Discrimination of Complete Hydatidiform MoleFrom Its Mimics by Immunohistochemistry of thePaternally Imprinted Gene Product p57KIP2

Diego H. Castrillon, M.D., Ph.D., Deqin Sun, M.S.,Stanislawa Weremowicz, Ph.D., Rosemary A. Fisher, Ph.D.,Christopher P. Crum, M.D., and David R. Genest, M.D.

The p57KIP2 protein is a cell cycle inhibitor and tumor sup-pressor encoded by a strongly paternally imprinted gene. Weexplored the utility of p57KIP2 as a diagnostic marker in hyda-tidiform mole, a disease likely the result of abnormal dosageand consequent misexpression of imprinted genes. Using amonoclonal antibody on paraffin-embedded, formalin-fixed tis-sue sections, the authors evaluated p57KIP2 expression in nor-mal placenta and in 149 gestations including 59 complete hy-datidiform moles, 39 PHMs, and 51 spontaneous losses withhydropic changes. p57KIP2 was strongly expressed in cytotro-phoblast and villous mesenchyme in normal placenta, all casesof partial hydatidiform moles (39 of 39) and all spontaneouslosses with hydropic changes (51 of 51). In contrast, p57KIP2

expression in cytotrophoblast and villous mesenchyme was ab-sent or markedly decreased in 58 of 59 complete hydatidiformmoles. In all gestations p57KIP2 was strongly expressed in de-cidua and in intervillous trophoblast islands, which served asinternal positive controls for p57KIP2 immunostaining. p57KIP2

immunohistochemistry can reliably identify most cases of com-plete hydatidiform mole irrespective of gestational age and isthus a useful diagnostic adjunct, complementary to ploidyanalysis, in the diagnosis of hydatidiform mole.Key Words: Hydatidiform mole—Genomic imprinting—Immunohistochemistry—p57KIP2.

Am J Surg Pathol 25(10): 1225–1230, 2001.

The histopathologic diagnosis of hydatidiform mole isdifficult, with significant interobserver variation in thedistinction between complete hydatidiform mole (CHM),partial hydatidiform mole (PHM), and spontaneous abor-

tion (SA).5,17 Although histologic criteria are used todistinguish CHM from PHM in most cases, histologicfeatures can have significant overlap, especially in earlygestations,12 and SAs can exhibit hydropic changes ortrophoblast hyperplasia that mimic hydatidiform mole.20

These difficulties have intensified in recent practice be-cause of the increasing use of prenatal screening tech-niques such as �-human chorionic gonadotropin (hCG)measurements and ultrasonography that permit earlierclinical recognition and termination of abnormal preg-nancies including hydatidiform moles.18,21 Because mostCHMs are diploid whereas most PHMs are triploid,ploidy analysis by flow cytometry or other methods is auseful diagnostic aid to identify PHMs15; however,ploidy analysis cannot be used to distinguish CHM fromnonmolar gestation. Analysis of polymorphic short tan-dem DNA repeats by polymerase chain reaction (PCR)can identify CHM and PHM but is technically cumber-some and difficult to standardize.2,8 Because CHMs areassociated with a high incidence of persistent disease andchoriocarcinoma whereas PHMs behave in a malignantfashion very infrequently,22 the lack of an adjunct test toconfirm the diagnosis of CHM is a significant limitationin current practice.

CHMs are derived exclusively from paternal DNA,whereas PHMs contain one maternally derived and twopaternally derived haploid genomes,11,16,26 suggestingthat both CHM and PHM are the result of abnormalexpression of imprinted genes.6 While the vast majorityof genes are expressed equally from maternal and pa-ternal copies, approximately 40 human genes areimprinted-expressed from only the paternal or maternalallele, with the other allele transcriptionally silent. Ge-nomic imprinting involves differential DNA methylationof maternal and paternal alleles leading to allele-specificexpression through complex mechanisms currently underinvestigation. The p57KIP2 gene is strongly paternally

From the Department of Pathology, Brigham and Women’s Hospitaland Harvard Medical School, Women’s and Perinatal PathologyDivision (D.H.C., D.S., C.P.C., and D.R.G.), and the CytogeneticsDivision (S.W.), Boston, Massachusetts, and the Department of CancerMedicine, Imperial College School of Medicine and Charing CrossHospital (R.A.F.), London, U.K.

Address correspondence and reprint requests to Diego H. Castrillon,MD, PhD, Department of Pathology, Brigham and Women’s Hospital,75 Francis Street, Boston MA 02115, U.S.A.; e-mail:dcastrillon@partners.org

The American Journal of Surgical Pathology 25(10): 1225–1230, 2001 © 2001 Lippincott Williams & Wilkins, Inc., Philadelphia

1225

imprinted, being expressed predominantly from the ma-ternal allele in most tissues.24 Because CHMs containonly paternal DNA, p57KIP2 should be underexpressed inCHMs. In contrast, because all other types of gestationsincluding PHMs contain maternal DNA, the pattern ofp57KIP2 expression should be essentially normal in non-CHMs. Recently, one study presented immunohisto-chemical evidence that p57KIP2 is differentially ex-pressed in molar versus nonmolar gestations.4 The cur-rent report describes the results of a comprehensive studydesigned to assess the value of p57KIP2 immunohisto-chemistry in the diagnosis of a wide range of molar andnonmolar gestations.

MATERIALS AND METHODS

Case Selection

Cases of CHM, PHM, and SA with hydropic changeswere randomly selected from the files of the PathologyDepartment of the Brigham and Women’s Hospital(1995–2000). Histologic evaluation of all cases was per-formed on routine sections stained with hematoxylin andeosin; diagnoses were made by gynecologic pathologistsusing published criteria.12,19 The diagnosis in each case(CHM, PHM, or SA) for the purpose of this study wastaken from the original pathology report. Ancillary stud-ies (flow cytometry on fresh tissue or karyotyping) hadbeen performed in the original diagnostic workup of 40%of cases diagnosed as hydatidiform moles. The gesta-tional age for CHMs ranged from 5 to 22 weeks and was<10 weeks in more than half of the cases; 24 of 59 CHMswere confirmed diploid or tetraploid by flow cytometryor cytogenetic analysis, whereas 35 CHMs were notevaluated. Included in our 59 cases of CHM were twoinvasive moles confirmed after hysterectomy and 10 re-currences in patients with previously diagnosed CHM.The gestational age for PHMs ranged from 7 to 19weeks; 16 of 39 PHMs were confirmed to be triploid,whereas 23 PHMs were not evaluated. Cases identifiedas SA were specimens noted to have hydropic change inthe original pathology report. Of the 51 cases of SA,eight cases had karyotypic abnormalities involvingsingle chromosome gains or losses. Mature and imma-ture placentas (three cases each) were randomly selectedamong normal term gestations or elective terminations.

Antibodies and Immunohistochemistry

Mouse monoclonal antibodies against the p57KIP2

protein were obtained from NeoMarkers/Lab VisionCorporation (Fremont, CA, USA). We used a ready-to-use preparation without dilution (catalog no. MS-1062-R7)or a concentrated preparation (catalog no. MS-1062-P1)

at 1:200 dilution with similar results. Only formalin-fixed, paraffin-embedded tissue was studied. Antigen re-trieval was performed at 93° in 10 mM sodium citratebuffer pH 6.0 for 30 minutes with a 10-minute cool-down. The detection system was StreptABC (DakoCorporation, Carpinteria, CA, USA) with diaminobenzi-dine as the chromogen. The slides were counterstainedwith hematoxylin. Only distinct nuclear staining wasscored as positive. For all cases the presence or absenceof nuclear staining was assessed in five cell types (vil-lous mesenchyme [VM], cytotrophoblast [CT], syncytio-trophoblast [ST], intervillous trophoblast, and decidua),blinded to the original diagnosis and independent of thehematoxylin and eosin histologic appearance.

DNA Analysis and Fluorescence InSitu Hybridization

DNA analysis and ploidy determination were per-formed on one case of CHM, as described in Results.Sections were cut from a block of paraffin-embedded,formalin-fixed tissue. An unstained 6-�m section wasoriented by referring to a paired consecutive sectionstained with hematoxylin and eosin, and areas of molarvilli and maternal decidua were identified. DNA wasthen prepared from this tissue, amplified by the PCR, andanalyzed as described.7 The microsatellite repeats ana-lyzed were D1S1656, D5S816, D7S440, D9S43,D10S179, D15S659, and D18S535 (Human GenomeDatabase, Toronto, Ontario, Canada: Hospital for SickChildren; Baltimore, MD, USA: Johns Hopkins Univer-sity, http://www.gdb.org). Fluorescent in situ hybridiza-tion of interphase nuclei from suspensions prepared from50-�m paraffin sections was performed essentially asdescribed.13

RESULTS

p57KIP2 Expression in Normal Placenta, SA WithHydropic Change, and PHM

In normal placenta p57KIP2 was strongly expressed inCT, VM, and in intervillous trophoblast islands (IVTI),whereas the ST was uniformly negative. In both imma-ture (first trimester) and mature (term) placenta, a sig-nificant proportion of CT, IVTI, and VM nuclei (at least30%) were strongly p57KIP2 positive (Fig. 1A, B). Wealso evaluated an early gestational sac (approximately 3weeks), which can be mistaken for a cavitated molarvillus. In this early gestation the VM was stronglyp57KIP2 positive; staining was weak in CT and absent inST (Fig. 1C). In fragments of gestational endometriumsecretory glands were negative, but many stromal (i.e.,decidual) cells were strongly positive, in roughly equal

D. H. CASTRILLON ET AL.1226

Am J Surg Pathol, Vol. 25, No. 10, 2001

intensity to that observed in the CT, IVTI, and VM(Fig. 1D).

In all SAs (51 cases) and all PHMs (39 cases), thestaining pattern was similar to that seen in normal pla-centa, with strong p57KIP2 expression in the CT and VM(Fig. 1E, F). In areas of stromal edema the density ofpositive VM cells was lower than in nonedematous areas,but the proportion of positive nuclei was unchanged. InPHMs two discrete populations of villi (sclerotic or

edematous) are usually present. p57KIP2 staining wassimilar in PHM villi irrespective of their morphology(not shown).

p57KIP2 Expression in CHM

In contrast to both PHM and SA, almost all cases ofCHM exhibited a striking lack of p57KIP2 expression inboth CT and VM (Fig. 1G). In the majority of CHM villi

FIG. 1. Examples of p57KIP2 immu-nostaining in molar and nonmolargestations and decidua, counter-stained with hematoxylin. (A) Imma-ture placenta (first trimester). Mostnuclei in the CT layer (arrow) werestrongly positive, as were a signifi-cant proportion of villous mesenchy-mal nuclei (arrowhead). IVTIs werestrongly and diffusely positive (rightthird of panel). In contrast, the ST(asterisk) was negative. (B) Matureplacenta at term with strong CT (ar-row) and VM (arrowhead) p57KIP2

positivity. (C) Villus from very earlygestational sac (2–3 weeks). Manymesenchymal cells (arrowheads)were strongly positive. (D) Maternaldecidua. A significant proportion ofdecidual cells were strongly p57KIP2

positive (arrowheads), whereas en-dometrial glands (asterisks) werenegative. Decidua can thus serve asan internal positive control forp57KIP2 immunostaining. (E) Hy-dropic spontaneous abortus. As innormal placenta, strong p57KIP2

positivity was observed in CT (ar-rows) and VM (arrowheads). (F)PHM. The pattern of p57KIP2 immu-nostaining was indistinguishablefrom that of nonmolar gestations,with strong CT (arrow) and VM (ar-rowheads) positivity. (G) CHM (15-week gestation). Most villi werenegative for p57KIP2 immunostainingin VM (arrowhead) and CT (arrow).Strong immunoreactivity in IVTIswas retained in complete moles(right third of panel), serving as aninternal positive control. (H) CT posi-tivity in a complete mole. Occasionalvilli in some cases contained posi-tive CT cells, typically only 1–3 cells.In this villus (the most extreme ex-ample among our cases) approxi-mately 12 positive cells (arrows)were present. The absence of

p57KIP2 positivity in villous mesenchymal cells (arrowheads) in this and other villi, however, was indicative of a completemole. (I) Early complete mole (5-week gestation). Lack of CT (arrows) and VM (arrowhead) positivity was also charac-teristic of early complete moles. Internal positive controls were present elsewhere on the slide. (J) Single complete molethat exhibited p57KIP2 positivity (10-week gestation). A strong signal was observed in CT (arrow) and VM (arrowheads).This staining pattern was indistinguishable from that observed in all partial moles and SAs.

DISCRIMINATION OF COMPLETE HYDATIDIFORM MOLE 1227

Am J Surg Pathol, Vol. 25, No. 10, 2001

expression of p57KIP2 in VM and CT was virtually ab-sent. However, focal CT positivity was present in someCHMs, with occasional villi exhibiting a small numberof positive CT cells (1–3 cells in most cases but up to 12cells in one villus in one case) (Fig. 1H). Despite this, theoverall scarcity of p57KIP2 positive CT cells in CHMcompared with PHM or SA (<0.1% of CT vs 30%) ap-pears to readily permit the discrimination of CHM fromPHM or SA in most instances. p57KIP2 was equally di-agnostic in early CHM (Fig. 1I) and other CHM variantsincluding invasive CHM and recurrent/persistent CHM(not shown). Notably, p57KIP2 CT and VM expressionwas absent or dramatically decreased even in CHM villiwith minimal trophoblast hyperplasia or stromal changes(e.g., Fig. 1I). In all cases most IVTI nuclei were stronglypositive and together with maternal decidua served as aninternal positive control for p57KIP2 immunostaining(Fig. 1G, left third of panel). In summary (with the ex-ception of the single case described below), CHM couldbe readily distinguished from PHM or SM by evaluationof p57KIP2 expression in the CT and VM.

The difference in p57KIP2 CT and VM positive immu-nostaining in CHM (1 of 59) versus non-CHM (90 of 90)in our cases was highly statistically significant (p value� 10−41, Fisher exact test) (summarized in Table 1). Oneof 59 cases diagnosed as CHM exhibited diffuse VM andCT positivity, indistinguishable from the staining patternof PHM and SA (Fig. 1J). This pregnancy, which repre-sented the mother’s third, was terminated at a gestationalage of 10 weeks; there was no antecedent history ofhydatidiform mole. All slides for this case were re-viewed, and the histologic features including significanttrophoblast hyperplasia and diffuse villous cavitationconfirmed CHM. Flow cytometry on fresh tissue hadindicated a diploid DNA content, consistent with aCHM. To confirm this result, we performed fluorescencein situ hybridization using probes corresponding to chro-mosomes X, Y, 8, 11, and 18. The results demonstratedthat the lesion was diploid for chromosomes X, 8, 11,and 18 (no Y signal was observed) again consistent witha CHM. To confirm that the lesion was androgenetic(and therefore a CHM), both the lesion (using DNA pre-pared from molar villi) and the patient (using DNA fromdecidual tissue) were genotyped for seven microsatellite

markers on different chromosomes. The lesion wasclearly androgenetic for four of the markers, having noalleles in common with the patient. Of the seven markersexamined, the CHM was homozygous for five markersbut heterozygous for two, indicating that this was a di-spermic CHM.

DISCUSSION

In CHMs, which lack nuclear DNA of maternal origin,strongly paternally imprinted genes should in principlebe significantly underexpressed or not expressed at all.Consistent with this view, we find that p57KIP2 is strik-ingly underexpressed in the CT and VM of CHM, but notin PHM and SA, and that this underexpression can serveas the basis for a reliable diagnostic test to identify CHM.Because PHMs have one set of maternally derived chro-mosomes, the underexpression of p57KIP2 in PHMshould be comparatively slight and probably not detect-able by immunohistochemistry. Although other pater-nally imprinted genes could serve as useful markers toidentify CHM, p57KIP2 seems especially well suited forthis purpose because: 1) it is strongly expressed in nor-mal placenta throughout gestation, and its staining pat-tern in CHM, PHM, and SA appears to be similarlyindependent of gestational age; and 2) internal positivecontrols for p57KIP2 immunostaining (IVTI and maternaldecidua) are always present in clinical specimens. Be-cause we evaluated a large number of cases and used acommercially available monoclonal antibody, our find-ings significantly expand upon an earlier study thatdocumented the expression of p57KIP2 in normal placen-tal cell types and suggested the potential utility ofp57KIP2 immunohistochemistry in gestational tropho-blastic disease.4

p57KIP2 is a potent cell cycle inhibitor and tumor sup-pressor, and lack of p57KIP2 activity can lead to loss ofcell cycle control and hyperproliferation.27 Therefore,underexpression of p57KIP2 likely contributes to themolar phenotype including histologic correlates suchas trophoblast hyperplasia. Consistent with this, theBeckwith–Wiedemann syndrome, which is associatedwith loss or point mutation of the maternal p57KIP2 al-lele, is characterized by fetal/placental overgrowth, pre-disposition to tumors, and placental abnormalities thatcan resemble those observed in hydatidiform mole.14

Furthermore, mouse embryos deficient for p57KIP2 ex-hibit placentomegaly with significant trophoblast hyper-plasia and dysplasia.23 However, the fact that patientswith Beckwith–Wiedemann syndrome do not developfrank hydatidiform moles strongly suggests that misregu-lation of other imprinted genes also contributes to themolar phenotype.

The expression of p57KIP2 in CHM IVTI, althoughuseful as an internal positive control, was unanticipated.

TABLE 1. Summary of p57KIP2 immunohistochemistryresults in different types of gestation

Type of gestation Positive CT and VM

Hydropic spontaneous abortion 52/52 (100%)Partial hydatidiform mole 39/39 (100%)Complete hydatidiform mole 1/59 (1.7%)

CT, cytotrophoblast; VM, villous mesenchyme.Total number of cases = 149.P value (CHM vs. non-CHM) <.000.

D. H. CASTRILLON ET AL.1228

Am J Surg Pathol, Vol. 25, No. 10, 2001

This expression likely reflects incomplete or relaxed im-printing of p57KIP2, as some imprinted genes are notexpressed exclusively from a single allele in all tis-sues.3,25 For example, the human GRB10 gene is im-printed in a highly isoform- and tissue-specific manner.In fetal brains most GRB10 variants are transcribed ex-clusively from the paternal allele, whereas in skeletalmuscle one splice variant is expressed from the maternalallele, and in other tissues all splice variants are trans-ribed from both parental alleles.3 Intriguingly, anotherpaternally imprinted gene expressed in placenta, H19(which encodes a nontranslated RNA), is expressed in asimilar pattern as p57KIP2 in normal placenta and is alsoexpressed in IVTI in CHM.1 The alternative explanationfor the observed expression of p57KIP2 in CHM IVTI,that of cross-reactivity of the monoclonal antibody witha related protein such as p27KIP1 or an unrelated protein,seems unlikely because of the very limited extent ofsimilarity between p57KIP2 and p27KIP1 at the amino acidlevel and because the staining in IVTI is nuclear, asexpected for p57KIP2. Similarly, the expression ofp57KIP2 observed in rare CT of CHM may also be due toincomplete or relaxed imprinting.

Of 59 cases diagnosed as CHM, only one exhibitedp57KIP2 expression in a significant proportion of CT andVM. Interestingly, the p57KIP2 staining pattern in thiscase was completely unlike all other CHMs in our seriesand indistinguishable from PHM and SA. It is clearly nota PHM, as the lesion was found to be diploid by twoseparate methodologies, and DNA analysis confirmedthat the lesion was androgenetic and therefore a CHM.The underlying basis for p57KIP2 expression in this CHMremains unclear. Although the lesion is diploid for acentromeric chromosome 11 probe by fluorescent in situhybridization (the p57KIP2 locus maps to chromosome11p15.5), it remains formally possible that a maternalchromosomal fragment containing the 11p15.5 regionwas retained, leading to p57KIP2 expression from a ma-ternal copy of the locus. Perhaps more likely, p57KIP2

was, for unknown reasons, abnormally imprinted in theontogenesis of this lesion. It is also notable that thislesion is XX and dispermic because 90% of CHMs aremonospermic and only about 5% of CHMs are XX anddispermic.9

Our results indicate that although p57KIP2 is a highlyspecific and sensitive marker of CHM, false-negativeresults are possible. The results of p57KIP2 immunostain-ing should thus be interpreted in conjunction with mor-phologic features and all other available information. Weidentified no false positives in our series, but additionalstudies will be required to determine the true specificityand sensitivity of p57KIP2 immunostaining. Becausep57KIP2 immunostaining is highly accurate in the diag-nosis of hydatidiform mole, confirmation of p57KIP2

false positives and false negatives by other methods in-cluding ploidy and DNA analysis will be necessary.

The monoclonal antibody against p57KIP2 used in thisstudy worked reliably on paraffin-embedded, formalin-fixed tissue, with strong immunoreactivity of internalpositive controls in the 149 archival cases studied. There-fore, this antibody appears to be well suited for use as adiagnostic test of clinical specimens as collected andprocessed in routine practice. p57KIP2 immunostaining,although complementary to flow cytometry, may offersome advantages. First, it does not require collection offresh tissue. Flow cytometry can be performed on par-affin sections,10 but the results can be difficult to inter-pret, particularly if abundant tissue of maternal origin ispresent. Second, although a triploid flow cytometry re-sult confirms the suspicion of a PHM, a diploid result canleave some doubt that contaminating maternal tissue hadbeen inadvertently sent for analysis. Because p57KIP2

immunostaining is an in situ technique performed onparaffin-embedded tissue, it largely avoids these prob-lems. Lastly, p57KIP2 positively identifies CHM, a lesionthat has a much higher risk of progression and chorio-carcinoma than PHM. In conclusion, our study showsthat p57KIP2 immunostaining is a practical and accurateadjunct in the diagnosis of hydatidiform mole. �

Addendum

Analysis of microsatellite DNA repeats in the singlep57KIP2 positive CHM has revealed that although it is diploidfor all markers on the 11q arm and other chromosomes (asexpected), at least one marker on chromosome 11p is triploid,suggesting that the aberrant expression of p57KIP2 in thisCHM is the result of retention of a portion of the maternalchromosome 11p containing the p57KIP2 locus.

Acknowledgments

The authors thank Cathy Quigley for expert help in immu-nohistochemical staining.

REFERENCES

1. Ariel I, Lustig O, Oyer CE, et al. Relaxation of imprinting introphoblastic disease. Gynecol Oncol 1994;53:212–9.

2. Bell KA, Van Deerlin V, Addya K, et al. Molecular genetic testingfrom paraffin-embedded tissue distinguishes nonmolar hydropicabortion from hydatidiform mole. Mol Diagn 1999;4:11–9.

3. Blagitko N, Mergenthaler S, Schulz U, et al. Human GRB10 isimprinted and expressed from the paternal and maternal allele in ahighly tissue- and isoform-specific fashion. Hum Mol Genet 2000;9:1587–95.

4. Chilosi M, Piazzola E, Lestani M, et al. Differential expression ofp57kip2, a maternally imprinted cdk inhibitor, in normal humanplacenta and gestational trophoblastic disease. Lab Invest 1998;78:269–76.

5. Conran RM, Hitchcock CL, Popek EJ, et al. Diagnostic consider-ations in molar gestations. Hum Pathol 1993;24:41–8.

6. Falls JG, Pulford DJ, Wylie AA, et al. Genomic imprinting: im-plications for human disease. Am J Pathol 1999;154:635–47.

DISCRIMINATION OF COMPLETE HYDATIDIFORM MOLE 1229

Am J Surg Pathol, Vol. 25, No. 10, 2001

7. Fisher RA, Khatoon R, Paradinas FJ, et al. Repetitive completehydatidiform mole can be biparental in origin and either male orfemale. Hum Reprod 2000;15:594–8.

8. Fisher RA, Newlands ES. Rapid diagnosis and classification ofhydatidiform moles with polymerase chain reaction. Am J ObstetGynecol 1993;168:563–9.

9. Fisher RA, Povey S, Jeffreys AJ, et al. Frequency of heterozygouscomplete hydatidiform moles, estimated by locus-specific minis-atellite and Y chromosome-specific probes. Hum Genet 1989;82:259–63.

10. Hedley DW, Friedlander ML, Taylor IW, et al. Method for analysisof cellular DNA content of paraffin-embedded pathological mate-rial using flow cytometry. J Histochem Cytochem 1983;31:1333–5.

11. Kajii T, Ohama K. Androgenetic origin of hydatidiform mole.Nature 1977;268:633–4.

12. Keep D, Zaragoza MV, Hassold T, et al. Very early completehydatidiform mole. Hum Pathol 1996;27:708–13.

13. Kuchinka BD, Kalousek DK, Lomax BL, et al. Interphase cyto-genetic analysis of single cell suspensions prepared from previ-ously formalin-fixed and paraffin-embedded tissues. Mod Pathol1995;8:183–6.

14. Lage JM. Placentomegaly with massive hydrops of placental stemvilli, diploid DNA content, and fetal omphaloceles: possible asso-ciation with Beckwith-Wiedemann syndrome. Hum Pathol 1991;22:591–7.

15. Lage JM, Popek EJ. The role of DNA flow cytometry in evaluationof partial and complete hydatidiform moles and hydropic abor-tions. Semin Diagn Pathol 1993;10:267–74.

16. McFadden DE, Kwong LC, Yam IY, et al. Parental origin oftriploidy in human fetuses: evidence for genomic imprinting. HumGenet 1993;92:465–9.

17. Messerli ML, Parmley T, Woodruff JD, et al. Inter- and intra-pathologist variability in the diagnosis of gestational trophoblasticneoplasia. Obstet Gynecol 1987;69:622–6.

18. Mosher R, Goldstein DP, Berkowitz R, et al. Complete hydatidi-form mole: comparison of clinicopathologic features, current andpast. J Reprod Med 1998;43:21–7.

19. Paradinas FJ, Browne P, Fisher RA, et al. A clinical, histopatho-logical and flow cytometric study of 149 complete moles, 146partial moles and 107 non-molar hydropic abortions. Histopathol-ogy 1996;28:101–10.

20. Redline RW, Hassold T, Zaragoza M. Determinants of villoustrophoblastic hyperplasia in spontaneous abortions. Mod Pathol1998;11:762–8.

21. Schorge JO, Goldstein DP, Bernstein MR, et al. Recent advancesin gestational trophoblastic disease. J Reprod Med 2000;45:692–700.

22. Seckl MJ, Fisher RA, Salerno G, et al. Choriocarcinoma and partialhydatidiform moles. Lancet 2000;356:36–9.

23. Takahashi K, Kobayashi T, Kanayama N. p57(Kip2) regulates theproper development of labyrinthine and spongiotrophoblasts. MolHum Reprod 2000;6:1019–25.

24. Taniguchi T, Okamoto K, Reeve AE. Human p57(KIP2) defines anew imprinted domain on chromosome 11p but is not a tumoursuppressor gene in Wilms tumour. Oncogene 1997;14:1201–6.

25. Weber M, Milligan L, Delalbre A, et al. Extensive tissue-specificvariation of allelic methylation in the Igf2 gene during mouse fetaldevelopment: relation to expression and imprinting. Mech Dev2001;101:133–41.

26. Zaragoza MV, Surti U, Redline RW, et al. Parental origin andphenotype of triploidy in spontaneous abortions: predominance ofdiandry and association with the partial hydatidiform mole. Am JHum Genet 2000;66:1807–20.

27. Zhang P, Liegeois NJ, Wong C, et al. Altered cell differentiationand proliferation in mice lacking p57KIP2 indicate a role in Beck-with-Wiedemann syndrome. Nature 1997;387:151–8.

D. H. CASTRILLON ET AL.1230

Am J Surg Pathol, Vol. 25, No. 10, 2001