Closing the circle of germline and stem cells: the Primordial Stem Cell hypothesis

Few animals have a cell lineage as well studied and understood as C. elegans[70,71]. The early divisions of the zygote (P0 cell, Figure 4) give rise to a series of germ plasm component-containing cells (the P lineage) which culminate in the P4 cell (Figure 4), which is the first PGC of C. elegans, with germ-only potential. P4 later divides in Z2 and Z3, which remain quiescent until hatching, and later proliferate to give rise to the adult germline. Each of the early divisions of the P lineage cells generates one somatic cell as well, therefore exhibiting mixed germ/soma potential. So does the zygote, which can be seen as the original PriSC in most animals. Therefore, the P lineage of C. elegans represents the PriSCs in this model, since all three cells (P1, P2 and P3) exhibit dual somatic and germ potential, and self-renewal for a very limited number of divisions.

Many organisms exhibit similar patterns in their early embryonic divisions. The generation of the 4d blastomere in molluscs for instance is reminiscent of this process of progressive restriction of germ plasm components into one cell, while generating somatic cells in each division. As has been discussed, however, other lineages of self-renewing germ plasm-containing cells can be generated in this process as well; such is the case of the MPGZ in P. dumerilii. These cells can undergo unlimited or limited self-renewal, regenerate the germ cells post-embryonically and drive asexual reproduction processes in many organisms.

The PriSC hypothesis provides, therefore, a very evolutionarily dynamic framework in which the regenerative capacities, the potential for asexual reproduction and the germline specification can be explained by the behavior of the PriSC population in each animal. Multiple transitions could be explained by only a few changes in the properties of PriSCs, and this could affect the life history of the organism. These changes, in the extent of self-renewal, proliferation or quiescence, and developmental potential, can be regulated by a small number of genes in an evolutionarily dynamic framework. A continuum of possibilities can be envisaged, enabling multiple transitions in different phyla. All three models proposed are present, for instance, in the spiralian group.

Possible examples of the three models here described can be found widespread throughout the animal tree of life. For example, it has been reported that in the sea urchin Strongylocentrotus purpuratus Vasa and other GMP components accumulate in the small micromere lineage [4,14,72-74]. The small micromere lineage is a distinct population of cells that gives rise only to adult tissues, most likely including the germ cells [23,75]. Therefore, these cells, which express GMP components and have mixed germ/soma potential would constitute the PriSCs in sea urchin, likely following a restricted model similar to the evolutionarily distant P. dumerilii.

Most ecdysozoan and deuterostome animals reproduce sexually. However, some examples of asexual reproduction can be found among these phyla. For instance, ascidians, such as Ciona intestinalis reproduce sexually, and follow a typical example of germline specification by preformation [76-78], but closely related species of colonial ascidians can reproduce asexually. It has been reported that the colonial tunicates Botryllus schlosseri and Botryllus primigenus possess hemoblasts, which are undifferentiated stem cells that have both somatic and germ potential (reviewed in [79]). Some populations of hemoblasts express GMP components [80] and can specify the germline both at embryonic development and during asexual reproduction. Therefore, colonial ascidians likely follow the unlimited PriSCs model, while solitary ascidians such as C. intestinalis follow the rudimentary PriSCs model. However, further work is needed to demonstrate the existence of totipotent stem cells in colonial ascidians, as has been suggested [79].

However, little is known about the expression of germ plasm components in many organisms. The majority of phyla were considered to have either a mixture of preformation or epigenesis, or epigenesis only, with the only exception being the nematodes and the rotifers, which specify their germlines exclusively by preformation [2]. Molecular data on the expression of germ plasm components or GMP components is needed in all these phyla in order to elucidate and properly trace the PriSCs and the GCs. Also, for technical reasons, GMP components are more commonly detected on the basis of mRNA, not protein. This can lead to confusion; for instance, vasa mRNA was not detected in the zygote of the cnidarian H. echinata[13], but the protein was, as it is maternally supplied. As mRNA distribution does not always reflect protein distribution, cross-reactive or species-specific antibodies should be produced and used to immunolocalize GMP components in the early stages of development.

The PriSC hypothesis postulates that GMP components and their morphological manifestation, nuage granules, are indeed part of a machinery to safeguard the genome and prevent the occurrence of mutation or transposition of selfish elements among others [38,39,81,82]. This machinery might, therefore, also be needed in cells that can propagate the genome to another individual, by either sexual or asexual reproduction. This fact explains its presence in pluripotent stem cells classically considered to be somatic, such as planarian neoblasts, echinoderm small micromere lineage cells and spiralian 4d lineage. Future studies should investigate if this is the case as well for other organisms. Combinations of lineage tracing studies and molecular labelling of GMP components will help elucidate the hypothetical presence of germ plasm components in zygote-derived germline-precursor pluripotent cells. Attention should be given to asexually-reproducing species belonging to mainly sexually-reproducing groups of animals, in order to understand if restricted or rudimentary PriSCs can evolutionarily revert to unlimited PriSCs. Such could be the case, for instance, of the colonial barnacle Polyascus polygenea, in which vasa-related mRNAs are expressed in somatic stem cells during the asexual reproduction process, in addition to GCs [83]. It would be important, then, to elucidate if these cells are derived from germ plasm component-containing cells during embryonic development and are, therefore, consistent with the PriSC hypothesis, or if the GMP component expression occurs de novo in somatic cells.

It is important to point out, however, that not all cells that can propagate the genome to future individuals might need GMP components. It is believed, for instance, that GMP components and nuage function are intimately linked to transcription, translation and regulation of gene expression processes in general [84]. PGCs, however, have been described as being quiescent in many organisms, from their specification to the onset of gametogenesis [85]. This fact would explain the reported disappearance of GMP components until the onset of gametogenesis in some molluscs, for instance.

Furthermore, many animals proceed through early developmental stages in a transcriptionally independent way [82,84,85]. Such is the case of C. elegans, in which transcription of zygotic genes in the germline starts in the Z2/Z3 cells [86]. P granules, the germ granules of C. elegans, a physical manifestation of the GMP, are present in the zygote and throughout early development. However, P granules in C. elegans are not likely to be functional during early development, since they are present at a cytoplasmic location. They only attach to the nuclear pores progressively in the P2 to P4 cells, before zygotic transcription starts [87]. Therefore, it can be hypothesized that their function in transcriptional and posttranscriptional control of gene expression might start at this point, and the P granules are only required as determinants in earlier stages. Such could be the case in several other organisms with transcriptionally inactive early embryogenesis.

It is also important to note that GMP components have been shown to play different roles in other somatic cell types, such as neurons [27,88,89]. The presence in neurons of a kind of RNA granule similar in composition to nuage granules might account for this presence and point to a similar function of these RNA granules. Common processes of stem cells, such as mitosis, can also be regulated by GMP components [73,90]. Furthermore, distinct separate subsets of GMP genes, such as piwi genes, might be needed in other cell types [91-93]: for instance, Piwi proteins are not only expressed in the germline of D. melanogaster but in the accompanying somatic cells, likely because transposons present in them can give rise to viral particles than can invade the germline [39]. In order to unravel this question, careful examination of gene expression patterns on a global scale is needed.

Since GMP genes can also be expressed in somatic tissues, it is important to understand the qualitative functional differences in germline and soma. For instance, although piwi expression is seen in D. melanogaster somatic cells, a much more intricate piRNA amplification cycle, the ping-pong cycle [94], occurs exclusively in the germline, with the involvement of the related proteins AGO3 and Aubergine, which are only found in the germline [38,39]. Therefore, qualitative differences in the gene expression patterns and the function of GMP components exist between germline and soma. The extent of these differences, however, remains to be explored. Thus, it would be interesting to test if the ping-pong cycle is truly germline exclusive in other animals and if it exists in PriSCs such as planarian neoblasts, for instance.

Several studies have already started to decipher in a global scale the expression profile of stem cells such as planarian neoblasts [46-49] and cnidarian I-cells [50]. Cnidarian I-cells were profiled along with the other, somatic, stem cells in the cnidarian Hydra magnipapillata. Interestingly, I-cells and neoblasts are closest at the gene expression level than neoblasts and any of the other two stem cell lineages of H. magnipapillata, suggesting that this resemblance could be due to the fact that they both represent lineages of PriSCs. It would, therefore, be interesting to investigate the potential contribution to forthcoming generations of the other two stem cell lineages of hydrozoans. No trace of somatic host cells was found by Wagner and co-workers in their key single neoblast transplantation experiment in planarians [29], but planarians lack other stem cell types of a somatic nature. Epidermal replacement is, for instance, undertaken by neoblast cells in triclad planarians and no mitoses are found in their epidermis. However, mitoses are found in the epidermis of many related groups [95], similar to hydrozoans.

Many organisms possess, then, true somatic stem cells. These stem cells, such as human epithelial or hematopoietic stem cells, never contribute to the next generation in sexually reproducing animals. Molecular profiling of these stem cells in many organisms and cross-comparison with the corresponding PriSCs will reveal if germ plasm components are needed simply for proliferation or stemness properties or are a direct consequence of the Weismann barrier limitation. Therefore, approaches to isolate or cultivate these stem cells are needed to perform global-scale gene expression surveys efficiently, both at the transcriptomic or the proteomic level.

I have discussed the presence of a population of pluripotent stem cells, the PriSCs, during embryonic development. However, amid the myriad of transitions that the different cell populations in an embryo go through, all regulated by intrinsic or extrinsic signals, some of them must affect the PriSCs as well. Only the zygote/PriSCs and the PriSCs/GCs transitions are depicted in the model but, nevertheless, evidence from different model systems teaches us that this is an oversimplification. For instance, the zygote of C. elegans gives rise to a series of cells (the P lineage), each a product of a subsequent cell division (Figure 4). Each of these steps could potentially be intrinsically or extrinsically regulated (by a ‘preformation-like’ or ‘epigenesis-like’ kind of event). It is clear in C. elegans that the selective retention of P granules, an intrinsic signal, drives this process, rendering C. elegans as a model of preformation. The Z2 and Z3 cells become dormant until the onset of gametogenesis, which is another developmentally regulated transition.

Similarly, during the embryonic development of P. dumerilii, several transitions are also observed. First, Vasa protein is deposited in the zygote as an asymmetrically distributed maternal component, even though the mRNA distributes ubiquitously. Therefore, the first zygotic cleavages and the progressive restriction of germ plasm components to the 4d blastomere are a preformation event, despite the classification of germline specification in P. dumerilii as epigenesis. This fact was already noted by Rebscher and co-workers [10,19]. Subsequently, this one gives rise to the MPGZ cells and the PGCs.

In the triclad S. polychroa, different populations of GMP expressing cells are also detected [8,20]: first, embryonic cells in the early development and then, smaller sized cells which resemble neoblast cells. These cells likely give rise to the neoblast cells and to the GCs, denoted by the GC-specific expression of nanos in the related species Schmidtea mediterranea[32-34], although neoblasts retain the ability to give rise to GCs at any time, probably through an epigenetic event.

Therefore, the classification of germline specification in two different types, epigenesis and preformation, is deemed oversimplified and incomplete by the PriSC hypothesis. I propose that both the zygote/PriSCs and the PriSCs/GCs transitions, along with those occurring in the PriSCs themselves, can be by either preformation or epigenesis. Remarkably, animals following the rudimentary PriSCs model are mostly believed to specify their germlines by preformation. This fact is explained by the rudimentary state and fate restriction of the PriSCs in these animals.

The PriSC hypothesis states that even though the continuity of the germline is considered to be interrupted in animals following epigenesis as a means of germline specification, there is still a continuity of the germ plasm that flows from zygote to PriSCs and then to GCs. Therefore, if PriSCs are included in the germline definition, then the continuity of the germline is not interrupted either. A growing number of studies in recent years have pointed to a germ plasm continuity observed at the molecular level in a wide variety of animals, even those classically considered to specify their germlines by epigenesis [8-10,13,14,16-23].

However, it is difficult to translate the classical term ‘germ plasm’ into a 21st century molecular definition. The closest approximation would be the use of the GMP as the molecular circuitry that drives the germ plasm. The GMP might not be, nevertheless, a united and indivisible block, but a collection of molecular circuits [82]. Not all of them might be needed in all stages. It should also be noted that germ plasm or nuage granules can vary in shape, size and number from organism to organism. For instance, large electron-dense granules were found in the PGCs of the enigmatic Chaetognath worms [96]. In contrast, tiny vasa- and nanos-positive granules were recently described during the early cleavage steps of the cephalochordate Branchiostoma floridae[22]. Interestingly, this new finding shows that germline specification likely occurs by preformation in cephalochordates, although it was previously believed to occur by epigenesis, due to the previous lack of evidence regarding the presence of a germ plasm in oocytes and early embryos. Therefore, a close examination of GMP components could likely reveal the continuity of the germ plasm in other organisms that are believed to specify their germline by epigenesis.

The PriSC hypothesis predicts that cells from the ICM and the epiblast, from which the PGCs are specified [97] are the mammalian PriSCs, and that they follow a restricted model (Figure 3E). Models similar to this have already been depicted in the literature (see for instance [98]). This notion poses, however, the caveat that GMP components are typically expressed in the PGCs of mammalian embryos but not before their specification [99]. However, several studies have already established parallels between ICM cells, or their in vitro derivatives, the embryonic stem cells (ESCs, [100,101]), and PGCs. Pluripotent stem cells, morphologically indistinguishable from ESCs were derived in vitro from PGCs without the use of reprogramming factors [102-104] long before these were used to reprogram non-pluripotent somatic cells into induced pluripotent stem cells (iPSCs [98,105]) for the first time [106]. It could be, therefore, that the known components of the GMP are just the tip of the iceberg of a wider genetic circuitry which would be present in both ESCs and PGCs. This wider GMP notion would include Oct4 as a master regulator in mammals [98]. Consistently, Oct4 is needed to reprogram somatic cells to iPSCs but not if PGCs or GCs are used, since they already express it. An Oct4-like gene has been found to be expressed in cnidarian I-cells and shown to reprogram cnidarian somatic cells [107,108]. Furthermore, several genes similar to Oct4 and numerous Oct4 targets and regulators have been shown to be expressed in planarian neoblasts [47], implying that the network could be conserved in cnidarians and in planarians.

In support of the evolutionary conservation of this wider GMP circuitry, it has been recently reported that there is extensive conservation of the mechanisms that govern pluripotency in mammalian ESCs and planarian neoblasts [47]. It is, therefore, tantalizing to hypothesize that this close relationship is due to the fact that they are both representatives of the PriSCs in different organisms and that the shared components might also be part of the GMP molecular circuitry, even though the expression of classical germline components, such as Vasa, does not start until later in mammals when the PGCs are specified. Furthermore, recent molecular profiling of planarian neoblasts [36,46-49] and cnidarian stem cells [50] reveals that planarian neoblasts share molecular components, such as epigenetic regulators, with cnidarian I-cells, which are the progenitors of the cnidarian GCs and, therefore, the representatives of the PriSCs in these organisms. Interestingly, epigenetic regulation is in turn also shared by mammalian ESCs and their in vivo counterparts. Consistently, the other two cnidarian stem cell types, however, do not show such a prominent conservation of the expression of epigenetic regulators [50]. Therefore, these data point to epigenetic regulation as another part of an expanded GMP that drives pluripotency in the PriSCs of cnidarians, planarians and mammals.

As I have discussed here, the root of the multiple controversies regarding germline specification and its phylogeny comes from the exclusion of the PriSCs from the germline definition, while the precursors of PGCs are generally included in the organisms believed to specify their germline by preformation. The inclusion of all PriSCs in a more generalized germline concept would, therefore, eliminate the root of these controversies. Thus, animals that reproduce both sexually and asexually would have a sexual-type germline and an asexual-type germline (Figure 5A). The widespread occurrence of unlimited PriSCs in basal organisms such as sponges and cnidarians suggests that this was the model used by the first metazoans. With the inclusion of the PriSCs in the germline definition, purely asexual organisms, such as the asexual strains of freshwater planarians, would therefore have a germline, though an asexual-type one (Figure 5B). However, animals exclusively reproducing sexually lost their asexual-type germline due to the restriction in fate and self-renewal of their PriSCs (Figure 5C).

The classical definition of germline in all animals has been biased by the consideration of sexually reproducing animals. As soon as non-conventional model systems with asexual reproduction or regenerative powers are considered the definition of germline starts to be controversial. For instance, planarian neoblasts, representatives of the unlimited PriSCs model, have classically been considered somatic stem cells since they give rise to somatic tissues. However, as I have discussed, the classical germ-to-soma boundary that excludes neoblasts from the germline (Figure 6A, blue line) also breaks germline continuity, while a germ plasm continuity is observed in these animals, flowing from the zygote to the PriSCs (embryonic cells and their derivatives, the neoblasts) and forth to the GCs. The PGCs, understood as the first cells with germ-only potential and their derivatives, would no longer be PriSCs and are, therefore, a distinct cell population. Similarly, controversy is posed by the classical germ-to-soma boundary as seen in animals that follow the restricted PriSCs model, such as P. dumerilii. In these animals this boundary (Figure 6B, blue line) separates the zygote from the 4d lineage and its derivative, the MPGZ. A similar exclusion of the small micromere lineage would occur if sea urchins are considered. However, the PriSCs of animals following the rudimentary PriSCs model have been classically considered to be part of the germline, despite their very limited, but existing, somatic potential (Figure 6C, blue line). This inclusion, while the closely related PriSCs with broader somatic potential are excluded, therefore, is at the root of the controversies explained above.

The proposed inclusion of the PriSCs in the germline concept is made on the basis that they can contribute in all animals to forthcoming generations. In the case of PriSCs of the unlimited model, exemplified by planarian neoblasts, they can do it by means of both asexual and sexual reproduction. Unlimited PriSCs can contribute to a next generation (indicated by green arrows): first, by being the cellular source during asexual reproduction; second, by generating germ cells; and third, by constantly regenerating them when lost (Figure 6A). It is also clear that PriSCs of the restricted model contribute to the next generation by generating the GCs during embryonic development (Figure 6B), just like the PriSCs of the rudimentary model (Figure 6C). However it remains unclear if restricted PriSCs can regenerate GCs when lost (Figure 6B). Many annelids, for example, can regenerate only posteriorly (while many others can regenerate both anteriorly and posteriorly and effectively reproduce asexually) [109,110]. It would be interesting to know if germ cells can be regenerated by cells from the PriSC lineage in these annelid species. These experiments would also be illustrative in other groups: similar to annelids, many triclads, for instance, do not possess the regenerative abilities of S. mediterranea and Dugesia japonica, the reference planarian model species, and hardly regenerate anterior fragments [111]. Therefore, they reproduce exclusively sexually. It would be interesting to know if neoblasts from these species can still regenerate GCs if these are lost.

I define primordial stem cells (PriSCs) as highly conserved stem cells that are predicted to be intercalated between the zygote and the germline during embryonic development and that are delineated by the classical concept of the Weismann barrier. The behavior of the PriSCs, in terms of self-renewal and somatic and germ potential is related to the mode of reproduction and the regenerative capacities in each animal. In the most classic developmental model systems, the PriSCs are rudimentary and primarily give rise to the PGCs, along with a few somatic cells during early embryonic development. However, recent evidence reveals that the PriSCs in other animals are not so restricted and have different levels of self-renewal and somatic potential. The comparison of different animal embryonic developmental data reveals that proliferative, self-renewing, somatic precursor PriSCs are likely related to the more restricted precursors of the PGCs in more classic developmental organisms. Non-rudimentary PriSCs can also be restricted in fate or give rise to populations of stem cells with unlimited self-renewal. All kinds of PriSCs have similar gene expression patterns, as evidenced at the morphological level by the presence of nuage-like RNA granules. This presence can be explained by the Weismann barrier concept, since all kinds of PriSCs can effectively drive reproduction processes by means of sexual or asexual reproduction. Therefore, the machinery needed in GCs to carry out its genome keeping function is also needed in the PriSCs. The inclusion of the PriSCs in a more generalized germline concept, by the use of the Weismann barrier concept, resolves numerous controversies. This germline concept consists of a germline cycle with PriSCs and GCs, each capable of driving reproductive processes by means of asexual or sexual reproduction.

Both germline continuity and germ plasm continuity are observed in the new germline concept, just as originally envisaged in the 19th century by August Weismann in his germ plasm theory.

ESC: Embryonic stem cell; GC: Germ cell; GMP: Germline multipotency program; I-cells: Interestitial cells; ICM: Inner cell mass; iPSC: Induced pluripotent stem cell; MPGZ: Mesodermal posterior growth zone; PGC: Primordial germ cell; PriSC: Primordial stem cells.

I thank the numerous and diverse people working in diverse fields and who helped devise and discussed the ideas presented here. Among others I am grateful to Paul Lasko, Mark Q. Martindale, Francesc Cebrià, Aziz Aboobaker, Ulrike Ziebold, Cristina González-Estévez, Daniel Felix and Nicole Rebscher for their helpful comments and suggestions. I am also indebted to the members of the Nikolaus Rajewsky laboratory for extensive discussion on these ideas. Among others, I thank Anna-Carina Jungkamp, Marvin Jens, Marlon Stoeckius, Catherine Adamidi, Pinar Önal and Dominic Grün for their helpful comments and insight. I also thank the anonymous reviewers whose comments strengthened this manuscript. Finally, I also thank Nikolaus Rajewsky for funding and his support in the writing of this manuscript.