Heterogeneity in the in vitro survival and proliferation of human seminoma cells

Smumry The in vitro culture conditions allowing survival and initial proliferation of murine primordial germ cells from 10.5 days post coitum embryos, which include the use of a murine embryonal fibroblast (STO) feeder, were applied to 21 human seminomas, composed of tumour cells which are considered as the malignant counterparts of human primordial germ cells. Cells from 18 seminomas attached poorly to STO, and only a few survived through day 10. In contrast, three seminomas showed a higher degree of attachment. Two of them showed initial proliferation and enhanced survival: 30 days for tumour SEI and 25 days for tumour SE3. Tumour SEI was more extensively studied, using the culture conditions allowing the derivation of pluripotent embryonic stem cells from 8.5 days post coitum munne primordial germ cells, which include the use of STO feeder, stem cell factor, leukaemia inhibitory factor and basic fibroblast growth factor. The presence of stem cell factor was necessary and sufficient for colonies of tumour cells to form during the first 3 days of culture. While the cell number decreased after day 3 in medium without fetal calf serum, it increased until day 9 in medium containing fetal calf serum. No reprogramming of SEI cells to pluripotent stem cells was observed. Our data indicate that seminomas form a tumour population with a heterogeneous in vitro behaviour not equivalent to that of 8.5-10.5 days post coitum munne primordial germ cells.

In humans, a unique histological entity of testicular germ cell tumours of adults (TGCTs) exists, namely seminomas (SEs), which are composed of tumour cells that are considered to be the malignant counterpart of human primordial germ cells (PGCs) (Mostofi et al., 1987). No animal model for these tumours is known, and they cannot be cultured in vitro for a prolonged period. To develop an in vitro culture system, we studied the use of Sertoli cell feeders and observed an enhanced survival (Berends et al., 1991). Because of the number of animals repeatedly needed as Sertoli cell donors, as well as the heterogeneity of the feeder preparations, we looked for alternatives. SE cells are indistinguishable from the cells of carcinoma in situ (CIS), the precursor of all TGCTs (Skakkebzk et al., 1987), except for their invasive behaviour (Skakkebmk et al., 1987;Oosterhuis and Looijenga, 1993). Both have morphological and immunohistochemical similarities to PGCs (Skakkebtk et al., 1987;Gondos, 1993), and like PGCs most SEs express the stem cell factor (SCF) receptor c-Kit (Strohmeyer et al., 1991;Murty et al., 1992). Therefore, we assumed that SE cells share microenvironmental requirements with PGCs. Using a murine embryonal fibroblast (STO) feeder, initial proliferation and survival up to day 6 has been described for murine PGCs isolated from 10.5 days post coitum (d.p.c.) embryos (Donovan et al., 1986). The use of soluble or membrane-bound stem cell factor (SCF), which has an important role in gametogenesis (see Witte, 1990, for review), and/or leukaemia inhibitory factor (LIF), which maintains the pluripotent phenotype of murine embryonal stem (ES) cells (Heath, 1992;Hilton, 1992) and embryonal carcinoma (EC) cells (Brown et al., 1992), allowed enhanced survival and proliferation (De Felici and Dolci, 1991;Dolci et al., 1991;Godin et al., 1991;Matsui et al., 1991). Further addition of basic fibroblast growth factor (bFGF), probably involved in the regulation of germ cell proliferation (Ueno et al., 1987;Suzuki et al., 1991), resulted in long-term prolifera-  (Matsui et al., 1992;Resnick et al., 1992). Under these conditions, plunrpotent ES cells can be derived from 8.5 d.p.c. PGCs (Matsui et al., 1992;Resnick et al., 1992). This is interesting in view of the linear progression model, which assumes the reprogramming of SE cells to pluripotent stem cells, subsequently giving rise to embryonic and/or extraembryonic tissues in non-seminomatous TGCTs (NS) (Oosterhuis and Looijenga, 1993). Therefore, we have now studied the survival of cells from 21 primary SEs in co-culture with STO cells. The effect of SCF, LIF and bFGF on the cells from the SEs with the longest survival on STO feeder was analysed.

Materials and methods
Tumour handling Twenty-one orchidectomy specimens from patients suspected of having a germ cell tumour were collected during surgery in collaborating hospitals. Macroscopically representative parts of the tumour and the adjacent normal parenchyma were partly snap frozen using liquid nitrogen, partly put in medium A [Dulbecco's modified Eagle medium (DMEM/ F12, with 100 kUI -' penicillin, 100 mg -' streptomycin, 40 mg 1-' gentamycin, 365 mg 1-L-glutamine; Gibco, Paisley, UK] and taken to the laboratory for further processing. Fresh representative samples of all components were used for imprints and subsequently fixed in 4% formalin (J.T. Baker, Deventer, The Netherlands) for paraffin embedding. After imprint and frozen section diagnosis of SEs [using a haematoxylin and eosin (H&E)-stained slide], the tumour was mechanically disaggregated at room temperature, using two crossed scalpel blades. Tissue fragments were allowed to settle in a SOml tube in 30ml of medium A. The supernatant, containing almost only single cells (as analysed by microscopy), was washed twice with medium A and either directly cultured or cryopreserved. To the cell suspension 10% (final concentration) dimethylsulphoxide (DMSO) (Merck, Darmstadt, Germany) was added slowly. The suspension was aliquoted, automatically frozen in a Kryo 10 Series 2 (Planer Biomed, Sunbury-on-Thames, UK) (-2C min-' to -5_C, -l1Cmin-' to -40 C, -5Cmin' to -160-C) and stored under liquid nitrogen.
Feeders STO cells were cultured in T25 fiasks (Costar, Cambridge, MA, USA) at 37C in a humid atmosphere with 5% arbon dioxide in air, using 5 ml of medium A containing 10% fetal calf serum (FCS) (Gibco) and subcultured once a week. The feeders were grown in 0.1% gelatin (Sigma, St Louis, MO, USA) coated six or 12-well tissue culture plastic plates (Costar) or T25 flasks and at confluence treated with 1O ig ml1' mitomycin C (Sigma) for 3 h. After a triple wash with phosphate-buffered saline (PBS) the feeders were kept in medium A containing 10% FCS and SE cells were inoculated on the next day.

Seminma-STO co-cultures
Single-cell suspensions from 21 SEs (coded SEI to SE21) were seeded onto STO feeder. Two tumours were cultured using the cryopreserved suspension. Nimeteen tumours were cuured using fresh cell suspensions, while from nine of these tumours cryopr suspensions were also used. SE cells were seeded in either a T25 flask (I07cells per flask) or a six-well plate (5 x 10I cells per well), in respectively 5 and 2 ml of medium A with 10% FCS. After overnight incubation at 34C in a humid atmosphere with 5% carbon dioxide in air, medium was taken off and half a volume of fresh medium was given to the culture. The old medium was spun down at 1000 r.p.m. for 5 min and half the volume was returned to the culture. Subsequently, half of the medium was changed every other day.
Seminoma proliferation in the presence ofgrowth factors Feeder-containing six-well plates were seeded with 5 x 10' cryopr SE cells per well after rapid thawing in a 3TC water bath (experiments 1, 2 and 3). In experiments I and 2 four wells (referred to as wells 1-4) were inoculated for each condition, while in experiment 3 wells were inoculated in duplicte (wells 3 and 4). The wells contained 2 ml of medium A, with or without 10% FCS, with or without 60 ng ml-' human recombinant SCF (provided by Amgen, Thousand Oaks, CA, USA) (experiments 2 and 3) or a combination of 60ngml'I SCF, lOngml ' human recombinant LIF (provided by Dr JK Heath, Department of Biochemistry, University of Oxford, UK) and 1 ng ml-' human recombinant bFGF (Gibco) (experiments 1, 2 and 3) (this combination of growth factors is referred to as SLB). SE cell number and colony size in wells 3 and 4 were counted on days 3 and 9 in experiment 1, on days 1, 3 and 9 in experiment 2 and on days 1, 3 and 6 in experiment 3. Cultures were kept at 34-C in a humid atmosphere with 5% carbon dioxide in air, while half of the medium was changed every day.
On day 1 of culture, 10 iM bromodeoxyuridine (BrdU) (Sigma) was added to wells 1 and 2 in experiments 1 and 2. After overnight incubation, cultures were fixed at room temperature in 70% ethanol for 15 min. The bottoms of the wells were cut into two sections using a hot scalpel blade. The sections were immunohistochemically stained either for cytokeratin, using 3,3'-diaminobenzidine tetrahydrochloride visualisation (Oosterhuis et al., 1989), or for PLAP and BrdU (Organon Teknika, Boxtel, The Netherlands), using a double-staining technique according to Hardonk and Harms (1990). PLAP was stained using an immunoperoxidase with 3-amino-9-ethyl carbazole (Sigma) visualisation, while BrdU was stained with an immuno alaline phosphatase with fast blue BB salt (Sigma) visuisation. This procedure was repeated for wells 3 and 4 (in all three experiments) after incubation with BrdU from day 8 to 9.
For all conditions, the cells present in three visual fields at a 320 x magnificaion were counted, using a Zeiss Axiovert phase contrast microscope (Zeiss, Germany), equipped with a Sony charge-coupled device (CCD) camera and screen (Sony, Japan), to allow evaluation of the observed cells by two individuals.

Statistical analysis
Welch's t-statistics (Sachs, 1982;Miller, 1986) was used to analyse the influence of serum and growth factors on SE colony number and size. Analysis was done for each culture condition separately, comparing the counts of the fixed time points, i.e. days 3 and 9 in experiment 1, days 1, 3 and 9 in experiment 2 and days 1, 3 and 6 in experiment 3. All calulations were performed using Stata software (release 3; Stata, Santa Monica, CA, USA).

Redls Twnour characterisation
Immunohistochemially all 21 tumours, which were located in the testis, were negative for AFP, while they showed consistent membranous staining for PLAP and c-Kit. Six tumours were negative for cytokeratin and hCG. In five tumours cytokeratin-expressing cells and in three tumours hCG-positive cells were detected. Seven tumours showed cytokeratin as well as hCG expression.

Seminoma-STO co-cultures
The cells of all tumours except three (SE1, SE2 and SE3) showed poor attachment to STO feeder, and only few cells survived through day 10 (not shown). While fresh and cryopreserved SE2 cells only showed enhanced attachment, fresh SEl and SE3 cells also survived for over 24 days.
Cryopreserved SEI cells survived for up to 15 days, while no cryopreserved suspensions of SE3 were available. In all SEI and SE3 cultures initial proliferation was observed (not shown).
Seminoma proliferation in the presence ofgrowth factors Because of the availability of cryopreserved cell suspensions and the better performance in culture of SEI, these cells were used to study the effect of growth factors. Therefore, SEI cells were cultured on STO feeder in DMEM/F12 with or without FCS, SCF or SLB. All data shown in the figures and presented in the text are from experiment 2; experments 1 and 3 yielded essentially similar results.
The morphology of SEl cells on STO feeder, in medium without FCS and growth factors or medium with FCS and SLB at day 9 of culture, is shown in Figure 1. In medium without FCS, feeder quality had morphologically declined and colonies were absent, while large colonies were present in medium with FCS. Under FCS-free conditions, SE cells were found on top of the STO cells, while they seemed to sink into the feeder layer when exposed to FCS-containing medium.
The mean colony size of SEI over time, under the various conditions, is shown in Figure 2 (first penrod) and days 3 -9 (second period) were calculated. Under FCS-free conditions, without growth factors, no change in total cell number was observed during the first period, while this number decreased during the second period (data not shown). The colony number significantly (P<0.01) decreased after day 3 (data not shown). During the entire experiment the mean colony size was constant (1.2 cells per colony). The use of SCF or SLB with FCS-free medium resulted in an increase in total cell number during the first period and a rapid decrease in this number after day 3 (data not shown). The colony number was constant until day 3, but decreased significantly (P<0.01) dunrng the second period. The mean colony size increased significantly (P < 0.01) until day 3 and decreased significantly (P<0.01) afterwards. In FCS-free medium no colonies with more than two cells were detected on day 9, irrespective of the presence of growth factors. In FCS-containing medium the colony number was constant throughout the experiment, irrespective of the presence of growth factors. In the absence of growth factors, total cell number (not shown) and mean colony size were constant until day 3 (mean size 1.6 cells per colony). Both increased during the second period. This increase was only significant (P<0.01) for the mean colony size (mean size 2.2 cells per colony at day 9). In the presence of SCF or SLB, total cell number (not shown) and colony size significantly (P<0.05) increased during the whole culture period.
In the presence of SCF, the mean colony size on days 3 and 9 was 2.0 and 3.2 cells per colony respectively, while in the presence of SLB the mean size was 2.1 and 3.3 cells per colony respectively. Figure 3 illustrates the range of the colony sizes under the various conditions on days 1, 3 and 9.
In the absence of FCS and growth factors only colonies of 1 and 2 cells were present on day 9, whereas the use of these 6 7 8 9 10 Fuge 2 Mean size of SEI colonies over time under the various culture conditions. Results are from experiment 2 as a representative example. Vertical bars indicate standard errors. The letters a -e indicate a significant (P<0.Ol in all cases, except for a and e on day 3 with P<0.05) increase or decrease in mean colony size as compared with the previous count of the colony size (FCS, fetal calf serum; GF, growth factor(s); SCF, stem cell factor; SLB, the combination of stem cell factor, leukaemia inhibitory factor and basic fibroblast growth factor).
additives resulted in the formation of much larger colonies (up to> 10 cells per colony) at this time point. Only in the cultures with SLB were a few colonies of up to 40 cells detected outside the counted fields.
No obvious morphological and immunohistochemical (PLAP and cytokeratin expression) changes were identified during the in vitro culture (not shown).

PLAP staining and BrdU incorporation
The counts of PLAP-positive cells confirmed the data obtained with morphological phase-contrast detection of SEl cells in culture. SEI cell cultures were incubated with BrdU from day 1 to 2 or from day 8 to 9 to detect DNA synthesis. The percentage of PLAP-positive cells showing BrdU incorporation ranged from 10% to 24% on day 2, while 0.4-17% of the cells were labelled on day 9. No differences were found for the vanrous culture conditions. On days 2 and 9, cells with incorporated BrdU were detected in colonies of all sizes.
We have previously shown (Berends et al., 1991) and confirmed in the present study that SE cells cultured without a feeder layer (on tissue culture plastic, in DMEM/F12 containing 10% FCS) die within 3 days (data not shown). This was also found for the cells (SEl) with the longest survival on STO feeder, even in the presence of SCF or SLB (not shown). Therefore, we conclude that SE cells need contact with a specific matrix, which might be provided by feeder cells, possibly through interactions of the membrane-bound form of SCF and the receptor c-Kit, in order to survive and proliferate. This is supported by the finding of the same survival of the SEI cells on STO with or without additives during the first 3 days of culture. Since the results from the a u . .
Ha,gI.usIt d inmd p ind m J PA RAObe eta present study are similar to those reported previously on the use of Sertoli cell feeders for SE culture (Berends et al., 1991), we conclude that the homogeneous STO feeders seem to form a good alternative to the use of Sertoli cell layers. The more extensively studied tumour SEl showed an increase in colony size during the first 3 days of culture on STO feeder, using both SCF and SLB, irespective of the presence of FCS. From days 3 to 9 colony size increased in the presence of FCS alone, or with added SCF or SLB.
Probably because of quality decli (morphological changes) of STO in FCS-free medium, the number and colony size of SEI cels decreased from day 3 onwards. Therefore, FCS seems to be necey to directly support the STO clls, while its effect on SE cells seems to be indirect and through the feeder layer.
During the first 3 days of culture, the colony number of SEl ceDs was constant for all conditions. This indicates that proliferation of SEI cells, for which the p ce of SCF was necessary and sufficient, caused the growth of the colonies, instead of chltering of the cells owing to (enhanced) cell motility. From day 3 onwards all cultures containing FCS showed an increase in colony size.
Three SEs (SEl, SE2 and SE3) had a plating efficiency on STO similar to that found for 8.5 dp.c. murine PGCs (30%) (Matsui et al., 1992), while two of them (SEl and SE3) initially proliferated, just like 10.5 d.p.c. murine PGCs (Donovan et al., 1986). However, 18 of the 21 SEs studied had a plating efficiency on STO of less than 1%. In spite of survival of the att clls from the SEs to about day 10, no proliferation was found. Apparently, SEs form a tumour population with a heterogeneous in vitro behaviour, differing in attachment to STO feeder cells and subsequent survival and proliferation. The SE cells with an attachment and initial proliferation simila to that of 8.5 d.p.c. murine PGCs showed no reprogrammig to pluripotent stem cells under any of the conditions appled, as judged by their unhanged morphology and continued expression of membrane-bound PLAP. Therefore, we conclude that the differentiation state of SE cells is not similar to that of 8.5-10.5 d.p.c. PGCs. However, this does not exclude a lnear progression model for CIS, SE and NS. The differences in the in vitro behaviour of SEs and murine PGCs might be related to the crucial role of the age (d.p.c.) of the latter in the ability to respond to growth factors: 11.5-12.5 d.p.c. murine PGCs do not proliferate when co-cultured with feeder cells in the presence of SCF or LIF (De Felici and Dolci, 1991;Matsui et al., 1991).
The heterogeneity in the in vitro behaviour did not correlate with the expression of the markers hCG and/or cytokeratins 8 and 18. The differentiation status of human TGCT cell lines has recently been described to correlate with the expression of distinct glycolipids, among others carrying the stagecific embryonic antigens 1, 3 and 4 (Wenk et al., 1994). Our preliminary results from an extensive study of the glycolipid profile of primary human TGCTs, especialy SEs (RA Olie et al. in preparation), revealed no distinct differentiation status of SEl, SE2 and SE3 as compared with the other SEs described here. The SCF receptor c-Kit was detced on all SEs. Recently, we found a possible explanation for the aberrant in vitro behaviour of SE1, SE2 and SE3. From 40 SEs analysed, including 17 of the 21 tumours described in this paper, these tumours were the only three containing an activated Nor K-ras gene (Olie et al., 1994). Intestingly, suppression of apoptosis by an activated ras gene has been reported (Arends et al., 1993), and we have indications that mechanical disociation of SE tissue results in apoptosis of the tumour cells (RA Olie et al., in preparation). These indications are in agreement with the findings by Frisch and Francis (1994), who recently reported on the induction and abrogation by an activated ras gene of apoptosis by disruption of cell-matrix interactions. Therefore, we conclude that a higher degree of attachment, alone or in combination with enhanced survival and initial proliferation of SE cells in vitro, might be related to the presence of an activated ras gene, which possibly interferes with the apoptotic pathway. , while an enhanced in vitro proliferative capacity is reported for adult acute myeloid leukaemia with a poor prognosis (L6wenberg et al.. 1993). In view of these and our findings we are currently investigating the prognostic relevance of the in vitro behaviour and presence of ras mutations in SE.

AckD
is This work was supported by the Dutch Cancer Society Grant DDHK 91-19. Purchase of the CCD camera and screen and of a biohazard flowhood was supported by the Nijbakker-Morra Foundation. We thank Amgen for providing SCF, Dr Heath for LIF, Dr Ullrich (Max Planck Institute for Biochemistry, Martinsried, Germany) for the c-Kit monoclonal antibody and Dr Mummery (Hubrecht Laboratory, Utrecht, The Netherlands) for providing STO. Wim van Putten is acknowledged for assistance in the statistical analysis of the data. Dennis van der Wel is thanked for printing the photos. Collaborating pathologists and urologists in the southwestern part of The Netherlands are thanked for providing the tumour samples. Riette de Bruijn is acknowledged for technical assistance.