, A. (1999) Curr Biol 9, R661-7. 777 Bird, A.P. (1980) Nucleic Acids Res 8, 1499-504. 788 Brandeis, M., Kafri, T., Ariel, M., Chaillet, J.R., McCarrey, J., Razin, A. and Cedar, H. (1993)

Emboo J 12, 3669-77. 799 Frank, D., Keshet, I., Shani, M., Levine, A., Razin, A. and Cedar, H. (1991) Nature 351, 239-

41. . 800 Macleod, D., Charlton, J., Mullins, J. and Bird, A.P. (1994) Genes Dev 8, 2282-92. 811 Neumann, B., Kubicka, P. and Barlow, D.P. (1995) Nat Gen 9, 12-3. 822 Brown, C.J., Ballabio, A., Rupert, J.L., Lafreniere, R.G., Grompe, M., Tonlorenzi, R. and

Willard,, H.F. (1991) Nature 349, 38-44. 833 Smrzka, O.W., Faé, I., Stöger, R., Kurzbauer, R., Fischer, G.F., Henn, T., Weith, A. and Barlow,

D.P.. (1995) Hum Mol Genet 4, 1945-52. 844 Dittrich, B., Buiting, K., Gross, S. and Horsthemke, B. (1993) Hum Mol Genet 2, 1995-9. 855 Hatada, I. and Mukai, T. (1995) Nat Genet 11, 204-6. 866 Jinno, Y., Sengoku, K., Nakao, M., Tamate, K., Miyamoto, T., Matsuzaka, T., Sutcliffe, J.S.,

Anan,, T., Takuma, N., Nishiwaki, K., Dteda, Y., Ishimaru, T., Ishikawa, M. and Niikawa, N. (1996)) Hum Mol Genet 5, 1155-61.

877 Sasaki, H., Shimozaki, K., Zubair, M., Aoki, N., Ohta, K., Hatano, N., Moore, T., Feil, R., Constancia,, M., Reik, W. and Rotwein, P. (1996) DNA Res 3, 331-5.

888 Moore, T., Constancia, M., Zubair, M., Bailleul, B., Feil, R., Sasaki, H. and Reik, W. (1997) Procc Natl Acad Sci USA 94, 12509-14.

899 Wevrick, R. and Francke, U. (1997) Hum Mol Genet 6, 325-32. 900 Huq, A.H., Sutcliffe, J.S., Nakao, M., Shen, Y., Gibbs, R.A. and Beaudet, A.L. (1997) Genome

Ress 7, 642-8. 911 Pearsall, R.S., Plass, C, Romano, M.A., Garrick, M.D., Shibata, H., Hayashizaki, Y. and Held,

W.A.. (1999) Genomics 55, 194-201. 922 Boccaccio, I., Glatt-Deeley, H., Watrin, R, Roeckel, N., Lalande, M. and Muscatelli, F. (1999)

Humm Mol Genet 8,2497-505. 933 Miura, K., Miyoshi, O., Yun, K., Inazawa, J., Miyamoto, T., Hayashi, H., Masuzaki, H.,

Yoshimura,, S„ Niikawa, N., Jinno, Y. and Ishimaru, T. (1999) J Hum Genet 44, 1-9. 944 Li , L.L., Szeto, I.Y., Cattanach, B.M., Ishino, F. and Surani, M.A. (2000) Genomics 63, 333-

40. . 955 Okamura, K., Hagiwara-Takeuchi, Y., Li , T., Vu, T.H., Hirai, M., Hattori, M., Sakaki, Y,

Hoffman,, A.R. and Ito, T. (2000) Genome Res 10, 1878-89. 966 Stadnick, M.P., Pieracci, F.M., Cranston, M.J., Taksel, E., Thorvaldsen, J.L. and Bartolomei,

M.S.. (1999) Dev Genes Evol 209, 239^18. 977 Sunahara, S., Nakamura, K., Nakao, K., Gondo, Y, Nagata, Y and Katsuki, M. (2000)

Biochemm Biophys Res Commun 268, 590-5. 988 Birger, Y, Shemer, R., Perk, J. and Razin, A. (1999) Nature 397, 84-8. 999 Kitsberg, D., Selig, S., Brandeis, M., Simon, I., Keshet, I., Driscoll, D.J., Nicholls, R.D. and

Cedar,, H. (1993) Nature 364,459-63. 1000 Knoll, J.H., Cheng, S.D. and Lalande, M. (1994) Nat Genet 6, 41-6. 1011 Kawame, H., Gartler, S.M. and Hansen, R.S. (1995) Hum Mol Genet 4, 2287-93. 1022 Bickmore, W.A. and Carothers, A.D. (1995) J Cell Sci 108, 2801-9. 1033 Watanabe, T., Yoshimura, A., Mishima, Y, Endo, Y, Shiroishi, T., Koide, T., Sasaki, H.,

Asakura,, H. and Kominarni, R. (2000) Hum Mol Genet 9, 3029-35. 1044 Hark, A.T. and Tilghman, S.M. (1998) Hum Mo! Genet 7, 1979-85. 1055 Greally, J.M., Gray, T.A., Gabriel, J.M., Song, L., Zemel, S. and Nicholls, R.D. (1999) Proc

Natll Acad Sci USA 96, 14430-5. 1066 Feil, R. and Khosla, S. (1999) Trends Genet 15,431-5. 1077 Marahrens, Y, Panning, B., Dausman, J., Strauss, W. and Jaenisch, R. (1997) Genes Dev 11,

156-66. . 1088 Pfeifer, K. and Tilghman, S.M. (1994) Genes Dev 8, 1867-74. 1099 Wutz, A. and Jaenisch, R. (2000) Mol Cell 5, 695-705. 1100 de los Santos, T., Schweizer, J., Rees, CA. and Francke, U. (2000) Am J Hum Genet 67,1067-

Chapterr 1

82. . 1111 Hurst, L.D., McVean, G. and Moore, T. (1996) Nat Gen 12, 234-7. 1122 Lyle, R., Watanabe, D., te Vruchte, D., Lerchner, W., Smrzka, O.W., Wutz, A., Schageman, J.,

Hahner,, L., Davies, C. and Barlow, D.P. (2000) Nat Gen 25, 19-21. 1133 Lee, J.T. (2000) Cell 103, 17-27. 1144 Schmidt, J.V., Levorse, J.M. and Tilghman, S.M. (1999) Proc Natl Acad Sci USA 96, 9733-8. 1155 Bell, A.C. and Felsenfeld, G. (2000) Nature 405,482-5. . 1166 Thorvaldsen, J.L. and Bartolomei, M.S. (2000) Science 288, 2145-6. 1177 Bass, B.L. (2000) Cell 101,235-8. 1188 Stancheva, I. and Meehan, R.R. (2000) Genes Dev 14, 313-27. 1199 Svoboda, P., Stein, P., Hayashi, H. and Schultz, R.M. (2000) Development 127,4147-56. 1200 Mette, M.F., Aufsatz, W., van der Winden, J., Matzke, M.A. and Matzke, A.J. (2000) Embo J

19,5194-201. . 1211 McGrath, J. and Solter, D. (1984) Cell 37,179-83. 1222 Surani, M.A., Barton, S.C. and Norris, M.L . (1984) Nature 308, 548-50. 1233 LaSalle, J.M. and Lalande, M. (1996) Science 272, 725-8. 1244 Paldi, A. and Jouvenot, Y. (1999) Results Probl Cell Differ 25, 271-82. 1255 Szabo, P.E. and Mann, J.R. (1996) Embo J 15, 6018-25. 1266 Nogami, M., Kohda, A., Taguchi, H., Nakao, M., Ikemura, T. and Okumura, K. (2000) J Cell

Scii 113, 2157-65. 1277 Ashe, H.L., Monks, J., Wijgerde, M„ Fraser, P. and Proudfoot, N.J. (1997) Genes Dev 11,

2494-509. . 1288 Potter, S.S. and Branford, W.W. (1998) Mamm Genome 9, 799-806. 1299 Tasheva, E.S. and Roufa, D.J. (1995) Genes Dev 9, 304-16. 1300 Li , A.W., Seyoum, G., Shiu, R.P. and Murphy , P.R. (1996) Mol Cell Endocrinol 118, 113-23. 1311 Murashov, A.K. and Wolgemuth, D.J. (1996) Brain Res Mol Brain Res 37, 85-95. 1322 Krystal , G.W., Armstrong, B.C. and Battey, J.R (1990) Mol Cell Biol 10, 4180-91. 1333 Plass, C, Shibata, H., Kalcheva, I., Mullins , L., Kotelevtseva, N., Mullins , J., Kato, R, Sasaki,

H.,, Hirotsune, S., Okazaki, Y., Held, W.A., Hayashizaki, Y. and Chapman, V.M. (1996) Nat Genett 14,106-9.

1344 Bielinska, B., Blaydes, S.M., Buiting, K., Yang, T., Krajewska-Walasek, M., Horsthemke, B. andd Brannan, C.I. (2000) Nat Gen 25,74-8.

1355 Constancia, M., Dean, W., Lopes, S., Moore, T., Kelsey, G. and Reik, W. (2000) Natt Genet 26, 203-6. .

1366 Smilinich, N.J., Day, CD., Fitzpatrick , G.V., Caldwell, G.M., Lossie, A.C., Cooper, P.R., Smallwood,, A.C., Joyce, J.A., Schofield, P.N., Reik, W., Nicholls, R.D., Weksberg, R., Driscoll,, D.J., Maher, E.R., Shows, T.B. and Higgins, M.J. (1999) Proc Natl Acad Sci USA 96, 8064-9. .

1377 Jackson-Grusby, L., Beard, C, Possemato, R., Tudor, M., Fambrough, D., Csankovszki, G., Dausman,, J., Lee, P., Wilson, C, Lander, E. and Jaenisch, R. (2001) Nat Genet 27, 31-9.

1388 Sado, T., Fenner, M.H., Tan, S.S., Tarn, P., Shioda, T. and Li , E. (2000) Dev Biol 225,294-303. 1399 El Kharroubi , A., Piras, G. and Stewart, C.L. (2001) J Biol Chem 276, 8674-80. 1400 Ng, H.H. and Bird , A. (1999) Curr Opin Genet Dev 9,158-63. 1411 Wolffe, A.P. and Guschin, D. (2000) J Struct Biol 129, 102-22. 1422 Obata, Y, Kaneko-Ishino, T., Koide, T., Takai, Y, Ueda, T., Domeki, I., Shiroishi, T., Ishino, F.

andd Kono, T. (1998) Development 125, 1553-60. 1433 Kato, Y, Rideout, W.M., 3rd, Hilton , K., Barton, S.C, Tsunoda, Y. and Surani, M.A. (1999)

Developmentt 126, 1823-32. 1444 Barlow, D.P. (1997) Embo J 16, 6899-905. 1455 Duhl, D.M., Vrieling , H., Miller , K.A., Wolff, G.L. and Barsh, G.S. (1994) Nat Genet 8,59-65. 1466 Michaud, E.J., van Vugt, M.J., Bultman, S.J., Sweet, H.O., Davisson, M.T. and Woychik, R.P.

(1994)) Genes Dev 8,1463-72. 1477 Kratochwil , K., von der Mark , K., Kollar , E.J., Jaenisch, R., Mooslehner, K., Schwarz, M.,

Haase,, K., Gmachl, I. and Harbers, K. (1989) Cell 57, 807-16. 1488 McClintock, B. (1992) The Dynamic Genome, Barbara McClintock' s ideas in the century of

genetics.. CSH Press, USA., 361-80.

41 1

1499 Martienssen, R. (1998) Trends Genet 14, 263-4. 1500 Hattori, M., Fujiyama, A., Taylor, T.D., Watanabe, H., Yada, T., Park, H.S., Toyoda, A., Ishii,

K.,, Totoki, Y., Choi, D.K., Soeda, E., Ohki, M., Takagi, T., Sakaki, Y., Taudien, S., Blechschmidt,, K., Polley, A., Menzel, U., Delabar, J., Kumpf, K., Lehmann, R., Patterson, D., Reichwald,, K., Rump, A., Schillhabel, M., Schudy, A., Zimmermann, W., Rosenthal, A., Kudoh,, J., Schibuya, K., Kawasaki, K., Asakawa, S., Shintani, A., Sasaki, T., Nagamine, K., Mitsuyama,, S., Antonarakis, S.E., Minoshima, S., Shimizu, N., Nordsiek, G., Hornischer, K., Brant,, P., Scharfe, M., Schon, O., Desario, A., Reichelt, I , Kauer, G., Blocker, H., Ramser, J., Beek,, A., Klages, S., Hennig, S., Riesselmann, L., Dagand, E., Haaf, T., Wehrmeyer, S., Borzym,, K., Gardiner, K., Nizetic, D., Francis, F, Lehrach, H., Reinhardt, R. and Yaspo, M.L. (2000)) Nature 405, 311-9.

1511 Dunham, I., Shimizu, N„ Roe, B.A., Chissoe, S., Hunt, A.R., Collins, J.E., Bruskiewich, R., Beare,, D.M., Clamp, M., Smink, L.J., Ainscough, R., Almeida, J.P., Babbage, A., Bagguley, C, Bailey,, J., Barlow, K., Bates, K.N., Beasley, O., Bird, CR, Blakey, S., Bridgeman, A.M., Buck, D.,, Burgess, J., Burrill , W.D., O'Brien, K.P. and et al. (1999) Nature 402,489-95.

1522 Walsh, CR, Chaillet, J.R. and Bestor, T.H. (1998) Nat Genet 20, 116-7. 1533 Kidwell, M.G. and Lisch, D. (1997) Proc Natl Acad Sci USA 94, 7704-11. 1544 Smit, A.F. (1999) Curr Opin Genet Dev 9, 657-63. 1555 Prak, E.T. and Kazazian, H.H., Jr. (2000) Nat Rev Genet 1, 134-44. 1566 Mi, S., Lee, X., Li, X., Veldman, G.M., Finnerty, H., Racie, L., LaVallie, E., Tang, X.Y.,

Edouard,, P., Howes, S., Keith, J.C, Jr. and McCoy, J.M. (2000) Nature 403, 785-9. 1577 Ripoche, M.A., Kress, C, Poirier, F. and Dandolo, L. (1997) Genes Dev 11, 1596-604. 1588 Brannan, CI. and Bartolomei, M.S. (1999) Curr Opin Genet Dev 9, 164-70. 1599 Schmidt, J.V., Matteson, P.G., Jones, B.K., Guan, X.J. and Tilghman, S.M. (2000) Genes Dev

14,, 1997-2002. 1600 Takada, S., Tevendale, M., Baker, J., Georgiades, P., Campbell, E., Freeman, T., Johnson,

M.H.,, Paulsen, M. and Ferguson-Smith, A.C. (2000) Curr Biol 10, 1135-8. 1611 Sleutels, F., Barlow,, D.P. and Lyle, R. (2000) Curr Opin Genet Dev 10, 229-233. 1622 Bird, A. (1999) Science 286, 2287-8. 1633 Tanaka, M., Puchyr, M., Gertsenstein, M., Harpal, K., Jaenisch, R„ Rossant, J. and Nagy, A.

(1999)) Mech Dev 87, 129-42. 1644 Frank, D., Mendelsohn, C.L., Ciccone, E., Svensson, K., Ohlsson, R. and Tycko, B. (1999)

Mammm Genome 10, 1150-9. 1655 Duvillie, B., Cordonnier, N., Deltour, L„ Dandoy-Dron, R, Itier, J.M., Monthioux, E., Jami, J.,

Joshi,, R.L. and Bucchini, D. (1997) Proc Natl Acad Sci USA 94, 5137-40. 1666 Yang, T., Adamson, T.E., Resnick, J.L., LeflF, S., Wevrick, R., Francke, U., Jenkins, N.A.,

Copeland,, N.G. and Brannan, C.I. (1998) Nat Genet 19, 25-31. 1677 Tsai, T.F., Jiang, Y.H., Bressler, J., Armstrong, D. and Beaudet, A.L. (1999) Hum Mol Genet 8,

1357-64. . 1688 Brambilla, R., Gnesutta, N., Minichiello, L., White, G., Roylance, A.J., Herron, C.E., Ramsey,

M.,, Wolfer, D.P., Cestari, V., Rossi-Arnaud, C, Grant, S.G., Chapman, P.F., Lipp, H.P., Sturani, E.. and Klein, R. (1997) Nature 390,281-6.

1699 Itier, J.M., Tremp, G.L., Leonard, J.F., Multon, M.C., Ret, G., Schweighoffer, F., Tocque, B., Bluet-Pajot,, M.T., Cormier, V. and Dautry, F. (1998) Nature 393,125-6.

1700 Lefebvre, L., Viville , S., Barton, S.C., Ishino, F., Keverne, E.B. and Surani, M.A. (1998) Nat Genett 20,163-9.

1711 Li, L., Keverne, E.B., Aparicio, S.A., Ishino, F., Barton, S.C. and Surani, M.A. (1999) Science 284,, 330-3.

1722 Gerard, M., Hernandez, L., Wevrick, R. and Stewart, C.L. (1999) Nat Genet 23, 199-202. 1733 Walther, T., Balschun, D., Voigt, J.P., Fink, H., Zuschratter, W, Birchmeier, C, Ganten, D. and

Bader,, M. (1998) J Biol Chem 273, 11867-73. 1744 Yu, S., Gavrilova, O., Chen, H., Lee, R., Liu, J., Pacak, K., Parlow, A.F., Quon, M.J., Reitman,

M.L.. and Weinstein, L.S. (2000) J Clin Invest 105, 615-23. 1755 Peters, J., Wroe, S.F., Wells, C.A., Miller, H.J., Bodle, D., Beechey, C.V., Williamson, CM.

andd Kelsey, G. (1999) Proc Natl Acad Sci USA 96, 3830-5. 1766 Guillemot, F., Caspary, T., Tilghman, S.M., Copeland, N.G., Gilbert, D.J., Jenkins, N.A.,

42 2

Chapterr 1

Anderson,, DJ., Joyner, A.L., Rossant, J. and Nagy, A. (1995) Nat Genet 9,235-42. 1777 Zhang, P., Liegeois, N.J., Wong, C, Finegold, M., Hou, H., Thompson, J.C., Silvennan, A.,

Haiper,, J.W., DePinho, R.A. and Elledge, SJ. (1997) Nature 387, 151-8. 1788 Yan, Y., Frisen, J., Lee, M.H., Massague, J. and Barbacid, M. (1997) Genes Dev 11, 973-83. 1799 Jiang, Y.H., Armstrong, D., Albrecht, U., Atkins, CM., Noebels, J.L., Eichele, G., Sweatt, J.D.

andd Beaudet, A.L. (1998) Neuron 21, 799-811. 1800 Schuster-Gossler, K., Simon-Chazottes, D., Guenet, J.L., Zachgo, J. and Gossler, A. (1996)

Mammm Genome 7, 20-4. 1811 Wroe, S.F., Kelsey, G., Skinner, J.A., Bodle, D., Ball, S.T., Beechey, C.V., Peters, J. and

Williamson,, CM. (2000) Proc Natl Acad Sci USA 97, 3342-6. 1822 Lee, Y.J., Park, C.W., Hahn, Y, Parte, J., Lee, J., Yun, J.H., Hyun, B. and Chung, J.H. (2000)

FEBSS Lett 472, 230-4. 1833 Jong, M.T., Gray, T.A., Ji, Y, Glenn, CC, Saitoh, S„ Driscoll, D.J. and Nicholls, R.D. (1999)

Humm Mol Genet 8, 783-93. 1844 Sutcliffe, J.S., Nakao, M., Christian, S., Orstavik, K.H., Tommerup, N., Ledbetter, D.H. and

Beaudet,, A.L. (1994) Nat Genet 8, 52-8. 1855 Rougeulle, C, Cardoso, C, Fontés, M., Colleaux, L. and Lalande, M. (1998) Nat Gen 19, 15-

6. . 1866 Schuster-Gossler, K., Simon-Chazottes, D., Guenet, J.L., Zachgo, J. and Gossler, A. (1996)

Mammm Genome 7, 20-4. 1877 Lee, J.T., Davidow, L.S. and Warshawsky, D. (1999) Nat Genet 21,400-4.

43 3

13.. Appendix 1

Abbreviationss used in the text Notee that gene loci follow the accepted nomenclature rules and are written in capitals to refer too human gene loci, and in lower case with an initial capital letter, for mouse loci.

Air Air as s bFGF bFGF Collal Collal Cop2g Cop2g Cdknlc Cdknlc CTCF CTCF Dlkl Dlkl Dnmt t DMR R HERV HERV Hsp70 Hsp70 IgP IgP Ins2 Ins2 GOI I GrblO GrblO GÜ2 GÜ2 H19 H19 ICE E IPW IPW KCNQ1 KCNQ1 KCNQIOTI KCNQIOTI MAGEL2 MAGEL2 Mash2 Mash2 Mest Mest M6P-Igf2r M6P-Igf2r

Movl3 Movl3 Nesp Nesp Nmyc Nmyc PAR1.5 PAR1.5 PegJ PegJ Peg3 Peg3 PW71 PW71 PWCR1 PWCR1 RasGrfl RasGrfl RNAi i RPS14 RPS14 SNRPN SNRPN TP53 TP53 Tsix Tsix U2afl-rsl U2afl-rsl UBE3A UBE3A Xist Xist ZFP127 ZFP127

Antisensee Igf2r RNA antisense e Basicc fibroblast growth factor Collagenn type I, alpha I chain. Coatomerr protein complex, subunit gamma 2 Cyctin-dependentt kinase inhibitor 1C (also called p57kip2) CCCTCC nucleotide binding factor Delta-likee 1 DNAA methyltransferase Differentiallyy Methylated Region Humann endogenous Retrovirus HeatHeat shock protein 70 Insulin-likee growth factor type 2 Insulinn II (mice have two insulin genes, humans only one) Gainn Of Imprinting Growthh factor receptor-bound protein 10 Genee trap locus 2 aa cDNA clone isolated from a fetal Hepatic library Imprintt Control Element Imprintedd in Prader-Willi region Potassiumm voltage-gated channel, KQT-like subfamily, member 1 Kcnqll overlapping transcript 1 MAGE-likeMAGE-like 2 MusMus musculus Achaete-Scute homologue 2 Mesodermm specific transcript Cation-independentt -Mannose-6-phosphate receptor, (alsoo known as Insulin-like growth factor type 2receptor) Molonyy Virus intergration 13 NeuroendocrineNeuroendocrine secretory protein Neuroblastomaa myc(avian myelocytomatosis related)-related oncogene Prader-Willi/Angelmann region-1 and 5 Paternallyy Expressed Gene 1 Paternallyy Expressed Gene 3 Prader-Willii 71 Prader-Willii critical region 1 Rass protein-specific guanine nucleotide-releasing factor 1 RNAA interference Ribosomall protein S14 Smalll nuclear ribonucleoprotein polypeptide N Tumorr protein p53 X-inactivation-specificc transcript-antisense U22 small nuclear ribonucleoprotein auxiliary factor-related sequence 1 Ubiquitinn protein ligase E3A XX inactive specific transcript Zincc finger protein 127 also known as ZN127 (now called Makorin 3)

44 4