immunostaining in a panel of 59 human germ cell tumors of diâ¬erent histological origins. We show that TDGF-1 expression is markedly elevated in a subset of ...
Oncogene (1997) 15, 927 ± 936 1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00
Expression of teratocarcinoma-derived growth factor-1 (TDGF-1) in testis germ cell tumors and its eects on growth and dierentiation of embryonal carcinoma cell line NTERA2/D1 Gustavo Baldassarre1, Annunciata Romano2, Franca Armenante1,5, Marco Rambaldi2, Iole Paoletti1, Claudia Sandomenico3, Stefano Pepe3, Stefania Staibano4, Gaetano Salvatore4, Gaetano De Rosa4, Maria Graziella Persico1 and Giuseppe Viglietto2 1
International Institute of Genetics and Biophysics, CNR, Via Marconi 12, 80125 Naples, Italy; 2Istituto Nazionale dei Tumori di Napoli, Fondazione Senatore Pascale, Via M. Semmola, 80131 Naples, Italy; 3Dipartimento di Endocrinologia ed Oncologia Sperimentale e Clinica and 4Istituto di Patologia, FacoltaÁ di Medicina e Chirurgia, UniversitaÁ Federico II di Napoli, Via Pansini, 80131 Naples, Italy
The teratocarcinoma-derived growth factor-1 (TDGF-1) gene codes for a 188-aminoacid glycoprotein that shares structural homology with the epidermal growth factor (EGF) family of growth factors. TDGF-1 is highly expressed in the undierentiated embryonal carcinoma stem cell line NTERA2 clone D1 (NT2/D1) and its expression is downregulated in response to dierentiating agents such as retinoic acid (RA) and hexamethylenbisacetamide (HMBA). To assess the role of TDGF-1 in the onset and/or progression of human germ cell tumors, we analysed TDGF-1 expression by Northern blot and immunostaining in a panel of 59 human germ cell tumors of dierent histological origins. We show that TDGF-1 expression is markedly elevated in a subset of human testicular germ cell tumors as compared to normal testes. TDGF-1 overexpression occurs in about 100% of tumors with non-seminomatous phenotype, such as embryonal carcinomas and malignant undierentiated teratocarcinomas. To address the questions of how TDGF-1 (previously called CRIPTO) may aect the growth and/or the dierentiation of embryonal carcinoma cells, we have characterized the eects of exogenous recombinant TDGF-1 protein on the proliferation rate and dierentiation potential of NT2/D1. Exogenous TDGF-1 protein stimulated DNA synthesis and cell proliferation in both undierentiated and dierentiated NT2/D1 cells. However, TDGF-1 protein treatment was unable to block dierentiation induced by both RA and HMBA. These results suggest that TDGF-1 growth factor may represent an autocrine growth factor that may be involved in the process of development of testicular neoplasms. Keywords: TDGF-1; teratocarcinoma; seminoma; cell dierentiation
Introduction Germ cell tumors (GCTs) of the testis represent a heterogeneous group of neoplasms that predominantly occur in young men. Their incidence has progressively Correspondence: MG Persico 5 Present address: Institute of Biological Chemistry, University of Verona, Strada Le Grazie, 8, 37134 Verona, Italy Received 21 March 1997; revised 7 May 1997; accepted 7 May 1997
increased in the twentieth century, since testis cancer is the most common carcinoma in the 15 ± 35 year-old age-group. Despite the advances in the managment of GCT, this disease and the accompanying treatment have a severe impact on such patients during the most productive years of their life. Histologically, GCTs comprise two main groups, seminomas and nonseminomas (Ulbright and Roth, 1994). The biological properties of these tumors make them a suitable system to study the relationship between dierentiation and transformation. However, the molecular basis of malignant transformation and expression of differentiated phenotypes by GCTs cells are poorly understood. Experimental evidence suggests that polypeptide growth factors may play a role in the growth of human tumors both in vitro and in vivo (Cross and Dexter, 1991; Salomon et al., 1995a). Growth factors have been grouped in a small number of families on the basis of their aminoacid sequence. One of these families comprises, among others, epidermal growth factor (EGF) (Carpenter and Cohen, 1990), transforming growth factor-a (TGF-a) (Derynck, 1988), amphiregulin (AR) (Plowman et al., 1990) and CRIPTO-1 (Ciccodicola et al., 1989). The TDGF-1 (previously named CRIPTO) gene has been isolated from the human embryonal carcinoma cell line NT2/D1 (Ciccodicola et al., 1989). It encodes a 2.0 kb mRNA expressed in undierentiated NT2/D1 cells (Ciccodicola et al., 1989). TDGF-1 can function as a dominant transforming gene. Overexpression of TDGF-1 full-length cDNA in the NOG-8 mouse mammary epithelial cell line (Ciardiello et al., 1991b) or NIH/3T3 ®broblasts (Ciccodicola et al., 1989) results in the in vitro transformation of these cells as it confers to TDGF-1transfected cells the ability to grow in soft agar. Complementary evidence of the causal involvement of TDGF-1 gene in tumor growth is furnished by Ciardiello and co-workers (Ciardiello et al., 1994) who have shown that TDGF-1 down-regulation by the use of antisense cDNA-expressing vector strongly reduces the anchorage-independent growth in vitro and tumor development in vivo of the colon carcinoma cell line GEO. TDGF-1 is overexpressed in 70% of human primary colorectal carcinomas, in 60% of hepatic metastasis from colon cancers (Ciardiello et al., 1991a) and in 75% of mammary carcinomas (Qi et
TDGF-1 in testis germ cell tumors G Baldassarre et al
al., 1994). Here we analyse the level of TDGF-1 expression in a panel of germ cell tumors. Sequence analysis of TDGF-1 reveals two distinct domains: a region of 6 cysteine residues (EGF-like) followed by a second cysteine rich sequence that do not present similarity to any known protein motif. The ®rst region can stimulate the proliferation of several cell types through an EGF receptor-independent pathway (Brandt et al., 1994). Here we demonstrate that TDGF-1 expression is associated with the poorly dierentiated tumorigenic phenotype of NT2/D1. Exogenous addition of recombinant protein containing only the EGF-like motif of TDGF-1 to undierentiated and to RA- or HMBA-dierentiated NT2/D1 cells is able to stimulate cell proliferation, but does not interfere with the dierentiation program induced by RA or HMBA.
Results TDGF-1 expression in human germ cell tumors We analysed 59 human testicular germ cell tumors of dierent histological types (32 seminomas, 19 nonseminomas and eight mature teratomas) and three normal testicular tissues for TDGF-1 gene expression by Northern blot hybridization. The results are summarized in the Table 1 and a representative experiment is reported in Figure 1. About 50% of the tumors analysed expressed high levels of TDGF-1 mRNA whereas no expression of TDGF-1 was found in normal testicular tissues. Interestingly, TDGF-1 was highly expressed in 100% (19/19) of the nonseminomas analysed whereas variable levels of TDGF1 mRNAs were detected in about 30% (10/32) of seminomas. No TDGF-1 expression was detected in mature teratomas. These ®ndings indicate that upregulation of TDGF-1 in GCTs is strictly related to the poorly dierentiated histological type and is associated with a higher degree of malignancy.
teratomas. In normal testicular tissue, no staining was obtained using anti-TDGF-1 antibodies, consistent with the results obtained by Northern blot analysis (Figure 2a and Table 1). TDGF-1 expression was detected in 9/9 non-seminomatous tumors analysed (Figure 2b and c), consistent with the results obtained by Northern blot hybridization. The analysis of tumor specimens with seminomatous phenotype revealed that the TDGF-1 protein was expressed only in anaplastic seminomas (Figure 2d and e). Finally, no expression of TDGF-1 protein was found in the four mature teratomas analysed (Figure 2f). Eects of dierentiating agents on TDGF-1 expression To verify whether the down regulation of TDGF-1 expression is associated with poorly dierentiated phenotype in testicular GCTs, as suggested by the analysis of tumor specimens, we used the teratocarcinoma-derived NT2/D1 cell line (Andrews, 1988) as a model system. NT2/D1 cells have been shown to dierentiate if treated with RA or HMBA. Recognizable neuronal cells appeared after 7 days of RA treatment (Andrews, 1984). Phorbol esters (TPA) also induce partial dierentiation in NT2/D1 cells, since TPA decreases the expression of the immunophenotypic dierentiation markers SSEA-3, a globo-series carbohydrate, and enhances the ganglio-
Normal Testis
928
Non seminomas 1
2
3
4
5
Seminomas 6 7
8
9 10 11 12
18S
18S
Immunostaining of normal and neoplastic tissues To analyse the tissue-distribution of the TDGF-1 protein, we performed immunohistochemical staining of sections obtained from normal and neoplastic testis specimens using anity-puri®ed antibodies against a TDGF-1 synthetic peptide (see Materials and methods). Immunoperoxidase staining was performed on two normal testicular tissues, 12 seminomatous tumors, nine non-seminomas and four well-dierentiated
Figure 1 Analysis of TDGF-1 gene expression in human germ cell tumors by Northern blot. Twenty micrograms of total RNA derived from normal testis and testicular neoplasms were loaded onto each lane. The position of 18S ribosomal RNA is shown. Upper panel: hybridization with TDGF-1 probe. Lane 1, normal testis; lanes 2 ± 4, embryonal carcinomas; lanes 5 and 6, teratocarcinomas; lane 7, anaplastic seminomas; lanes 8 ± 13, classic seminomas. Lower panel: hybridization of the same ®lter with the b-actin probe
Table 1 Analysis of TDGF-1 expression in testicular GCTs Tumor type Non-seminomas Embryonal Carcinomas Teratocarcinomas Choriocacinomas Seminomas Seminomas Anaplastic Seminomas Mature Teratomas Normal Testis a
Sample number
Northern Blota
Immunostainingb
19 11 6 2 32 24 8 8 3
19/19(100%) 11/11 6/6 2/2 10/32(31%) 2/24 8/8 0/8(0%) 0/3(0%)
9/9(100%) 6/6 3/3 1/1 4/12(35%) 1/9 3/3 0/4(0%) 0/2(0%)
Number of tumor samples of increased TDGF-1 expression compared to normal testes as detected by Northern blot. bNumber of tumor samples showing staining with TDGF-1 antibodies
TDGF-1 in testis germ cell tumors G Baldassarre et al
929
Figure 2 Analysis of TDGF-1 protein expression in normal testis and testicular germ cell tumors. Immunostaining analysis on dierent GCT samples was performed with anti-TDGF-1 puri®ed IgGs. (a) section of normal testis tissue negative to immunostaining. (b and c) two dierent non-seminoma samples positive to immunostaining (b embryonal carcinoma. c teratocarcinoma). (d) classical seminoma negative to immunostaining reaction. (e) anaplastic seminomas strongly positive to immunostaining. (f) mature cystic teratoma negative to immunostaining
series carbohydrate antigens GD2 and GD3 (Kurie et al., 1993). Furthermore, TPA cooperates with RA in inducing neuronal dierentiation of NT2/D1 cells. HMBA also induces dierentiation of NT2/D1 cells although the identity of the resulting cells remains as yet unclear. However, most of the HMBA-dierentiated NT2/D1 cells exhibit a dierent morphology compared to RA-dierentiated NT2/D1 cells (Andrews et al., 1990). A few neurons can sometimes be detected in HMBA-induced cultures, but they are much less prominent than in RA-induced cultures. Thus, it appears that distinct pathways of dierentiation are activated in NT2/D1 cells by RA, TPA and HMBA (Andrews, 1988). The eect of HMBA, RA and TPA on the TDGF-1 transcript is shown in Figure 3. A tenfold reduction in the steady-state level of TDGF-1 mRNA is detected
after 6 h of HMBA treatment (Figure 3a, lane 3) and a complete loss of TDGF-1 transcript is observed starting from one day of HMBA treatment (Figure 3a, lanes 4, 5 and 6). The 2.0 kb TDGF-1 transcript is detected at high levels in undierentiated, highly proliferating NT2/D1 cells (Figure 3b, lane 7). However, when cells are induced to dierentiate with 10 mM RA (Figure 3b, lanes 1, 3 and 5) a ®vefold reduction in the steady-state level of TDGF-1 mRNA is detected after 24 h of RA treatment (Figure 3b, lane 2 and Ciccodicola et al., 1989) and a complete loss of TDGF-1 transcript is observed starting from 3 days after RA treatment (Figure 3b, lanes 4 and 6). The mRNA expression of TDGF-1 gene was also downregulated by the activation of the protein kinase C (PKC) in NT2/D1 cells, since a fourfold reduction in the level of TDGF-1 transcript
TDGF-1 in testis germ cell tumors G Baldassarre et al
a
b
+HMBA 0
1h
6h
24h
48h
7d
RA
1d —
3d +
—
NT2-D1
930
7d +
—
+
18S
18S
18S
18S
c
d
6h
18h
+TPA 0
1.5h
3h
RA TPA
8h
— +
— —
+ —
+ +
+ —
+ +
18S
18S
18S
18S
Figure 3 Regulation of TDGF-1 mRNA expression by RA, TPA and HMBA in the EC NT2/D1 cell line. (a) Upper panel: lane 1, untreated NT2/D1; lanes 2 ± 6, NT2/D1 treated for 1 and 6 h and 1, 2 and 7 days with 5 mM HMBA. (b) Upper panel: lanes 1, 3, 5: NT2/D1 treated for 1, 3 and 7 days with 0.1% DMSO; lanes 2, 4, 6: NT2/D1 treated for 1, 3 and 7 days with 10 mM RA; lane 7, untreated NT2/D1. (c) Eect of TPA on TDGF-1 mRNA expression in the EC NT2/D1 cell line. Upper panel: lane 1, untreated cells; lanes 2, 3, 4: NT2/D1 cells treated for 1.5, 3 and 8 h with 100 ng/ml of TPA, respectively. (d) Eect of simultaneous activation of RA and PKC pathways on TDGF-1 mRNA expression in the EC NT2/D1 cell line. Upper panel: lane 1, NT2/D1 treated for 3 h with 100 ng/ml TPA; lane 2, untreated cells; lane 3, NT2/D1 cells treated for 6 h with 10 mM RA; lane 4, NT2/D1 cells treated for 6 h with 10 mM RA plus 100 ng/ml of TPA in the last 3 h; lane 5, NT2/D1 cells treated for 18 h with 10 mM RA; lane 6, NT2/D1 cells treated for 18 h with 10 mM RA plus ng/ml of TPA in the last 3 h. The lower panels show the same ®lters hybridized with a 32 P-labeled b-actin probe
was observed after TPA administration (100 ng/ml) (Figure 3c). Finally, cooperative eect of RA and TPA on the down-regulation of TDGF-1 expression was observed after 6 h (Figure 3d). These results show that the expression of TDGF-1 is strictly associated with the undierentiated state of NT2/D1 cells; however TDGF-1 expression is completely abolished when cells are induced to terminally dierentiate, suggesting that expression of TDGF-1 may be a marker of the undierentiated state of NT2/ D1 cells. Eects of recombinant GST-Tdgf-1 protein on the proliferation of undierentiated NT2/D1 cells Brandt and co-workers (1994) demonstrate that two dierent refolded 37- and 47-aminoacid long peptides, containing the EGF-like domain from human TDGF-1 protein are able to stimulate growth of normal or transformed human mammary epithelial cell lines in EGF-receptor independent manner (Brandt et al.,
1994). Thus, to determine the eects of TDGF-1 protein on the proliferation of undierentiated or dierentiating NT2/D1 cells, we constructed a recombinant GST-fusion protein (see Materials and methods) containing the EGF-like region of the murine Tdgf-1 protein (Dono et al., 1993). Previous reports showed that NT2/D1 cells are unable to grow in low serum concentration (Dmitrovsky et al., 1990a). In fact, the growth rate of these cells in medium containing 1 ± 2.5% FCS is strongly reduced (data not shown). To determine the eects of GST-Tdgf-1 protein on the proliferation of undifferentiated NT2/D1 cells, we performed a [3H]thymidine incorporation assay on undierentiated NT2/D1 cells in low serum medium. As shown in Figure 4a the addition of GST-Tdgf-1 fusion protein to undifferentiated NT2/D1 cells starved in low serum substantially increases the growth rate of NT2/D1 cells in a dose dependent manner, whereas cells treated with GST-protein alone showed no signi®cant changes in their thymidine uptake. Interestingly, the ability of the
TDGF-1 in testis germ cell tumors G Baldassarre et al
931
Figure 4 Eect of the EGF-like domain of TDGF-1 on the incorporation of thymidine and cell growth into undierentiated NT2/ D1 cells. (a) [3H]thymidine incorporation in NT2/D1 undierentiated cells. Approximately 5.06104 cells/well were plated into 24well cell culture plates as described in Materials and methods. The result expressed in percent increase over control of thymidine incorporation represent the average (+SD) of two experiments performed in quadruplicate. (b) Anchorage-dependent growth of undierentiated NT2/D1 cells, in absence (&) or presence of 1 (*) and 10 ng/ml (~) of GST-Tdgf-1 fusion protein or 10 ng/ml of GST protein alone (&). The amount of GST-Tdgf-1 fusion protein indicated corresponds only to the Tdgf-1 only. Approximately 7.06104 cells/well were plated into 6-well cell culture plates as described in Materials and methods. The results represent the average (+SD) of three experiments, each performed in duplicate
EGF-like domain of TDGF-1 fusion protein to increase [3H]thymidine incorporation is higher respect to EGF used at the same dose as a positive control. Anchorage-dependent growth experiments were performed to test the ability of the GST-Tdgf-1 fusion protein to support the proliferation of NT2/D1 cells. Undierentiated NT2/D1 cells were plated in 2.5% serum and in the presence of GST-Tdgf-1 fusion protein or control GST protein for 6 days. When treated with GST-Tdgf-1 fusion protein, undifferentiated NT2/D1 cells exhibited a dose dependent increase in growth rate in monolayer up to twofold as compared with untreated or GST-treated cells (Figure 4b). In summary, these ®ndings demonstrate that the EGF-like domain of TDGF-1 is mitogenic for undierentiated NT2/D1 cells, and suggest that TDGF-1 may represent an autocrine growth factor for this embryonal carcinoma cell lines. Eect of recombinant GST-Tdgf-1 protein on the proliferation of dierentiating NT2/D1 cells Miller and co-workers (1994) demonstrate that RA or HMBA aect dierentially the growth rate of NT2/D1 cells. In fact, RA-treated cells continued to grow at rate only slightly reduced respect to control, whereas HMBA-treatment resulted in a greater growth inhibition (Miller et al., 1994). Since HMBA caused a complete inhibition of cell growth during four days from the beginning of HMBA treatment (Miller et al., 1994 and data not shown), NT2/D1 cells were grown for 2 days in the presence of 5 mM HMBA, and subsequently supplemented with 1 and 10 ng/ml of recombinant GST-Tdgf-1 protein in fresh medium. A dose-dependent increase in [3H]thymidine incorporation was observed in NT2/D1 cells treated with TDGF-1
fusion protein. In fact, incorporation of 140% and 180% over control was obtained after treatment with 1 ng/ml or 10 ng/ml TDGF-1 protein, respectively, whereas no signi®cant increase in [3H]thymidine incorporation was obtained when only GST was added (Figure 5a). Similar results were obtained in a [3H]thymidine incorporation assay performed by simultaneous adding HMBA and GST-Tdgf-1 fusion protein to the medium (data not shown). In anchorage-dependent growth experiments, HMBA-treated NT2/D1 cells also showed a two- to threefold increase in growth rate when GST-Tdgf-1 fusion protein were added to the culture medium. NT2/ D1 cells were plated and grown in medium supplemented with 5 mM HMBA for 6 days in the presence of 1 or 10 ng/ml of GST-Tdgf-1 fusion protein or GST alone. After 6 days, cells treated with 10 ng/ml of GST-Tdgf-1 fusion protein were increased twofold respect to control or GST-treated cells (Figure 5b). Similar results were obtained when cells were grown for 4 days with HMBA in the absence of TDGF-1 protein and then treated with TDGF-1 protein in fresh medium (data not showed). We also tested the ability of GST-Tdgf-1 protein to stimulate the growth of RA-dierentiating NT2/D1 cells. After 6 days of RA-treatment NT2/D1 cells were plated at high density without RA and in presence of TDGF-1 fusion protein or GST at the concentration of 1 and 10 ng/ml. As shown in Figure 5c and d treatment with 1 or 10 ng/ml of TDGF-1 is able to stimulate the [3H]thymidine incorporation and the growth of RA-dierentiating NT2/D1 cells in a dosedependent manner. These ®ndings demonstrate that the EGF-like domain of TDGF-1 is able to exert mitogenic eects also on dierentiating NT2/D1 cells, suggesting that the addition of exogenous TDGF-1 may recruit in
TDGF-1 in testis germ cell tumors G Baldassarre et al
932
Figure 5 Eect of the EGF-like domain of TDGF-1 on the incorporation of thymidine and the monolayer growth of dierentiating NT2/D1 cells. (a) [3H]thymidine incorporation in NT2/D1 HMBA-dierentiating cells. Approximately 5.06104 cells/ well were plated into 24-well cell culture plates as described in Materials and methods. The result expressed in c.p.m./well represent the average (+SD) of two experiments each performed in quadruplicate. (b) Ancorage-dependent growth of HMBA-dierentiating NT2/D1 cells, in absence (&) or presence of 1 (*) and 10 ng/ml (~) of GST-Tdgf-1 fusion protein or 10 ng/ml of GST protein alone (&). The amount of GST-Tdgf-1 fusion protein indicated corresponds to the Tdgf-1 only. Approximately 2.06105 cells/well were plated into 6-well cell culture plates as described in Materials and methods. The result represent the average (+SD) of three experiments, each performed in duplicate. (c) [3H]thymidine incorporation in NT2/D1 RA-dierentiating cells. Approximately 5.06104 dierentiating cells/well were plated into 24-well cell culture plates as described in Materials and methods. The result expressed in c.p.m./well represent the average (+SD) of two experiments each performed in quadruplicate. (d) Ancorage-dependent growth of RA-dierentiating NT2/D1 cells, in absence (*) or presence of 1 (*) and 10 ng/ml (&) of GST-Tdgf-1 fusion protein or 10 ng/ml of GST protein alone (D). Approximately 2.06105 cells/well were plated into 12-well cell culture plates as described in Materials and methods. The result represent the average (+SD) of three experiments, each performed in duplicate
cycle cells that are committed to dierentiate but that are not yet terminally dierentiated. Alternatively, it may be possible that NT2/D1 cultures induced to dierentiate with RA or HMBA contain a fraction of cells which may retain the capability to respond to TDGF-1. Eects of recombinant GST-Tdgf-1 protein on the dierentiation of NT2/D1 cells We have demonstrated that recombinant TDGF-1 protein is able to stimulate the proliferation of undierentiated NT2/D1 cells. Our study shows that TDGF-1 protein is also able to enhance DNA synthesis and growth rate of dierentiating NT2/D1 cells. We next investigated whether TDGF-1 protein was able to interfere with RA- or HMBA-induced dierentiation of NT2/D1 cells. Several dierentiation markers recognized by speci®c antibodies have previously been shown to be dierentially expressed in undierentiated or in differentiated NT2/D1 cells (Andrews et al., 1990; Miller et al., 1994): SSEA-3 is a MAb that recognizes a glycolipid surface antigen that is expressed by the majority of
undierentiated NT2/D1 cells (Shevinsky et al., 1982); the A2B5 MAb binds to a neuron-speci®c cell surface molecule that is induced following RA treatment (Andrews, 1984); the VINIS-53 MAb recognizes an antigen associated to dierentiation induced by HMBA (Andrews et al., 1990). In order to determine the eect exerted by exogenous TDGF-1 protein on the differentiation program triggered by RA or HMBA on NT2/ D1 cells, we performed immuno¯uorescence and ¯ow cytometry analysis. NT2/D1 cells were induced to dierentiate with RA or HMBA for 6 days, in the presence of 10 ng/ml of GST-Tdgf-1 protein or GST. The dierentiation programs induced by RA or HMBA, in the presence of TDGF-1 was subsequently monitored by analysis of the immunophenotypic changes of SSEA3, A2B5 and VINIS-53 MAbs. As described in Figure 6a, expression of the SSEA-3 marker was diminished from 78% to 15% following HMBA treatment, whereas the VINIS-53 antigen raised from 1 ± 98%, and the A2B5 marker raised from 2 ± 18% of cells, thus showing that HMBA-treated NT2/D1 cells have dierentiated, but that fewer cells have undergone the neuronal pathway. However, TDGF-1 was unable to modify the terminal
TDGF-1 in testis germ cell tumors G Baldassarre et al
933
Figure 6 Flow cytometry analysis in HMBA- (a) and RA-dierentiated (b) NT2/D1 cells for expression of cell surface antigens detected by SSEA-3, A2B5 and VINIS-53 MAbs
dierentiation of NT2/D1 cells induced by HMBA, as shown by the expression of the dierentiation-speci®c immunophenotypic markers. Furthermore, following RA treatment, the SSEA-3 marker was diminished from the 78% to 25%, whereas the A2B5 marker raised from 2 ± 75% (Figure 6b), thus demonstrating that neuronal dierentiation occurred. As with HMBAtreated cells, treatment with 10 ng/ml of GST-Tdgf-1 fusion protein did not aect the dierentiation program induced by RA in NT2/D1 cells. These results suggest that although TDGF-1 enhances cell growth in dierentiating NT2/D1 cells, it is unable to antagonize the dierentiation action of RA or HMBA. Discussion In this paper we determined the expression of the TDGF-1 gene in testicular GCTs by Northern blot hybridization and immunostaining. We show that TDGF-1 gene expression is up-regulated in 50% of testicular GCTs analysed whereas TDGF-1 is not expressed in normal testicular tissues. However, the expression of TDGF-1 gene is limited to the poorly dierentiated testicular neoplasms since TDGF-1 transcript is detected in 100% of anaplastic seminomas (8/8), of embryonal carcinomas (11/11) and teratocarcinomas (6/6). Conversely, only 2/24 classic seminomas and 0/8 mature teratomas express TDGF-1 mRNA and protein. The ®nding that TDGF-1 is expressed preferentially in non-seminomatous tumors is of particular interest, since non-seminomas appear to be more clinically aggressive, and usually present a worse prognosis compared to seminomas, suggesting that TDGF-1 expression may be correlated with a higher malignant phenotype in germ cell tumors. Experimental evidences suggest that tumor development and progression are multistep processes due to the accumulation of genetic alterations, such as loss of tumor suppressor genes and/or activation of oncogenes
and growth factors (Salomon et al., 1995b). The molecular events underlying the development of testicular cancer are poorly characterized. The inactivation of the tumor suppressor DCC gene (Murty et al., 1994) localized on chromosome 18q21 or of an unidenti®ed tumor suppressor gene localized on chromosome 12q (Murty et al., 1992; Rodriguez et al., 1992) have been proposed as an early event. Moreover, mutation and/or overexpression of genes of the ras family could also be involved in GCTs tumorigenesis (Bos, 1989; Dmitrovsky et al., 1990b). The results shown in this study suggest that changes in the expression level of the TDGF-1 gene can also be added to the genes that might contribute to the onset or progression of testicular cancers. Moreover, our results support the idea that TDGF-1 up-regulation in testicular cancers may represent a late event during tumor progression since it occurs predominantly in anaplastic seminomas and undierentiated non-seminomas. Tumor cells generally exhibit a decreased requirement for exogenously-supplied growth factors when compared to their normal counterparts. This relaxation in growth factor dependency may be due in part to the ability of transformed cells to synthetize and respond to growth factors that can regulate their proliferation through autocrine, paracrine, juxtacrine or intracrine pathways (Sporn and Roberts, 1992). Accordingly, it has been suggested that dierentiation of the embrional carcinoma cell line NT2/D1 occurs in parallel with the loss of the mRNA expression of several growth factors, such as TGF-a, bFGF and k-FGF (Dmitrovsky et al., 1990a; Miller et al., 1994). Upon RA treatment, NT2/ D1 cells dierentiate into a neuronal lineage (Andrews, 1984) whereas upon HMBA treatment, NT2/D1 cells develop a distinct morphologic and antigenic phenotype (Andrews et al., 1990; Miller et al., 1994). Dierentiation of NT2/D1 induced by RA and HMBA is accompanied by arrest in the cell growth, reduced cloning eciency and loss of tumorigenicity in nude mouse (Andrews, 1988).
TDGF-1 in testis germ cell tumors G Baldassarre et al
934
Our study provides straightforward evidence that the activation of pathways that trigger the dierentiation program of the multipotential NT2/D1 cells is accompanied by the disappearance of the transcript encoding for the TDGF-1 gene product, independently of the speci®c pathway activated, suggesting that the decrease in TDGF-1 mRNA expression may be a critical event for NT2/D1 cell dierentiation in vitro. Furthermore, the observation that the complete loss of TDGF-1 mRNA expression in cells treated with HMBA or RA occurs before terminal dierentiation (6 and 24 ± 48 h, respectively), suggests the possibility that the suppression of TDGF-1 expression may be necessary for dierentiation to occur. In analogy with the proposed mechanism of action of the other members of the EGF-family (EGF, TGFa), it has been suggested that TDGF-1 protein may act as an autocrine and/or paracrine growth factor for normal and transformed cells (Brandt et al., 1994). The results shown here suggest that the TDGF-1 gene product may function as a growth-stimulating factor for tumor cells of testicular neoplasms, and extend previously reported observations that NT2/D1 cells infected with anti-TDGF-1 anti-sense retrovirus have a longer doubling time, a signi®cantly reduced cloning eciency in semi-solid medium and reduced tumorigenicity (Baldassarre et al., 1996). To address the relationship between the expression of TDGF-1 and the growth and the dierentiation of embryonal carcinoma cells, we produced a recombinant GST-Tdgf-1 fusion protein, containing the EGFlike domain encoded by the TDGF-1 gene. We determined the eects of such recombinant protein on the growth and dierentiation of NT2/D1 cells. Native TDGF-1 protein consists of two major glycoproteins of 28 and 36 kDa (Brandt et al., 1994); however, it has recently been shown that refolded 37- or 47-aminoacid long peptides containing the TDGF-1 EGF-like domains are biologically active (Brandt et al., 1994). In vitro experiments with NT2/D1 cells demonstrate that exogenous addition of the EGF-like domain of TDGF-1 to undierentiated NT2/D1 cells stimulates cell growth; such ®nding together with the high levels of TDGF-1 mRNA and protein expression found in undierentiated NT2/D1 cells, suggest the existence of a TDGF-1 autocrine mitogenic pathway which may drive proliferation of NT2/D1 cells in their undifferentiated state. Interestingly, the existence of such TDGF-1-dependent autocrine stimulation of embryonal carcinoma cells may not be only limited to NT2/D1 cells, but may be common to most teratocarcinomaderived cell lines. In fact, all cell lines tested (F9, 833 KE, 1411 H and PAI 1) express high levels of TDGF-1 (Ciccodicola et al., 1989 and data not shown). In addition, the EGF-like domain of TDGF-1 is also able to exert mitogenic eects on NT2/D1 cells that have been induced to dierentiate with RA or HMBA, suggesting that exogenous TDGF-1 may recruit in cycle G1-arrested cells that are committed to differentiate but that are not yet terminally dierentiated (if, for instance the receptor pathway for TDGF-1 is still on). TDGF-1 may select, among dierentiating cells, only the fraction which retain the capability to respond to TDGF-1. However, TDGF-1 was not able to antagonize the dierentiating eects exerted by RA and HMBA, as shown by the observation that no
substantial change in the immunophenotypic markers was observed after treatment of RA- or HMBA-treated cells with TDGF-1 protein. These results suggest that although TDGF-1 enhances cell growth of differentiating NT2/D1 cells, it is unable to antagonize the dierentiation action exerted by RA or HMBA, thus uncoupling the dierentiation program from the proliferation arrest. Interestingly, a previous work reported that the constitutive expression of TGF-a in NT2/D1 antagonizes the antiproliferative action exerted by RA without blocking the maturation eects (Baselga et al., 1993); moreover, exogenous bFGF is able to stimulate the growth of RA-dierentiated TERA 2 cells (Tienari et al., 1995). However, the in vivo relevance of such growth factors has not yet been investigated, since the expression of bFGF and TGF-a in GCTs has not been thoroughly investigated in these tumors. Conversely, as suggested by the results shown in this paper, the high-level expression of TDGF-1 in non-seminomatous human tumors together with in vitro experiments, TDGF-1 may be involved in the growth and dierentiation of germ cell tumors. Therefore, this study suggests that TDGF-1 may represent a growth modulator for NT2/D1 cells, which may act in concert with other growth factors (bFGF, TGF-a) to maintain NT2/D1 cellular proliferation and tumorigenic potential. Unlike other members of the EGF-family, the human TDGF-1 protein lacks a conventional hydrophobic signal peptide and a transmembrane domain, suggesting that the TDGF-1 protein may be secreted through a non-classical secretory pathway. This poorly characterized mechanism of protein secretion, which has been suggested to operate in TDGF-1-expressing CHO cells (Brandt et al., 1994), has already been described for other growth factors, such as the ®broblast growth factor (D'Amore, 1990). Alternatively, it has been reported that TDGF-1 protein may be associated with the membrane (Brandt et al., 1994), thus raising the possibility that in embryonal carcinoma NT2/D1 cells the native human TDGF-1 protein can function as a mitogen through a juxtacrine or intracrine mechanism. However, the ®nding that exogenous TDGF-1 protein is able to stimulate proliferation of NT2/D1 cells, suggests that TDGF-1 protein is indeed able to bind and signal through a surface receptor. In conclusion, in this study we report that TDGF-1 mRNA and protein are expressed in the more malignant histotype of germ cell tumors, such as anaplastic seminomas, embryonal carcinomas and teratocarcinomas but not in dierentiated teratomas. Moreover, in an in vitro model system, such as the NT2/D1 cell line where transformation and differentiation are mutually exclusive pathways, we demonstrate that TDGF-1 mRNA expression is invariably associated with poorly dierentiated tumorigenic phenotype. Recombinant TDGF-1 protein was able to stimulate proliferation of both undierentiated or dierentiating NT2/D1 cells, but was not able to aect the dierentiation program triggered by RA or HMBA in NT2/D1 cells. These results identify TDGF1 as a growth modulator of embryonal carcinoma cell line in vitro, and a potential autocrine growth factor for poorly dierentiated tumors of the germ cells.
TDGF-1 in testis germ cell tumors G Baldassarre et al
Materials and methods Tissue samples Tumor samples were obtained after resection from 59 patients who had undergone surgery at the Istituto Nazionale Tumori, `Fondazione Pascale', Naples, Italy, and at the Institute of Pathology of the UniversitaÁ Federico II, Naples, Italy and from the Cooperative human tissue network Northwestern Division (Columbus, OHIO, USA). Tumor bioptic specimens, whenever possible, were divided into two equally representative parts: one immediately frozen into liquid nitrogen until RNA extraction was performed, the other ®xed into 4% paraformaldehyde for 15 h at 48C, embedded in paran and then processed for immunoperoxidase staining. Immunoperoxidase staining The anti-TDGF-1 rabbit antiserum used in this study has been previously described (Saeki et al., 1992). Immunohistochemistry was performed using anti-TDGF-1 antiserum at a concentration of 1 mg/ml. Control reactions were performed with normal rabbit immunoglobulins (IgGs) at a concentration of 1 mg/ml. Incubation with anti-rabbit IgG and avidin-biotin-peroxidase complex was carried out according to the supplier's instructions (Vectastain; Vector Laboratories) followed by counterstain with Harris Haematoxylin. Cell culture and treatment The NT2/D1 cell line was grown in Dulbecco's modi®ed Eagle medium (DMEM) (Sigma Inc.) supplemented with 10% heat inactivated fetal calf serum (FCS) (Seromed). RA (Kodak) was solubilized in dimethyl sulphoxide (DMSO) and used at the ®nal concentration of 10 mM. Hexamethylen bisacetamide (HMBA) (Sigma) was solubilized in phosphate buered saline (PBS) and used at ®nal concentration of 5 mM. Dierentiation of the NT2/D1 cell line was performed as previously described (Andrews, 1984; Miller et al., 1990). Cells were plated at a dilution of 1.26106 per 10 mm culture dish and exposed to 10 mM RA, 5 mM HMBA or to the solvent. RNA extraction was carried out at 0, 1, 6, 18, 24, 48, 72 h and 7 days of treatment. Treatment of NT2/D1 cells with 12-O-tetradecanoylphorbol-13-acetate (TPA) was carried out as follows: cells were grown in DMEM supplemented with 10% FCS and exposed to TPA (100 ng/ml) for 1.5, 3 and 8 h. Cooperation between RA and PKC pathway was performed as follows: NT2/D1 cells were plated at a dilution of 1.26106 per 10 mm culture dish and exposed to 10 mM RA for 6 and 18 h; TPA (100 ng/ml) was added in the last 3 h of treatment. At the indicated time cells were harvested and RNA extracted. RNA extraction, Northern blotting and hybridization Total cellular RNA was isolated from cultured cell lines as described previously (Chomczynski and Sacchi, 1987). RNA was extracted from frozen specimens by the CsCl cushion method (Chirgwin et al., 1979) with minor modi®cations. Northern blots were performed essentially as described (Sambrook et al., 1989) using nylon Hybond-N membranes (Amersham Inc.) according to the manufacturer's instructions. All cDNA probes were radio-labeled with a random prime synthesis kit (Amersham Inc.). Hybridization reactions were performed at 428C in 50% formamide, 5% Denhardt's, 56SSPE, 0.2% SDS and 100 mg/ml of denatured sonicated salmon sperm DNA, with 26106 c.p.m./ml of hybridization solution. Filters were washed at 608C twice in 26SSC, 0.2% SDS twice for 30 min and subsequently, for stringent washes, twice for 30 min each in 0.26SSC, 0.1% SDS. Filters were
air-dried and exposed to autoradiographic ®lm for 5 days in the case of TDGF-1 probe and 24 h in the case of b-actin probe. The TDGF-1 probe used in this study was a 0.9 kb EcoRI fragment of the plasmid speci®c for the human TDGF1 cDNA (Ciccodicola et al., 1989). The b-actin probe is described in Tokunaga et al (1986). Construction and puri®cation of GST-Tdgf-1 fusion protein Construction and puri®cation of GST-Tdgf-1 protein were performed essentially as described (Smith and Johnsin, 1988). To express EGF-like motif of TDGF-1 protein mouse Tdgf-1 cDNA (Dono et al., 1993) encoding amino acids 43 ± 95 (VRDRSFQFVPSVGIQNSKSLNKTCCLN GGTCILGSFCACPPSFYGRNCEHDVR) was isolated using PCR (forward 5'-CCGGGATCCGTACGCGAT CGGTCTTTCC-3' and reverse 5'-CTCAAGCTTAGCGA ACATCATGTTCAACAG-3'), cloned into the bacterial expression plasmid pGEX 2T and sequenced. GST and GST-fusion protein production was induced in log-phase growing bacteria upon treatment with 1 mM isopropyl b-Dthiogalactopyranoside (IPTG) for 2 h. Bacteria were recovered by centrifugation, resuspended in PBS pH 7.2 and lysed by sonication. After adding Triton X 100 to 1% (vol/vol), lysates were clari®ed by centrifugation at 8000 r.p.m. in a Sorvall SS34 rotor. The resulting supernatant was incubated with glutathione agarose (Sigma Inc.) for 15 min at room temperature with gentle rotation. The resin was washed with excess PBS and incubated with 10 m M glutathione (Sigma) for 30 min at room temperature. Purity of eluted proteins was assessed by electrophoresis through 12% SDS polyacrylamide gel followed by staining with Comassie blue. GST-Tdgf-1 fusion protein but not GST protein alone was recognized in Western blot experiments by anti-mouse-Tdgf-1 polyclonal antibody as previously described (Brandt et al., 1994). [3H]thymidine incorporation assay NT2/D1 cells were plated at density of 5.06104 cells/well into 24-well cell culture plates (Falcon) in DMEM containing 10% FCS. After 24 h, the medium was replaced with medium containing 1% FCS and cells were starved for 48 h, DMEM-1% FCS was replaced and GSTTdgf-1 fusion protein or GST protein were added or not to the cells in presence of 1 mCi/ml of [3H]thymidine for 18 h. For RA-dierentiating NT2/D1, cells were plated at density of 1.26106 cells/dish into 100 mm-dish cell culture plates (Falcon) in DMEM containing 10% FCS plus 1075 M RA. After 6 days of treatment, cells were trypsinized and plated at density of 5.06104 cells/well into 24-well cell culture plates in DMEM containing 10% FCS and 1 mCi/ml of [3H]thymidine in presence or not of treatments for 18 h. For HMBA-dierentiating NT2/D1, cells were plated at density of 5.06104 cells/well into 24well cell culture plates in DMEM containing 10% FCS plus 5 mM HMBA. After 48 treatments were added in presence of 1 mCi/ml of [3H]thymidine for 18 h. Cells were then washed with PBS and precipitated with 10% trichloroacetic acid (TCA). After solubilization in 0.5 N NaOH contining 1% SDS, the incorporated radioactivity was counted in a b-counter scintillator. Each experiment was performed in quadruplicate and the results are the mean of two dierent experiments. Proliferation assay NT2/D1 cells were plated at density of 7.06104 cells/well into 6-well cell culture plates (Falcon) in DMEM containing 10% FCS. After 18 h the medium was replaced with medium containing 2.5% FCS and treatments were added. At indicated times, cells were harvested and counted. Each
935
TDGF-1 in testis germ cell tumors G Baldassarre et al
936
experiment was performed in duplicate and the results are the mean of three dierent experiments. For HMBAdierentiating NT2/D1, cells were plated at density of 2.06105 cells/well into 6-well cell culture plates in DMEM containing 10% FCS plus 5 mM HMBA in presence or not of indicated treatments. At indicated times cells were harvested and counted. Each experiment was performed in duplicate and the results are the mean of three dierent experiments. For RA-dierentiating NT2/D1, cells were plated at density of 1.26106 cells/dish into 100 mm-dish cell culture plates (Falcon) in DMEM containing 10% FCS plus 1075 M RA. After 6 days of treatment, cells were trypsinized and plated at a density of 2.06105 cells/well into 6-well cell culture plates in DMEM containing 10% FCS in presence or not of treatments. At the indicated times cells were harvested and counted. Each experiment was performed in triplicate and the results are the mean of two dierent experiments. Flow-cytometric analysis NT2/D1 cells treated or not with RA, HMBA and TDGF1 protein were analysed for expression of cell surface antigens detected by the SSEA-3 (Shevinsky et al., 1982),
A2B5 (Andrews, 1984) or by the VINIS-53 (Andrews et al., 1990) monoclonal antibodies (MAbs), as previously described (Baselga et al., 1993). All MAbs were gently provided by Dr PW Andrews. Speci®city of staining was assessed using an isotopic control at the same dilution as the other antibodies. Speci®c ¯uorescence was mesured on a FACScan ¯ow cytometer (Becton Dickinson, San Jose, CA) interfaced with Helwett Packard (Palo Alto, CA) computer for data analysis. Acknowledgements We thank Mrs M Terracciano for her technical assistance and Miss A Secondulfo for correcting and typing the manuscript. We thank Dr WJ Gullick for the generous gift of the anti-TDGF-1 antibodies and Dr PW Andrews for providing SSEA-3, A2B5 and VINIS-53 MAbs. This work was supported by grants from Associazione Italiana Ricerca sul Cancro (AIRC) to GV and MGP, Ricerca Corrente, Ministero della SanitaÁ to GV, Progetto Finalizzato `Ingegneria Genetica', CNR, Istituto Superiore di SanitaÁ, Ministero della SanitaÁ, `Progetto A.I.D.S. 1995 Roma - Italy' and `Progetto Finalizzato Policentrico Coordinato - 1995', Ministero della SanitaÁ to MGP.
References Andrews PW. (1984). Develop. Biol., 103, 285 ± 293. Andrews PW, Nudelman E, Hakomori S and Fenderson B. (1990). Dierentiation, 43, 131 ± 138. Andrews PW. (1988). Biochim. Biophys. Acta. 948, 17 ± 36. Baldassarre G, Bianco C, Tortora G, Ruggiero A, Moasser M, Dmitrovsky E, Bianco AR and Ciardiello F. (1996). Int. J. Cancer, 66, 538 ± 543. Baselga J, Maerz WJ, Moy D, Miller Jr, WH, Castro L, Reuter VE, Yao TJ, Masui H and Dmitrovsky E. (1993). Oncogene, 8, 3257 ± 3263. Bos JG. (1989). Cancer Res., 49, 4682 ± 4689. Brandt R, Normanno N, Gullick WI, Lin JH, Harkins R, Schneider D, Wallace-Jones B, Ciardiello F, Persico MG, Armenante F, Kim N and Salomon DS. (1994). J. Biol. Chem., 269, 17320 ± 17328. Carpenter G and Cohen S. (1990). J. Biol. Chem., 265, 7709 ± 7712. Chirgwin JM, Pryziyba AE, MacDonald RJ and Ruttner WJ. (1979). Biochemistry, 18, 5294 ± 5299. Chomczynski P and Sacchi N. (1987). Anal. Biochem., 162, 156 ± 159. Ciardiello F, Kim N, Saeki T, Dono R, Persico MG, Plowman GD, Garrigues J, Redke S, Todaro GJ and Salomon DS. (1991a). Proc. Natl. Acad. Sci. USA, 88, 7792 ± 7796. Ciardiello F, Dono R, Kim N, Persico MG and Salomon DS. (1991b). Cancer Res., 51, 1051 ± 1054. Ciardiello F, Tortora G, Bianco C, Selvam MP, Basolo F, Fontanini G, Paci®co F, Normanno N, Brandt R, Persico MG, Salomon DS and Bianco AR. (1994). Oncogene, 9, 291 ± 298. Ciccodicola A, Dono R, Obici S, Simeone A, Zollo M and Persico MG. (1989). EMBO J., 8, 1987 ± 1991. Cross M and Dexter TM. (1991). Cell, 64, 271 ± 280. D'Amore PA. (1990). Cancer Metastasis Rev., 9, 227 ± 238. Derynck R. (1988). Cell, 54, 593 ± 595. Dmitrovsky E, Moy D, Miller Jr, WH, Li A and Masui H. (1990a). Oncogene Res., 5, 233 ± 239. Dmitrovsky E, Murty VVVS, Moy D, Miller WH, Nanus D, Albino AP, Samaniego F, Bosl GJ and Chaganti RS. (1990b). Oncogene, 5, 543 ± 548. Dono R, Scalera S, Paci®co F, Acampora D, Persico MG and Simeone A. (1993). Development, 118, 1157 ± 1168.
Kurie JM, Brown P, Salk E, Scheinbrg D, Birrer M, Deutsch P and Dmitrovsky E. (1993). Dierentiation, 54, 115 ± 122. Miller Jr, WH, Maerz WJ, Kurie J, Moy D, Baselga J, Lucas DA, Grippo JF, Masui H and Dmitrovsky E. (1994). Dierentiation, 55, 145 ± 152. Miller Jr, WH, Moy D, Li A, Grippo JF and Dmitrovsky E. (1990). Oncogene, 5, 511 ± 517. Murty VVVS, Houldsworth J, Baldwin S, Reuter V, Hunziker W, Besmer P, Bosl G and Chaganti RSK. (1992). Proc. Natl. Acad. Sci. USA, 89, 11006 ± 11010. Murty VVVS, Li R-G, Houldsworth J, Bronson DL, Bosl GJ and Chaganti RS. (1994). Oncogene, 9, 3228 ± 3231. Plowman GD, Green JM, McDonald VL, Neubauer MG, Disteche CM, Todaro JG and Shoyab M. (1990). Mol. Cell. Biol., 10, 1969 ± 1981. Qi CF, Liscia DS, Normanno N, Merlo G, Johnson GR, Gullick WJ, Ciardiello F, Saeki T, Brandt R, Kim N, Kenney N and Salomon DA. (1994). Br. J. Cancer, 69, 903 ± 910. Rodriguez E, Mathew S, Reuter V, Ilson HD, Bosl JG and Chaganti RSK. (1992). Cancer Res., 52, 2285 ± 2291. Saeki T, Stromberg K, Qi CF, Gullick WJ, Tahara E, Normanno N, Ciardiello F, Kenney N, Johnson GR and Salomon DS. (1992). Cancer Res., 52, 3467 ± 3473. Salomon DS, Brandt R, Ciardiello F and Normanno N. (1995a). Critical Reviews in Oncology-Hematology, 19, 183 ± 232. Salomon DS, Kannan S, Kenney N, Brandt R, Normanno N and Ciardiello F. (1995b). Current Opinion in Endocrinology and Diabetes, 2, 500 ± 509. Sambrook J, Fritsch EF and Maniatis T. (1989). Molecular cloning: a laboratory manual, Cold Spring Harbour, New York, Cold Spring Harbour Laboratory Press. Shevinsky LH, Knowles BB, Damjanov I and Solter D. (1982). Cell, 30, 697 ± 705. Smith DB and Johnsin KS. (1988). Gene, 67, 31 ± 40. Sporn MB and Roberts A. (1992). Annu. Intern. Med., 117, 408 ± 414. Tienari J, Alanko T, Sakseta O, Vesterinen M and Lehtonen E. (1995). Dierentiation, 59, 193 ± 199. Tokunaga K, Taniguchi H, Yoda K, Shimizu M and Sakiyama S. (1986). Nucleic Acids Res., 14, 2829. Ulbright TM and Roth LM. (1994). Diagnostic surgical pathology, Raven Press Ltd., New York, 46, 1885 ± 1943.