Expression of Transforming Growth Factor-a inHepatoblastoma

Andras Kiss, M.D.

Agota Szepesi, M.D.

Gabor Lotz, M.D.

Peter Nagy, M.D.

Zsuzsa Schaff, M.D.

First Institute of Pathology and Experimental Can-cer Research, Semmelweis University of Medicine,Budapest, Hungary.

Presented in part at the 32nd Annual Meeting ofthe European Association for Study of the Liver,London, United Kingdom, April 9–12, 1997.

Supported in part by National Science Foundationgrants T016077 and 023579.

Address for reprints: Zsuzsa Schaff, M.D., FirstInstitute of Pathology and Experimental CancerResearch, Semmelweis University of Medicine, Ul-loi Str. 26, H-1085 Budapest, Hungary.

Received October 14, 1997; revision received Feb-ruary 4, 1998; accepted February 4, 1998.

BACKGROUND. Transforming growth factor-a (TGF-a) is a potent stimulator of cell

proliferation in the liver and in liver tumors; however, its significance and associ-

ation with hepatocyte proliferation remains unclear.

METHODS. Expression of TGF-a and proliferation markers, such as proliferating cell

nuclear antigen (PCNA) and cyclin A, were studied and correlated with each other

in samples of tumor and surrounding liver tissue taken from nine patients with

hepatoblastoma. An avidin-biotin-peroxidase immunohistochemical method was

used for detection of TGF-a, PCNA, and cyclin A, and in situ hybridization was

used to detect TGF-a mRNA.

RESULTS. Two types of tumor cells of epithelial origin were distinguished based on

the expression of TGF-a protein and RNA. The more differentiated “fetal” pheno-

type had a high expression of TGF-a and correlated with a low expression of

proliferation markers. The less differentiated “embryonal” phenotype had low

TGF-a expression and high proliferation activity.

CONCLUSIONS. The expression of TGF-a is associated with a certain morphologic

phenotype of tumor cells in hepatoblastoma; higher expression can be detected in

more differentiated tumor cells. The negative correlation between the expression

of TGF-a and proliferation markers suggests that the less differentiated embryonal

cells do not depend on growth stimulation provided by TGF-a. Cancer 1998;83:

690 –7. © 1998 American Cancer Society.

KEYWORDS: hepatoblastoma, transforming growth factor-a, cell proliferation,cyclin A.

Hepatoblastoma (HB) is a distinctive tumor of the liver in infantsand children that consists of cells resembling hepatocytes of

varying degrees of maturation. Morphologically, HB resembles theliver cells of the embryo and fetus.1–3 It is a rare tumor: The annualincidence is 1 in 1,000,000 children,4 yet it is the most commonpediatric primary malignancy of the liver.

HBs can contain several cell and tissue components of epithelialand mesenchymal origin and are classified accordingly as epithelial,mixed (epithelial and mesenchymal), and not otherwise specified.1,4

Based on the epithelial components, different patterns (or subclassi-fications) have been distinguished as fetal, embryonal,1 macrotrabe-cular,4 and small cell or anaplastic5 types by using routine histologicalstainings and electron microscopy. It has been demonstrated that thehistomorphologic appearance of certain types of HB is associatedwith differences in the cytokeratin pattern and different expression ofcertain oncofetal antigens.6,7

Transforming growth factor-a (TGF-a) is one of the major intra-hepatic growth factors. It is a polypeptide that stimulates the prolif-eration of epithelial tissue and may play a role in neoplastic transfor-

690

© 1998 American Cancer Society

mation.8 –11 It is likely that most if not all types ofnormal epithelial cells synthetize TGF-a,12–14 similarto several malignant tumors.15–18

Overexpression of TGF-a was found in the regen-erating liver,19,20 in liver tumors in rats,21,22 and inbenign23 and malignant23–25 liver tumors in humans.Increased expression of TGF-a was detected in hepa-titis B virus (HBV)-infected cells in colocalization withHBV surface antigen (HBsAg).23,24 Expression ofTGF-a might be related to cell proliferation and/ordifferentiation.25,26

The proliferation rate of the different epithelialcell types can be estimated by the expression of cellcycle-related proteins, such as proliferating cell nu-clear antigen (PCNA) and cyclin A. PCNA, the auxiliaryprotein of DNA polymerase d, which is involved inDNA replication, is a widely accepted proliferationmarker in histology.27 The PCNA level is very low in G0

and early G1 cells and increases to a maximum in theS-phase, when it becomes associated to the DNA rep-lication sites and stays high in continuously prolifer-ating cells.28

Cyclin A protein is required for the onset of DNAreplication in all mammalian cells, and it is also in-volved in the regulation of mitosis.29 Cyclin A proteinis absent in resting cells. Its synthesis begins shortlybefore the S-phase, and its level remains high until theend of mitosis, when it is degraded.30 The protein is

detectable in the cell nuclei, and it is a reliable indi-cator that proliferating cells are in the S-G2-M phasesof the cell cycle.31 The goal of this study was to inves-tigate the expression of TGF-a in human HBs withspecial reference to different cell types and prolifera-tive activity.

MATERIALS AND METHODSPatientsNine cases of HB were studied (Table 1). The patientswere treated in different hospitals between 1977 and1994 and were all life-long residents of Hungary. Themean age was 22.5 months (range from 2 months to 11years) at the time of diagnosis, and the patients com-prised two males and seven females. Patient sera weretested for HBsAg and antibody to the hepatitis B coreantigen (anti-HBc; Ausria II; Abbott Laboratories, Chi-cago, IL) and anti-hepatitis C virus (Ortho) since 1990(six cases). All sera tested were negative for the viralmarkers.

No cirrhosis was noted in any of the patients. Thetumors initially were suspected because of increasingabdominal circumference or from physical examina-tion, and they were confirmed by abdominal ultra-sound and computed tomography. Fine-needle aspi-ration biopsy was performed in two patients, surgicalbiopsy and removal of the whole tumor were per-formed in seven cases after obtaining the consent of

TABLE 1Clinical Data, Histopathology, and Expression of TGF-a, PCNA, and Cyclin A in Hepatoblastoma

Cases Sex/ageType ofbiopsy Histological type

Expression of TGF-a

Immunohistochemistry In situ hybridization

PCNAimmunohistochemistry

(LI)

Cyclin Aimmunohistochemistry

(LI)

Tumor

Livera

Tumor

Livera

Tumor

Livera

Tumor

LiveraFetal Embr Fetal Embr Fetal Embr Fetal Embr

1 M/11 yrs Needle Epithelial 11 1/2 NA 11 1/2 NA 0.1 0.9 NA 0.01 0.1 NA2 F/2 mos Surgical Mixed 11 1/2 1/2 11 1/2 2 0.1 0.85 0.001 0.05 0.3 0.13 M/19 mos Surgical Epithelial

(clear cell)11 NA 11 11 NA 2 0.05 NA 0 0.0025 NA 0

4 F/2 mos Needle Mixed 111 1/2 NA 11 1/2 NA 0.05 0.9 NA 0.002 0.1 NA5 F/7 mos Surgical Epithelial 111 1/2 1/2 11 2 2 0.3 0.9 0 0.02 0.074 06 F/11 mos Surgical Epithelial

(clear cell)11 NA 1 NA NA NA 0.05 NA 0.001 0.002 NA 0

7 F/1 yr Surgical Epithelial 111 2 1 11 2 2 0.04 0.85 0 0.005 0.05 08 F/1 yr Surgical Epithelial 111 1/2 1 NA NA NA 0.05 0.9 0.001 0.005 0.1 09 F/4 mos Surgical Epithelial 111 1/2 1 NA NA NA 0.1 0.9 0 0.01 0.1 0Normal

liver2 2 0 0

TGF-a: transforming growth factor-a; PCNA: proliferating cell nuclear antigen; LI: labeling index; embr: embryonal; NA: material not available.a Surrounding, tumor free liver.

TGF-a in Hepatoblastoma/Kiss et al. 691

the guardians. A small amount of surrounding nontu-morous liver was available in seven cases. Three pa-tients (cases 1, 2, and 8) died, and the other six arealive and tumor free, having been checked in 1997.The range of survival of the six patients who are aliveis 3–10 years (mean, 5.3 years).

ControlsNegative control tissue was from a normal liver thatwas obtained from a patient from the United States, a1-year-old white female who died of causes unrelatedto the liver (kindly provided by the Liver Tissue Pro-curement and Distribution System, a service contractof the National Institute of Health, Bethesda, MD).Normal human epidermis from a patient without liverdisease (kindly provided by Dr. D. E. Kleiner, NationalCancer Institute, National Institutes of Health) wasused as a positive control for TGF-a.

Histology and ImmunohistochemistryThe tissue was fixed immediately after removal in 10%buffered Formalin and embedded in paraffin. Sectionsmeasuring 3–5 mm were stained with hematoxylin-eosin, diastase/periodic acid-Schiff, Mallory, andPrussian blue stains to establish the diagnosis.

Immunohistochemistry was performed on depar-affinized 3–5 mm sections. The primary antibodieswere a monoclonal antihuman TGF-a (AB-2; OncogenScience, Inc., Manhasset, NY) in a 1:800 dilution, amonoclonal anti-HBsAg (Medix Biotech Inc., FosterCity, CA) in 1:100 dilution, a monoclonal anti-PCNA(clone PC10; Dako, Glostrup, Denmark) in 1:2000 di-lution, and a monoclonal anti-cyclin A (NovocastraLaboratories, Newcastle, United Kingdom) in 1:100dilution. An avidin-biotin-peroxidase complex tech-nique (Vectastain ABC Elite Kit; Vector Laboratories,Burlingame, CA) was used for the detection of theimmunohistochemical reaction. The following treat-ments were necessary before applying the primaryantibodies: 1) The endogeneous peroxidase activitywas blocked with 0.3% H2O2 in methanol (in all cases);2) three minute microwave treatments of the slideswere performed in a 0.1 M citrate buffer, pH 6.0,before applying the anti-PCNA and anti-cyclin A; 3)sections were digested with proteinase K [2 mg/mL inphosphate-buffered saline (PBS), pH 7.4; BoehringerMannheim, Penzberg, Germany) for 10 minutes be-fore applying the anti-TGF-a. The incubation with theprimary antibodies was performed overnight at 4°C.The peroxidase activity was revealed after incubationwith a 0.1% 3,39-diaminobenzidine tetrahydrochloride(Polyscience Inc., Warrington, PA) solution containing0.03% H2O2. Sections were then counterstainedslightly with hematoxylin.

The results of the immunohistochemical stainingfor TGF-a were evaluated as 11/21/31 based on theintensity of the reaction read by three independentobservers. Nuclear staining of the PCNA and cyclin Aantibodies was evaluated in at least 1000 cells in fivehigh-power fields (340). Hepatocytes with definitereddish-brown staining were considered positive. Thenuclear labeling index (LI) was calculated as the ratioof positive hepatocyte nuclei and the total number ofcells counted.

In Situ HybridizationThe in situ hybridization was performed as describedpreviously.32 In brief, after permeabilization with PBSwith 5 mM MgCl2 and 0.2 M Tris with 0.1 M glycine,the sections were digested with 1 mg/mL Proteinase K.Prehybridization for 4 hours was followed by 16 hoursof hybridization at 45°C in the presence of 35S-labelledriboprobe. The sections were washed at increasingstringency, processed for autoradiography, and ex-posed for 3 weeks. The result of the autoradiographywas evaluated by light- and darkfield microscopy. A310 base-pair fragment of the 59 end of the TGF-a

cDNA was used as a probe subcloned in pGEM3 vector(Promega), as described previously.11 Serial sectionswere hybridized with the sense strand of the appliedRNA probe as negative control.

RESULTSThe nine cases of HB were classified as epithelial ormixed types by histology (Table 1). The epithelial partsof the tumors consisted of embryonal and fetal tumorcells (Fig. 1). In two cases, only the fetal cells (clear cellvariant) were present. All nine HB cases stained pos-itively for TGF-a by immunohistochemistry (Fig. 2)compared with the normal control liver, which wasnegative or very slightly stained for TGF-a. Samplesfrom surrounding tumor free livers were available inseven cases. They stained less intensely than the fetaltumor cells in six cases and stained equally in one. Inthe liver samples surrounding the tumors, the stainingintensity was stronger than in the embryonal cells inthree cases and was equal in two cases.

Furthermore, there was a difference in the TGF-a

immunostaining among the areas of the same tumor.Groups of cells corresponding to fetal type cells on theroutinely hematoxylin and eosin stained slides (Fig. 1)produced a very intense immunohistochemical reac-tion, whereas large areas or foci consisting of embry-onal type cells stained weakly or not at all for TGF-a

(Fig. 2). The immunohistochemical reaction for TGF-a

was observed only in the epithelial cell components ofthe tumors; the mesenchymal cells were always neg-ative.

692 CANCER August 15, 1998 / Volume 83 / Number 4

TGF-a mRNA detection by in situ hybridizationwas accomplished in six cases. Signals were detectedin the epithelial tumor cells in all six cases available.The density of the grains indicating TGF-a expression,however, was different within the tumors; it was verystrong in areas of fetal type cells, whereas it was muchweaker or negative in embryonal type tumor cells (Fig.3). The differences in expression of the TGF-a tran-scripts can be demonstrated even better by darkfieldmicroscopy (Fig. 4). This distribution of TGF-a mRNAcorresponded well to the immunohistochemical reac-

tions. The areas showing high expression of TGF-a byimmunohistochemistry (Fig. 2) and in situ hybridiza-tion (Figs. 3,4) were found to match those areas of thetumor occupied by fetal type cells (Fig. 1).

PCNA and cyclin A immunostaining was positivein the cell nuclei of the tumor cells; however, a signif-icant difference in the LI of both antibodies was de-tected between the embryonal and fetal types of tu-mor cells. The PCNA LI of the embryonal type tumorcells was 0.89 6 0.024 (mean 6 standard deviation;SD) and was 0.093 6 0.081 (mean 6 SD; P , 0.001) in

FIGURE 1. More differentiated fetal cells (F) and less

differentiated embryonal cells (E) in hepatoblastoma (case

7; H & E, 3200).

FIGURE 2. Transforming growth factor-a (TGF-a) im-

munohistochemical staining. The fetal cells (F) are

strongly positive in contrast to the weak staining of the

embryonal cells (E; 3200).

TGF-a in Hepatoblastoma/Kiss et al. 693

the fetal type tumor cells (Fig. 5). The same differencewas observed comparing the percentage of cyclin A-positive nuclei in the different areas of the tumors(Fig. 6): The cyclin A LI of the embryonal cells was0.12 6 0.015 (mean 6 SD), whereas that of the fetalcells was 0.012 6 0.08 (mean 6 SD; P , 0.002). ThePCNA positivity was very low (.0.1%), and the cyclinA positivity was negative in the surrounding tumorfree liver (Table 1).

In the normal control liver sections, no reaction oronly very weak staining of TGF-a was seen by immu-

nohistochemistry. A very weak signal was detected byin situ hybridization, and the normal liver tissue wasnegative for PCNA and cyclin A. The epithelial com-ponents of the skin, used as positive control, stainedstrongly for TGF-a.

DISCUSSIONHepatoblastoma, like Wilms’ tumor of the kidney,contains epithelial components in different stages ofmaturation.6 In addition, HB contains other types oftissues as mesenchymal components, osteoid mate-

FIGURE 3. In situ hybridization with a TGF-a probe.

Large numbers of grains are located in the fetal cells (F),

whereas very low numbers are located in embryonal cells

(E; H & E, 3200).

FIGURE 4. In situ hybridization with a TGF-a probe

from the same field of view shown in Figure 3 by

darkfield microscopy. The brighter area represents the

fetal (F) type of tumor cells, and the dark area represents

the embryonal (E) type of tumor cells (3200).

694 CANCER August 15, 1998 / Volume 83 / Number 4

rial, and neuroepithelium, and it has been distin-guished clearly from hepatocellular carcinoma as adifferent entity.1 Two basic histologic types of HB havebeen described: epithelial and mixed.1 The latter con-sists of epithelial and mesenchymal elements, whichmay be of predictive value.3 The prognostic value ofhistologic typing seems to be based on the epithelialcomponents.33 However, others have not found suchdifferences in the prognosis.34 More extended studieshave demonstrated that the stage of HB at the time of

presentation is a more significant factor than the his-tologic type or DNA content.4

A further distinction has been made based on theepithelial components of HB into fetal or embryonaltype cells.1 Clear and dark cell variants of the fetal typehave been observed that are associated partly with theglycogen content; however, the physiological signifi-cance of these phenotypes is not known.1 Embryonaltype cells of HB are less differentiated, poorly defined,and more irregular than fetal type cells of HB.1

FIGURE 5. Immunostaining for proliferating cell nu-

clear antigen. Ninety percent of the nuclei of the embry-

onal cells (E) are positive compared with the few positive

cells in the area occupied by the fetal cells (F; 3200).

FIGURE 6. Immunostaining for cyclin A. Several nuclei

(approximately 10%) stained positively in the embryonal

area (E) compared with the fetal area (F; 3200).

TGF-a in Hepatoblastoma/Kiss et al. 695

Our study demonstrated a high expression ofTGF-a protein and mRNA in the fetal type cells of HB,whereas the embryonal type cells showed low levels ofor no TGF-a protein and mRNA expression. However,a “mirror image” was detected by using proliferationmarkers, such as PCNA and cyclin A. The stronglyTGF-a-positive fetal cells had much lower proliferativeactivity than the embryonal cells, which expressed lowlevels of TGF-a and showed a high LI of proliferationmarkers.

The high expression of TGF-a in fetal type cells ofHB can be regarded as a sign of differentiation. TGF-a

is highly expressed in prenatal hepatocytes from em-bryonic day 20,20 being the highest at birth and thendecreasing progressively.35 Usually, only a very lowlevel of TGF-a is detected in normal liver,23,24 which isin agreement with our recent observation. It is higherin regenerating liver20 and in several tumors, includingliver carcinoma.23,24 The low expression of TGF-a inthe embryonal type cells of HB might be explained bythe less differentiated phenotype in contrast to themore differentiated fetal cells. A similar observationwas made concerning TGF-a expression in hepatocel-lular carcinoma among tumor areas with varying de-grees of differentiation.25

Our data showing that the increased TGF-a pro-duction and the highly proliferating cell population inHB are different can be explained by different mech-anisms. TGF-a produced by the fetal cells may stim-ulate the proliferation of the embryonal cells in anendocrine fashion. It is more likely, however, that theless differentiated embryonal cells do not depend ongrowth stimulation provided by TGF-a in contrast tothe fetal cells, which might be regulated by TGF-a inan autocrine way.

Our data demonstrate that the expression ofTGF-a is associated with a certain morphologic phe-notype or functional state of cells in HB. The “fetal”and “embryonal” type cells—as they have been namedby Ishak and Glunz1— can be differentiated clearlyfrom each other based on their expression of TGF-a,which is associated conversely with proliferation. Fur-ther studies are needed to throw light on whether thehigher expression of TGF-a is a significant factor intumor growth, a possible autocrine stimulator, or onlya “marker” signaling the stage of differentiation of thetumor cell population.

REFERENCES1. Ishak GK, Glunz RP. Hepatoblastoma and hepatocarcinoma

in infancy and childhood. Report of 47 cases. Cancer 1967;20:396 – 422.

2. Gonzalez-Crussi F, Manz JH. Structure of a hepatoblastomaof pure epithelial type. Cancer 1972;29:1272– 80.

3. Gonzalez-Crussi F, Upton MP, Maurer HS. Hepatoblastoma:

attempt at characterization of histologic subtypes. Am J SurgPathol 1982;6:599 – 612.

4. Conran MR, Hitchcock LC, Waclawiw AM, Stocker JT, IshakGK. Hepatoblastoma: the prognostic significance of histo-logic type. Pediatr Pathol 1992;12:167– 83.

5. Kasai M, Watanabe I. Histologic classification of liver-cellcarcinoma in infancy and childhood and its clinical evalu-ation: a study of 70 cases collected in Japan. Cancer 1970;25:551– 63.

6. Pascual A, Manivel CJ, Wick RM, Hagen K, Dehner PL.Hepatoblastoma: an immunohistochemical and ultrastruc-tural study. Hum Pathol 1987;18:1025–35.

7. Van Eyken P, Sciot R, Callea F, Ramaekeres F, Schaart G,Desmet JV. A cytokeratin-immunohistochemical study ofhepatoblastoma. Hum Pathol 1990;21:302– 8.

8. Aaronson SA. Growth factor and cancer. Science 1991;254:1146 –53.

9. Hoffmann B, Paul D. Basic fibroblast growth factor andtransforming growth factor-a are hepatotrophic mitogens invitro. J Cell Physiol 1990;142:149 –54.

10. Ismail T, Howl J, Wheatley M. Growth of normal humanhepatoocytes in primary culture: effect of hormones andgrowth factors on DNA synthesis. Hepatology 1991;14:1076 –82.

11. Jhappen C, Stahle C, Harkins RN. TGFa overexpression intransgenic mice induces liver neoplasia and abnormal de-velopment of mammary gland and pancreas. Cell 1990;61:1137– 46.

12. Coffey RJ, Derynck R, Wilcox JN. Production and autoinduc-tion of transforming growth factor-a in human keratino-cytes. Nature 1987;328:817–20.

13. Gottlieb AB, Chang CK, Posnett DN. Detection of transform-ing growth factor-a in normal, malignant and hyperprolif-erative human keratinocytes. J Exp Med 1988;167:670 –5.

14. Liscia DS, Merlo G, Giardiello F. Transforming growth fac-tor-a messenger RNA localization in the developing adultrat and human mammary gland by in situ hybridization.Dev Biol 1990;140:123–31.

15. Muller W, Borchard F. Expression of transforming growthfactor-a in gastric carcinoma and normal gastric mucosacells. Cancer 1992;69:2871–5.

16. Derynck R, Goeddel DV, Ullrich A. Synthesis of messengerRNAs for transforming growth factors receptor by humantumors. Cancer Res 1987;47:707–12.

17. Mydlo JH, Michaeli J, Cordon-Cardo C. Expression of trans-forming growth factor a and epidermal growth factor recep-tor messenger RNA in neoplastic and nonneoplastic humankidney tissue. Cancer Res 1989;49:3407–11.

18. Smith JJ, Derynck R, Korc M. Production of transforminggrowth factor-a in human pancreatic cancer cells: evidencefor a superagonist autocrine cycle. Proc Natl Acad Sci USA1987;84:7567–70.

19. Castilla A, Prieto J, Fausto N. Transforming growth fac-tors-b1 and a in chronic liver disease. N Engl J Med 1991;324:933– 40.

20. Evarts RP, Nakatsukasa H, Marsden ER. Expression of trans-forming growth factor-a in regerating liver and during he-patic differentiation. Mol Carcinogen 1992;5:25–31.

21. Mead JE, Fausto N. Transforming growth factor-a may be aphysiological regulator of liver regeneration by means of anautocrine mechanism. Proc Natl Acad USA 1989;86:1558 –62.

22. Derynck R. Transforming growth factor-a. Cell 1988;54:593–5.

696 CANCER August 15, 1998 / Volume 83 / Number 4

23. Schaff ZS, Hsia CC, Sarosi I, Tabor E. Overexpression oftransforming growth factor-a in hepatocellular carcinomaand focal nodular hyperplasia from European patients. HumPathol 1994;25:644 –51.

24. Hsia CC, Axiotis CA, Di Bisceglie AM. Transforming growthfactor-a in human hepatocellular carcinoma and coexpres-sion with hepatitis B surface antigen in adjacent liver. Can-cer 1992;70:1049 –56.

25. Morimitsu Y, Hsia CC, Kojiro M, Tabor E. Nodules of less-differentiated tumor within or adjacent to hepatocellularcarcinoma: relative expression of transforming growth fac-tor-a and its receptor in the different areas of tumor. HumPathol 1995;26:1126 –32.

26. Masuhara M, Yasunaga M, Tanigawa K, Tamura F, Ya-mashita S, Sakaida I, et al. Expression of hepatocyte growthfactor, transforming growth factor-a, and transforminggrowth factor-b1 messenger RNA in various human liverdiseases and correlation with hepatocyte proliferation.Hepatology 1996;24:323–9.

27. Hall AP, Levison AD, Woods LA, Yu W-CC, Kellock BD,Watkins AJ, et al. Proliferating cell nuclear antigen (PCNA)immunolocalization in paraffin sections: an index of cellproliferation with evidence of deregulated expression insome neoplasms. J Pathol 1990;162:285–94.

28. Morris FG, Mathews BM. Regulation of proliferating cellnuclear antigen during the cell cycle. J Biol Chem 1989;264:13856 – 64.

29. Walker HD, Maller LJ. Role for cyclin A in the dependence ofmitosis on completion of DNA replication. Nature 1991;354:314 –7.

30. Zindy F, Lamas E, Chenivesse X, Sobczak J, Wang J, FesquetD, et al. Cyclin A is required in phase in normal epithelialcells. Biochem Biophys Res Comm 1992;182:1144 –54.

31. Girald F, Strausfeld U, Fernandez A, Lamb NJ. Cyclin A isrequired for the onset of DNA replication in mammalianfibroblasts. Cell 1991;67:1169 –79.

32. Nagy P, Bisaard HG, Thorgeirsson SS. Expression of hepatictranscription factors during liver development and oval celldifferentiation. J Cell Biol 1994;126:223–33.

33. Haas EJ, Muczynski AK, Krailo M, Ablin A, Land V, ViettiJT, et al. Histopathology and prognosis in childhoodhepatoblastoma and hepatocarcinoma. Cancer 1989;64:1082–95.

34. Schmidt D, Wischinger P, Leuschner I, Sprenger E, Lan-genau E, Schweinitz von D, et al. DNA analysis in hepa-toblastoma by flow and image cytometry. Cancer 1993;72:2909 –14.

TGF-a in Hepatoblastoma/Kiss et al. 697