The natural history of type I GCT, emerging from numerous clinical and pathological observations [55, 94, 103–109], is that regardless of anatomical site, they begin during embryonic life as immature teratoma, probably arising from one pluripotent progenitor cell, which may evolve to mature teratoma with trilaminar derivatives. However, an immature teratoma component may contain foci of YST easily overlooked on microscopic examination [106, 110–113], which may eventually overgrow the original teratoma (Fig. 3.7). In fetuses and neonates, and in prenatally resected tumors, YST is virtually always associated with immature teratoma, while pure YST is rare [94, 113]. Thus, these tumors come in three histological variants: pure (immature) teratoma; pure YST, whereby the original teratoma component is probably overgrown by the more aggressive YST component; and combinations of (immature) teratoma and YST. The younger the infant, the more often an immature component is present and the lesser the chance that an overt YST component has developed, irrespective of gender (Table 3.2). YST can take the form of both intraembryonic endodermal derivatives, such as the primitive gut and liver, and extraembryonic structures such as allantois and yolk sac [114], reason for Nogales to advocate the name primitive endodermal tumor instead of YST (Chap. 6).

The frequency of the different histological variants differs per anatomical site; however, in population-based registries, teratomas are the most frequent at all sites. In fact, the large majority of type I GCT have a favorable course regardless of degree of immaturity, with the exception of high-grade immature teratomas of the ovary. A YST component, overall present in about 5–10 % of the cases at birth [94, 109], is the only predictor of recurrence at any site [106, 109]. Type I GCT typically lack a seminomatous component, EC, and choriocarcinoma, which are indicative for a type II GCT (see below). Choriocarcinoma may rarely occur in infants, in association with an intracranial type I teratoma [115] or metastatic from placental/gestational choriocarcinoma [116, 117].

Type I GCT may contain OCT4-positive cells, usually in higher-grade immature teratoma components [118, 119], which however are negative for SOX2 and CD30. This is in contrast to EC cells, the stem cells of type II GCT, which typically express these two proteins in addition to OCT4 and other pluripotency proteins, such as NANOG and STELLAR [57, 120]. These OCT4-positive cells may be the stem cells of type I GCT. The lack of expression of CD30 may be explained by the cells being diploid. It was shown that in vitro ESC cells, the normal counterparts of EC cells, only start to express CD30 when they become aneuploid [121]. The rarity of these stem cells in type I GCT suggests that they do not readily self-renew but are rather poised to differentiation, particularly into the various somatic lineages, explaining the usually benign character of these tumors. In fact, the precursor cells seem to be in the primed state.

Type I GCT may contain SOX2-positive cells; however, these are not the OCT4-positive putative stem cells [122], shown in Fig. 3.8.

The age distribution of GCT shows a neonatal peak in the sacrococcygeal area, retroperitoneum, mediastinum, head and neck, brain (apart from the pineal gland), and testis. In the ovary, the early peak is missing; however, GCT do occur from birth through adulthood without interruption. The tumors represented by the early peak are type I GCT, rare tumors, most often occurring in the fetus, neonates, and children under the age of two and seldom beyond age six, with an overall predilection for girls, mainly due to the about 3.5:1 female to male ratio of sacrococcygeal tumors [76]. This skewed sex distribution may be due to global demethylation of PGC taking place earlier in males than in females. As a result, in males, the PGC are more fragile and prone to apoptosis upon mis-migration, whereas in females, the PGC are more robust and therefore have a greater chance to undergo reprogramming, giving them survival advantage outside a proper niche [57]. Such a mechanism would explain the overall slightly increased risk of extragonadal type I and perhaps also type 0 GCT in females [123], not answering the question of why mainly in the sacrococcygeal region.

Exact incidence figures are hard to get because in most cancer registries, only malignant GCT and not teratomas are included. The best approximation is achieved by combining cancer registry data with population-based figures on all GCT obtained in centers specializing in the treatment of these tumors [55, 94, 103, 105, 107, 109, 114]. The Netherlands and Belgian National Cancer Registries report, respectively, 5.2 and 5.4 malignant GCT per million children less than 15 years of age [77]; this figure is 4/million in Germany [107]. Assuming that over half, in fact, up to 70 %, of the type I GCT in children are teratomas, the overall incidence would be 1–1.5/100,000 [77].

Multiplicity of type I GCT is exceedingly rare: one child with a bilateral pure YST of the testis [124]; four cases of bilateral teratoma of the testis [125–127], two of which were brothers with Klinefelter’s syndrome [127]; no bilateral stage I ovarian YST [128, 129]; 1–2 % of ovarian immature teratomas is bilateral [130, 131]; and no published multiple extragonadal cases to the best of our knowledge.

Combinations in one individual of type I GCT with other GCT types do occur: type I GCT may rarely be combined with fetus in fetu (type 0 GCT) among others in the cranial region [132]; ovarian type I GCT may be combined with a type IV GCT in the same ovary and in 11 % in the contralateral ovary [130].

Familial cases of type I GCT have not been described for the testis, apart from the two brothers, both infants, with Klinefelter’s syndrome, mentioned above, with bilateral testicular teratomas [127]. In view of the rarity of bilateral testicular type I teratoma [125, 126] as well as Klinefelter’s syndrome, this is probably not a chance occurrence, suggesting that this syndrome is a risk factor also for type I GCT, in addition to being an established risk factor for type II GCT of the mediastinum and brain.

Ovarian type I GCT may cluster with dermoid cysts of the ovary (type IV GCT). Since the latter may have a familial component, this is probably also true for the type I GCT [130, 133–135].

Finally, there is the family of a mother with an immature teratoma of the ovary coexisting with a newborn baby with an intracranial immature teratoma [132, 136]; Poremba et al. excluded that the tumors were clonally related. Giambartolomei et al. [137] retrieved four families from the literature in which an ovarian type I GCT was combined with one or more type II GCT of the testis (five cases) or ovary (one case).

Type I GCT are most often localized in extragonadal sites along the midline of the body: the sacral region, retroperitoneum (cranially of the kidneys), stomach, anterior mediastinum, heart, head and neck, and brain [94, 138]. This peculiar distribution along the midline, including the brain, is attributed to the migration route of PGC during embryonic development [52, 54, 56, 57]. Others explain it by the relative abundance of ESC (for review [55]) or NSC [139] along the midline of the developing embryo. Type I GCT occur also in the testis, the second most frequent site after the sacral region, and in the ovary. They have never been described in dysgenetic gonads in keeping with the different pathogenesis of type I and type II GCT of the gonads, the latter being derived from transformed, virtually always aneuploid gonocytes in the naïve state, while the former probably originate through direct reprogramming of essentially normal, still methylated, and pre-erased diploid gonocytes to ESC in the primed state.

Type I (immature) teratoma, either pure or combined with YST, is virtually always diploid, lacking chromosomal rearrangements. However, YST of type I, pure or combined with teratoma, is most often aneuploid, usually (near)diploid, with multiple gains and losses of (parts) of chromosomes. Involved in gains are 1q, 3, 3p, 8q24, 12p13, 20q, and 22; involved in losses are 1p (1p36), 4, 4q, 6q (6q24-qter), 16q, and 20p. Overrepresentation of the whole of 12p or the region 12p11.2–p12.1, typical for type II GCT, is not a feature of type I GCT. As mentioned above, the distal part of 12p and in particular 12p13 may be overrepresented ([111, 140–150], for review [151]). Some of these changes, such as gain of 1q and loss of 1p and 6q, are shared by type II YST and may be related to the phenomenon of progression/differentiation toward YST rather than being specific for type I GCT [152].

Although highly speculative, for some of the chromosomal gains and losses, possibly involved genes have been suggested; gain of 8q24 has been associated with amplification of MYC [142]; gain of 12p13 might involve the pluripotency genes STELLAR, NANOG, and GDF3 [38]; and loss of 1p36 [140] might involve CHD5, a tumor suppressor gene deleted from 1p36.31 in neuroblastoma [153].

Sporadic case reports describe specific balanced chromosomal translocations in type I GCT. Two infantile sacral teratoma cases had constitutional balanced translocations involving 12q13 probably affecting different genes [146, 154]. In one of the two patients, the genes involved in the translocation t(12;15)(q13;q25) were identified as SUMO-/Sentrin-specific protease 1 gene (SENP1) and the embryonic polarity-related mesoderm development gene (MESDC2) [155]. The resulting fusion protein SEME interferes with the function of MESDC2 as a chaperone for the WNT co-receptors LRP5 and/or LRP6. It is suggested that in both patients, the constitutional translocation was predisposing to the development of the sacral teratoma. In an intrathoracic mature teratoma in a 15-year-old girl, a balanced chromosomal translocation, t(8;22)(p21;q12), was the sole cytogenetic aberration. It resulted in fusion of the genes PPP2R2A and CHEK2, supposedly the initiating event in this teratoma [156].

A genome-wide association study involving type I and II GCT suggests that a variant in BAK1 involved in suppression of apoptosis [157] is associated with gonadal GCT of both types. Type I GCT were not associated with variants in KITLG, SPRY4, and DMRT1 (doublesex and mab-3 related transcription factor 1), which confer an increased risk for type II GCT of the testis [158].

The Wnt/beta-catenin [159] and the TGFbeta/BMP signaling [160] pathways are strongly expressed in type I and II YST, compared to seminoma/dysgerminoma, EC, and choriocarcinoma ([161] for review). Methylation of APC and LOH at 5q21-22 suggests that APC might be involved in the activation of the Wnt pathway [162]. In general, the transcriptome of pediatric YST is enriched for genes associated with a differentiation and proliferation phenotype as compared to seminomatous (type II) GCT [163]. It is likely that the expression patterns of the various GCT types are mostly determined by the cell type(s) present and much less by pathways activated in the process of tumorigenesis. Seminomatous GCT and EC express pluripotency genes, and the various extraembryonic and somatic tissues express genes characteristic for the involved cell lineages [161, 163].

Type I GCT usually have a biparental GI pattern as in somatic cells. In a small proportion, GI is partially erased, in particular in type I GCT of the testis and ovary [144, 164–167]. These findings on GI support the hypothesis that extragonadal type I GCT may originate from PGC, which are pre-erased or partially erased, corresponding to the methylation status of the genome (see Sect. 3.2.3). Reprogramming of these cells will produce an ESC with the developmental potential of the primed state. It does not exclude their derivation directly from ESC in the primed state, which are also characterized by a biparental GI pattern. Also, a somatic cell with induced pluripotency (iPSC), for example, by reactivation of pluripotency genes, in particular OCT4, as demonstrated for human NSC [72], could theoretically be the precursor of type I GCT. This has been suggested by Scotting and co-workers for GCT of the brain [139, 168, 169], without making the essential distinction between type I and type II GCT [86]. Such iPSC would endow the derived tumors with their own GI pattern [56], as it is stably transmitted to daughter cells [75, 170]. Any degree of loss of imprinting in a type I GCT tumor would suggest that the tumor is derived from a germ cell precursor [166], unless, as reported [169], NSC may also have partial loss of imprinting.

Apart from humans, pluripotent tumors have been most extensively studied in the mouse. Relevant for human testicular type I GCT are the spontaneous [80, 81] and experimental [171] testicular teratomas in the 129 mouse strain. The spontaneous ovarian teratomas in LT mice [172] are probably a model for human ovarian type I GCT. Teratomas derived from pre- and postimplantation embryos transplanted to various organs, in particular the testis [173] or the kidney [174], might be a model for extragonadal type I GCT in man. Spontaneous extragonadal teratomas in mice [83, 175, 176] are too rare to be practically useful for animal experiments.

The difference between the spontaneous and experimental testicular teratomas as compared to the embryo-derived teratomas is that the gonocytes from which the testicular tumors originate are committed to the germ lineage and not themselves pluripotent [61]. They have to be reprogrammed before being able to form pluripotent tumors [60], a process similar to what happens in the human type I GCT, and to reprogramming of a somatic cell to an iPSC, by converting the nucleus from nullipotent to pluripotent [61].

These different mouse models have a similar developmental potential; when fully developed, they are mainly composed of mature somatic tissues derived from the three germ layers. Immature teratoma and EC cells are less frequent and often minor components; rarer still are extraembryonic lineages. Late takes of embryo transplantation under the kidney capsule consist of parietal YST and occasionally trophoblastic giant cells [177]; these tumors are most often aneuploid [178], just as human type I YST (Fig. 3.9). The observation that in chimeric blastocysts polyploid murine ESC only give rise to extraembryonal lineages (yolk sac and placenta), while the embryo proper is derived from the diploid ESC [179, 180], might explain the restricted developmental potential of aneuploid tumor cells in type I GCT: probably, they are no longer capable to form somatic tissues.

The testicular teratomas in the mouse models originate when a luminal gonocyte or a prespermatogonium in its niche is reprogrammed to an ESC in the primed state, either directly or via an EGC that apparently loses its naïve-state developmental potential [60, 61, 80, 181, 182]. Initially, when proliferating within the seminiferous tubules, the tumor cells stay undifferentiated, as EC cells. When the EC distends and disrupts the tubular wall and invades the testicular interstitium, it starts to differentiate into immature somatic tissues, which gradually develop into mature teratomas [81, 183]. A minority of the tumors will maintain immature teratoma and EC and will be retransplantable in syngeneic hosts. Rare tumors are pure EC from the start, which can readily be transplanted. This same evolution is seen in tumors derived from pre- or postimplantation embryos up to E8. The percentage of tumors with an EC component depends on the strain of the transplanted embryos and is usually higher than in the testicular teratomas [184]. These mouse models, as well as the ovarian teratomas in LT strain mice, resemble human type I GCT. They have the same cells of origin (PGC/gonocytes and ESC), histological evolution, and developmental potential. Whatever the cells of origin, they seem to have or acquire the primed state in view of the developmental potential of the derived tumors.

Probably the most important lesson to be learnt from these models is that disruption of the microenvironment of the pluripotent cell itself suffices to initiate a pluripotent tumor. For gonocytes/prespermatogonia in the developing testis of 129 strain mice, this principle is demonstrated by genital ridge transplantation, as will be discussed in the following.

In 129/Sv mice carrying the loss of function steel mutation (steel or Kitlg is the mouse homolog of KITLG), spontaneous testicular teratomas occur in about 4 % of the animals, twofold the spontaneous rate (2 %) in 129 strain mice lacking this mutation. When from the same animals the genital ridges are transplanted, teratomas develop in over 80 % of the grafted genital ridges, often at multiple sites [171]. This is counterintuitive: loss of PGC with the steel mutation, and even more so by the procedure of the genital ridge transplantation, increases the yield of teratomas. The steel mutation, in the membrane-bound Kitlg [185], and the transplantation procedure are not carcinogenic events acting on the PGC but rather factors that disturb the niche of the PGC, promoting reprogramming of the surviving PGC. Apparently, cell-intrinsic mechanisms for repression of the developmental capacity of gonocytes/prespermatogonia, such as those via Blimp1, Prdm14, and AP2ɣ [186–188] and Dmrt1 [189], are not sufficient when the restraints of the normal tubular environment are disturbed, like Nanos2-/Dmrt-dependent GDNF signaling by Sertoli cells [189]. Remarkably, only male genital ridges produce teratomas; female genital ridges never do [173], probably because in the female genital ridge, the germ cells are blocked in meiosis I and few in numbers. In the male genital ridges, the germ cells are more numerous, premeiotic, and arrested in G0/G1 of mitosis [12]. Contrasting patterns of Dnd1 expression in female and male gonads may also contribute to the different susceptibility to teratoma formation [190].

As for embryo-derived tumors, perfectly normal embryos may turn into teratomas when transplanted into a testis or a kidney [173, 184]. EC cells derived from such tumors, when introduced into the ICM of a blastocyst, can contribute to the normal tissues of the resulting chimeric mouse, demonstrating that the tumor cells when restored to their proper environment may normalize, as well as cause malignant tumors [191].

The importance of genetic factors conditioning the micro-milieu of the niche was demonstrated by crossing susceptibility genes for testicular teratomas into 129 strain mice. As already mentioned, in the original strain, about 2 % of the mice had spontaneous testicular teratomas, introducing the loss of function steel mutation, which reduces the number of PGC and spermatogenesis [192], doubled this percentage [171], and by adding the ter mutation, one third of the mice developed spontaneous testicular teratomas [81]. The gene ter is a recessive gene that causes germ cell deficiency in mice, and in 129/Sv-ter mice, it also enhances the yield of teratomas. Male 129/Sv-ter mice, homozygous for the ter mutation, are sterile and almost always have teratomas, often bilateral [193]. The ter mutation occurs in the Dnd gene, expressed in fetal gonads [190]; in mice, Dnd isoform α is necessary for viability of germ cells including PGC from E8 and for viability of embryos [194]. Specifically, in 129 strain mice homozygous for this mutation, PGC die apoptotically or when they escape apoptosis may be reprogrammed to ESC, which form teratomas. It is even more likely the other way round that some PGC escape apoptosis because they have been reprogrammed [190]. As a corollary, the proneness of 129 mice – and not of other strains [194] – to form teratomas is due to the ease with which PGC of 129 mice PGC are reprogrammed to an ESC in the primed state. This may be due to incompetence of 129 strain mice to adequately suppress reprogramming to pluripotency in germ cells, a process in which among others Dmrt1 expressed in PGC/gonocytes is involved [189].

The variants in KITLG that increase the susceptibility for testicular type II GCT in humans [195, 196] do not seem to affect the incidence of type I GCT [158].

Although derived from PGC, KIT mutations are probably exceptional in type I GCT. In support of this contention, none of the pure immature and mature teratomas of the brain studied by Wang [197], almost certainly type I GCT, had KIT mutations and also very rarely other mutations. This is indeed remarkable since KIT is the crucial survival and proliferation factor for PGC.

In the mouse, germ cells that do not reach the genital ridges die through apoptosis caused by the proapoptotic protein Bax. In Bax-null embryos, large numbers of ectopic (extragonadal) germ cells fail to die [57]. A similar mechanism of impairment of apoptosis of mis-migrated PGC might enhance the development of extragonadal human type I GCT; however, this has not been demonstrated.

The available evidence points to the pathogenesis of type I GCT being foremost “developmental” and not driven by somatic mutations. This implies that the p53-dependent DNA damage response is intact in these tumors, explaining their favorable response to cisplatin-based chemotherapy, just like type II GCT.

The most likely cells of origin of extragonadal type I GCT are mis-migrated PGC, as these cells have been demonstrated along the midline of the body, indeed in large numbers at the typical sites of these tumors [52]. Most of these PGC die apoptotically; probably only those that are reprogrammed to an ESC manage to survive outside the niches in the gonads, the thymus, and the midline of the brain suitable for PGC. Reprogramming occurs when the mechanisms, with a key role for SOX17, BLIMP1, and OCT4 [35], maintaining the phenotype and suppressing the developmental potential of PGC, break down, probably because of lack of a suitable niche. Since the PGC are pre-erased, reprogramming will result in an ESC in the primed state capable of forming immature somatic tissues that will usually differentiate to fully mature teratoma. YST and very rarely choriocarcinoma are the only other components, which develop from tumor cells that have become aneuploid.

In the testis and ovary, type I GCT originate when diploid, pre-erased gonocytes, and oogonia are reprogrammed to ESC in the primed state due to failure of control of developmental potential by germ cell-intrinsic (DMRT1 in addition to SOX 17, BLIMP1, and OCT4) and niche factors (such as GDNF) [189]. Reprogramming to an ESC can occur directly or via an EGC in which the naïve state is rapidly dismantled [182].

Pathogenesis is mainly developmental; somatic mutations probably play a minor role.

Sacrococcygeal type I GCT, with a frequency of 1/35,000 live births, constitute about 40–50 % of extragonadal type I GCT and are the most frequent neonatal tumor. They are rarely diagnosed beyond the age of 2 years and virtually do not occur after age six [76, 109, 198, 199]. The fact that there are practically no GCT at all in the sacral region past the age of six is in accordance with the absence of type II GCT at this anatomical site. The rare sacrococcygeal type I teratomas in adults probably had their inception before birth and went undetected [55]. There is a strong predilection for girls with a male to female ratio of 1:3.5.

Other congenital disorders occur in up to 25 % of patients with sacrococcygeal type I GCT, including trisomy 21/Down’s syndrome (implying a higher risk for type I GCT in Down’s syndrome), genitourinary malformations, congenital hip dislocation, esophageal atresia and congenital heart disease [114, 198], and duplication of pelvic organs attributable to hindgut twinning [200]. There is a well-documented association with multiple pregnancies, either within the same pregnancy or as a family history of multiple pregnancies [94, 96].

The evolution of these tumors is typical for type I GCT. Starting as immature teratomas prior to birth, they become more mature with time. When completely removed at this stage, which entails removal of the coccyx bone in continuity with the tumor ([114] for review), the child is cured. Incomplete or delayed surgery may allow the tumor to recur as mature or immature teratoma, or by means of tumor progression, to develop a YST component in the primary tumor or in a recurrence. A YST component is found in 5–10 % of the tumors removed before the age of 2 months; thereafter, this figure increases rapidly, and by the age of three, most sacrococcygeal type I GCT are malignant, in principle due to progression to YST [94, 201, 202]. The tendency for malignancy seems somewhat greater in males than in females [201]. Rarely, a somatic-type malignancy may develop such as Wilms’ tumor [203, 204]. Also the type I teratomas of adults may in some 10 % develop a malignant component [205]. Metastases can be local or visceral and are usually composed of YST or less frequently immature teratoma [198, 199]. In contrast to teratoma, YST is aneuploid with the chromosomal aberrations characteristic for YST progression in type I GCT, as discussed.

Sacral teratomas so highly developed that they have a vertebral axis should according to the definition of Willis [206] be classified as parasitic twins. A somewhat less strict definition [96, 98] considers a sacral teratoma with clearly developed limbs as a parasitic twin; in view of the site of attachment, they should be classified as a parasitic pygopagus [96]. Indeed, a personal case, published as sacral teratoma with a classical clinical history, including recurrence as YST (with the characteristic chromosomal aberrations) upon incomplete surgery [207], should probably be reclassified as a parasitic pygopagus, a conjoined twin parasite attached to sacrococcygeal area (Fig. 3.10). This case illustrates the continuum between twinning and the development of a type I GCT and the difficulty pinpointing the cells of origin of these growths. Indeed, some deem it possible that all extragonadal teratomas have originated as twins [96], and at the other end of the spectrum, others consider them as derived from mis-migrated PGC, which have a preference for the rostral and caudal part of the sympathetic nervous system [52]. In between are those who favor the idea that they are derived from an ESC.

Cases like ours [207], and an almost identical one reported by Chen et al. [208], blur the distinction between parasitic twin and teratoma or rather between type 0 and type I GCT. The two types of GCT may be derived from the same or different precursor cells in the 2C, respectively, primed state.

In the retroperitoneal region, all GCT under the age of six are probably type I GCT [76]. Perhaps some may be poorly organized included twins (type 0 GCT), as the retroperitoneum is the most common site of fetus in fetu [95].

Five to ten percent of extragonadal type I GCT occur in the retroperitoneum, most of them in the left or right suprarenal region consistent with lateral migration of PGC toward the gonadal ridges [52]. The sex distribution is about equal when several smaller series are combined [104, 209, 210]. Over 10–20 % of tumors are partly or wholly composed of YST, of the remainder, about half have an immature teratoma component and half are completely mature teratomas. The relatively high figure for YST is probably due to the fact that most of the tumors are diagnosed a couple of months after birth.

In postpubertal males, retroperitoneal GCT are virtually always metastatic from unrecognized testicular type II GCT [108, 211–213]. In postpubertal females, retroperitoneal GCT are very rare, usually benign, and probably type I GCT that have remained undetected until after puberty.

Of the type I GCT, 2–3 % are located in the stomach with a male-to-female ratio of 1:3.7; progression to YST is rare [94]; however, like the type I GCT of the neck, they may metastasize in the form of immature teratoma [94].

The mediastinal type I GCT constitute 2–3 % of the total, most are located in the anterior mediastinum, originating in the thymus [114], and only rarely in the posterior mediastinum. There is a slight preponderance of females [94, 138]. Progression to YST occurs in up to 30 % probably due to surgery several months after birth [214].

Type I GCT of the heart are relatively frequent, 4–7 % of the total, most often located in the pericardial cavity, attached to the great vessels at the base of the heart, and only rarely within the heart itself, very much in accordance with the sites where mis-migrated PGC are found [52]. Males and females are equally affected. Progression to YST occurs in about 5 % [94].

Type I GCT of the head and neck occur in less than 40,000 live births and constitute 10–20 % of all extragonadal type I GCT; the sex distribution is roughly equal. Anatomical localizations are the neck including the thyroid gland (35 %); face (8 %); oro- and nasopharynx and surrounding structures, in particular hard palate and nasopharynx (45 %); and orbit (12 %) [94, 138, 215–217].

They develop during embryonic life and are often diagnosed before birth. The histology is most often mature teratoma, about one third of the cases contain immature teratoma. Immature neural tissue may rarely metastasize to regional lymph nodes and the lungs and on very rare occasions spontaneously mature [218]. About 3 % of the tumors present as pure YST or as teratomas with microscopic foci of YST [94, 216]. In the series of 16 cases described by Lack [215], there were three YST, respectively, in the oropharynx, the nasopharynx, and the floor of the mouth. In two cases, surgery was not carried out immediately after birth but after 6 and 10 months, respectively. Progression to a somatic-type malignancy may occur, in particular neuroblastoma [218]; squamous cell carcinoma has been reported as well [219].

Progression to YST and metastasis did not occur in 51 cases occurring in the neck [217], probably because surgery is done shortly after delivery, preventing the tumors to progress. The low progression rate might raise the suspicion that many of the teratomas are in fact parasitic cephalopagus [96]. Indeed the oral mature teratoma, mentioned before, diagnosed prenatally in a female baby, most likely was a disorganized dizygotic twin [102], which in retrospect should have been classified as epignathus or more formally as parasitic cephalopagus. This is yet another example of a case that blurs the distinction between type 0 and type I GCT.

Oro-nasopharyngeal and cervical teratomas are associated with other congenital disorders in 12 and 6 %, respectively [94].

The highly aggressive sinonasal pluripotent tumors in adults [220–222] are often characterized by chromosomal translocations and will be discussed as type VI GCT.

Intracranial type I GCT constitute about 10–15 % of all type I GCT with an equal sex ratio and 3 % associated with YST [94, 109, 132]. In one third of the cases, the size of tumor obscures the original anatomical localization. When the site can be determined, it is most often cerebral hemisphere (25.5 %), followed by the suprasellar region (23 %), third ventricle (5.6 %), and pineal region (4.4 %) [132]. The tumors may extend into the orbit, neck, face, mouth, or pharynx [94, 132, 138].

Under the age of six, GCT of the testis are practically always of type I [76], amounting to 5–10 % of all type I GCT [94, 109, 223]. Eleven out of the 19 tumors described by De Backer et al. [223] were teratomas, confirming that teratomas are more frequent than YST in unbiased institutional registries [224]; indeed, under the age of 1.5 years, no YST was diagnosed. Four of the 11 teratomas had immature areas; however, none of the tumors was combined with YST or a raised serum alpha-fetoprotein (AFP). Mixed type I GCT, combining teratoma with YST, are rare but do occur also in the testis [224].

In view of the supposed pathogenesis of type I GCT, it is remarkable that mixed tumors are so rare in the testis, at least ten times less frequent than pure YST [224–226]. The presence of immature teratoma may increase the risk of progression toward YST [106, 198]. Probably, when progression occurs early, in a microscopic immature teratoma, the tumor appears as pure YST at clinical presentation; progression in an established teratoma results in a mixed type I GCT combining teratoma with YST. Teratomas may very rarely, also by way of progression, develop PNET as a somatic-type malignancy [227].

Type I GCT of the testis are not associated with germ cell neoplasia in situ (GCNIS) [228] and testicular dysgenesis syndrome (TDS) [229] and do not share the risk factors of testicular type II GCT nor their increasing incidence. Familial susceptibility for prepubertal YST has not been demonstrated [230, 231]. Unlike the type II GCT of the testis, there is no association with single nucleotide polymorphism (SNP) variants of KITLG, SPRY4, and DMRT1, among others. There seems to be an association with a SNP variant of BAK1 [158], suggesting that resistance to apoptosis of primitive germ cells might play a role in the pathogenesis of prepubertal GCT. This is in line with the hypothesis that testicular type I GCT originate through reprogramming of a diploid, methylated, pre-erased, premeiotic PGC to an ESC in the primed state.

In the ovary, the early neonatal peak in the age distribution of GCT, representing type I GCT, is not apparent [76]; however, it is unlikely that they do not exist. Rather their age distribution is probably broader and overlaps with types II and IV, as shown below.

Among 158 reviewed cases of pure and mixed dysgerminomas of the ovary, by definition type II GCT, the youngest was 4 years old; 6 % were in the age group 0–9 years and 41 % between 10 and 19 [232]. A review of 517 dermoid cysts, type IV GCT of the ovary, showed an almost Gaussian age distribution with no cases under age 10 and 1.5 % under age 15 [233]. From these figures, it can be deduced that the large majority of the 66 pediatric patients through age 15, reported by De Backer et al. [234], had a type I GCT. Six tumors were purely cystic, thus probably type IV GCT, and 12 were type II GCT on histological grounds, leaving 48 type I GCT. Of these, three were pure YST, consistent with the rate of about 5 % YST in other anatomical sites. This makes the ovary the second most frequent site of type I GCT after the sacrococcygeal region, accounting for 15–25 % of all type I GCT.

Apparently, teratomas of the ovary can be of three types: I, II, and IV and taking the type VI teratomas associated with clear cell carcinoma of the ovary (Chap. 6) also into account, four types. The overlapping age distributions and morphological resemblance may pose problems separating them. A morphologically typical dermoid cyst in a patient over 10 years of age is almost certainly a type IV GCT. A solid teratoma or a pure YST, or the combination of the two under age five, is most probably a type I GCT. Any GCT with a dysgerminoma, EC, or choriocarcinoma component, with or without other components, is a type II GCT regardless of age. Teratomas associated with epithelial cancers of the ovary are of type VI. Cases composed of teratoma and/or YST over age 5 could be type I or type II GCT. Separating these malignant GCT is probably not so important clinically. However, for a (partly) solid pure teratoma, in a patient over 5 years, it is crucial to make the distinction, since a type I teratoma is benign, whereas a type II teratoma is malignant. In such cases, the diagnosis needs (cyto)genetic confirmation.

Like for the testicular ones, the assumption is that ovarian type I GCT originate through reprogramming of a diploid, methylated, pre-erased, premeiotic PGC to an ESC in the primed state. Such mitotic germ cells persist in the periphery of the ovary through week 20 gestational age [235, 236]. Oogonia can be present in the cortex of the ovary in the two first years of life before they are finally cleared [237].

Type II GCT are malignant tumors that come in two variants: first, seminomas (named dysgerminoma in the ovary; germinoma in the brain; seminoma or germinoma in the mediastinum), which are homogeneous neoplasms composed of neoplastic PGC/gonocytes, the default development of type II GCT; second, non-seminomas, which are caricatures of embryonic development, including both somatic and extraembryonic lineages [244]. Non-seminomas arise when a neoplastic PGC/gonocyte is reprogrammed to become an EGC in the naïve state, or, in pathological terms, when a seminomatous cell is reprogrammed to a totipotent EC cell, the stem cell of non-seminomas [245], as originally demonstrated for mouse EC cells by Kleinsmith and Pierce [246]. EC cells may give rise to all lineages of embryogenesis: YST (secreting AFP) and choriocarcinoma (secreting beta-human chorionic gonadotropin (β-HCG)) represent the extra-embryonic tissues; immature and mature teratomas represent somatic tissues of the three germ layers of the embryo in varying degrees of maturation; occasional primitive germ cells represent the germ lineage [247] (Fig. 3.12). These elements are characterized by lineage-specific mRNA [248, 249] and protein expression profiles [161]. In non-seminomas, so-called embryoid bodies can be encountered, which strongly resemble 10-day-old, early presomite human embryos. They show the same expression patterns of both mRNA and proteins as during normal development, like OCT4. Beyond that particular stage, corresponding to the time that in a pregnancy implantation is completed [250], development becomes disorganized, with embryoid bodies turning into patches of EC, YST, trophoblastic giant cells/choriocarcinoma, or teratoma, and disorderly combinations thereof. A possible explanation is that the neoplastic embryo lacks the biparental imprinting pattern of the zygote that is required for proper development of extraembryonic structures and concomitant vascular supply. Mature teratoma may be highly differentiated at the tissue level and even contain organoid structures closely resembling the gut, bronchi, etc., but never fully developed organs as in type 0 GCT or hair and teeth as in type IV GCT. The complete gamut of differentiation lineages, in particular the capacity to develop both embryonic and extraembryonic lineages, the germline competence, and the high capacity of self-renewal of its stem cells (EC cells) characterize type II GCT indeed as totipotent, apparently derived from precursor cells in the naïve state [251].

The mechanism of reprogramming of a seminomatous tumor cell (including the cells of GCNIS) is unknown. It is likely that microenvironmental factors play an important role, suggested by the observation that in the cryptorchidism, the percentage of seminoma depends on the location of the testis: about 90 % in abdominal, about 80 % in inguinal, and about 50 % in scrotal position (both after spontaneous or surgical/hormonal correction of cryptorchidism) [252–254]. The age of clinical presentation of the tumor was the same as in patients with scrotal tumors without a history of cryptorchidism [253]. Recently, it was suggested that interstitial stromal factors like NOGGIN might inhibit bone morphogenetic protein (BMP) in the tumor cell, whereupon reprogramming is initiated via NODAL signaling in two stages [255]. During a maturation phase, a fast-acting NODAL loop stimulates its own activity and temporarily inhibits BMP signaling. During the stabilization phase, a slow-acting NODAL loop, involving WNT signaling [159], reestablishes BMP signaling and the pluripotency circuitry [255]. This is in line with the observations on Cripto, the co-receptor for Nodal [256, 257], which is highly expressed in GCNIS, seminoma, EC, and YST, associated with hypomethylation of the promoter and absent in teratoma where the promoter is hypermethylated [257].

Interestingly, inhibition of BMP is the opposite mechanism from initiation of germline specification in the mouse embryo via expression of Bmp4 ([35] for review). This NODAL-mediated mechanism of reprogramming implies that the tumor cells are exposed to interstitial stromal cells, which is not the case in the intratubular environment, suggesting that within the seminiferous tubule, other factors are involved in reprogramming of GCNIS or intratubular seminoma cells. Moreover, stromal factors are usually not sufficient as primary seminoma is reprogrammed in only about 15 %, giving rise to a mixed non-seminoma with a seminoma component. It seems there is more to be learned about reprogramming of a seminomatous precursor cell to a totipotent EGC.

In all anatomical sites, over half of all primary type II GCT are pure seminomatous tumors. In fact, the younger the patient population, the higher the proportion of seminoma: in dysgenetic gonads and the brain about 80 %, in the ovary 60 %, in the mediastinum 55 %, and in the testis about 50 %. Reprogramming continues even in metastatic seminoma of the testis: 44 % of seminoma metastases eventually develop non-seminoma components [258]. Thus, in the natural history of testicular type II GCT, only 30 % maintain their seminoma phenotype until demise of the patient. These observations suggest that reprogramming is a chance event accumulating over time, whereby in a non-scrotal testis, the chance of reprogramming is deminished, as discussed above.

Seminomatous tumors are by definition pure; the only cells other than neoplastic gonocytes are scattered trophoblastic cells occurring in less than 10 % of the cases [244]. Dysgerminomas in the ovary, mediastinal seminomas, and germinomas of the brain may also contain trophoblastic giant cells in a small percentage [243, 259, 260]. Non-seminomas are often composed of more than one differentiation lineage, in all possible combinations including seminoma, so-called mixed non-seminomas. EC is almost always present and may be the only component, like its derived lineages, thus accounting for pure EC, YST, choriocarcinoma, and teratoma. The frequency of EC attests to the high capacity for self-renewal of these totipotent stem cells of non-seminomas and likely explains the more rapid evolution and earlier clinical manifestation of non-seminomas than seminomas. This is well documented for the testicular type II GCT, where the age distribution for seminomas peaks at 35 years and for the non-seminomas at 25 years. Mixed non-seminomas with a seminoma component, in which reprogramming is delayed because it occurs in already invasive seminoma, peak at the median age of 30 in between non-seminoma and seminoma [261, 262] (Fig. 3.13). Primary type II GCT of the brain, mediastinum, and ovary show the same order in age distribution: for brain, the mean age for germinomas, mixed tumors with a germinoma component, and EC is, respectively, 18, 15, and 12 years [263]; for mediastinum, the mean age for seminomas is about 30 [264] and for non-seminomas 25 years [265]; and for the ovary, the median age of dysgerminoma is 22 years [266], for EC 14 years [267], and for mixed non-dysgerminomas with a dysgerminoma component in between.

Somatic tissues of non-seminomas may progress to form somatic-type malignancies that closely resemble their somatic counterparts, as will be discussed per primary site (see also Chap. 12) (for review [244, 268]).

Seminomas and non-seminomas may metastasize to regional lymph nodes and from then on to distant organs, so-called visceral metastasis, in order of frequency: lungs, liver, brain, and bone [258]. Choriocarcinoma has a propensity for blood-borne metastases, which may cause the first clinical manifestation of the tumor [244]. Seminoma cells may at metastatic sites be reprogrammed to non-seminoma in up to 44 %, as mentioned [258]. In non-seminoma, EC cells are the principal metastatic cells; likewise, in somatic cancers, cancer stem cells are the ones that frequently metastasize [269]. This is microscopically apparent as tumor emboli are virtually always composed of EC cells [244]. At the site where these tumor stem cells will eventually lodge, they may resume differentiation, mimicking the histology of the primary tumor. This phenomenon is demonstrated in mouse models of type I GCT, which in this respect are also valid for type II GCT [270] (Fig. 3.14). The level of differentiation at the metastatic site is in general less than in the primary tumor [271], which may be related to the different microenvironment or due to selection of more metastatic EC cells. In support of the latter, the more distant the metastases, the lesser the level of differentiation; in particular, visceral metastases rarely contain teratoma components but often consist only of EC cells that apparently in the metastatic process have progressively been selected toward stemness and loss of differentiation capacity [272].

Over 90 % of type II GCT occur in the testis [76], being the most frequent cancer of males aged 25–45 in Western white Caucasian populations [273]; the remaining develop in dysgenetic gonads/ovary (about 4 %) and in the extragonadal sites: anterior mediastinum/thymus and brain midline/pineal gland (about 3 %) [274]. The youngest age of presentation is in dysgenetic gonads/ovary (from age four with a broad age distribution) [232, 266], followed by brain (children from age 10 to adolescence) [263], mediastinum, and testis (adolescents and adults) [264, 265]. Remarkably, overall the peak age is about 30 for testicular type II GCT: 15 years later than the peak age for those of the brain and 5 years later than for those in the ovary and mediastinum. Type II GCT occur in patients in whom puberty has started or is completed, except for rare cases associated with disorders of sex development (DSD) [275], Down’s syndrome [276], and Klinefelter’s syndrome [277].

In each of the extragonadal sites, males greatly outnumber females with regard to type II GCT [76]. Apparently, type II GCT is very much a disease of adolescent and adult males, probably related to the presence of the TSPY gene on the Y chromosome, as will be explained later on.

The overall global incidence is 1.5/100,000 with a 20-fold difference between areas with the lowest and highest incidence [273]. The global incidence differences and the rising incidence are attributable to the testicular type II GCT [273, 278]. Geographic incidence differences for the other anatomical sites are dwarfed by those of the testis [76].

Testicular and ovarian type II GCT are bilateral, respectively, in 3–5 % [279, 280] and 10–15 % [281, 282]. Rarely, gonadal tumors may be combined with extragonadal type II GCT in the same individual, like in the patient who had a testicular seminoma and a germinoma of the pineal gland [283] and a patient with GCNIS of the testis simultaneous with a mediastinal non-seminoma [212].

Type II GCT have a strong familial component: over 5 % of patients with a testicular type II GCT have a relative with a similar tumor [284]; an estimated 25 % of testicular cases is due to familial susceptibility [285, 286]. Familial clustering is also documented for ovarian tumors, and testicular may cluster with ovarian type II GCT [137]. In one family, a woman with an ovarian dysgerminoma had a brother with a mediastinal EC [287], suggesting a common etiology in these cases. The only difference between sporadic and familial cases is a younger age of clinical manifestation: 2–3 years for testicular tumors [288] and about 7 years for ovarian cases [137].

Gonadal type II and I GCT may cluster in families as discussed under the type I GCT. An intriguing combination was reported by Heimdal et al. [284] of two brothers, one with a non-seminoma (type II) and the other with a spermatocytic tumor (type III). Spermatocytic tumor being so rare, this combination is probably not by chance but due to a pathogenetic commonality.

From the epidemiological data, it appears that type II GCT occur only in the testis, ovary, dysgenetic gonads, anterior mediastinum, most likely arising in the thymus, and midline of the brain with a preference for the pineal gland [76]. Type II GCT localized in the retroperitoneum are not primary tumors as suggested [289] but metastases from unrecognized primary testicular tumors [108, 211–213].

The occurrence of type II GCT in the mediastinum and the midline of the brain is explained by the migration route of PGC, also the explanation proposed for the anatomical distribution of type I GCT [54]. Clearly, the anatomical distribution of type II GCT is much more limited than for type I GCT. Probably initiation and development of the former require specific conditions of the microenvironment, only offered by certain cell types in these sites [56]. The PGC/gonocytes giving rise to type II GCT are hypomethylated, erased, and apoptosis prone and therefore in need of specific supportive cells for their survival: Sertoli cells in the testis, granulosa cells in the dysgenetic gonad/ovary, and perhaps equivalent cells in the thymus and pineal gland. What these supportive cells probably have in common is expression of soluble and membrane-bound KITLG that may activate the KIT receptor expressed on PGC/gonocytes, thereby enhancing their survival and proliferation [290–292]. Moreover, assuming a similar role for AKT in PGC/gonocytes as in EC, KIT signaling may through phosphorylation of OCT4 by AKT be involved in maintenance of the undifferentiated PGC phenotype [293].

Type I GCT occur at all sites of type II GCT and in addition in many more anatomical localizations along the midline of the body and occasionally in organs outside the midline. It seems that the requirements for the development for a type I GCT are less demanding than for a type II GCT, as mentioned. This may be explained by the PGC/gonocytes giving rise to type I GCT still being in an earlier stage and therefore still methylated, pre-erased, or rarely partially erased, and thus less fragile than the more mature, hypomethylated, erased PGC from which the type II GCT originate. Finally, and perhaps most importantly in view of the animal models of type I GCT discussed earlier, the PGC giving rise to type I GCT, because they lack a proper niche, probably do not survive as such but only if reprogrammed to pluripotent, primed state-ESC.

Testicular type II GCT are virtually always peritriploid [294, 295], whereas in the ovary [296], mediastinum [297], and brain [197, 298], type II GCT may be (near)diploid or (near)tetraploid, reportedly in up to 50 % in the brain [197]. The consistent peritriploidy of the testicular tumors is probably due to the older age of clinical manifestation than at the other sites and the long preceding period of intratubular development with concomitant karyotype evolution.

Regardless of anatomical site, type II GCT are characterized by gain of (parts of) the short arm of chromosome 12, usually in the form of an isochromosome of 12p (i(12p)) [299, 300] (Fig. 3.15). In the testis, it occurs in virtually 100 % [295, 301–304], in the ovary/dysgenetic gonad in about 75 % [296, 305, 306], in the mediastinum in 87 % [307, 308], and in the midline of the brain in 60 % [309, 310]. Also just the more proximal parts of 12p may be involved, as an amplicon, in particular 12p11.2-p12.1, specifically in the invasive components [311–314]. It looks as if the proportion of tumors with 12p gain is inversely related to the age of clinical presentation. The very high proportion of 12p gain in testicular tumors is probably, like their peritriploidy, due to the long period of intratubular karyotype evolution.

Isochromosome 12p arises from an erroneous centromeric division during mitotic anaphase preceded by tetraploidization [315] and is of uniparental origin [316]. Among the genes involved in the 12p aberrations are NANOG, STELLAR, GDF3, and EDR1, necessary for maintaining pluripotency; cyclin D2 and KRAS providing proliferative advantage; genes involved in glucose or glycolytic metabolism, including GLUT3, GAPDH, and TPI1 for energy metabolism in a low-oxygen environment [304, 313, 317, 318]; and genes involved in suppression of apoptosis such as EKI1, SOX5, and DAD-R [312, 313]. Expression of these genes maintains the PGC/gonocyte-phenotype of the tumor cells and allows them to survive and proliferate in the proper niches.

With rare exceptions [306], the tumors have over- and underrepresentation of (parts of) chromosomes other than 12p. Gains involve chromosomes X, 7, 8, 12, and 21 and losses the chromosomes Y, 1p, 11, 13, and 18. The overall pattern is consistent with early tetraploidization of the tumor cells, possibly as a result of malfunction of the mitotic-meiotic switch [319], followed by net loss of chromosomes due to nonrandom losses and gains of (parts of) chromosomes [294, 320]. The large stretches involved, often entire chromosome arms, suggest that aberrant meiotic division may have a role in the evolution of the chromosomal aberrations [197].

Type II GCT are chromosomally instable, probably due to their hypomethylated [321] and polyploid genome, and therefore subject to a continuous reallocation of chromosomal material between chromosomes (Fig. 3.16). Upon cell division, this may cause unequal distribution of chromosomal material over the two daughter cells, resulting in different gene dosage [304]. Chromosomal instability likely drives tumor progression of type II GCT, exemplified by the increasing gain of entire chromosome 12, 12p (among others KRAS), 12q (KITLG, located on 12q) [296], and 4q12 (KIT) [322], which renders the neoplastic gonocytes ultimately feeder cell independent and endows them with invasive capacity. Interestingly, in non-seminoma where because of reprogramming KITLG is no longer advantageous to the tumor, 12q13-q22 including KITLG is often deleted [323]. Chromosomal instability may also drive cell fate decisions in the tumor cells, e.g., by influencing the balance of signals promoting germline (BMP) versus embryonal phenotype (NODAL) [255], and thereby the reprogramming of a seminomatous cell to a totipotent EC cell. Finally, in a non-seminoma, it could be involved in lineage determination, e.g., by tipping the balance of normally maternally and paternally expressed genes favoring somatic and extraembryonic/trophoblastic differentiation, respectively.

Mutations and amplifications of oncogenes are rare in type II GCT, with 0.5 mutations per Mb [295] lower than in any other solid cancer of adults [324]. KIT is most frequently mutated and mainly involved in seminomatous GCT: seminoma of the testis, about 30 % [295, 324–329]; dysgerminoma of the ovary, up to 50 % [326, 330–332]; seminoma of the mediastinum, 38 % [333]; and germinomas of the brain, over 50 % [197, 334]. In non-seminomatous tumors, KIT mutations are rare, less than 1.5 %; the same low figure was reported in gonadoblastoma and derived germinomas [332]. Functional studies have shown the KIT mutations to be activating [326, 327], occurring predominantly in the activation loop (exon 17, usually in codons 816, 820, 822, 823, and 825) of the second TK domain [335].

In seminomatous tumors, mutations of KIT appear to be only one of the mechanisms of activation of KIT signaling and its downstream pathways: KRAS/RAF/MEK/ERK and AKT/mTOR. In fact, in these tumors, KIT signaling is always activated either by upregulation of expression or genetically by mutation or amplification [322]. In the TCam-2 seminoma cell line, siRNA-induced reduction of KIT expression reduced the viability of the cells, although only marginally [329].

The quoted studies report that activating KRAS mutations occur in a few percent of type II GCT, more or less at the same rate in seminomatous and non-seminomatous GCT. Earlier studies had found higher figures in testicular seminomas, with subclonal activating KRAS and NRAS mutations in 40 % [325] and activating KRAS mutations in 2/15 cases (15 %) [336] (Fig. 3.17).

KIT and KRAS mutations are mutually exclusive in type II GCT of the brain [334] and perhaps also in those of the testis [324]. Consistent with this observation, “large-scale” gain of 12p (harboring KRAS) seems also to be mutually exclusive with mutations of KIT both in seminomas of the testis [295] and germinomas of the brain [197]. This phenomenon may explain why in non-testicular seminomatous tumors, where gain of 12p is less frequent, KIT mutations are more common.

The fact that KIT mutations are predominantly found in seminomatous tumors and only rarely in non-seminomas suggests that they, like upregulation and amplification of the gene, are in general involved in the progression of seminomatous tumors, rather than in initiation of type II GCT. Further evidences that KIT mutations are most often related to progression of seminomatous tumors include:

The report that virtually all bilateral type II GCT, including non-seminomas, had KIT mutations, usually in codon 816 and often with the same mutation on both sides, has strongly incriminated KIT mutations as an initiating event in migrating PGC prior to their reaching the genital ridges [338]. Later studies could not reproduce these findings neither in bilateral tumors of the testis [283, 322] nor of the ovary [330]. Other studies did show more (14/22, 64 %) [339] or less (2/7, 28 %) [340] preference of KIT mutations for bilateral tumors. Whole exome sequencing of 42 testicular type II GCT [295] demonstrated KIT mutations in 3/9 bilateral cases (33 %) and in 3/33 unilateral cases (9 %). It seems, after all, that bilateral cases do have KIT mutations at a somewhat higher rate than unilateral cases, in support of a possible initiating role of this genetic event. Notably, the most frequent mutation in bilateral cases and unilateral cases is probably different, respectively, Y823D and D816V [339]. Also in favor of initiation, the same KIT mutation (A816V) was found in the testicular seminoma and the pineal germinoma of the same patient [283]. This pathogenetically revealing case demonstrates two significant points: initiating KIT mutations may occur in migrating PGC and extragonadal type II GCT of the brain may be derived from mis-migrated PGC.

Indeed, upregulation of KIT signaling [341] and mutant KIT [342] can transform cells; thus, it is very likely that KIT mutations may occasionally be the initiating event in PGC, which during migration depend on KIT signaling for survival and proliferation.

It has been proposed that when a KIT mutation is the initiating event, the transformed PGC will preferentially develop as seminoma [327]. The rarity of KIT mutations in non-seminomas then could be due to the unlikeliness that a seminomatous tumor cell with a KIT mutation is reprogrammed to an EC cell. If this were true, one would expect to find KIT mutations preferentially in GCNIS adjacent to seminomas with a KIT mutation and vice versa; this has not yet been investigated. This hypothesis would also predict the absence of KIT mutations in the seminoma-component of mixed non-seminomas where reprogramming of an invasive seminoma cell has given rise to the non-seminoma component. In the brain 4/8 (50 %) mixed GCT including a germinoma component showed a KIT mutation [197], suggesting that a KIT mutation is compatible with reprogramming in a seminomatous GCT.

Tumor suppressor genes seem to play a modest role, perhaps not surprising in view of the early tetraploidization in the pathogenesis of type II GCT, making loss of heterozygosity less likely to develop. In addition, inactivation of tumor suppressor genes by promoter hypermethylation is rare in type II GCT [343, 344].

The overwhelming male predominance of type II GCT may be explained by the role of co-expression of OCT4 and TSPY [345–347] in their pathogenesis, as it appears from the study of these GCT in DSD patients [290, 291], cryptorchid testis [348], and complete androgen insensitivity (CAIS) [349] (Sect. 3.6.2.1). Notably, TSPY is present in about 35 copies [350] in the GBY region of the male-specific region on Y [351].

Similarly, in the proper niches in the thymus and midline of the brain where sporadic mis-migrated, but not yet demonstrated PGC [52], may survive, co-expression of these two proteins may drive transformation of PGC. In view of the predominance of males, this pathogenetic mechanism is probably the most important but not the only one to explain the origin of type II GCT, as they do occur in phenotypically normal females [352], whereby the possibility of constitutional or chimeric mosaicism for the GBY region of Y has to be kept in mind [353]. In normal females, somatic mutations, in particular KIT mutations, may indeed be the initiating event of type II GCT as illustrated by the study of Hersmus et al. [332], where KIT mutations were found in 53 % of ovarian dysgerminomas of normal women and in only 6 % of dysgerminomas originating in gonadoblastoma in DSD patients.

It seems that type II GCT have at least two fundamentally different pathways of origin: first and foremost, if GBY/TSPY is present, the “developmental pathway,” due to disturbed maturation of PGC leading to co-expression of embryonal proteins in particular OCT4 and early differentiation genes in particular TSPY (followed by overexpression of KITLG in the supportive cells) [290, 291, 348, 349]; second, and much rarer, the “somatic mutation pathway,” in which mutations in oncogenes, in particular in KIT and RAS, are initiating events. In view of the mutation frequency of these two genes in non-seminomatous tumors, the “somatic mutation pathway” seems to occur in less than 2 % of the cases. The frequent mutations in KIT in seminomatous GCT are in general progression related, as discussed. In normal females, obviously lacking GBY/TSPY, initiation of type II GCT by somatic mutations is probably a more common mechanism.

A brief summary of the progression of type II GCT, with emphasis on the various roles of KIT activation, is at this point appropriate. Early progression is sustained by upregulation of KITLG, followed by tetraploidization, whereby probably gain of chromosome 12, with KITLG on 12q and among many others KRAS on 12p, is most significant. In the in situ precursor lesions prior to the development of an invasive tumor, best studied in GCNIS, progression is driven by nonrandom gains and losses of parts of chromosomes harboring genes that promote feeder cell independence, most conspicuously gain of 12p [315]. Activation of KIT (gene on 4q12) signaling via upregulation, mutation, or amplification of the gene is essential in the progression to seminoma. Upon reprogramming of a seminomatous cell into an EC cell, factors favoring seminomatous cells are no longer relevant, as exemplified by the loss of 12q, harboring KITLG in non-seminomas [323].

Mutations common in adult cancers, like in p53, appear in type II GCT when they acquire resistance to cisplatin-based chemotherapy [354] or progress to somatic-type malignancies, most often in (late) recurrences [244, 268, 355]. Of interest is the observation that expression of a specific set of miRNAs, i.e., 372 and 373, might function as an alternative for inactivation of p53 in the pathogenesis of type II GCT [356].

Except for teratoma components, type II GCT, including GCNIS and gonadoblastoma, are characterized by global demethylation, erasure of parental imprinting, and the presence of permissive histone modifications [87, 357–364]. Only Alu repeats have been reported to be methylated in non-seminomas [344].

Specifically the developmentally important miRNAs, miR-371-373 as well as miR-302 and miR-367 [356, 365], which have a crucial role in development of embryonic stem cells, are highly expressed in type II GCT, including GCNIS [366].

Global demethylation and this typical expression pattern of miRNAs are part of the phenotype of PGC, underscoring the origin of type II GCT from these germ cells committed to totipotency and confirming the phenotypic similarity of PGC and the cells of the precursors GCNIS [367–370] and gonadoblastoma [290–292], respectively, in the testis and the dysgenetic gonad/ovary.

Type II GCT are the solid tumors in adults with the highest sensitivity to DNA-damaging agents, where, e.g., for testicular primary tumors, cure rates >80 % are achieved in disseminated disease [371]. Both seminoma cells and EC cells, the stem cells non-seminoma, are probably highly accessible for DNA-targeting drugs because of their open chromatin structure [344, 360, 372]. EC cells are a factor two to four more sensitive to cisplatin than the clonogenic cells of cell lines derived from common adult cancers [372]. Seminomas are exquisitely sensitive to radiotherapy and cisplatin-based chemotherapy. The typical failure of seminoma cells to repair radiation-induced double-stand breaks, requiring homologous recombination, may be related to premeiotic and embryonic characteristics of the neoplastic gonocyte. A significant change brought about by the reprogramming from seminoma to non-seminoma is that this high radiosensitivity is lost.

Progression from a gonocyte to a type II GCT, particularly in the “developmental pathway,” requires fewer genetic changes than in the development of virtually any other type of cancer. Significantly, the most stringent barrier to immortalization and thereby carcinogenesis is progressive shortening of telomeres due to repression of telomerase activity [373]. This barrier is absent in PGC and ESC, as telomerase is indeed active, and thus immortality is part of their normal [374] and neoplastic phenotype, except for mature teratoma [375–377]. Accordingly, mutations are rare in testicular type II GCT [295, 324]; there is no selective pressure for loss of function of p53 [356], and the DNA-damage response remains intact (for review [167, 354, 378]). In fact, it is hypersensitive, reflecting the physiological situation in germ cells and embryonal cells, characterized by inducibility of wild-type p53 jointly with the absence of p21-induced cell cycle arrest [379], whereby cells with damaged DNA are not repaired but rather eliminated due to a low threshold for apoptosis. This preference of apoptosis over DNA repair is a physiologic mechanism protecting against propagation of repair errors via germ cells into the next generation or via ES cells into the developing embryo.

The open chromatin structure and the hypersensitive germ cell/embryonal phenotype get lost upon somatic differentiation into teratoma [380, 381, 382] of which the adult tissue stem cells are well equipped to survive DNA damage by prolonged G1 and G2 arrest and proficient repair (for review [167, 354, 378]). The phenomenon that residual teratoma after chemotherapy is usually associated with primary tumors with a teratoma component demonstrates that it results from selective survival of somatically differentiated cells rather than induction of somatic differentiation due to chemotherapy [271, 383].

Primary and acquired resistance is relatively rare in type II GCT, probably because of the low mutation rate in these tumors [324, 378]. Mutations involved in resistance have been found in BRAF, correlated with MSI [384], and in the DNA repair gene XRXX2, promoting cisplatin resistance in animal studies [295]. Also mismatch repair deficiency was found correlated with treatment failure [384]. Further molecular mechanisms of (cisplatin)resistance, partly the same as occurring in spontaneous somatic differentiation, are: somatic differentiation accompanied by downregulation of OCT4 (e.g., as a result of hypoxia or treatment with retinoic acid); failure to induce the apoptotic factors Puma and Noxa; changes in the expression levels of miRNAs such as miR-17, miR-106b, and miR-302a or miR-371-373; elevated levels of MDM2 and cytoplasmic translocation of p21 by phosphorylation; activation of the PDGFRβ/PI3K/pAKT pathway [354]. Jointly, these molecular mechanisms explain only the minority of therapy-resistant cases. Development of resistance usually occurs without obvious deviation from typical GCT morphology, mainly as YST but also as EC, choriocarcinoma, and seminoma [385].

A final important mechanism causing resistance of type II GCT is further progression of teratoma and YST due to accumulation of mutations commonly found in adult cancers. Twenty percent of late relapses of testicular type II GCT, defined as recurrences more than 2 years after initial complete response, contain histological elements resulting from further tumor progression with morphologies not typical of type II GCT [386], so-called somatic-type malignancies, which may be derived from teratoma or YST (see also Sect. 3.6.2.2).

In addition to 12p aberrations [387–389], the somatic-type malignancies may have the genetic characteristics of their somatic counterparts, like 2q37 rearrangements in rhabdomyosarcoma, p53 mutations in sarcomas, and t(11;22) translocations in PNET [355]. A somatic-type malignancy confined to a stage I non-seminoma, occurring in about 5 %, does not adversely affect prognosis [390].

Cisplatin-based chemotherapy of non-seminoma is, in fact, a model for stem cell therapy of a solid tumor showing that eradication of the stem cell population does not necessarily cure the patient [268, 391]. The surviving committed stem cells and differentiated cells may possess or acquire clonogenic potential and appear as therapy-resistant recurrence.

Type II GCT are probably unique for humans, as no convincing examples of spontaneous or induced type II GCT have been reported in animals. Neoplasms of primitive germ cells in mice [392] and fish, e.g. experimentally induced in zebrafish [393–395], have been reported; however, reprogramming of a neoplastic primitive germ cell into a totipotent GCT, a key feature of type II GCT, has to the best of our knowledge not been described.

The only experimental model for seminoma is the TCam-2 cell line, derived from a human primary testicular type II GCT with a seminoma component, which can be propagated in vitro and as xenograft in immune-compromised mice [396] due to a BRAF mutation making it more apoptosis resistant than seminoma cells normally are [397, 398]. TCam-2 cells were at the molecular and epigenetic level characterized as seminoma cells [329, 398], including the expression of OCT4 in combination with SOX17 [399]. As expected, TCam-2 cells resemble human PGC, cells committed to totipotency, whose fate is determined by SOX17 and by BLIMP1 that represses differentiation into endodermal and other somatic lineages in PGC by repressing SOX2 [35]. Indeed, by inhibition of BMP, TCam-2 cells could be reprogrammed to an EC phenotype (among others expression of SOX2 instead of SOX17; genome-wide DNA methylation) via NODAL signaling [255, 400, 401], demonstrating in vitro what was first hypothesized from pathological observation [245].

The absence of type II GCT in mice may be explained by the molecular mechanisms of specification and epigenetic modification of the early germline being different in mice and humans. Particularly relevant could be that Oct4 in murine PGC is co-expressed with Sox2, like in mouse ESC, whereas OCT4 in human PGC is co-expressed with SOX17 [35]. Moreover, as mentioned before, epigenetic reprogramming of mouse PGC during embryonic development takes place in 24 h [8], whereas in humans, this process takes several weeks [6, 236]. This situation in humans creates a longer time frame for neoplastic transformation of early germ cells with a totipotent developmental potential, which in addition have the obstacle that they have to switch from SOX17 to SOX2 before being able to revert to an EGC phenotype. The very frequent step of tetraploidization early in the pathogenesis of type II GCT suggests that it is important for maintenance and survival of totipotent tumor cells in the naïve state.

In mice, and possibly also in other animals, the time frame for generating neoplastic totipotent cells is short, probably not long enough for the necessary steps, including polyploidization, to give rise to a tumor of transformed PGC/gonocytes. In mice, PGC/gonocytes can only give rise to neoplasms if they are directly reprogrammed to a diploid, pluripotent ESC in the primed state, probably because it expresses Oct4 in tandem with Sox2, as in ESC. As mentioned before, PGC that are not reprogrammed die apoptotically. The various teratoma models in mice are not representative for type II GCT but rather for type I GCT (Sect. 3.5.1.6). Several features of the models that do help to understand type II GCT are addressed throughout this chapter, where appropriate.

In humans, the PGC/gonocyte phenotype is indeed fairly stable in type II GCT, as at each anatomical site, over half are seminomas, the default state of these tumors. At the same time, it is obvious from the biology and histology of these tumors that reprogramming of a seminomatous cell (GCNIS or seminoma, primary or metastatic) to a totipotent EGC-like cell giving rise to a non-seminoma is a regular, though poorly understood event. To study this phenomenon, more typical seminoma cell lines would be helpful.

Non-seminoma cell lines, derived from in vivo reprogrammed seminomatous precursors, have been less difficult to establish than seminoma cell lines and are readily available. The first clonal cell line, and probably best characterized and intensely studied in vitro and in xenografts, is NT2D1, derived from Tera-2 [402]. It has been widely used, among others non-seminoma cell lines, to study differentiation of EC cells as a model for human embryonic development before human ESC became available [403, 404].

Because of the paucity of models for type II GCT, their pathogenesis has been studied primarily in human tumors in a multidisciplinary approach, combining epidemiology, pathology, (cyto)genetics, cell biology, and molecular approaches, as will be discussed per anatomical site in the following paragraphs.

Much of the pathogenesis of type II GCT has been learnt from the study of gonadal dysgenesis in DSD and its typical type II precursor lesion gonadoblastoma, in which the inception of these GCT can be closely followed. It has provided crucial insight into the interactive role of supportive cells and tumor cell-intrinsic factors in the pathogenesis of type II GCT.

Again, the important lesson to be learned from gonadoblastoma is that there are probably two pathogenetic pathways for the origin of dysgerminoma: the first, as in DSD, merely by disturbance of the normal development of the gonad, the “developmental pathway,” related to GBY enabling co-expression of TSPY and OCT4, and the second, much rarer pathway, the somatic mutation pathway involving mutations, particularly in KIT. The somatic mutation pathway is probably more frequent in normal females, as discussed earlier (Sect. 3.6.1.4) [332].

The large incidence differences among ethnic groups within the same society, e.g., Caucasians and blacks in the USA [76], demonstrate the importance of genetic factors. On the other hand, the geographic pattern of increasing incidence of testicular type II GCT, the changing incidence among certain immigrant populations [440–442], points to an important causative role of environmental factors, associated with modern, Western lifestyle that probably favor hypovirilization of the developing male embryo [167, 370, 443]. Indeed, the young age of the patients, the bell-shaped age distribution, the decreased risk for testicular type II GCT in Danish men born during World War II [444] and similar birth cohort effects in other populations, and the evidence that type II GCT are derived from gonocytes; all these observations point to their initiation during embryonal development.

Factors increasing risk with a lower OR, in the order of 1.3 or less, which also one way or another may relate to disturbed hormonal conditions in utero are maternal bleeding, low birthweight, short gestational age, twin, tall stature, and being first born child. Sibship size is protective, the more siblings the lower risk, OR 0.80, as is late puberty, OR 0.81 [453]. There is sufficient evidence to support a relation between testicular type II GCT and three widely used hormone disruptive organochlorine insecticides (dichlorodiphenyldichloroethylene, cis-nonachlor, and trans-nonachlor), which may have a hypovirilizing influence on a developing male embryo in a specific window of development (masculinization window) [454].

The two identified occupational risks (fire fighting and aircraft maintenance) [455] and cannabis smoking [456] obviously affect males postnatally, in adolescence and adulthood. It might be assumed that these risk factors modify the development of already existing precursor lesions. Cannabis smoking is particularly interesting because it selectively increases the risk of non-seminoma, suggesting that it might promote reprogramming of a seminomatous precursor cell.

Indeed, the small size of affected families, usually a father and a son or two brothers, and the high risk in monozygotic compared to dizygotic twins [457] are consistent with multiple autosomal recessive low-penetrance susceptibility genes [286, 288, 461–463].

The polymorphic gene variants increasing susceptibility to type II GCT shown in recent genome-wide association studies (GWAS) are also consistent with genetic susceptibility being the result of multiple common, relatively low-penetrance gene variants. The first variants (KITLG, BAK1, SPRY4) were demonstrated in 2009 by two independent studies [195, 196]. As of 2015, over 30 variants of genes for proteins which are plausibly involved in the biology of testicular type II GCT have been published, e.g., KIT/KITLG signaling (KITLG, SPRY4, BAK1, PDE11A) in relation to PGC survival and proliferation; DMRT1 variants, involved in sex determination and regulation of meiotic division; genes involved in telomere maintenance, testis differentiation, and sex determination (such as HPGDS), among others [288, 376, 464–467]. Together, these account for about 15 % of the excess familial risk to brothers of testicular type II GCT patients and 22 % of the excess to sons of testicular type II GCT patients [468]. Remarkably, these variants have no association with the established phenotypic risk factors: family history, cryptorchidism, inguinal hernia, age at diagnosis, and bilateral testicular type II GCT [469]. TDS, including cryptorchidism, hypospadias, male infertility, impaired testicular development, and testicular type II GCT [229] may be associated with variants of TGFBR3 and BMP7 [470] and variants in genes involved in steroid signaling but not with the established risk variants for testicular type II GCT.

Importantly, different distribution of variants in KITLG and AR in Caucasian and black populations may partly explain the 20-fold ethnic difference in the incidence of testicular type II GCT [195, 196].

It seems that some of the variants primarily target the PGC/gonocyte, in particular the KIT/KITLG pathway and others primarily the supportive cells (Sertoli and Leydig cells in the testis), and that homozygosity for risk alleles in the two pathways confers the highest risk. For example, men with testicular type II GCT have a 14 times higher chance than controls to be homozygous for the two risk alleles, KITLG (PGC/gonocyte survival and proliferation) and DMRT1 (testicular development) [464].

In keeping with the above observations, the association between increased risk for testicular type II GCT and impaired fertility might be due to variants within DMRT1 [464], comparable to the role of Dmrt1 in fertility and testicular teratoma formation in 129Sv mice [189]. Similarly, there might be a parallel between the loss of function mutation of steel/Kitlg in this mouse model, impairing fertility and enhancing teratoma formation [185] and variants in KITLG in men. Variants in KITLG resulting in loss of function may disturb the function of Sertoli cells and the germ cell niche, thereby impairing fertility and promoting initiation of type II GCT.

Specifically, the disturbed Sertoli cells/niche might interfere with downregulation of OCT4 in gonocytes relocated from the center of the tubules to the prespermatogonial niche, thereby creating a window for co-expression of OCT4 and TSPY, assumed to respectively protect the germ cells from apoptosis and to stimulate their proliferation [84, 477]. Co-expression of these proteins is in due time accompanied by increased expression of KITLG by the stromal component and/or via an autocrine loop, which supposedly further stimulates the neoplastic transformation. The association of a higher risk for testicular type II GCT with SNP variants in KITLG [195, 196], for example, might be mechanistically explained by interference with KIT signaling in Leydig cells and, as mentioned above, a disturbed interaction between Sertoli cells and gonocytes [348, 377]. Interestingly, one of the likely related variants within KITLG concerns a binding site for p53 [478], whereby the expression level of this allele of KITLG might increase under conditions of stress during early development, such as TDS/DSD [415], via upregulation of p53.

As in dysgenetic gonads, the neoplastic transformation of gonocytes does not require somatic mutations but rather results from a disturbed timing of expression of critical embryonal and differentiation proteins during development, as discussed. Therefore, type II GCT of the testis can be considered developmental tumors [370]. Indeed, somatic mutations are rare and limited to KIT and KRAS [295, 322–326, 490], whereby activating KIT mutations are found in about 25 % of seminomas [327] and rarely in non-seminomas [340]. In seminomas, the KIT pathway is always activated via mutation or amplification of KIT or via overexpression of the protein [322, 491]. KRAS mutations are in a few percent found both in seminomas and non-seminomas. Mutation of KRAS does not seem to occur in combination with high-level amplification of the gene [324] or overexpression of the protein [329]. In testicular type II GCT, KIT and RAS mutations may be mutually exclusive [324].

The much higher frequency of activating KIT mutations in seminoma than non-seminoma suggests that this genetic event does not take place in an early stage of tumor evolution shared by seminoma and non-seminoma. Rather, as discussed before, mutation of KIT is part of progression of seminoma, as is amplification of the gene, and upregulation of its expression [322] resulting in KIT activation, characteristic for seminoma. Since it is related to seminoma progression, it is understandable that non-seminomas rarely harbor KIT mutations, for they offer no advantage after reprogramming of a seminomatous tumor cell into an EC cell. In the rare non-seminomas with a KIT mutation, reprogramming probably took place in a seminomatous tumor cell that already had acquired a KIT mutation. Indeed, one of the first two type II GCT in which an activating KIT mutation was demonstrated was an ovarian mixed dysgerminoma/YST, with the mutation in both components [326].

The observation that KIT mutations are usually found in seminomas that lack large-scale 12p amplifications [295] could mean that for seminomas, KIT mutations and 12p gain are alternative pathways of tumor progression.

The occurrence of the same KIT mutations in bilateral testicular tumors demonstrates that a KIT mutation can be the initiating event taking place in PGC prior to their arrival in the gonadal ridges [283, 339, 340]. Even more convincing, as already referred to, the same KIT mutation was found in the testicular seminoma and the pineal germinoma of the same patient [283].

GCNIS stays more or less dormant until resuming further progression upon hormonal stimulation at puberty [492]. In prepubertal GNCIS, about a quarter of the tumor cells express DMRT1, a key regulator of the mitosis-meiosis switch, whereas in adult cases, this figure drops to a few percent. GCNIS cells that express DMRT1 are not mitotically active and considered dormant [486].

In so-called isolated GCNIS that has not yet given rise to an invasive type II GCT [493] and in GCNIS confined to the spermatogonial niche [304], there is no overrepresentation of 12p (Fig. 3.20). Gain of 12p coincides with the GCNIS cells becoming feeder independent, apparent from their leaving the spermatogonial niche and their capacity to float in the lumen of the seminiferous tubules [304], usually adjacent to an invasive GCT [493] (Fig. 3.21). The next step in the “default development” of a testicular type II GCT is the formation of intratubular seminoma, whereby the GCNIS cells completely fill and distend the seminiferous tubules and oust the Sertoli cells. Intratubular seminoma may contain lymphocytes like invasive seminoma. Though highly proliferative it virtually never shows necrosis, which is practically always present in intratubular non-seminoma. Apparently, intratubular seminoma cells, like normal gonocytes, are well adapted to the intratubular low-oxygen environment [244].

The trigger for invasive growth of intratubular seminoma is not known. Morphologically, invasion of seminoma has different appearances: one whereby intratubular seminoma extends the seminiferous tubule to the point of breaching the tubular wall [262] and another, microinvasive seminoma in which the tumor cells appear as single cells in the interstitial stroma of the testis [244, 494–496], and combinations of both. Just like intratubular seminoma [262], microinvasive seminoma can be found adjacent to a non-seminoma in about 20 % [494]. Microinvasive growth may be due to expression of matrix metalloproteinase 9 and plasminogen activator, urokinase, by the tumor cells [495].

Upon invasion, seminoma invariably elicits an inflammatory host response, usually composed of lymphocytes, macrophages, plasma cells, and often a granulomatous reaction. Its significance remains controversial. Recently, it was suggested that it is not involved in active immune surveillance [497]. An earlier study demonstrated clonally expanded cytotoxic T cells and evidence of specific and functional T-cell responses operating in seminoma, indicating that the inflammatory infiltrate is indeed involved in the immunological control of the tumor; however, class I MHC molecules could not be demonstrated on the seminoma cells [498], making them invisible to cytotoxic T cells. From histology, the infiltrating lymphocytes seem capable to cause complete regression of seminoma, leaving the scar of a so-called burnt-out tumor. GCNIS and intratubular seminoma may undergo regression as well, whereby the tubules become atrophic and in the end completely fibrosed. This host reaction, and probably the older age of the patients, explains why GCNIS is usually much less extensive in association with seminoma than non-seminoma and even absent in up to 15 % of the cases [262]. The few lymphocytes accompanying GCNIS adjacent to non-seminoma, as opposed to the many adjacent to seminoma, suggest that the host response is indeed elicited by the seminoma, which upon invasion disturbs the mechanisms of immune privilege in the testis [499]. GCNIS, composed of tumor cells that are phenotypically similar to seminoma cells, is probably secondarily involved. It seems less likely that the GCNIS cells themselves, within the intact immunologically privileged testis, trigger a reaction of the host.

The host response is probably clinically relevant in view of the 10 years difference of the median age of presentation of seminoma in patients with AIDS and the general population: respectively, 25 and 35 years. In AIDS patients, seminoma and non-seminoma present at the same age, 25 years, also the age of presentation of testicular non-seminoma in the general population. In addition, a higher risk for disseminated seminoma has been reported in patients with AIDS, suggesting a protective role of an intact immune system [500]. Invasive non-seminoma also contains inflammatory cells; however, the surrounding parenchyma is much less involved than with seminoma, probably explaining why GCNIS is often very extensive and rarely absent [262].

With sporadic exceptions, intratubular non-seminoma is composed of pure EC that is partly necrotic and often calcified. The invariable necrosis indicates that EC cells are less adapted to the intratubular low-oxygen environment than GCNIS/seminoma cells. It is conceivable that hypoxia-induced factors like MET trigger invasion of intratubular non-seminoma. Remarkably, only upon invasion, the EC cells start to differentiate and display the totipotent nature of naïve-state EGC due to differentiation-inducing factors in the stroma of the testis and/or loss of differentiation-inhibiting factors in the intratubular microenvironment (Fig. 3.23). Possible stromal factors are TGF-β, FGF, and BMP, which may derepress differentiation-promoting genes by removing polycomb repressive complexes recruited to the promotor sites of these genes by the pluripotency proteins OCT4, SOX2, and NANOG [502]. Noteworthy, spontaneous and experimental teratomas in 129 mice also start as intratubular EC and begin to differentiate when the seminiferous tubules are extended and disrupted [81, 183] (Fig. 3.24).

The ratio between dysgerminoma and non-dysgerminoma, and thus the rate of reprogramming of dysgerminoma, is difficult to establish because usually epidemiologic studies do not specify the histology to the degree that the distinction between type I, II, and IV GCT is possible. The report by Smith et al. [504] is a notable exception: among 1262 malignant ovarian GCT, the 449 pure immature teratomas and at least 110 pure YST (60 % of 183, based on Kraggerud [296]) should be considered as type I GCT, while the 37 teratomas with malignant degeneration should be type IV GCT with a somatic-type malignancy. That leaves 666 type II GCT: 414 dysgerminomas (62 %) and 252 non-dysgerminomas (73 pure YST, 67 mixed GCT, 52 EC, 27 choriocarcinomas, and 33 EC plus teratoma, so-called teratocarcinomas). The percentage of pure dysgerminoma is higher than the slightly over 50 % seminomas within testicular type II GCT. It suggests that reprogramming in ovarian dysgerminoma, which is convincingly documented [505–508] and also apparent from the mixed GCT with a dysgerminoma component [504], occurs less frequently than in GCNIS and seminoma of the testis. The spectrum of histological types in ovarian and testicular non-seminomatous GCT is similar; however, the distribution is different: mixed non-seminoma is the most frequent histology in the testis and pure YST in the ovary; pure choriocarcinoma seems to be more frequent in the ovary than in the testis. In mixed tumors, the percentage of tumors combining dysgerminoma with just immature teratoma or YST seems to be higher than the combination of seminoma with only immature teratoma or YST in the testis (Chap. 6).

Type II GCT of the ovary may arise before puberty, likely in DSD [275, 513] (see also Sect. 3.6.3). DSD may remain unrecognized, as in the case of two phenotypically normal females having a 46,XX karyotype, who well after puberty were diagnosed, respectively, with gonadoblastoma with a germinoma [514] and gonadoblastoma with a mixed type II GCT [515]. Ninety-five percent of ovarian type II GCT become manifest after puberty in women with a normal 46,XX karyotype [516], a couple of years earlier than in males, in accordance with the earlier onset of puberty in females [232, 266, 517]. Age of presentation is in the typical order: first, the pure non-dysgerminomas, followed by the mixed non-dysgerminomas/dysgerminomas, and then the pure dysgerminomas with a median age of the latter close to 20 years [511]. Dysgerminomas and non-dysgerminomas are bilateral in over 6 % [129, 131, 518–520], a slightly higher figure than for bilateral type II GCT of the testis, probably due to the contribution of gonadoblastoma-related tumors, as 40 % of gonadoblastomas are bilateral [405]. In fact, in a retrospective study, one out of three bilateral dysgerminomas was associated with bilateral gonadoblastoma [521].

Familial cases are rare: among 18 families retrieved from the literature with at least one female with a GCT [137, 522, 523], eight involved only females and ten both females and males (in about 0.2 % of the pedigrees of familial testicular cancer, a female member has a GCT [524]). The information on the histology of the tumors provided in the reviews and the quoted original case reports allows the distinction between GCT of types I and II. In 12 of the families, at least one case concerned a dysgerminoma, combined with various type II GCT of the gonads in close relatives of both sexes; remarkably, one relative had a mediastinal EC. Also noteworthy, in three families, type II GCT clustered with at least one ovarian type I GCT: three pure (immature) teratomas and in two families with pure YST, which may have been of type I. (The remaining family consisted of a woman with an ovarian type I immature teratoma and her baby with a type I immature teratoma of the brain [136].) These families demonstrate two significant points: type II GCT of the ovary may cluster with similar testicular and extragonadal GCT and also with type I GCT of the ovary.

Less certain is that a disturbed hormonal milieu in the mother increases the risk of malignant ovarian GCT in daughters, comparable to the role of hormone disruption in the etiology of TDS and type II GCT of the testis [229, 525]. Reported risks are maternal use of exogenous hormones during pregnancy (OR 3.6), maternal obesity (OR 2.7), early regular menstruation after menarche (OR 1.8), and age at index pregnancy under 20 years (OR 2.8) [526]. Other studies have also reported association of malignant ovarian GCT with reproductive risk factors such as parity, use of contraceptives, ages at first and last births, and time since last birth [527–529]. A more recent study could not link levels of circulating sex hormones with risk of ovarian germ cell cancers [530]. Most studies do not distinguish between type I and II GCT, making the results difficult to interpret.

As discussed above, type II GCT of the ovary lack the strong familial risk of those of the testis. Sporadically, type II and I GCT cluster in families. Both tumor types are so rare that the clustering is probably not by chance, rather these families have susceptibility for both type I and II GCT, pointing to a common cell of origin in different states of developmental potential. The obvious target cell is the PGC/gonocyte, and the familial susceptibility factor might have bearing, for example, on resistance to apoptosis of PGC, such as variants in BAK1, which are associated with a higher risk for type II GCT [195, 196] and perhaps also for type I GCT [158].

In one of the families reported by Huddart [524], a male with a testicular type II GCT clustered with a female with bilateral dermoid cysts (type IV GCT) of the ovary. More recently, a similar family was identified, with a father having a seminoma and his daughter metachronous bilateral dermoid cysts [531]. In both families, it may be a chance occurrence in view of the high frequency of dermoid cysts of the ovary, with bilaterality in 10–15 % of the patients [233, 532] (Chap. 6). It is intriguing though that in both females, the dermoid cysts were bilateral, because in familial cases of dermoid cysts where laterality was stated, 11/28 (39 %) were bilateral, and among the patients who were twins or triplets, 9/12 (75 %) were bilateral [134, 135]. One of identical twins [533] had over the years seven dermoid cysts removed from her left and one from her ovary. These case histories are suggestive of a genetic risk factor for bilaterality of type IV GCT of the ovary, which perhaps could also increase the risk for type II GCT.

As for the other chromosomes in dysgerminoma, the most common changes are gains from chromosome arms 1p (33 %), 6p (33 %), 12q (75 %), 15q (42 %), 20q (50 %), 21q (67 %), and 22q (58 %); gains of the whole of chromosomes 7 (42 %), 8 (42 %), 17 (42 %), and 19 (50 %); and losses from 13q (58 %) [535]. The strong predominance of gains over losses might be explained by the (near)tetraploidy of most dygerminomas.

Somatic mutations of KIT in exon 17 codon 816 have been found in 27–33 % of dysgerminomas [330–332] and in codon 822 in an additional 20 % [332], adding up to activating mutations in 53 %, always in unilateral and not in bilateral cases. Among 16 DSD patients with GBY in their genome who developed a dysgerminoma, only one case had a KIT mutation, in codon 820; the same mutation was found in the gonadoblastoma from which the dysgerminoma was derived [332]. This may be part of the explanation for the absence of KIT mutations in bilateral dysgerminoma, as bilateral tumors are probably in a substantial proportion derived from gonadoblastoma. KIT mutations are absent in other malignant ovarian GCT, i.e., immature teratoma, YST, and tumors combining these two components. Gain of 12q in dysgerminoma may be related to the localization of KITLG on 12q22, with possible involvement of KITLG in an autocrine loop with KIT. KRAS has hardly been investigated in ovarian type II GCT; in two studied dysgerminomas, it was not mutated [536].

Five percent of ovarian type II GCT develop in the context of DSD by the “developmental” pathway, related to the presence of GBY and co-expression of OCT4 and TSPY, consistent with the low rate (0.6 %) of KIT mutation in these tumors [332]. In a molecular analysis of 45 malignant ovarian GCT in patients without signs of DSD, 32 can be classified as type II GCT. In four of these (13 %, two dysgerminomas and two immature teratomas), TSPY was demonstrated in tumor tissue [353]. This would mean that the origin is “developmental,” in association with TSPY, in about 20 % of ovarian type II GCT (5 % DSD and 13 % clinically silent mosaicism for TSPY).

As discussed in the general section on (cyto)genetics of type II GCT (Sect. 3.6.1.4), the KIT mutations found in up to 50 % of dysgerminomas [330, 332] may be initiating, but they are probably most often progression related. In view of the absence of KIT mutations in type II non-germinomatous tumors of the ovary [330], it is likely that the situation is not much different from the testis and that also in the ovary, very few type II GCT are caused by somatic mutations. Thus the initiation pathway in over 80 % of ovarian type II GCT in 46,XX phenotypically normal females remains to be explained.

It is tempting to speculate that in the ovary, the majority of type II GCT develop in the context of mild dysgenesis, comparable to TDS of the male, mainly caused by imbalances of factors regulating gonadal development. It has been shown in a mouse model that downregulation of FOXL2, the key protein for maintenance of the female identity of the gonad, results in reprogramming of the ovary in the male direction, even in adults. Granulosa cells are transformed into Sertoli cells forming tubular structures. Remarkably, if in this situation the expression of FOXL2 is restored, the ovarian identity is repaired [537]. If this would happen during gonadal development, even only transiently, in the stage before oocytes become arrested in the prophase of meiosis I, it would create a hypovirilized testis-like environment favoring disturbance of maturation oogonia/gonocytes and a window for the development of a type II GCT. However, due to the absence of GBY/TSPY, at a much lower rate than in males or in DSD patients carrying GBY in their genome, who have an up to 70-fold risk of developing a type II GCT as compared to normal females [275]. Indeed, insight into the factors involved in the development and maintenance of sexual identity of the gonads (DMRT1, FOX9, and SOXL2) lends credence to the hypothesis that as yet pathogenetically unexplained type II GCT of the ovary have their origin in mild forms of ovarian dysgenesis, possibly even transient, which leave no obvious further phenotypical traces (Fig. 3.25). If this hypothesis were true, it would mean that the large majority ovarian type II GCT have a “developmental” origin.

The much lower incidence of type II GCT in the ovary as compared to the testis has been explained by the lower number of susceptible germ cells in the ovary, and the fact that they are blocked in meiosis I, whereas the gonocytes are more numerous in the testis, are arrested in mitosis [509]. This is probably only part of the explanation in view of the epidemiology of the type II GCT of the mediastinum and brain. One might assume that in these sites, the number of target cells is similar in men and women; moreover, the mis-migrated PGC are blocked in meiosis I in both genders [58, 59]. Yet type II GCT of the mediastinum [241] and brain [539] are much more frequent in males than females. As alluded to earlier, this may be due to men having the GBY region on Y and thus being able to express TSPY in combination with OCT4 in the critical stage of neoplastic transformation of erased PGC. This suggestion is consistent with the high risk of developing a type II GCT in dysgenetic gonads containing GBY/TSPY [275].

Also relevant in this context is the observation that type I GCT have a roughly similar frequency in the ovary and the testis and that the sex distribution, apart from the sacral region, is equal for most extragonadal sites, in particular for the mediastinum and brain [94]. It suggests that indeed the number of susceptible cells, probably mis-migrated, pre-erased PGC for type I GCT, is similar in males and females and that GBY has no role in the pathogenesis of type I GCT.

Of note, in patients older than six [76, 165], pure (immature) teratoma, pure YST, and mixed GCT combining the two can only with certainty be classified as type I or type II GCT by analysis of 12p status. This is clinically relevant as a type I teratoma is benign and type II teratoma is malignant.

Most intriguingly, 41 patients with Klinefelter’s syndrome who had a type II GCT based on histology and secretion β-HCG showed a very different distribution of histological types from the general adult male population, with seminoma 0 %, mixed non-seminoma 44 %, EC 10 %, choriocarcinoma 15 %, YST 2 %, and (immature) teratoma (29 %) [277]. Not only there were no pure seminomas but also only two of the 12 mixed GCT had a seminoma component. Choriocarcinoma/trophoblastic differentiation, either in pure form or as part of a mixed tumor based on histology or elevated serum β-HCG, was very common: 35/41 (85 %). Considering that over 20 % [540, 541], of all mediastinal GCT, are diagnosed in patients with Klinefelter’s syndrome, this may partly explain the histological differences between testicular and mediastinal type II GCT in adolescent and adult males.

Ten to twenty percent of mediastinal non-seminomas develop a solid somatic-type malignancy [542, 543]. The distribution of histologies is largely similar as in the testis, sarcomas being the most frequent type and among them rhabdomyosarcoma ranking first [430, 543]. In the mediastinal cases, angiosarcomas are more frequent than in the testis [241, 544]. The associated GCT are most often mixed non-seminomas with a (immature) teratoma component and rarely pure YST or seminoma [430]. Patients with Klinefelter’s syndrome may develop somatic-type malignancies other than hematopoietic malignancies.

Hematologic malignancies develop in 2–6 % [543, 545–547], sometimes combined with a sarcoma [388]. These hematological somatic-type malignancies are uniquely associated with mediastinal YST, either pure or as part of a mixed non-seminoma [265, 543, 547–549]; the most frequent types being megakaryoblastic leukemia, followed by malignant and benign histiocytosis and myelomonoblastic leukemia among many other types encompassing virtually all hematopoietic lineages [543]. Except associated with mediastinal non-seminomas, hematologic malignancy has been reported only once, in association with a suprasellar dysgerminoma [550]. Hematologic malignancies are usually diagnosed at a median time of 6 month after primary treatment [549]; about 40 % are synchronous with the primary mediastinal non-seminoma [543]. The rate is comparable in patients with and without Klinefelter’s syndrome, as about 20 % of the hematologic malignancies are diagnosed in patients with this syndrome, indicating that it is indeed the mediastinal localization that predisposes to this somatic-type malignancy. These hematologic malignancies are not treatment related [546, 549] but a peculiar biologic characteristic of mediastinal non-seminomas with a YST component [241, 547].

Added up, the solid and hematologic somatic-type malignancies develop in a quarter of mediastinal non-seminomas, which is about sixfold of the 3–6 % in primary testicular non-seminomas [390]. The higher rate of somatic-type malignancies is possibly due to the larger size at surgery of the non-seminomas in the mediastinum than in the testis [551] and due to the fact that surgery is always preceded by chemotherapy. Indeed, in originally testicular non-seminomas, the percentage of somatic-type malignancies increases to 8 % in postchemotherapy retroperitoneal lymph node dissection specimens [429] and to >20 % in late relapses [386], approaching the rate of somatic-type malignancies in mediastinal non-seminomas.

The poorer prognosis of mediastinal than testicular non-seminomas may be because of the larger size at clinical manifestation and the overall higher rate of somatic-type malignancies, which are largely resistant to the chemotherapy given for non-seminomas [354].

In a large Danish cohort, mediastinal non-seminoma was the only cancer for which Klinefelter’s syndrome conferred a higher risk compared to males without this syndrome, with a relative risk of 67 [554]. Over 20 % of mediastinal type II GCT are associated with Klinefelter’s syndrome [540, 541], meaning that roughly half of the mediastinal type II non-seminomas occur in this context, as pure seminomas do not seem to occur in Klinefelter’s syndrome. The age distribution of Klinefelter patients with mediastinal GCT is bimodal. However, the early peak (between age 4 and 10) is later than in typical type I GCT; moreover, the tumors in the early peak are not the usual type I GCT but more like type II GCT, based on the histological composition and the secretion of β-HCG causing precocious puberty [277]. Remarkably, the two mixed GCT with a seminoma component occurred in two 8-year-old boys. Klinefelter’s syndrome also increases the risk of type II GCT of the brain [555, 556].

Sporadically, mediastinal seminomas have been diagnosed in individuals with Down’s syndrome [557, 558]. Neurofibromatosis 1 [559, 560] and Li-Fraumeni syndrome [561, 562] are not associated with mediastinal type II GCT.

There is one report of a patient with mediastinal type II non-seminoma associated with only GCNIS in one testis, suggesting that the tumors were two independent primary type II GCT [212]. Indeed, contrary to retroperitoneal type II GCT, which are metastatic from unrecognized testicular tumors, mediastinal type II GCT are normally not associated with GCNIS of the testis [211, 563]. König et al. [564] report a mediastinal non-seminoma (“teratocarcinoma”) and a metachronous pituitary stalk germinoma in a patient with Klinefelter’s syndrome, probably both related to the underlying syndrome.

There is one patient, mentioned earlier, with a mediastinal EC who had a sister with an ovarian dysgerminoma [287], demonstrating that in some families, mediastinal type II GCT may cluster with type II GCT at other anatomical sites. To the best of our knowledge, clustering with other GCT types has not been reported [137].

Mediastinal type II GCT are indeed primary tumors, as testicular type II GCT only rarely metastasize to the anterior mediastinum and only simultaneous with metastases in the visceral mediastinum [568].

Results of karyotyping are in agreement with these ploidy data, with chromosome numbers being (near)diploid [307, 570–572], (near)tetraploid [573], (near)diploid plus (near)tetraploid [572], and hypertriploid [574]. Karyotyping, CGH analysis [165], and FISH [308] show i(12p) as the most common structural aberration in type II GCT of the mediastinum; other recurrent changes are gains of chromosomes 21 and X and loss of chromosome 13, similar to type II GCT of other sites. In the study by Schneider [165], two patients had Klinefelter’s syndrome, as did the case karyotyped by Mann [571], partly explaining the extra copies of X. Different from testicular cases is that gain of 12p may be lacking [571, 574] and that fewer chromosomes are involved in gains and losses, again consistent with less extensive karyotype evolution. Mediastinal non-seminomas in patients with Klinefelter’s syndrome may [165] or may not have i(12p) [571].

The demonstration of gain of 12p (in particular i(12p)) in solid and hematologic somatic-type malignancies, often in combination with genetic hallmarks of the somatic cancer, proves their origin from the GCT [575, 576].

Three out of eight mediastinal seminomas (38 %) had activating KIT mutations (exon 17 and codons 818, 820, and 822) [333]; 1/13 seminomas (8 %) had a KRAS mutations (exon 1, codon 13) [336]. Essentially the same pattern of somatic mutations as demonstrated for testicular seminomas [295]). Non-seminomas of the thymus have not been studied for the presence of mutations.

There are no published studies addressing the epigenetics of mediastinal type II GCT.