Proc. Natl. Acad. Sci. USAVol. 86, pp. 7480-7484, October 1989Genetics

A model for embryonal rhabdomyosarcoma tumorigenesis thatinvolves genome imprinting

(allele inactivation/cancer genetics)

HEIDI SCRABLE*, WEBSTER CAVENEE*t, FERESHTEH GHAVIMIt, MARK LOVELL§, KENNETH MORGANt¶,AND CARMEN SAPIENZA*tII*Ludwig Institute for Cancer Research, 687 Pine Avenue West, Montreal, PQ, Canada H3A lA1; tCenter for Human Genetics, McGill University, Montreal,PQ, Canada H3A iB1; tMemorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021; §Department of Pathology, University ofVirginia Health Sciences Center, Charlottesville, VA 22908; and ¶Department of Epidemiology and Biostatistics, McGill University, Montreal, PQ,Canada H3A 1A2

Communicated by Alfred G. Knudson, Jr., June 15, 1989 (received for review February 15, 1989)

ABSTRACT Embryonal rhabdomyosarcomas (malignantpediatric tumors of striated muscle oriin) have been shown toarise from cells that are clonally isodisomic for loci on chro-mosome lip. We determined the parental origin of alleles inthis genomic region in familial and sporadic cases of this diseaseand found that isodisomic chromosome lip alleles in eachtumor were of paternal origin. We have developed a modifi-cation of Knudson's two-hit model from these data that iscapable of explaining the preferential allele retention and ofresolving the apparent contradiction between such specific andearly events in several embryonal tumors and discrepancies inthe inheritance of predisposition in some of these diseases.

Models of tumorigenesis that require as few as two events incancer initiation (1-3) have provided a strong theoreticalframework within which to interpret characteristic geneticchanges observed in a number of human tumors (4-12).Knudson's model, which invokes the successive inactivationof alleles at a single locus, has been particularly useful indetermining the underlying mechanisms by which manytumors are formed (1, 2). Little is known of the molecularnature of the initial, predisposing events in this pathway,although they occasionally involve detectable alterations innucleotide sequence (mutations, deletions, inversions, ortranslocations) (13, 14).The possibility that the first event need not be due to such

an alteration is suggested by recent studies of the phenom-enon of genome imprinting, a process most convenientlydescribed as an epigenetic, gamete-of-origin-dependent, al-lele-inactivation process (15). In the mouse, its effects areobserved as gamete-of-origin-dependent mutant phenotypes(16-18), developmental failure of zygotes containing onlymaternal or paternal genetic contributions (19, 20), anddifferences in the expression (21) or methylation (21-24) ofhemizygous transgenes. In the human, complete hydatidi-form moles are the result of zygotes that contain onlypaternally derived genetic information (25).There are two important differences between alleles inac-

tivated by imprinting and those inactivated by nucleotidesequence changes. The first is that imprinted alleles need notcarry a "mutation" in the classical sense, even though suchepigenetic changes may have the same effect on phenotype.The second is that genome imprinting must necessarily affectalleles in a gamete-specific way. Thus, mutant alleles thatshow a preference for inheritance from only one sex suggestits involvement.Rhabdomyosarcomas are tumors that are believed to arise

from undifferentiated mesenchyme and that resemble devel-

oping striated muscle. They occur at the rate of 1.3-4.5 casesper million children per year (26). Tumors of the embryonalsubtype arise from precursor cells that are isodisomic [i.e.,exhibit loss of heterozygosity (4)] at loci on chromosome 11(5, 12). In some instances, these allele losses involve only themost distal markers on the short arm (12) suggesting that a"rhabdomyosarcoma locus" maps to chromosome 11, bandp15.5-pter (12).Rhabdomyosarcoma can be clinically associated with

Wilms tumor (27), a pediatric tumor of the kidney that alsocontains alterations of chromosome lip (5-8). Further, thealterations observed in Wilms tumors are associated withpreferential retention of only paternal chromosome 11 alleles(28-30).We have analyzed genetic marker loci on chromosome lip

for parental origin of alleles retained in rhabdomyosarcomas.These data give rise to a model that can explain the nonran-dom nature of allele loss, the nonsyntenic relationship be-tween predisposing and tumor-specific loci, and the unusualinheritance patterns observed in some familial cancer cases,as well as the clinical association between rhabdomysosar-coma and Wilms tumor.

MATERIALS AND METHODSDNA Extraction and Southern Hybridization. DNA extrac-

tion from peripheral blood leukocytes and tumor tissue,restriction endonuclease digestion, agarose gel electropho-resis, DNA blot hybridization, and autoradiography wereperformed as described (12, 31, 32).

Probes. Alleles at the HRASJ locus in Fig. 1 A, C, and Fand Fig. 2B were determined by hybridizing Msp I-digestedDNA with a 1-kilobase-pair (kb) Msp I fragment of probepEJ6.6 (33). Data in Fig. 1B were obtained by hybridizing RsaI-digested DNA with the probe pHins310, homologous to theinsulin locus (34). Genotypes in Fig. 1 D and E were deter-mined at HBBC by hybridizing probe JW151 (35) to HindIII-digested DNA.Probe pADJ-762, an arbitrary DNA segment, reveals re-

striction fragment length polymorphisms at the Dl1S12 locus(Fig. 2B) in Taq I-digested DNA (36). Alleles at DI1524 weredetermined in either Rsa I- or BamHI-digested DNA withprobe pE4B-TGH2 (37). Alleles at IGF2 (Fig. 2B) weredetermined with a cDNA probe that reveals alleles of 0.9 and0.7 kb in Msp I-digested DNA (38). Probe p20.36 revealsalleles of 2.6 kb and 1.9 plus 0.7 kb at the PTH locus in DNAdigested with Pst 1 (39). The LDHA probe, pLDH-1, revealsalleles of 2.3 and 1.9 kb in Taq I-digested DNA (40). ProbepHC9 is homologous to the CALCA locus, with alleles of 8.0and 6.5 in Taq I-digested DNA (41).

I1To whom reprint requests should be addressed.

7480

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

Proc. Natl. Acad. Sci. USA 86 (1989) 7481

12T81 122 181 ',22 'el 118 22, '22

kb1.9-__

a

0j7kb

3.0-r ___1.8 -1.6- -

kb2.5- _w _P1.9-_.r

1.0-_0 4_

16

975

r.

18 1612 119 97 75

18 20 189 10 9

3 4 3

i0kb8.0_7.2-

'Tokb8.0- _M7.2-

_. _4.

kb2.1 -ur _1.7

1.0- as a_

4

4

6

I1

94-

.22 16

.13 11

.10 9-8 7.6 5

.2 1

16 I8*0 9

56

4360 4

.6

,3

10

9

4

6

S

II16 18. E22II 13- -139 9. 10

7 8

I QQ2::2

18 16. .209 10. 10

90Il6 5,l6

3 4'..4

4 31 !.3

QO

-5--;

65: 1

s::

19 17 17 ,2017 17 17: ,20

15 14 14 15 14 14 15--15

3 144 14 143 3ft (50

111 tr~~wFIG. 1. Chromosome 11 gentoypes in six sporadic embryonal

rhabdomyosarcoma pedigrees. Autoradiograms show genotypes atHRASI in Rhabd 26 pedigree (A), at INS in Rhabd 31 pedigree (B),at HRASJ in Rhabd 45 pedigree (C), at HBBC in Rhabd 6 pedigree(D), at HBBC in Rhabd 33 pedigree (E), and at HRAS1 in Rhabd 48

RESULTSChromosome 11 Alleles Remaining in the Tumor in Sporadic

Cases of Rhabdomyosarcoma Always Come from the Father.We compared genotypes from normal and tumor tissue inaffected individuals and normal tissue from their parents. Thefirst tumor, Rhabd 26, was found as a nasopharyngeal massin a 3-year-old girl. A mitotic recombination event during thegenesis of the tumor left all but the most distal informativelocus on lip (HRASI) heterozygous, as previously reported(12). The father, who is heterozygous for the 1.9- and 0.85-kballeles, contributed the 1.9-kb allele to his daughter, and it isthis allele that remains in the rhabdomyosarcoma (Fig. 1A).The proband from the second pedigree was a 5-year-old

boy with rhabdomyosarcoma of the nasopharynx (Rhabd 31,Fig. 1B). The child inherited the 3.0-kb INS allele from hisfather and the 1.6-kb allele from his mother. The tumorretains only the paternally derived 3.0-kb allele. Similarbehavior was observed at the HRASI locus, where the tumorcontained only a 2.5-kb allele, originating from the father.Additional loci at which paternal derivation of the remainingallele could be established were Dl IS12, HBBC (Gg and Ag),as well as the D11524 locus on llq.The proband from the third pedigree (Fig. 1C) was an

11-month-old girl with an embryonal rhabdomyosarcoma ofthe thigh. Compared to her normal DNA at HRASI, whereshe is heterozygous for the 2.5-kb maternal allele and the1.9-kb paternal allele, the tumor is homozygous for the 1.9-kbpaternal allele. Other loci exhibiting the same pattern werePTH and HBBC, Gg.The fourth pedigree (Fig. 1D) consists of mother and the

proband, who was diagnosed with embryonal rhabdomyosar-coma of the pelvis and retroperitoneum when he was 6months old. The 7.2-kb allele ofthe Gg portion of the P-globinlocus (HBBC) in the proband's normal tissue was inheritedfrom his mother, who is homozygous for this allele. There-fore, the 8.0-kb allele remaining in the rhabdomyosarcomamust be derived from his father. Similarly, at DJ1S12 and atPTH in 11p15.4, the mother is again homozygous for theallele not found in the tumor.The fifth case we examined (Rhabd 33, Fig. 1E) includes a

tumor that was examined after therapy. The cells of thistumor retain primarily the paternally derived chromosomelip, which carries the paternal 7.2-kb allele at HBBC.The final case was a congenital embryonal rhabdomyosar-

coma of the heel in a male infant. Again, alleles at HRASIdemonstrated that the 2.1-kb allele in the tumor came fromthe father, the 1.0-kb maternal allele having been lost (Fig.1F).

Mitotic Events in Familial Rhabdomyosarcoma Result inLoss ofMaternal Alleles at Loci on Chromosome lip. We haveanalyzed the family presented in Fig. 2A, in which twosiblings were diagnosed with embryonal rhabdomyosarcomabefore they were 3 years ofage (arrows). We compared tumorDNA and normal DNA from the younger of the two childrenand found that there was loss of alleles at loci on lip in thetumor. Fig. 2B shows that at LDHA in lipi5 and at the linkedloci HBBC (not shown), DllS12, IGF2, and HRASI in11p15.5, one of the codominantly inherited alleles present inthe normal tissue is missing in tumor DNA. At each locus,

pedigree (F). Lanes contained DNA from the father (open square),mother (open circle), proband (filled symbol), and tumor (T). Sizes(in kilobase pairs) of allelic restriction fragments are given to the leftofeach autoradiogram. Other 11p loci are shown diagrammatically atright: 1 and 2 are alleles at DIIS24; 3 and 4 are alleles at PTH; 5 and6 are alleles at HBBC Gg; 7 and 8 are alleles at HBBC Ag; 9 and 10are alleles at D11512; 11, 12, and 13 are alleles at INS; 14 and 15 arealleles at IGF2; and 16-22 are alleles at HRAS. Solid chromosome,paternal; dashed chromosome, maternal.

B

C

D

E

F

I x

Genetics: Scrable et al.

Iif 1

1

Proc. Natl. Acad. Sci. USA 86 (1989)

6ilTkb

8.3- -M -

2.3- w*~ 3.2- 5m1 9- ".,

LDHA

6iiTkb

DIlS12

OTO.. .._kb

1.9-

0.9Ua " 0.85- @__@@_

0-0, ,os

IGF2 HRASI

FIG. 2. Familial rhabdomyosarcoma. (A) Pedigree of a familywith two siblings affected with rhabdomyosarcoma (arrows). RCC,renal cell carcinoma. (B) Alleles at loci on chromosome lip innuclear family and tumor (T).

this circumstance arose through the elimination of the ma-ternal allele and the retention of the paternal allele. Thus, atHRASI the affected daughter inherited the 1.9-kb allele fromher mother; in the tumor only the paternal 0.85-kb alleleremains. At the IGF2 locus in llpl5.5, the father is hetero-zygous for the 0.9- and 0.7-kb alleles, and the mother ishomozygous for the 0.9-kb allele. The tumor is homozygousfor the 0.7-kb allele, again demonstrating the paternal deri-vation of the tumor chromosomes 11. Similarly, at the otherloci shown in Fig. 2B (LDHA and Dl S12) and at HBBC (datanot shown), segregation analysis demonstrated the parentalorigin of the alleles in the child's normal DNA. At all loci,tumor alleles were of paternal derivation.Embryonal rhabdomyosarcoma occurs at the rate of =:4O-6

per year. Given the improbability of the appearance of thisphenotype in two members of this family, and the evidencefor somatic events on lip tightly associated with the pheno-type, we sought to determine whether specific markerscosegregated with the disease in this family. We examinedthe genotypes at all marker lip loci in 17 family membersspanning four generations.For data analyses we assumed the most favorable case; a

similar mutational basis for the development of the clinically

associated tumors rhabdomyosarcoma, breast cancer, lungcancer, and prostatic cancer. We further applied the analysisunder various assumptions of penetrance, including an in-completely penetrant autosomal dominant and autosomalrecessive predisposition. Maximum likelihood estimates ofrecombination fractions were computed using the LODSCOREprogram of the LINKAGE package, version 4.7 (42). We foundno evidence of linkage of any specified phenotype to anychromsome lip locus, but neither could we exclude linkage.Whether this was due to inadequacies of family structure,individuals available for sampling, information content ofgenetic markers, or application of inappropriate genetictransmission models remains to be determined.

DISCUSSIONThe accumulated data on parental origin of isodisomic chro-mosome 11 alleles in embryonal rhabdomyosarcoma andWilms tumor (refs. 28-30; this study) and chromosome 13alleles in osteosarcoma (43) indicate a preference for reten-tion of paternal alleles in these tumors. These observationsare not predicted by either Knudson's or Comings' originalmodels (1-3). Modification of these models to include differ-ential genome imprinting, however, gives rise to a usefulmodel within which to interpret these data.

Ifa tumor-suppressor allele were subject either to mutationor epigenetic inactivation by passage through the male germline (Fig. 3), then any cell that contained such an allele wouldbe functionally hemizygous at the affected locus. Thus, ineither a genetically or an epigenetically caused hemizygote,only one additional event would be required to produce a nullphenotype.The model in Fig. 3 is similar to Knudson's original model

(1) in that both models require two events. The first event isthe inactivation of one allele at a tumor-suppressor locus andthe second event is the inactivation or loss of the remainingfunctional allele. However, the model in Fig. 3 differs fromKnudson's model in three respects: (i) the predicted maplocation of predisposing mutations in some inherited cancers,(ii) the preferential retention of paternal chromosomes intumor tissue, and (iii) the nature of the initial allele-inactivation event.Knudson's model has generally been interpreted to imply

that an alteration in the nucleotide sequence of a tumor-suppressor locus has occurred as the first event, but it doesnot detract from the conceptual validity of the model if thefirst event is epigenetic (44). The molecular mechanism bywhich epigenetic allele inactivation is achieved is not known.Several studies indicate that DNA methylation may be in-volved (21, 24, 45), but such correlative findings are notuniversal (22, 23, 30, 46) and are not crucial to the predictionsof the model in Fig. 3.

If both genetic and epigenetic inactivation events arepossible, the existence of two classes of familial tumors ispredicted. In the first class (1), tumor-suppressor alleles thatcarry alterations in nucleotide sequence will be inherited asthe predisposing mutation. In such families, the disease willbe genetically linked to markers on the chromosome thatcarries the tumor-suppressor allele. The sex of the parentfrom whom the defective allele is inherited is not predicted tohave an effect. This appears to be the case for some retino-blastoma families (1, 13), familial adenomatous polyposis(10), and bilateral acoustic neurofibromatosis (47). In thesecond class, an epigenetically inactivated tumor-suppressorallele will be inherited as the predisposing mutation. Becausethe inactivation of this allele need not be dependent on theallele itself but will reflect the activity of the gene or genesinvolved in generating or maintaining the genome imprint, theinheritance of the tumor phenotype will not be linked to thetumor-suppressor locus (Fig. 3).

A2 3 4 5 6 m mt tIE

XT06 w t~~~ Ra 9 10 11 12oat410~cl, 8

n ' b

breast ni./ breast 11 12 13 14

tumorof1 oat cellt breast a~t~ bb6p 6

3RHABD2 RHABD

BO0O0

kb _ ___w _a _--

am_

_ _.@

7482 Genetics: Scrable et al.

Proc. Natl. Acad. Sci. USA 86 (1989) 7483

Rd QRd

inactivationby mutation

0

somatic

Rd 0 rd mosaicism Rd [ Rd

normal

rd E rd L rd rd 1 [ rd

heterozygous hemizygous isodisomictumor tumor tumor

inactivation byI

unlinked IMP activity6?

_.9

somaticmosaicism Rd' Rd

d? d dd

Rd4 rd Rdi Rd4 Rd'

heterozygous hemizygous isodisomictumor tumor tumor

FIG. 3. Model of tumorigenesis invoking genomic imprinting as an alternative first step in the attainment of nullizygosity at therhabdomyosarcoma locus (Rd). Ei, Active locus; s or *, inactive locus; Rd, wild-type locus; rd, mutant locus; IMP, imprinting.

Although we could not rigorously test this prediction in thepresent study, Grundy et al. (48) and Huff et al. (49) haveanalyzed Wilms tumor pedigrees for linkage between diseasepredisposition and markers on chromosome 11. Despitestrong cytogenetic evidence for the chromosome lp locationof the Wilms tumor locus, no linkage could be demonstratedbetween disease predisposition and any marker on chromo-some lp in either of these studies. Such results are unex-pected from Knudson's original model but are predicted bythe modification proposed in Fig. 3.These two classes of mutations, allele inactivation by

nucleotide sequence alteration and allele inactivation bygenome imprinting, are also expected in sporadic cases.While it is not possible to make a definitive statementregarding the relative importance of each class of mutation tothe generation of human tumors, the nonrandom retention ofpaternal alleles in several pediatric tumors is highly signifi-cant. These include 15 out of 17 Wilms tumors (refs. 7, 28-30,and 50; unpublished data), 6 out of 6 embryonal rhabdomyo-sarcomas (this study), and 9 out of 10 osteosarcomas (43). Foreach of these tumors, we calculated the likelihoods of sampleconfigurations as extreme or more so than the observed databy using a binomial distribution with equal probability ofmaternal or paternal origin. The probabilities were approxi-mately 0.002, 0.03, and 0.02, respectively.The model in Fig. 3 is a direct extension ofprevious models

used to explain the gamete-of-origin-dependent behavior ofimprinted alleles in the mouse (15, 18, 46). A conceptuallysimilar model has been invoked to describe the preferentialretention of paternal alleles in Wilms tumor (50), but thismodel fails to make the distinction between genes thatimprint and genes that are imprinted, and therefore fails topredict the lack of linkage between predisposition to thedisease and markers on chromosome lip. Instead, the modelrequires a chromosome lp location for a Wilms-specifictrans-regulator or dominant Wilms transforming gene. Thismodel appears to be ruled out by the data ofGrundy et al. (48)and Huff et al. (49).Because alleles that carry an imprint are presumed to be

inactivated as a result of the imprinting process (15), themodel in Fig. 3 has no requirement that such an allele befurther mutated in order to exhibit a null (tumorigenic)phenotype. While it is possible that the generation or main-tenance of an imprint renders an allele more likely to sufferan alteration in nucleotide sequence (43, 51), one must

presume that this is not always the case. A strong argumentin favor of this contention is that genome imprints in themouse are reversible simply by passage of imprinted allelesthrough gametogenesis of the opposite sex. In addition, thespontaneous regression and mixed normal cell/tumor cellhistology ofsome tumors are observations that are difficult torationalize within the confines of a model invoking gamete-of-origin dependent mutational differences (43, 51) but thatare easily accommodated within the framework of an epige-netic inactivation model (Fig. 3).

Published observations on the parental origin of chromo-some lp alleles in a familial Wilms tumor (48) and chromo-some 1 alleles in a pheochromocytoma from a multipleendocrine neoplasia type 2 family (52) are also consistentwith this model. In both of these reports, the isodisomicchromosomes present in tumor tissue were derived from theunaffected father rather than from the affected mother. Thesedata imply that the predisposing mutation in these familiesmay act in trans on a tumor-suppressor allele in a nonrandomfashion with respect to parental origin. These data, as well asdata reported for the mouse (46, 53), imply that the genomeimprints observed in somatic tissue are not finally establisheduntil after fertilization (46, 53) and that affected individualsare mosaics of cells that bear imprinted alleles and cells thatdo not. The frequency with which such mosaics are createdmay be influenced by stochastic factors, resulting in sporadicdisease, and/or genetic factors, resulting in inherited disease(15, 46).

If gamete-of-origin-specific imprints in humans affect mul-tiple loci, as they do in mice (54), then inheritance of anaberrant imprinting allele that influences the allocation ofcells with imprinted loci to the embryonic lineage, or thecreation of such cells from functionally diploid cells, mightresult in the cosegregation of more than one type of diseasetrait (15, 55). Thus, the present modification of the originalmodels (1-3) of predisposition to human cancer invokes bothgenetic and epigenetic changes at critical loci and may servewell as a framework within which to understand this vitalprocess.

We gratefully acknowledge the support of Dr. Beatrice Lampkin,Director, Hematology/Oncology, Children's Hospital Medical Cen-ter, Cincinnati, Ohio, who has continuously provided us with tumortissue used in these studies. We are grateful to Karen Peterson forunpublished data and to Drs. Karen Arden, Paul Grundy, DavidJames, Alan Peterson, Ross McGowan, Lois Mulligan, and Karen

Genetics: Scrable et al.

Alfik

Proc. Natl. Acad. Sci. USA 86 (1989)

Peterson for critical comments on the manuscript. We acknowledgethe assistance of Drs. Bradley Rodgers, R. Beverly Raney, andKaren Bringelsen, as well as Nancy Kashlak, Pat Walka, andMarianne Brown, for assistance in obtaining specimens from patientsand their families. We thank Drs. G. Bell, G. Bruns, D. Housman,J. Hoppener, M. Jansen, H. Kazazian, H. Mayer, R. Weinberg, andR. White for recombinant DNA probes. We thank Linda Sapienzaand Robert Derval for artwork. M.L. is a Scholar of the College ofAmerican Pathologists.

1. Knudson, A. G., Jr. (1971) Proc. Natl. Acad. Sci. USA 68,820-823.

2. Knudson, A. G., Jr., & Strong, L. C. (1972) J. Natl. CancerInst. 48, 313-324.

3. Comings, D. E. (1973) Proc. Natl. Acad. Sci. USA 70, 3324-3328.

4. Cavenee, W. K., Dryja, T. P., Phillips, R. A., Benedict,W. F., Godbout, R., Gallie, B. L., Murphree, A. L., Strong,L. C. & White, R. L. (1983) Nature (London) 305, 779-784.

5. Koufos, A., Hansen, M. F., Lampkin, B. C., Workman,M. L., Copeland, N. G., Jenkins, N. A. & Cavenee, W. K.(1984) Nature (London) 309, 170-172.

6. Orkin, S. H., Goldman, D. S. & Sallan, S. E. (1984) Nature(London) 309, 172-174.

7. Reeve, A. E., Housiaux, P. J., Gardner, R. J. M., Chewings,W. E., Grindley, R. M. & Millow, L. J. (1984) Nature (Lon-don) 309, 174-176.

8. Fearon, E. R., Vogelstein, B. & Feinberg, A. P. (1984) Nature(London) 309, 176-178.

9. Lundberg, C., Skoog, L., Cavenee, W. K. & Nordenskjold, M.(1987) Proc. Natl. Acad. Sci. USA 84, 2372-2376.

10. Solomon, E., Voss, R., Hall, V., Bodmer, W. F., Jass, J. R.,Jeffreys, A. J., Lucibello, F. C., Patel, I. & Rider, S. H. (1987)Nature (London) 328, 616-619.

11. Naylor, S. L., Johnson, B. E., Minna, J. D. & Sakaguchi,A. Y. (1987) Nature (London) 329, 451-454.

12. Scrable, H. J., Witte, D. P., Lampkin, B. C. & Cavenee,W. K. (1987) Nature (London) 329, 645-647.

13. Strong, L. C., Riccardi, V. M., Ferrell, R. E. & Sparkes, R. S.(1981) Science 213, 1501-1503.

14. Balaban, G. H., Gilbert, F., Nichols, W., Meadows, A. T. &Shields, H. J. (1982) Cancer Genet. Cytogenet. 6, 213-221.

15. Sapienza, C. (1989) Ann. N.Y. Acad. Sci., in press.16. Cattanach, B. M. & Kirk, M. (1985) Nature (London) 315,

496-498.17. Searle, A. G. & Beechey, C. V. (1985) in Aneuploidy, eds.

Dellarco, V. L., Voytek, P. E. & Hollaender, A. (Plenum, NewYork), pp. 363-376.

18. McGrath, J. & Solter, D. (1984) Nature (London) 308, 550-551.19. McGrath, J. & Solter, D. (1984) Cell 37, 179-183.20. Surani, M. A. H., Barton, S. C. & Norris, M. L. (1984) Nature

(London) 308, 548-550.21. Swain, J. L., Stewart, T. A. & Leder, P. (1987) Cell 50,

719-727.22. Reik, W., Collick, A., Norris, M. L., Barton, S. C. & Surani,

M. A. H. (1987) Nature (London) 328, 248-251.23. Sapienza, C., Peterson, A. C., Rossant, J. & Balling, R. (1987)

Nature (London) 328, 251-254.24. Hadchouel, M., Farza, H., Simon, D., Tiollias, P. & Pourcel,

C. (1987) Nature (London) 329, 454-456.25. Szulman, A. E. & Surti, U. (1984) in Human Trophoblast

Neoplasms, eds. Patello, R. A. & Hussan, P. 0. (Plenum, NewYork), pp. 135-145.

26. Enzinger, F. & Weiss, S. (1983) Soft Tissue Tumors (Mosby,Saint Louis), pp. 338-378.

27. Miller, R. W., Fraumeni, J. F. & Manning, M. D. (1964) N.Engl. J. Med. 270, 922-927.

28. Schroeder, W. T., Chao, L.-Y., Dao, D. D., Strong, L. C.,Pathak, S., Riccardi, V., Lewis, W. H. & Saunders, G. F.(1987) Am. J. Hum. Genet. 40, 413-420.

29. Mannens, M., Slater, R. M., Heyting, C., Bliek, J., de Kraker,J., Coad, N., de Pagter-Holthuizen, R. & Pearson, P. L. (1988)Hum. Genet. 81, 41-48.

30. Williams, J. C., Brown, K. W., Mott, M. G. & Maitland, N. J.(1989) Lancet i, 283-284.

31. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) MolecularCloning:A Laboratory Manual (Cold Spring Harbor Lab., ColdSpring Harbor, NY).

32. Feinberg, A. D. & Vogelstein, B. (1983) Anal. Biochem. 132,6-13.

33. Shih, C. & Weinberg, R. (1982) Cell 29, 161-169.34. Bell, G. I., Karam, J. H. & Rutter, W. J. (1981) Proc. Natl.

Acad. Sci. USA 78, 5759-5763.35. Antonarakis, S. E., Boehm, C. S., Giardina, P. J. & Kazazian,

H. H. (1982) Proc. Natl. Acad. Sci. USA 79, 137-141.36. Barker, D., Holm, T. & White, R. L. (1984) Am. J. Hum.

Genet. 36, 1159-1171.37. Glaser, T., Gerhard, D., Payne, C., Jones, C. & Housman, D.

(1985) Cytogenet. Cell Genet. 40, 643.38. Jansen, M., van Schaik, F. M. A., van Tol, H., Van den

Brande, J. L. & Sussenbach, J. S. (1985) FEBS Lett. 179,243-246.

39. Schmidtke, J., Pape, B., Krengel, U., Langenback, U., Coop-er, D. N., Brayel, E. & Mayer, H. (1984) Hum. Genet. 67,428-431.

40. Gerhard, D. S., Bruns, G. A. & Housman, D. E. (1987) Cyto-genet. Cell Genet. 46, 619.

41. Hoppener, J. W. M., Steenbergh, P. H., Zandberg, J., Bakker,E., Pearson, P. L., Geurts Van Kessel, A. H. M., Jansz, H. S.& Lips, C. J. M. (1984) Hum. Genet. 66, 309-312.

42. Lathrop, G. M., Lalouel, J. M., Julien, C. & Ott, J. (1984)Proc. Natl. Acad. Sci. USA 81, 3443-3446.

43. Toguchida, J., Kitshizaki, M., Sasaki, S., Nakamura, Y.,Ikenaga, M., Kato, M., Sugimot, M., Kotoura, Y. & Yama-muro, T. (1989) Nature (London) 338, 156-158.

44. Moolgavkar, S. H. & Knudson, A. G. (1981) J. Natl. CancerInst. 66, 1037-1052.

45. deBustros, A., Nelkin, B. D., Silverman, A., Ehrlich, G.,Boiesz, B. & Baylin, S. B. (1988) Proc. Natl. Acad. Sci. USA85, 5693-5697.

46. Sapienza, C., Tran, T. H., Paquette, J., McGowan, R. &Peterson, A. (1989) Prog. Nucleic Acid Res. Mol. Biol. 36,145-157.

47. Wertelecke, W., Rouleau, G. A., Superneau, D. W., Farehard,L. W., Williams, J. P., Haines, J. L. & Gusella, J. F. (1988) N.Engl. J. Med. 315, 278-283.

48. Grundy, P., Koufos, A., Morgan, K., Li, F. P., Meadows,A. T. & Cavenee, W. K. (1988) Nature (London) 336, 374-376.

49. Huff, V., Compton, D. A., Chao, L.-Y., Strong, L. C., Geiser,C. F. & Saunders, G. F. (1988) Nature (London) 336, 377-378.

50. Wilkins, R. J. (1988) Lancet, i, 329-331.51. Reik, W. & Surani, M. A. (1989) Nature (London) 338, 112-

113.52. Mathew, C. G. P., Smith, B. A., Thorpe, K., Wong, Z., Royle,

N. J., Jeffries, A. J. & Ponder, B. A. J. (1987) Nature (Lon-don) 328, 524-526.

53. Sanford, J. P., Clark, H. J., Chapman, V. M. & Rossant, J.(1987) Genes Dev. 1, 1039-1046.

54. Solter, D. (1988) Annu. Rev. Genet. 22, 127-146.55. Li, F. P. & Fraumeni, J. F. (1982) J. Am. Med. Assoc. 247,

2692-2694.

7484 Genetics: Scrable et al.