POU5F1 (OCT3/4) Identifies Cells with Pluripotent Potential in Human Germ Cell Tumors¹

oct3/4, also known as otf3 or pou5f1, is a member of the POU family of transcription factors, which is expressed in pluripotent mouse and human embryonic stem and germ cells, including PGCs⁴ (1, 2, 3, 4, 5, 6). Expression of this gene is down-regulated during differentiation (7). Furthermore, knocking out the pou5f1 gene in mice causes early lethality because of lack of inner cell mass formation (8) because pou5f1 is critical for self-renewal of embryonic stem cells (9). Interestingly, pou5fl has been linked to the capacity of proper outgrowth of somatic cell clones (10). During human development, expression of POU5F1 is found at least until the blastocyst stage (11) in which it is involved in gene expression regulation. The protein activates transcription via octamer motifs located distally or proximally from transcriptional start sites (12). POU5F1 binding sites have been identified in various genes, including fibroblast growth factor 4 and the 1.5-kb alternative promoter of the platelet-derived growth factor α receptor (13). The data indicate that pou5f1/POU5F1 functions as a master switch in differentiation by regulating cells that have, or can develop, pluripotent potential.

We have previously demonstrated that POU5F1 transcripts are found in a specific set of human testicular GCT of adolescents and young adults (TGCT): the seminomas and embryonal carcinomas (14). In addition, the precursor lesions of TGCT, known as CIS (15), also express POU5F1 (14). These lesions are composed of cells that are considered to be the malignant counterpart of an embryonic germ cell, most likely a PGC (15, 16, 17). Interestingly, these cell types are in principle pluripotent or even multipotent (Ref. 18, for review). In contrast, no expression was found in the differentiated components of nonseminomas, i.e., teratomas, yolk sac tumors, and choriocarcinomas (14). Indeed, expression of POU5F1 has been reported in embryonal carcinoma cells lines, and down-regulation of expression is found upon differentiation (4, 13).

In contrast to our finding of a specific expression pattern of POU5F1 in TGCT, expression of this gene has recently been reported in nonmalignant adult human tissues (19), as well as in a number of carcinoma cell lines (20). This latter finding was interpreted as the result of aberrant reactivation of embryonic genes during the process of malignant transformation. However, the conclusion that POU5F1 is expressed in these cells was based solely on reverse transcription-PCR results, which can be misleading because of the presence of multiple POU5F1 pseudogenes (Refs. 14, 19, 21, own unpublished observations). We are not aware of any previous studies reporting analysis of POU5F1 protein expression in normal and malignant human tissues to clarify whether the mRNA is translated to functional POU5F1 protein. Therefore, an extensive immunohistochemical screening for POU5F1 protein expression was done in various types of GCT at different sites and in a set of >3600 tumors of >100 different types using multitumor tissue microarrays. POU5F1 immunoreactivity was detected only in cells of CIS/gonadoblastoma, seminoma/germinoma/dysgerminoma, and embryonal carcinoma. These results convincingly demonstrate that the presence of POU5F1 protein is related to pluripotent capacity of human GCT and that reactivation of its expression is not a common mechanism in cancer. Conclusively, POU5F1 is a distinctive immunohistochemical marker to identify tumor cells resembling embryonic/primordial germ and embryonic stem cells.

The (T)GCT not included in the tissue arrays were collected in the southwestern part of the Netherlands in collaboration with urologists and pathologists. Representative parts of the tumor (and adjacent tissue, if available) were snap frozen in liquid nitrogen and were fixed in 10% formalin for paraffin embedding. Tumors were diagnosed according to the WHO classification (22), as previously described (23), supported by immunohistochemistry using antibodies directed against germ cell/placental alkaline phosphatase, α-feto protein, human chorionic gonadotropin, the stem cell factor receptor c-KIT, and cytokeratin (CAM5.2). The testicular tumors included 35 seminomas, 50 nonseminomas [with 14 embryonal carcinoma components, 21 teratoma components (6 mature, 5 immature, and 10 mixed), 18 yolk sac tumor components, and 5 choriocarcinoma components], and 10 spermatocytic seminomas. In addition, CIS containing testicular parenchyma (n = 16, including both adjacent to seminoma and nonseminoma) and embryonic testes of different developmental stages, i.e., from 17 to 40 weeks and 28 weeks postpartum, were included. These latter samples have been reported before (24). Moreover, 3 gonadoblastomas, found in dysgenetic gonads, of which 2 also contained dysgerminoma, as well as 4 GCT of the midline of the brain of adults (including 1 germinoma, 1 embryonal carcinoma, and 2 mixed differentiated nonseminomas) were included. To investigate possible difference in presence of POU5F1 between chemotherapy-sensitive and -resistant GCT, a series of 34 patients with known clinical course, including 12 high-risk patients that were relapse-free after high-dose chemotherapy and 22 refractory cases, were investigated. Part of this series has been reported before (25, 26). The clinical parameters are indicated in Table 1. Patients were considered refractory when progression or relapse occurred despite adequate initial and salvage treatment, including high-dose chemotherapy with autologous stem-cell transplantation.

To expand the series of both GCT and non-GCT, three different multitumor tissue microarrays were investigated. One was generated at the Department of Pathology of the University of Basel (Basel, Switzerland), and included 3273 interpretable individual tumors of >100 different types, represented on seven different slides (Table 2 and Fig. 2). The second was generated at the Division of Pediatric Pathology of the Johns Hopkins Medical Institution (Baltimore, MD) and included 84 (T)GCT of newborn, infants, and adolescents (Table 3), represented on four different slides. Both arrays contained various positive and negative controls. The third array, generated at the Department of Pathology, Erasmus Medical Center (GE Rotterdam, the Netherlands), included 48 esophagus tumors and 100 prostate cancers (50 progressing and 50 nonprogressing), as well as benign prostatic hypertrophy (n = 18). The result of this latter array is included in Table 1.

Immunohistochemistry with anti-POU5F1 antibodies was performed on paraffin-embedded tissue sections of 3-μm thickness. The sections were incubated overnight at 4°C with a polyclonal goat anti-POU5F1 antibody (C20, sc 8629; Santa Cruz Biotechnology, Santa Cruz, CA), directed toward the COOH terminus of the protein, diluted 1:8000 (final concentration, 0.025 μg/ml). Subsequently, a biotinylated horse-antigoat secondary antibody was applied to the sections, and the bound antibody complex was visualized using the horseradish peroxidase avidin-biotin complex method. Double fluorescence staining was performed using the polyclonal goat-anti-POU5Fl antibody and a monoclonal mouse-anti-c-KIT antibody (final concentration 2 μg/ml; Neomarkers, Fremont, CA) on frozen tissue sections. c-KIT is a known marker for CIS and seminomas (27). Secondary antibodies were labeled with FITC and CY3 (Dako Diagnostics, Glostrup, Denmark, and Jackson Immunoresearch, West Grove, PA, with a final concentration of 5 μg/ml), respectively. The stainings were also performed separately. Every experiment was accompanied by appropriate positive and negative (without primary antibody) controls.

The human cell lines Tera2 (28), 2102Ep (29), and NCCIT (30), all nonseminoma derived, were cultured and split under conventional conditions (37°C, 5% CO₂). Cell lines were cultured until 80% confluency, and differentiation was induced with retinoic acid as described previously (30). Immunohistochemistry was performed on cytospin preparations as described above.

On the basis of mRNA analysis (14), we expected that POU5F1 protein is present in specific different histological subtypes of TGCT. We first analyzed the presence of POU5F1 by immunohistochemistry on a series of TGCT of various histological types. Representative images of representative stainings are shown in Fig. 1. All tumor cells of seminomas and embryonal carcinomas showed a nuclear staining (Fig. 1, A and B), whereas all nontumor cells were negative. A similar pattern of staining was found in the gonadoblastomas, dysgerminomas and germinomas, and embryonal carcinomas of the brain (see also below). In contrast, all teratomas, both mature and immature, choriocarcinomas, and yolk sac tumors were negative (Fig. 1, C–E), as were the differentiated nonseminomas of the brain (data not shown). The staining intensity in the positive cases varied between moderate and high. None of the spermatocytic seminomas showed a positive staining (Fig. 1,F). We also tested a series of CIS containing testicular parenchyma samples, both adjacent to seminoma and nonseminoma. All CIS cells, identified by a double staining for c-KIT, irrespective of the histology of the invasive tumor, were positive for POU5F1 (Fig. 1 G and inset). In contrast, no protein could be detected in any stage of spermatogenesis (see below). This makes POU5F1 one of the most informative immunohistochemical markers to identify CIS cells that our laboratory has tested.

To investigate whether a difference in presence of POU5F1 exists between chemosensitive and chemoresistant GCT, immunohistochemistry was done on a well-characterized series of GCT with known clinical outcome (Refs. 25, 26 and Table 2). The results indicate that the staining pattern is determined by histology, i.e., positive in CIS, seminoma and nonseminoma, irrespective of sensitivity to chemotherapy.

Although POU5F1 could be detected by immunohistochemistry in CIS cells and cells of seminoma and embryonal carcinoma (see above), no staining was observed in seminiferous tubules containing normal spermatogenesis (Fig. 1,G), neither on frozen nor formalin-fixed, paraffin-embedded tissue sections. In contrast, embryonic germ cells in embryos and fetuses were occasionally positive. The highest number and most intensely stained germ cells were found in gonads between week 17 and 24 of gestational age (Fig. 1 H and see below). Earlier stages of development were not available for investigation. Only sporadic positive cells were identified in the samples obtained from developmental stages weeks 28, 35 and 37, whereas no staining was found in samples of 38 and 40 weeks gestational age and after birth. Germ cells are present in these samples, as identified by staining for various markers, including c-KIT and VASA (24, 31). These data indicate that POU5F1 is present in germ cells at least from the 17^th to 37^th week of gestational age, after which the protein is no longer detectable by immunohistochemistry.

In the nonseminomatous TGCT-derived cell line Tera2, expression of the POU5F1 is only observed in undifferentiated cells, whereas induction of differentiation of embryonal carcinoma cells with retinoic acid results in down-regulation of POU5F1 expression. The potential of Tera2 cells to differentiate has been described before (4, 14). Here, we confirm this finding on the protein level by immunohistochemistry (Fig. 1 I, left and right panel). In contrast, both the cell lines NCCIT and 2102Ep showed POU5F1 expression and protein before and induction with retinoic acid differentiation (data not shown). This is in accordance to the finding that these cell lines, in contrast to Tera2, do not show complete differentiation after exposure to retinoic acid (28, 30).

To investigate whether POU5F1 protein can be detected in other tumors than GCT, we performed immunohistochemistry using the same antibody on three different multitumor tissue microarrays. One contained a large series of tumors of different origin (Table 1), one of the other contained specifically prostate and esophageal carcinomas (of which the results are also included in Table 1), and one contained mainly GCT of newborn and infants (Table 3). None of the control tissues, both adult and embryonal (with one exception, see below), included in the arrays showed positive staining for POU5F1. Of the 3439 tumors (from the first arrays), all testicular seminomas and nonseminomas (containing either a seminoma or an embryonal carcinoma component), as well as two dysgerminomas and one gonadoblastoma were positive, confirming the results presented above. Representative examples of the first arrays are shown in Fig. 2. In addition, one clear cell carcinoma of the kidney and two lung carcinomas (one squamous cell carcinoma and one undifferentiated large cell carcinoma) were positive, suggesting that POU5F1 protein can be found in a small percentage (1–2%) of other tumors.

The multitumor tissue microarray of the (T)GCT confirmed the specificity of the POU5F1 immunohistochemistry for the histological types CIS/gonadoblastoma, seminoma/dysgerminoma, and embryonal carcinoma, as described above (Table 3). None of the immature and mature teratomas and yolk sac tumors showed a positive staining. Again, embryonic germ cells, included as arrays in the array, were also positive (data not shown). These data establish POU5F1 as highly specific marker for normal and malignant embryonic germ cells and embryonic stem cells, both in precursor (CIS) lesions and invasive tumors.

Human GCTs comprise a heterogeneous group of neoplasms, which are predominantly found along the midline of the body. This localization can be explained by the migration route of the PGC during embryogenesis from the yolk sac to the genital ridge (32). The histology and clinical behavior of the various types of GCT depends on the sex of the patient, the age at clinical diagnosis, the anatomical location, and histology of the tumor (Ref. 33 for review). Overall, four different entities can be distinguished: (a) the teratomas and yolk sac tumors of newborn and infants; (b) the seminomas/dysgerminomas/germinomas and nonseminomas of the gonads (both ovary and testis, and dysgenetic gonad), anterior mediastinum, and midline of the brain; (c) the dermoid cyst (mature cystic teratoma) of the ovary; and (d) the spermatocytic seminoma of the elderly testis. From a biological point of view, the pluripotent tumors (entity b) are the most intriguing. Histologically, they are subdivided into seminomas (known as dysgerminoma when identified in the ovary and dysgenetic gonad, and germinoma when arising in extragonadal sites) and nonseminomas (Ref. 34 for review). The seminoma-like tumors are composed of cells with a similar morphology as CIS cells. The nonseminomas can contain different histological elements, including embryonal carcinoma as the stem cell component, teratoma (representing somatic differentiation), yolk sac tumor, and choriocarcinoma (representing extraembryonic differentiation). The occurrence of embryoid bodies in nonseminomas (22), as well as certain patterns of gene expression, including POU5F1 (4, 13, 14, 35, 36, 37, 38), illustrates the similarities between developmental patterns in the embryo and nonseminomas. This study aimed at a further understanding of the nature of the pluripotent (T)GCT compared with the other types of GCT and tumors of nongerm cell origin.

Here, we convincingly show that testicular CIS, seminoma, and embryonal carcinoma cells are consistently positive for POU5F1, irrespective of chemosensitivity or chemoresistance. Gonadoblastomas and associated dysgerminomas, as well as germinomas and embryonal carcinomas of the midline of the brain, showed a similar pattern. In contrast, none of the differentiated nonseminomatous derivatives (teratoma, yolk sac tumor, and choriocarcinoma) stained positive, irrespective of anatomical localization. This suggests that loss of protein expression is because of down-regulation of gene expression upon differentiation/maturation. These data are in accordance with earlier findings in mice, indicating that pou5f1 is indeed a marker for embryonic stem cells and germ cells (1, 2, 3) and that expression is lost upon additional differentiation and maturation (7). The fact that we observed a homogeneously positive staining for POU5F1 in the embryonal carcinoma components of both pure and mixed nonseminomas (and representative cell lines) supports the model that the encoded protein is crucial but not sufficient for pluripotency. This has also been implied by the recent finding that POU5F1 overexpression had no effect on HeLa cells (39). Apparently, particular auxiliary proteins are needed. In accordance with the histology-dependent pattern of POU5F1 staining, both mature and immature teratomas, as well as the yolk sac tumors of newborn and infants, are negative. This is also true for the spermatocytic seminomas. Both tumor types have a different pathogenesis from TGCT (22, 24, 40, 41, 42, 43, 44, 45, 46, 47). The absence of POU5F1 in these tumors is in agreement with their inability to generate pluripotent stem cells.

Our findings within in vivo tumor samples and representative cell lines are concordant with previous findings in mice (8, 9), showing that pou5f1 is expressed in pluripotent cells and is down-regulated upon differentiation. Therefore, we conclude that POU5F1 is involved in regulation of differentiation of human cells in line with a report on human embryos (6). No POU5F1 was detectable in adult testicular samples with normal spermatogenesis by immunohistochemistry, which is consistent with findings in mice using fluorescent protein-tagged pou5f1 expression analysis (48). However, a positive staining by immunohistochemistry has been reported in mouse spermatogonia A (7). This difference might be because of sensitivity differences of the methods applied and/or differences in protein level. The finding of POU5F1 in human embryonic germ cells and the down-regulation of expression before spermatogonia A formation supports an embryonic initiation of the pathogenesis of TGCT (15, 16, 27, 49, 50, 51, 52).

We did not find POU5F1 protein in any of the differentiated components of nonseminomas or in the majority (>99.9%) of 3340 nongerm cell tumors. This indicates that reactivation of POU5F1 expression and translation, as previously suggested based on cell line data (20), is not a frequent finding in cancer. The positive tumors were one kidney carcinoma (of 50 clear cell carcinomas) and two lung carcinomas (1 of 50 squamous and 1 of 47 large cell carcinomas analyzed). We do not currently have any plausible biological explanation for POU5F1 expression in these tumor types. Nevertheless, rare occurrence of POU5F1 immunoreactivity in kidney and lung must be kept in mind when performing immunohistochemistry for POU5F1 to demonstrate a pluripotent GCT in these organs. We think that the discrepancy between our study and the previous report suggesting reactivation of POU5F1 occurring as part of the malignant transformation process (20) can be attributable to the different techniques used. The mRNA expression analyses by Monk and Holding (20) were performed using reverse transcription-PCR with primers that may amplify both POU5F1 and different intron-less POU5F1-related sequences (unpublished observations).

In conclusion, this study demonstrates the usefulness of POU5F1 immunohistochemistry to support that TGCT originate from PGC and have pluripotent potential. However, POU5F1 is also present in nullipotent embryonal carcinoma and thus does not predict whether an embryonal carcinoma is able to generate various lineages of differentiation. Overall, POU5F1 is a highly specific immunohistochemical marker to confirm the diagnosis of CIS/gonadoblastoma, seminoma/dysgerminoma, and embryonal carcinoma. In addition, these data support the hypothesis that (T)GCT are excellent models to study the molecular mechanisms of pluripotency and differentiation, and therefore, a deeper understanding of these mechanism might also reflect into the rapidly evolving field of stem cell biology. Conversely, advances in the rapid evolving field of stem cell biology regarding the control of differentiation might also provide innovative therapeutic alternatives in (T)GCT in the future.

We thank Ronald de Krijger (Department of Pathology, Erasmus Medical Center) for sharing the embryonic tissue samples.