A number of interesting points have emerged from the study of primordial germ cells in Bax ^-/- embryos. First, and shown previously (Stallock et al., 2003), at E10.5 midline germ cells that are not eliminated by apoptosis do not continue their migration to the genital ridges, but remain in ectopic locations. During the period from E9.5, when germ cells do migrate from the midline towards the genital ridges, to E10.5, when they do not, the embryo grows almost two-fold in size. This supports a model in which chemoattractants such as SDF1 (Molyneaux et al., 2003) are short range signals, and embryo growth generates populations of germ cells along the migratory route, which are out of range of signals from the genital ridges, and are eliminated by withdrawal of survival factors. Second, the loss of Bax is not sufficient on its own for the formation of EGCTs. In this study, we examined postnatal Bax^-/- mice for the presence of these, but did not find any (data not shown). This suggests that an additional oncogenic event is required for germ cell tumor formation. Although defects in the apoptotic pathway may not be a primary cause of EGCTs, they may synergize with other causes, because a recent report shows a correlation between mutations in Bax, or low Bax protein expression, and the localization of pediatric GCTs in the sacro-coccygeal region (Addeo et al., 2007). Third, and shown in this paper, is the fact that extragonadal germ cells that are not eliminated by apoptosis form a heteroge- neous population of cells, depending on position in the embryo and gender. EGCTs are found in a large number of locations along the cranio-caudal axis, from the pineal gland to the coccyx. They are always in the midline. Time of incidence, and type of tumor, seem to correlate with position along the cranio-caudal axis, and gender. Sacrococcygeal tumors are more common in infancy, more common in females, and are always teratomas or yolk sac tumors (or a mixture of the two), and make up the majority of type I GCTs (Oosterhuis and Looijenga, 2005, Oosterhuis et al., 2007). They are thought to arise from early migratory germ cells that, based upon their widespread erasure of imprinting and sex chromosome studies, have not entered meiosis (Schneider et al., 2001, Wagner et al., 1997). In contrast, mediastinal tumors are more common in adolescence, more common in males, and are seminomas, or tumors derived from them including teratomas, and are grouped with type II GCTs (Oosterhuis and Looijenga, 2005). These tumors are thought to arise from germ cells of a slightly later stage and share cytogenetic malformations with post-pubertal gonadal GCTs (Schneider et al., 2002). Since EGCTs are thought to arise from extragonadal PGCs, can our observations on the locations and behavior of extragonadal germ cells in the mouse explain these correlations? To a certain extent, they do. We found a previously undescribed population of extragonadal germ cells in the tail. These apparently do not enter the hindgut from the primitive streak (or migrate posteriorly from the gut into the tail mesentery). They retain more primitive features of the germ line, have a lower incidence of entry into meiosis, and retain more primitive germ cell features in the female than the male. Since the mouse tail corresponds to the coccyx in the human, these data suggest the hypothesis that germ cells in the sacral/tail region, which have not entered meiosis, might be the population of origin for type I sacrococcygeal tumors in human neonates. This idea is supported, but by no means proven, by the coccygeal sample from a single patient in which Oct4-positive cells were found in the epithelium adjacent to a tumor. It is possible that this cell was not a germ cell, since Oct4 would also be retained by primitive epiblast cells. It would be interesting to know if epiblast cells that are not germ cells can also colonize the tail. This could be tested using transgenic mice in which a reporter is expressed under the control of a promoter which drives expres- sion in the epiblast, but not in germ cells. Further study with antibodies specific to human germ cells, and which distinguish them from primitive epiblast cells, would also be useful. Which- ever is the case, it indicates that a larger study would be of interest, especially of younger time points, including embryonic tissues, in which tumors have not yet formed. In contrast, extragonadal PGCs in the abdominal region, though much more numerous than those in the tail, were found to differentiate at a much higher incidence which supports the finding that far fewer type I EGCTs occur in the abdominal retroperitoneum. Further- more, the increased incidence of meiotic entry of more cranial germ cells provides an explanation for the rare cases of infantile mediastinal teratomas that show meiotic imprinting patterns (Schneider et al., 2001). However, there are features of human EGCTs that are not explained by studying the extragonadal PGCs in the mouse. We did not find any extragonadal germ cells in the brain, nor in the thymus, the sites where these tumors can be found. There are a number of potential reasons. First, germ cells may simply not colonize these areas in the mouse. Ectopic germ cells left in the midline in Bax^-/- embryos did not migrate very far, and rapidly lost their motility. In the human embryo germ cells could migrate further, due to a longer gestation period. Alternatively, more distant migration and colonization could be due to additional mutations which enable them to enter and exit the vasculature, as in the chicken embryo. Second, germ cells that do migrate into anterior structures could be killed off by Bax-independent death pathways. Third, PGCs that enter those regions could be signaled to differentiate into cell types that lose the markers we used. Fourth, we cannot exclude the possibility that human EGCTs in those locations don’t come from germ cells at all, but are derived from other stem cell populations. Further studies are required to distinguish these possibilities. The delayed differentiation of PGCs in the sacral/tail region suggest the importance of this extragonadal niche in EGCT formation. Repression of the retinoic acid (RA) signaling pathway in this niche may be important for maintenance of PGCs in an immature state. When present, RA induces differentiation and Oct4 inactivation in GCT cell lines (Looijenga et al., 2003), and is the trigger for activation of meiosis in vivo (Bowles et al., 2006, Koubova et al., 2006). The prevailing model of meiosis regulation, which is supported by our findings, is that meiotic germ cell entry is inhibited in the testis by the presence of a cytochrome P450 enzyme, Cyp26b1 (Bowles et al., 2006, Koubova et al., 2006). During development, a rostral/caudal gradient of retinoic acid (RA) signaling is established, with RA levels absent at the tail due to degradation by a similar enzyme, Cyp26a1 (Molotkova et al., 2005). This gradient may explain the maintenance of Oct4 and repression of Scp3 activation observed in germ cells in the sacral/ tail region. Further study of this pathway in this region may provide a better understanding of the mechanisms of type I EGCT formation. Lastly, we still do not know the final fates of extragonadal germ cells lacking Bax at postnatal stages. There are several possibili- ties. First, they may be eliminated by a Bax-independent mecha- nism. For example, the extrinsic pathway, activated by Fas, is generally Bax-independent, and is known to be activated postna- tally in testicular germ cells, and may be downstream of Steel/kit signaling at that time (Sakata et al., 2003). Second, they may die by a non-apoptotic mechanism such as autophagy. Third, they may differentiate into Oct4 and MVH-negative tissues. In non- seminomatous GCTs including teratomas, germ cells have vari- able or no expression of Oct4 or Vasa (Honecker et al., 2006, Zeeman et al., 2002). Lastly, they could differentiate into somatic lineages. This occurs in teratomas, and in Drosophila embryos, loss of expression of nanos causes germ cells to turn on somatic cell markers (Hayashi et al., 2004). This issue will only be resolved by germ cell specific lineage analysis of Bax ^-/- germ cells. This experiment awaits the availability of a germ cell-specific mouse Cre line.