Totipotency continuity from zygote to early blastomeres: a model under revision

Introduction

The single-cell embryo or zygote in metazoans created from the fusion of an oocyte with a spermatozoon has the property to generate a new organism, i.e. produce all the necessary cells, including the germ cell lineage, and – in mammals – the transient extraembryonic annexes (i.e. placenta). This property was named totipotency more than 100 years ago (see below). This is the most complete definition which could be qualified as the organism’s or reproductive totipotency at the animal level compared to other definitions based on cellular or molecular features. The first part of the review describes the multiple facets of the term ‘totipotency’ (Box 1) and related issues raised by its meaning (Box 2).

The first zygotic division in some metazoans (e.g. nematodes) is visibly asymmetric, generating a smaller and a larger blastomere, while in other metazoans (e.g. ascidians), this looks like an equal division which, nevertheless, produces two different daughter cells or blastomeres, as demonstrated by the subtle distinct left–right laterality of the embryos they generate once separated (i.e. mosaic development; for further discussion, see Lemaire 2009). Mosaic embryos (Box 3) were thought to have derived from oocytes organized as a patchwork of distinct territories containing various determinants, for example, molecules and/or organelles, which can be equally or unequally partitioned at mitosis. Consequently, only the zygote is totipotent in organisms defined as mosaic, anticipating a connection between the mode of partition of such determinants and totipotency. The situation in other organisms, including mammals, was thought to be different by virtue of the regulative property of their development (Box 3). This property should allow each two-cell blastomere to produce a new complete organism, as stated or implied in numerous publications (Saxen 1989, Balakier & Cadesky 1997, Council of Europe 2003, Chuva de Sousa Lopes & Mummery 2004, Van de Velde et al. 2008, Mitalipov & Wolf 2009, Li et al. 2010, Lu & Zhang 2015, Chazaud & Yamanaka 2016). Moreover, rabbits, pigs and sheep were occasionally born from blastomeres isolated after the third zygotic division (eight-cell stage; Moore et al. 1968, Willadsen 1980, Saito & Niemann 1991). This led to a generalization and oversimplified view of those experimental data whereby loss of totipotency after one zygotic division was considered implausible, and both two-cell blastomeres were considered as equally totipotent.

In fact, new functional and molecular data produced with mouse embryos (Casser et al. 2017, 2018) point to a need to reevaluate the long-held assumption of equal totipotency in two-cell stage blastomeres. This will be the goal of the second part of this review, which will reexamine the original data on the reproductive totipotency of single blastomeres from various mammals (mouse, rat, cow, sheep, goat and monkey) and from other organisms when relevant. We focus on those experiments that tested the developmental abilities of early mammalian blastomeres taken in isolation in a cell-autonomous way (i.e. excluding nuclear transfer and chimera production). Isolation is indispensable: If kept together in the same zona pellucida, the sister blastomeres would compete for space, raising issues concerning, for example, the asynchrony and order of cell division which their isolation prevents. Formally, one could still wonder whether isolated blastomeres behave as they would in the native embryo (Gulyas 1975, Gardner 1996, Edwards & Beard 1997). This question was addressed by experiments analyzing the cell lineage of sister blastomeres kept together and distinctly labeled. Since their progeny cell contribution at a later stage (i.e. the blastocyst stage) was equivalent to that obtained with separated blastomeres (Mihajlovic et al. 2015, Casser et al. 2017), isolated blastomeres can be considered as acceptable models of the blastomeres embedded in intact mammalian embryos.

In this review, after having revisited historical and recent data, we call for a revised model in which totipotency continuity in two-cell stage blastomere is influenced partially by the natural partition of components/ingredients that are nonuniformly distributed in the zygotes (mosaic construction) and partially by regulative mechanisms. As a result, a viable embryo can consist of totipotent and non-totipotent blastomeres already at the two-cell stage.

The multiple facets of totipotency

Totipotency is usually defined by the ability of a single cell to give rise to all the cells of a new organism. From a teleological point of view, which defines a concept considering its purpose, it is clear that zygotes are totipotent because the goal of the one-cell stage embryo is to produce an entire new organism by cell division and differentiation. After the first zygotic division, however, a second teleological argument can be raised that conflicts with the first argument. Most mammals are viviparous and produce few embryos in the highly sheltered environment of the mother’s body, where the risk of injury to one blastomere is minimal (Edwards & Beard 1997). Therefore, it is not apparent why totipotency in mammals should last over multiple cleavage stages. So, are the first two blastomeres still both totipotent? Strictly speaking, the answer is yes if – and only if – both of them can operate the same functions as the zygote in terms of generating an entire new organism. This corresponds to the meaning Wilhelm Roux originally gave to the German term ‘Totipotenz’ (Roux 1893) later adopted as ‘totipotence’ in English (Morgan 1901). This terminology was created to describe experiments conducted in invertebrate or lower vertebrate species for which isolated blastomeres had been able to give rise to premetamorphosis stages, such as pluteus larvae or tadpoles (Willier & Oppenheimer 1964). These experiments were taken as bona fide evidence for totipotency, although they demonstrated only a first level of developmental regulation. There is a second layer of complexity in eutherian mammals owing to viviparity; thus, the mammalian blastomere should only be granted with a status of totipotency if its full development in utero is observed, including the generation of all the cell lineages (extraembryonic and embryonic tissues, including the germ cell lineage). This facet of totipotency can be qualified as reproductive, developmental or organismal (Box 1, Fig. 1A).

View Full Size

Definition of totipotency and interrelation of the concepts of totipotency with regulative and mosaic development. (A and B) Schematize the two main definitions of totipotency. (C, D and E) Schematize the interrelations. (A) A totipotent cell is an embryonic cell or blastomere able to generate a new entire organism. This also corresponds to reproductive totipotency (see text for detail). (B) A totipotent cell is a cell or embryonic cell or blastomere able to differentiate into all the different cell types existing in the organism. (C) At the first cleavage, if the two-cell blastomeres are both totipotent, they can both give rise to an entire organism (in green) by regulative development. (D) At the first cleavage, if the two-cell blastomeres are both not totipotent due to mosaic construction of the zygote, they fail to produce an entire organism after having initiated development (in blue or red). (E) The mosaic construction of the zygote may allow the transmission of all the critical determinants for totipotency and successful development into only one (green) of the two daughter cells while the other undergoes unsuccessful development (in red). The successful blastomere can undergo development by itself thanks to regulative properties.

Citation: Reproduction 158, 2; 10.1530/REP-18-0462

• Download Figure
• Download figure as PowerPoint slide

Because this definition of totipotency entails gestation and full development, which in some mammalian species are impractical or unethical to test, its content was extended and ‘totipotency’ was employed in a more permissive way, depending on the biological context. Totipotency, for example, was used to qualify a cell which simply contributes to ‘all’ tissues in chimeras or differentiates into ‘all’ cell types where in practice not ‘all’, but a representative set of cell types can be considered, specifically including both embryonic and extraembryonic cell types in mammals (Suwinska et al. 2008; Fig. 1A and B). We would like to group the latter ones into cellular totipotency to distinguish between those possible definitions.

In addition, the search for a molecular definition of totipotency led to an even more permissive interpretation: A cell would be defined as totipotent if it expressed in high levels of certain transcripts or totipotent markers. Because embryos at the two-cell stage were taken as a gold standard for the totipotent stage in mice, transcripts characteristic of such embryos (Macfarlan et al. 2012) were considered as totipotency markers. However, expression of some marker genes is a simplistic criterion that is unlikely to satisfy a complex state as totipotency. To date, there is no clear molecular understanding of what it takes for a blastomere to be totipotent (see paragraphs titled ‘Possible mechanisms contributing to totipotency discontinuity’ and ‘In search of molecular correlates and origins of unequal blastomere potency’).

As a matter of fact, the term ‘totipotency’ has branched off into multiple flavors and directions (discussed in a recent review, Condic 2014), but none of them has been able to reflect consensus. Why is defining totipotency so challenging? This can be partly attributed to the following three considerations. Firstly, this property evolves as embryonic development progresses and a single, unique and rigid definition simply might not work. Secondly, this property must be empirically tested according to chosen criteria that are used for its definition (either reproductive, or cellular or molecular definition of totipotency, see above), and this choice may vary from one study to another. Thirdly, historical literature had associated totipotency with the concepts of regulative or mosaic development, but those concepts have themselves rapidly evolved. As more recent discoveries attract more attention, older knowledge can be forgotten even if it can still provide a useful compass. Consequently, those words (totipotency and regulative and mosaic development, Fig. 1C, D and E) are diversely combined and can correspond to various biological realities, as we will discuss in the next paragraph.

Revisiting totipotency in light of the historical and debated concept of mosaicism vs regulation

At the end of the 19th century, two scientists, Hans Driesch and Wilhelm Roux, designed experiments to address the problem of early embryo totipotency. They used two different animal species and methodologies. Driesch’s experiment separated the blastomeres of a European green sea urchin (Echinus microtuberculatus, later renamed Psammechinus microtuberculatus) two- or four-cell stage embryo and showed that those isolated blastomeres could generate completely albeit smaller larvae, thanks to regulative development (Driesch 1964 (1892)). Two- and four-cell blastomeres were, therefore, considered as totipotent, and the successful development of those blastomeres was considered as regulative (Fig. 1C). Unfortunately, experiments by Driesch and even more recent ones by others were embellished in the collective memory: Only 5–10% of the two-cell blastomeres led to successful complete development (Cameron et al. 1996). Thus, the development of monozygotic (MZ) twins or quadruplets was rarely observed, and when this was the case, a mix of normal and abnormal embryos was clearly visible. Those observations suggest that sister blastomeres are not as equally totipotent as once thought (Plough 1927, Marcus 1979).

Roux had chosen embryos from the green frog (Rana esculenta) for his experiments, and he tested the ability of one of the two-cell blastomeres to develop after damaging the second blastomere. The lack of complete development of the untouched blastomere was mistakenly interpreted as distinct potency of each blastomere and viewed as evidence for mosaicism (Fig. 1D).

In fact, the outcomes in Roux’s experiment were influenced by the presence of the injured blastomere, as later results showed (McClendon 1910). McClendon reported in his paper how he was able to eliminate one blastomere by aspirating its content so that the remaining blastomere could develop in a rather normal tadpole. He discussed the way in which frog zygote cleaved, bisecting the gray crescent – a less pigmented area opposite the sperm entry point – equally to produce ‘two totipotent halves’. Although McClendon had already understood, with others, the importance of the partition of the gray crescent, he did not envision any link between this process and the mosaic construction of the zygote. By contrast, he was convinced that regulative mechanisms could reformulate the developmental processes so that the single isolated blastomere could produce an entire tadpole (McClendon 1910).

The elegant simplicity of early experiments conducted in sea urchin and frog proved to be a challenge in mammals. An experimental study of mammalian totipotency was not possible without two important techniques: (1) Cell culture that was necessary to evaluate totipotency in terms of blastocyst formation (stage of the primary regulation with the formation of the inner cell mass, ICM, and the trophectoderm, TE) and biomarker gene expression and (2) embryo transfer into the uterus of a foster mother that was necessary to demonstrate the complete formation of a new organism as required by the reproductive definition of totipotency. Mammalian embryo culture became a reliable method only in the 1960s, thanks to the work of Whitten, McLaren, Chang, Brinster and Biggers (reviewed in Biggers 1998) and embryo transfer, which was pioneered by Walter Heape as early as 1890, was not reliable for many years (Biggers 1991). In 1942, 50 years after Driesch and Roux, Nicholas and Hall were able to separate blastomeres of two-cell rat embryos to obtain complete but not fully normal embryos (Nicholas & Hall 1942). In the 1950s, the laboratory of Albert Dalcq and Jacques Mulnard was working on the enzymatic activities characterizing the early development of rat embryos. They observed a diffuse phosphatase activity (diffuse reaction) which was unevenly located in blastomeres, leading them to think that it could correlate with ICM-TE segregation. Their findings could be interpreted as a possible molecular explanation for the early transition from totipotency to early differentiation (Mulnard 1955; reviewed in Mulnard 1986). Although those data were criticized later by Solter and collaborators (Solter et al. 1973; Box 3), the diagram of ‘segregation theory’ shown in Fig. 2A is instructive, as it prefigures later studies on two-cell diversification with the hypothesis of heterogeneously localized egg components (Fig. 2A).

View Full Size

Transmission of egg (either unfertilized or fertilized) components to the two-cell embryo. (A) Scheme adapted from Fig. 7 shown in Mulnard (1986) and related studies using rat embryo. Enzymatic activity (Mulnard’s diffuse reaction) identified in a cytoplasmic region of the fertilized egg is allocated to one of the two-cell blastomeres or partly to the two blastomeres, depending on the orientation of the first cleavage plane. This enzymatic activity is ultimately inherited by the cells forming the inner cell mass (ICM) of the blastocyst and not the cells of the trophectoderm (TE). Thus, the first differentiation event during development, ICM versus TE, is correlated to this enzymatic activity. (B) Schema adapted from Seidel and related studies using rabbit embryos (Seidel 1960). Seidel’s scheme hypothesizes that components stored in the egg can be transmitted in different ways depending on the 1st cleavage plane (1, 2, 3 or 3′). Developmental capability of the two-cell blastomere is illustrated by the type of blastocyst this single embryonic cell can give rise to: A normal blastocyst with ICM and TE, an abnormal blastocyst lacking ICM or a didermic blastocyst made of a double layer of cells without defined ICM. The dashed line in (A) and (B) labeled 1 represents the axis of the fertilized egg or zygote.

Citation: Reproduction 158, 2; 10.1530/REP-18-0462

• Download Figure
• Download figure as PowerPoint slide

Still in the 1950s, Friedrich Seidel and Andrzej Tarkowski used blastomere lethal injury (Roux’s experimental approach) independently to determine the totipotency of early blastomeres in rabbit and mouse, respectively. Rabbits could still be born when one of the two- or four-cell stage blastomeres was kept alive following ablation of the other blastomere(s) and transplantation into a foster female (Seidel 1952, 1960). Seidel’s careful analysis of the blastocysts (stage of primary regulation) identified three types of rabbit embryos: (1) The normal blastocyst with well-formed ICM, (2) the blastocyst made of a double layer of epithelial-like cells and (3) the blastocyst without ICM. Seidel hypothesized that those three types of embryos were obtained as a consequence of the egg organization, whose components (e.g. a ‘Bildungszentrum’, the German term for organization center) were allocated unequally to the two-cell stage blastomeres (Fig. 2B). Hence, Seidel suspected two-cell (or even four-cell) blastomere diversification. Unfortunately, his study was published in German, which limited its spread across the international scientific community: Die Entwicklungsfähigkeiten isolierter Furchungszellen aus dem Ei des Kaninchens Oryctolagus cuniculus (The developmental abilities of isolated embryonic cells from the egg of the rabbit Oryctolagus cuniculus; Seidel 1960). While cited in the English-written scientific literature (Denker 1976, Gardner 1996, Edwards 2005, Suwińska 2012), Seidel’s work probably did not receive all the attention it deserved. Tarkowski investigated the ability of the mouse two-cell blastomeres to develop into blastocysts and then into offspring (Tarkowski 1959). He described in his seminal report published in Nature how this capacity is variable and, similar to Seidel, he observed some blastocysts without ICM. Furthermore, it can be deduced from his data on postimplantation development that some embryos lacked the ability to build a functional placenta; thus, the totipotency of the blastomere tested was not confirmed in those cases (Tarkowski 1959).

Two conclusions could be drawn from those pioneer studies. Firstly, they revealed that the blastomere tested for its totipotency could not always be developmentally successful for technical or biological reasons, and this observation might have hampered any strong, definitive conclusion about its status. Secondly, several investigators (e.g. McClendon in frog, Seidel in rabbits) had envisioned the correct partition of key determinants in daughter blastomeres as critical for the maintenance of their totipotency. Hence, their work already supported the idea that the initial mosaic composition of the zygote could be followed by regulative processes that would lead to a new organism from a single blastomere (definition of organism’s/reproductive totipotency). However, the existing literature does not clearly reflect this functional association between the three words (totipotency and regulative and mosaic development), as demonstrated by a rather unproductive PubMed search made on February 9, 2019 (a combination of the three terms returned no items while the combined key words ‘totipotency’ and ‘regulative’ returned six items). We would like to reinvigorate this association in our current discussion featuring totipotency, and regulative and mosaic development as three interrelated concepts which should be integrated together with a fourth one, the organization of the egg materials to be inherited by the zygote after fertilization, i.e. mosaic construction of the zygote (Fig. 2).

Totipotent and non-totipotent blastomeres found in early embryos: imperfect regulation or mosaicism?

Additional studies since the first experiments in mammals have addressed the question of early blastomere totipotency. The equal totipotency of each blastomere composing an early embryo has particularly remained an open issue, because its testing requires the dissociation of the embryo and an isolation of each blastomere to prove its full development (twin, quadruplet). This is a complex task because blastomeres must be separated without damaging them and, even more importantly in viviparous species, development to term must occur in utero and requires the formation of extraembryonic tissues. Therefore, from a technical point of view, the experiment requires embryo transfer in the foster mother, and from an embryologic point of view, it implies the success of a first level of regulative development or ‘primary regulation’ (from the original German term ‘Gestaltungregulation’, F. Seidel). Starting from the single isolated blastomere, the ‘primary regulation’ enables the production of embryonic and extraembryonic tissues (ICM and TE). Because it is impossible to achieve full development in the absence of this first level of regulative development, the totipotency of two-cell blastomeres can be judged based on either blastocyst formation and biomarker gene expression (bona fide totipotency) or full development (retrospective-reproductive totipotency). Therefore, we would like to focus on studies that kept track of the native pair of blastomeres and scored development in order to answer two distinct but complementary questions. The first question asks how often two-cell blastomeres of the same embryo are equally totipotent, while the second question asks how prevalent totipotency still is after the two-cell stage.

How often are two-cell blastomeres of the same embryo equally totipotent? Isolated two-cell blastomeres exhibited a high frequency of blastocyst development (92%), similar to non-manipulated embryos, in our recent study in mice (Casser et al. 2017). By contrast, in this and in another four studies, 30 ± 3.7% of the pairs of isolated two-cell blastomeres or their derivative blastocysts were able to develop successfully into pairs of liveborn mice after transplantation into the genital tract of foster mothers (Tsunoda & McLaren 1983, Togashi et al. 1987, Wang et al. 1997, Sotomaru et al. 1998, Casser et al. 2017) (Fig. 3A, studies 1–5). The relatively low occurrence of equal totipotency in pairs of two-cell blastomeres may not be blamed on technical issues, since a high number of embryos (over 500 in total) were used for embryo transfer in different laboratories. As Seidel pointed out, not all blastocysts are developmentally competent (Seidel 1960), and this can be assessed by their cellular composition, via scoring for the presence/absence of ICM or by a more sophisticated approach using molecular markers for cell subpopulations (TE and epiblast or primitive endoderm in the ICM). These subpopulations arise from two consecutive phases of cell lineage commitment, ICM vs TE (1st phase) and epiblast vs primitive endoderm (2nd phase) (De Paepe et al. 2014). Using molecular markers (Casser et al. 2017), the authors were able to determine that only 27% of embryos had a presumptively functional epiblast in both members of the pair (≥4 NANOG-positive cells, Morris et al. 2012). This low percentage could be a consequence of the artificial separation of the blastomeres, which could have impinged on their adaptive/regulative abilities. In order to exclude the effect of artificial separation, each blastomere of an intact two-cell embryo was injected with a different fluorescent dye to perform lineage tracing at the blastocyst stage. Again, only a small group of embryos (22%, 8/36) exhibited a balanced contribution of both blastomeres to the formation of the epiblast (Casser et al. 2017). While pursuing a distinct scientific question, another research group using a similar experimental approach reported that only a small proportion of embryos (33%, 9/25) had a concordant ability to form epiblast cells from both blastomeres of the undivided two-cell stage (Mihajlovic et al. 2015). All together, these data collected from mouse embryos led us to doubt that the two-cell blastomeres could always be equally totipotent.

View Full Size

Ability of blastomeres to develop as singlets or twins after transfer of embryos derived from individual pairs of two-cell stage blastomeres (A) and ability of individual blastomeres from various stages to exhibit totipotency (B). (A) Distribution of initiated pregnancies for twins created by blastomere separation at the two-cell stage. Data from nine studies (1 to 9; T = total) are shown and number of embryos for each category is indicated in each block. Study 1: Casser et al. (2017). Study 2: Togashi et al. (1987). Study 3: Tsunoda and McLaren (1983). Study 4: Wang et al. (1997). Study 5: Sotomaru et al. (1998). Study 6: Hashiyada (2017). Study 7: Willadsen (1979). Study 8: Matsumoto et al. (1989). Study 9: Mitalipov et al. (2002). The total number of embryos pulled from all the studies/categories is shown. (B) Graph showing the percentage of blastomeres exhibiting totipotency (y axis) either by achieving full development or by generating stem cell line (*) calculated at various embryonic stages and as a function of the percentage of the volume of the blastomere (x axis) used perform the experiments. The approximate ratio of the blastomere volume to the total embryo volume is indicated in parentheses. Data summarized in (B) are extracted from the following studies. Mouse: two-cell data from (A), four-cell and eight-cell data from Tarkowski and Wroblewska (1967) and from Wakayama et al. (2007). Sheep: Willadsen (1981). Rabbit: Moore et al. (1968). Pig: Saito & Niemann (1991). MZ: monozygotic pair (twins).

Citation: Reproduction 158, 2; 10.1530/REP-18-0462

• Download Figure
• Download figure as PowerPoint slide

One must be cautious before translating findings obtained in mouse embryos directly to other mammalian species and establishing their general validity, because Mus musculus development exhibits some specificities whose impact on totipotency is not yet clear. The duration of the two-cell stage, for example, is extended in mouse embryos, which later form a tridimensional egg cylinder with the epiblast developing inside (‘inversion’ of the germ layers) as opposed to the more flat embryonic disc observed in other mammalian species (Johnson & Alberio 2015). In addition, comparative gene expression analysis of mouse and human embryos reveals distinct patterns in the appearance and localization of lineage-specific transcription factors (Niakan & Eggan 2013). Unfortunately, studies reporting relevant experiments of blastomere isolation carried out in non-murine mammalian species are scarce. After Seidel’s work in rabbit, only one study used this species to determine the developmental competence of isolated blastomere up to the eight-cell stage (Moore et al. 1968). The percentage of developmental success decreased with the embryonic stage and the authors concluded that ‘increasing cell stage of the parent egg might suggest that every blastomere is not totipotent’. However, this conclusion was not formally demonstrated, as pairs or quadruplets of isolated early blastomeres had not been tested simultaneously (Moore et al. 1968). After Dalcq’s work in rat, one study reported the complete development of 9 out of 19 pairs of rat two-cell blastomeres (Matsumoto et al. 1989). In cattle, three out of ten twin pairs produced by two-cell blastomere separation and culture were born using either Japanese black or Holstein cattle (Hashiyada 2017). In goat, two out of five pairs created at the two-cell stage were born as pairs (Tsunoda et al. 1984). In sheep, five out of ten pairs created at the two-cell stage and transferred were born as MZ pairs (Willadsen 1979). In nonhuman primates, no MZ pairs but one singlet were born after separation of four two-cell embryos (Mitalipov et al. 2002) (Fig. 3A, studies 6–9).

Overall, taking into account all the mammalian species examined and all the experiments with at least some positive data, a proportion ranging from 14.7 to 40.7% (mean = 27.7 ± 13%, n = 353) of the two-cell embryos produced liveborn MZ pairs after blastomere isolation and transfer to foster mother (Fig. 3A). We feel that the gap between the expected 100% and the actual 27.7% should be smaller if both sister blastomeres were indeed totipotent, even though it is unlikely that all the blastomeres from all embryos could achieve a normal development (e.g. not all concepti have a normal set of chromosomes). Thus, the answer to the first question is ‘no’, not all two-cell blastomeres are equally totipotent. This impacts directly on the second question (how prevalent is blastomere totipotency after the two-cell stage?), as we know that a substantial proportion of two-cell blastomeres have already lost their totipotency after the first zygotic cleavage. Conversely, totipotency might be restrained to only one of the sister blastomeres in mammalian two-cell embryos, while this possibility needs to be further evaluated at later stages.

The second question to address is whether blastomere totipotency is substantial past the two-cell stage. The ability of all the blastomeres of a single four-cell stage mammalian embryo to develop autonomously into liveborn animals was documented only in two studies (Willadsen 1981, Johnson et al. 1995). It, thus, appears that this ability is very difficult to demonstrate when the embryo has more than two blastomeres, because all of them have to be preserved alive and undamaged. Nevertheless, reproductive totipotency could be tested successfully for some of the blastomeres at the four-cell stage in cow and monkey (Johnson et al. 1995, Chan et al. 2000) and up to the eight-cell stage in rabbit, sheep and pig (Moore et al. 1968, Willadsen 1980, Saito & Niemann 1991). In humans, the intentional testing of blastomere totipotency by complete development of liveborn babies is precluded by ethical considerations, as this would be assimilated to an attempt at human cloning (Editorial 1994). Serendipitously, Veiga and collaborators obtained a living birth after transfer of a human embryo in which only one blastomere out of four had survived following freezing/thawing (Veiga et al. 1987). There is no report of single blastomeres producing liveborn offspring after the eight-cell stage in mammals. Only one case of ultrasound-detected early pregnancy from a 16-cell stage blastomere in pig could be mined in the literature (Saito & Niemann 1991; Fig. 3B).

Even when blastomere totipotency was assessed at an intermediate or primary stage of regulation (e.g. peri-blastocyst), the rate of developmental success was low. In mice, we found two reports analyzing how dissociated four-cell embryos developed to the peri-/early postimplantation stage (Tarkowski & Wroblewska 1967, Rossant 1976). An average of 30% of the isolated blastomeres were able to form blastocysts, while only 1 out of 25 could form a proper embryo after implantation (Rossant 1976). When the blastomeres were isolated from eight-cell stage embryo, the percentage of blastocysts decreased significantly (0% in Rossant 1976; 11% in Tarkowski & Wroblewska 1967). More recent studies aimed to determine whether this might be due to commitment/restriction of blastomeres to a distinct cell lineage (Tabansky et al. 2013). Tabansky and colleagues used a variant of the multicolor Brainbow system to generate distinct color labeling of each blastomere at the four-cell stage and trace their progenies at the blastocyst stage (Tabansky et al. 2013). All the colors were visible in the different lineages for 30% of the blastocysts developed, revealing that the contribution of the four blastomeres to ICM or to the entire blastocyst was homogeneous. By contrast, colors in the remaining two-thirds of the blastocysts were differently distributed among cells from the ICM and in the whole blastocyst. Thus, the four blastomeres in this larger group did not contribute equally to the ICM lineage, and this could be explained by the fact that they were not all equal in terms of cellular totipotency (Tabansky et al. 2013). In humans, Van de Velde and colleagues dissociated six embryos at the four-cell stage and one of them gave four blastocysts containing visible ICMs, which, however, included a variable number of NANOG-positive cells (Van de Velde et al. 2008), again highlighting the unequal potential of the four-cell blastomeres.

Lastly, without prejudice to the fundamental distinction between genuine development in utero and models of embryonic development in vitro (blastocyst stage and derivative stem cells), we consider that it may be possible to gather some information about totipotency from those stem cells. We refer specifically to the blastomere’s ability to yield embryonic stem (ES) cells in vitro. Generally considered as pluripotent when derived the conventional way from whole blastocyst-stage embryos, ES cells seem to have an expanded potential when derived from individual blastomeres prior to the stage of ICM-TE segregation (Yang et al. 2017). Here, expanded potential means that ES cells contribute not only to the ICM after transplantation in a blastocyst, but also to the TE of chimeric embryos. We note that the derivation rate of ES cells from single mouse blastomeres correlates with the progressive loss of full developmental ability or totipotency in blastomeres isolated from various stages: 69% ES cell derivation at the two-cell stage, 40% at the early four-cell stage, 22% at the late four-cell stage and 14% at the eight-cell stage (Wakayama et al. 2007; Fig. 3B).

Those studies together established that blastomere totipotency still exists after the two-cell stage, but it is also clear that the totipotency of single mammalian blastomeres is already no longer uniformly present at the two-cell stage and it is further reduced after the two-cell stage. This raises the question as to which mechanisms underlie the loss of blastomere equality, leading to totipotency discontinuity.

Possible mechanisms contributing to totipotency discontinuity

Several mechanisms may contribute to the emergence of interblastomere differences and the variation in the manifestation of totipotency. They can be analyzed from the regulative and/or mosaic point of view (Fig. 4), and they are not mutually exclusive. Nevertheless, their relative importance may vary over time, adding complexity to the developmental events we described above.

View Full Size

Possible explanations for the origin of interblastomere differences leading to unequal potency. (A) Asynchrony of developmental processes, such as mRNA degradation/deadenylation or EGA (embryonic genome activation) due to lack of coupling between sister blastomeres (Boni et al. 1999, Brison et al. 2014). (B) Unequal partition at 1st mitosis: (i) incomplete diffusion of the sperm content that is inherited preferentially by one blastomere (Piotrowska & Zernicka-Goetz 2002), (ii) unequal distribution of molecules such as the less abundant mRNAs (Shi et al. 2015), (iii) unequal distribution of cytoplasmic organelles (Zheng et al. 2016, Gao et al. 2017), (iv) unequal segregation of chromosomes.

Citation: Reproduction 158, 2; 10.1530/REP-18-0462

• Download Figure
• Download figure as PowerPoint slide

Firstly, several biological processes can occur at variable speeds and, thereby, asynchronously between sister blastomeres, such as (de)adenylation or degradation of maternal mRNAs (Fig. 4A). Since zygotic or embryonic genome activation (i.e. ZGA or EGA) are controlled partially by maternal mRNAs, it follows that ZGA-EGA can also take place asynchronously in sister blastomeres. This is accompanied by a reorganization of embryonic chromatin. It was shown, for example, that chromatin mobility is higher in two- than in eight-cell mouse embryos (Boskovic et al. 2014), so this parameter could also vary slightly between the two-cell stage blastomeres. These observations suggest that ZGA-EGA participates in the genesis of the first interblastomere differences. However, such a role is difficult to uphold in species other than mice. Indeed, EGA occurs at various stages depending on the species: two-cell stage in the mouse (Flach et al. 1982) but two- to four-cell stage in rat (Zernicka-Goetz 1994), four-cell stage in pigs (Hyttel et al. 2000), four- to eight-cell stage in human (Braude et al. 1988), six- to eight-cell stage in rhesus monkeys (Schramm & Bavister 1999) and 8- to-16 cell stage in cow, sheep and rabbit (Crosby et al. 1988, Plante et al. 1994, Brunet-Simon et al. 2001). In mice and humans, a transient nuclear localization of the otherwise mitochondrial enzyme pyruvate dehydrogenase precedes EGA and is essential for epigenetic remodeling (Nagaraj et al. 2017), suggesting a possible metabolic regulation of EGA and perhaps also a metabolic influence on totipotency. Hence, totipotency and its continuity could be determined by the variability occurring in the regulation of major early developmental processes as maternal mRNA degradation or/and EGA. There is a prominent regulative dimension in those scenarios.

A second possible cause of interblastomere differences is the unequal allocation of molecules that could be of any kind (Fig. 4B). This unequal partition can be initiated as early as the time of fertilization of the oocyte. Full diffusion of the sperm content within the oocyte cytoplasm could be incomplete so that it is inherited preferentially by one blastomere (Piotrowska & Zernicka-Goetz 2002) (Fig. 4Bi). Even if this is still debated, sperm entry by itself could also contribute to the inequality between blastomeres by influencing the position of the first zygotic cleavage plan (Gray et al. 2004, Motosugi et al. 2005). Nevertheless, interblastomere differences are also observed in parthenogenetic two-cell embryos, which lack sperm contribution, so other mechanisms than those mediated by sperm entry need to be evoked (Piotrowska & Zernicka-Goetz 2002).

As has already been mentioned, the first cleavage plane is critical in mediating an unequal partition of egg content between the two-cell blastomeres. The way this first cleavage is operated relies directly on the positioning of the first spindle, which needs to be centered and oriented. The associated mechanisms remain incompletely understood, but they seem to be based on stochastically established forces acting on the cytoskeleton (Chaigne et al. 2017, Salle et al. 2018). Conversely, this leads to the hypothesis that totipotency depends on the inclusion of cellular material from all axial levels of the oocyte (Gardner 1996) or fertilized egg (Figs 2 and 4Bii). Accordingly, this view considers mosaicism as the predominant concept versus regulation.

Regardless of the mosaic or regulative cause of unequal totipotency, an inevitable question is about the nature of the elements that need to be inherited by the blastomeres to maintain their totipotency. They can be a combination of any cellular material, including ions, lipids, coding and noncoding transcripts, proteins, cytoplasmic domains with bound transcripts and/or proteins as well as membranous or non-membranous organelles (Fig. 4B). The distribution of organelles is mediated by actin-based cytoplasmic streaming. Such a flux can become heterogeneous, causing asymmetric partition of cytoplasmic components (Ajduk et al. 2011). In addition to this mechanism, the allocation of most subcellular elements is yet to be examined in detail. Zheng and colleagues did not find any difference in the number or activity of mitochondria between the two 2-cell blastomeres in mice (Fig. 4Biii). We can also include here other elements such as the chromosomes, whose partition is controlled mainly by the spindle assembly checkpoint (i.e. SAC). Deficiency in such a checkpoint leads to chromosome missegregation causing aneuploidy (Fig. 4Biv). If such errors happen at the first division of the zygote, this immediately generates two imbalanced sister blastomeres (e.g. one monosomic and one trisomic) that present differences of gene expression (due to gene copy number differences) and are limited upon separation regarding full development (monosomies are always lethal except for the X chromosome, and only trisomies of chromosomes 12 to 14, 16, 18 and 19 allow survival to term in mice; Gropp 1982). Unlike chromosome segregation errors at the first cleavage, errors would be less detrimental when occurring at later stages, given the lower proportion of cells affected. The key question is how many of the cases of sister blastomere inequality can be explained reasonably by aneuploidy. The answer corresponds to the incidence of aneuploidy in first cleavage mouse embryos. Five different strains of mice were studied to determine that aneuploidy was observed in only 1.1–4.3% of the embryos produced by in vivo and in vitro fertilization after gonadotropin stimulation (Fraser & Maudlin 1979). Experimental abolition of the SAC by reversine, a pharmacologic inhibitor of a key kinase MPS1, increased aneuploidy frequency with aneuploid blastomeres suffering impaired mitotic proliferation and less developmental potential reflected in the production of fewer NANOG-positive cells (Bolton et al. 2016). NANOG-positive cells are critical to build the ICM at the blastocyst stage and, subsequently, the new organism itself. In the absence of any manipulation, two-cell blastomeres show an asynchronous cleavage, with the faster blastomere dividing up to 1 h before the other one (VerMilyea et al. 2011, Balbach et al. 2012). This gap could be related to prolonged mitosis (a sign of activated SAC), but this remains to be demonstrated experimentally. The percentage of aneuploid cells in human embryos is much higher than in mice, at least after in vitro fertilization (Vanneste et al. 2009), and a similar situation is present in bovine (Destouni et al. 2016, Tsuiko et al. 2017).

In this paragraph, we have touched upon global mechanisms inducing inequality between blastomeres. Again, we insist on the fact that the mechanisms are not mutually exclusive and that they can have a different impact at the molecular level. Several of the mechanisms cited, for example, dealt with the transcripts, which can be either degraded if they are of maternal origin or produced from the new combination of chromosomes inherited by the zygote and transmitted properly or not to blastomeres, which are initiating EGA. The less abundant transcripts are at a higher risk of unequal partition (Shi et al. 2015; further reviewed in White et al. 2018).

In search of molecular correlates and origins of unequal blastomere potency

Since experiments clearly demonstrated that the two-cell stage blastomeres frequently exhibit different developmental success reflecting the presence or absence of totipotency (Fig. 3A, % singlet), it should be feasible to find a profile of gene expression correlating with this difference. However, instead of the two-cell stage, molecular investigations dealt firstly with the four-cell stage. Perhaps this preference was induced by the meticulous observations of Gardner, who described distinct categories of four-cell stage mouse embryo configuration (tetrahedral, planar) with specific and predictable relationships to the partition of cytoplasm from the different positions along the animal-vegetal axis of the zygote (Gardner 2002). Shortly thereafter, another study found that the progenies of two-cell stage blastomeres contribute to both embryonic and extraembryonic tissues in the fetus, whereas the progenies of four-cell stage blastomeres sometimes contribute only to the one or the other (Fujimori et al. 2003). This finding was also confirmed more recently (Li et al. 2017). Therefore, we will first discuss the information provided in the four-cell stage studies before reanalyzing the data obtained at the two-cell stage.

Single-cell RNA sequencing methods were exploited in mice to allow genome-wide appreciation of interblastomere differences in mRNA abundance at the four-cell stage. After removing between-embryo variation, 13 genes presented within-embryo bimodality (Biase et al. 2014). Transcript abundance for most of these genes ranged from near zero FPKM (fragment per kilobase of transcript per million reads) in one blastomere to 10³–10⁶ FPKM in another blastomere of the same embryo, indicating a widespread variation among sister blastomeres. However, it is not yet entirely clear when those differences were acquired: Investigators were able to distinguish the two sister pairs forming the four-cell embryo but not which one was generated by the first-dividing vs second-dividing two-cell stage blastomere (Goolam et al. 2016, Biase et al. 2018).

Other studies focused on histone modifications, which can be taken as a proxy for the regulation of gene expression. In four-cell stage embryos exhibiting a tetrahedral shape, the signal detected by microfluidics-based Q-PCR of the mRNA of the histone H3 arginine 26 methyltransferase PRDM14 or by immunofluorescence of H3 methylation at Arginine 26 (H3R26me2) was significantly higher in one blastomere, and this blastomere had its fate oriented toward ICM (Torres-Padilla et al. 2007, Guo et al. 2010, Burton et al. 2013). Arginine 26 is also a methylation substrate of PRMT4/CARM1, whose immunodetection revealed differences between four-cell stage blastomeres (Torres-Padilla et al. 2007). Levels of H3R2me varied between blastomeres in a similar way to H3R26me, not only in zygotic but also in cloned mouse embryos (Liu et al. 2012). Interblastomere H3R16me2 and H3R2me differences were specific, since the methylation of H4R3, which is a target of a different methyltransferase (PRMT1), was equivalent between sister blastomeres. Such specific histone modifications can impact the expression of genes encoding critical transcription factors, such as NANOG and SOX2 (Torres-Padilla et al. 2007), and, potentially, the function of these factors. Consequently, SOX2 targets, such as Sox21, are heterogeneously expressed and this induces a diversification in four-cell blastomere fate (Goolam et al. 2016). When the DNA occupancy time of SOX2 was studied along with OCT4 and SOX21, it was found that the DNA-binding activity of SOX2 correlated with the allocation of the corresponding four-cell stage blastomeres to ICM lineage (White et al. 2016). Those results together suggest that ‘heterogeneous gene expression, as early as the four-cell stage, initiates cell fate decisions by modulating the balance of pluripotency and differentiation’ (Goolam et al. 2016) and that ‘mammalian development is to some extent mosaic from the four-cell stage onward’ (Condic 2016).

In addition to mice, four-cell stage molecular diversification was also studied in other mammals. When chromatin modifications were examined in four-cell bovine (H3R2me2) and rabbit (acetylated histone H4K5) embryos, significant interblastomere differences were noticed, comparable with those seen in the mouse (Chen et al. 2012, Sepulveda-Rincon et al. 2016). In human embryos, Chavez and colleagues noticed marked differences in chromatin modifications (H3-S28P and H4-R3me2) as early as the four-cell stage (Chavez et al. 2014). At the transcript level, PCR amplification of Oct4 and beta-HCG mRNA in human embryos revealed interblastomere differences, suggesting that blastomere potency was not equal as early as the four-cell stage (Hansis 2006). Chavez and colleagues confirmed that variable levels of Oct4 transcripts exist between blastomeres in eight-cell stage human embryos (Chavez et al. 2014). They showed in their study that another important marker gene of epiblast lineage, Nanog, had abundant transcripts in each blastomere of an eight-cell embryo, but was present in other cases only in two cells or even absent. Differences in the counts of NANOG-positive cells after immunofluorescence became apparent also in blastocysts derived from isolated human blastomeres. The four blastocysts derived from a single four-cell human embryo, for example, contained from 2 to 6 NANOG-positive cells (Van de Velde et al. 2008).

In view of the above, we can draw an interim conclusion that blastomere diversification or departure from totipotency in molecular terms is definitely supported at the four-cell stage in multiple mammalian species. Individual blastomeres from the four-cell stage embryo would then be considered insufficiently regulative to fulfill the requirements of reproductive totipotency. When considering those studies, however, we must also invite a related question: If the transcript/protein level or chromatin modifications are different among four-cell stage blastomeres, would it be possible that such interblastomere variability was already latent at the two-cell stage and simply became detectable at the four-cell stage?

As it has been introduced earlier, the two-cell blastomeres of the same embryo can exhibit different developmental success; therefore, there is a quest for molecular elements to explain this functional difference and, subsequently, to define a molecular signature of totipotency. Embryonic stages for which some extent of reproductive totipotency has been described (two-to-eight-cell stages) are also associated with the lowest complexity of blastomere transcriptome (Fig. 5). This lowest complexity was described by Piras and colleagues who calculated the Shannon entropy of the transcriptome for mouse and human eggs and preimplantation embryos (Piras et al. 2014). Therefore, it is not far fetched to consider that a comparison of studies that analyzed the genome-wide transcriptome (respecting the original pair associations of the two-cell blastomeres) would faciltate the discovery of useful molecular candidates. We found seven studies carried out on mouse embryos but only five presented a sufficient level of linear correlation of the transcriptomes within pairs to be further analyzed. Half (50%) of the pairs considered exhibited at least a two-fold difference in the mRNA abundance of 17% of the genes. The expression level of only 100 (= 1%) of the mRNAs shared by the studies was consistently different in four studies out of five, and only five transcripts were found consistently differently expressed in the paired transcriptomes in all five studies, based on the same criterion as above (Tead1, 1700010D01Rik, Fstl5, Pnrc2 and Zfp472). Thus, our literature search revealed that about 70% of sister blastomeres exhibit different developmental potential, while they have only very few transcriptomic differences (less than 1% of their commonly expressed genes) in common. In other words, interblastomere differences identified at the transcriptomic level are not reproducible from one study to the other. Lack of reproducibility is a well-known problem (Munafo et al. 2017) and is often considered as a criticism (claiming that it is due to, for example, poor experimental design or data manipulation). However, in the context of this review, this low level of reproducibility may be genuine and due to multiple factors.

View Full Size

Reproductive totipotency along developmental stages is inversely proportional to the complexity and variability of the transcriptome (Entropy H, as described in mice, see Piras et al. 2014). Data for the level of complexity and variability of the transcriptome were extracted from the Piras et al. paper. Data for reproductive totipotency are taken from Fig. 3 (this paper).

Citation: Reproduction 158, 2; 10.1530/REP-18-0462

• Download Figure
• Download figure as PowerPoint slide

How can low reproducibility relate to genuine data? This can happen, for example, when the origin of the trait examined is multifactorial but only some of those factors are included in the analysis. Firstly, transcripts are only one type of the possible mediators of totipotency, and protein level and localization, organelle partition, metabolic activity, and so on should also be considered. Thus, we should not expect transcripts to embody all of the interblastomere variability. Secondly, housekeeping genes, which totipotency might also rely on, were excluded from our analysis (Casser et al. 2018). Thirdly, irreproducibility might be genuine: The best case in point is the very low probability of finding an egg genetically identical to another egg, given the complexity of meiotic recombination. Lastly, the studies we analyzed worked with different mouse strains and this is known to generate additional molecular variation.

Despite their small number, we ask whether we can learn something from the five gene candidates identified in all five studies or from the larger group (100 mRNAs) selected in four out of five of them. So far none of the five transcripts have been recognized as developmentally relevant genes. Nevertheless, the larger group included Eomes, Tead1, Phlda2 and Cops3, which play significant roles in cell lineage specification or proliferation. In fact, fetal formation in mice required a minimum of four NANOG-positive cells at the time of implantation (Morris et al. 2012), and the number of NANOG-positive cells might also play a comparable role in human development (Van de Velde et al. 2008, Noli et al. 2015). Thus, it could be speculated that differences in gene expression between blastomeres are conducive to a variable number of cells finally committed to such a NANOG-defined cell lineage (e.g. ICM). This is consistent with the functional observations made in MZ twin blastocysts (Casser et al. 2017). Altogether, this would suggest that part of what makes a blastomere totipotent, that is totipotency signature, would lie in the ability to regulate the NANOG-positive cell number (Fig. 6).

View Full Size

Developmental view of totipotency-pluripotency transition. Totipotency is an important characteristic of the fertilized egg or zygote that is conserved or lost partially/totally with the first cleavage and subsequent embryonic divisions. Embryonic cells evolve to pluripotency, as observed in stem cells, exhibiting several properties such as self-renewal and high differentiation capacity. Reproductive totipotency is particularly associated with the later production of a sufficient number of NANOG-positive cells in the stem cell compartment of the epiblast (Morris et al. 2012).

Citation: Reproduction 158, 2; 10.1530/REP-18-0462

• Download Figure
• Download figure as PowerPoint slide

Regarding the molecular definition of totipotency, it was also proposed that because two-cell stage blastomeres can be totipotent, the common transcripts between those embryos and a small subpopulation of ES cells with a profile of gene expression that is reminiscent of the two-cell stage would correspond to the totipotency molecular signature. First, it is critical to remember that totipotent and pluripotent cells are often confused with stem cells, but stem cells have the property to self-renew and to divide indefinitely, which is not the case for the cells of the native embryo. The list of genes whose transcripts were used to define the two-cell-like state of ES cells and that were considered among the candidate genes for totipotency markers in mice includes: MERVL, and Eif1a-like genes, Zscan4b-Zscan4f, Zfp352, Tdpoz1-5 (Morgani & Brickman 2014, Wu et al. 2017). Those genes were not part of the common differences found between blastomere at the two-cell stage (Casser et al. 2018), simply because selection criteria (two-fold difference of mRNA level in at least 50% of the pairs) were not fulfilled. Functionally, two cell-like stem cells can contribute to both embryonic and extraembryonic lineages in chimeras; therefore, they could be called ‘totipotent’ according to the more permissive definition of totipotency, which is expressly not taken as the default in this review. Moreover, in global terms, the two-cell-like stem cells might be transcriptionally more similar to blastocyst cells than blastomeres from the two-cell stage (Kolodziejczyk 2016). A similar approach in humans revealed a set of 61 genes that might constitute a molecular totipotency signature distinguishing the blastomeres of day 3 embryos from pluripotent cells (Galan et al. 2013).

Independently of the genome-wide studies reanalyzed above, Zheng and collaborators found that mtrRNAs located outside of mitochondria were differently abundant in the cytoplasm of sister mouse blastomeres at the end of the two-cell stage (Zheng et al. 2016). This interblastomere imbalance was observed in 10–20% of the two-cell embryos examined. When Zheng and colleagues tested the functional importance of 16S mtrRNA by gain- and loss-of-function, they found that those mtrRNAs seemed to drive the blastomere fate toward the ICM (Zheng et al. 2016). Mechanistic understanding of the role of 16S mtrRNA warrants further investigation even if it seems unlikely that a single RNA could be in charge of regulating totipotency versus lineage commitment.

Being aware that linear correlation is generally poor between transcript and protein level (Schwanhausser et al. 2011) and that there is a two-fold variation in the amount of protein and nucleic acid contained in the different blastomeres of an individual eight-cell mouse embryo (Cheung et al. 2013), the single-cell proteome analysis of sister blastomeres is likely to be a key experimental approach to capture the molecular determinants of blastomere totipotency (Virant-Klun et al. 2016). Older and more recent studies contribute data supporting interblastomere differences at the protein level, for example, SNAI1/SNAI2 (Bell & Watson 2009) and GADD45a (Rosas et al. 2018). It will be necessary to analyze the composition of each blastomere at the two-cell stage and identify critical proteome requirements supporting totipotency, in order to make further progress, but, of course, it is impossible to combine a destructive assay with a proof of function for the same cells. Cellular biopsy may help, but because we suspect a dose of mosaicism, removing a subcellular volume of the blastomeres for analysis does not appear a safe way to infer the average blastomere composition. An interim solution to find the molecular origin of the early diversification occurring between two-cell blastomeres in 70% of the embryos could be to use modeling of the proteomic data collected from reproducibly bisected oocytes (i.e. along the animal-vegetal axis) or from single blastomeres whose relationship to the orientation of the zygotic cleavage plane is known. A comparison between the daughter cells obtained from the two-cell blastomere that cleaved first and second could also be productive. Transcriptomic datasets for all blastomeres of the same four-cell embryo are available (Goolam et al. 2016, Biase et al. 2018), but their origin from first-divided or second-divided two-cell blastomere was not recorded. Moreover, the fact that cleavage axes have different orientations in different embryos could result in multiple ways of maternal molecule partition, compromising the reproducibility and comparability of these data.

Conclusions and revised model: mosaic construction of zygotic totipotency

In this review, we revisited the multifaceted concept of totipotency, starting from the historical notions and confronting them with the data available on the development of MZ twins and gene expression profiles of sister blastomeres, to achieve a modern integrative view. Among various definitions, we specifically defined totipotency as the ability of one cell to give rise to a new complete organism. It should be underlined that this definition overlaps with the definition of developmental biology itself (study of processes by which a one-cell stage embryo or zygote develops into a fully structured organism). Thus, totipotency is a key concept in our understanding of the developmental program through which metazoan organisms are generated.

In mammals, it was generally accepted that early blastomeres were equally (toti)potent before showing overt signs of molecular differentiation at the four-cell stage (Torres-Padilla et al. 2007). However, our review documented that only about one-third of two-cell embryos achieve the full development of monozygotic twins after experimental separation of the blastomeres performed in various mammalian species (mouse, cow, sheep, rat and monkey). It is important to stress ‘full development’ (to term), which is the compass we used throughout the review. Hence, two-cell blastomeres have a common disposition to form an organism, but its extent is not necessarily equal. This pattern would go unnoticed in the intact embryo, because the ‘superior’ or more potent blastomere would mask the deficit of the other blastomere (provided the deficit is relevant for development). As a first conclusion of this review, we should recognize that most mammalian two-cell embryos have an inner imbalance of totipotency between the sister blastomeres and that this does not impair their development as a whole. Given the numbers we captured in this review (thousands of two-cell embryo bisections and hundreds of embryo transfers), the first conclusion seems rather solid, and we consider that further research would be more productive if used not to double-check whether the imbalance really exists, but to illuminate its molecular nature.

To acknowledge this imbalance is to recognize that there is no strict maintenance of totipotency at or beyond the two-cell stage, and this raises the question whether the loss is regulated or stochastic. We posit that a regulated process would have been reflected in a more homogeneous pattern of molecular differences between blastomeres than the pattern we actually saw. In teleological terms, we consider that there is no compelling reason for mammalian two-cell embryos to maintain totipotency in both blastomeres, because (1) the sheltered environment of the mother makes cell damage improbable and thereby a second totipotent blastomere – to compensate for damage – superfluous and (2) the natural tendency of increase in entropy is better satisfied when the sister blastomeres are not equal. In an evolutionary perspective, this inequality is also a source of diversity, which is a substrate of Darwinian selection. Accordingly, it is tempting to consider that the maintenance of totipotency through cleavage is no developmentally programmed event, and that there is no preordained mechanisms to ensure totipotency in both blastomeres. As a result, either both or just one could be totipotent, fortuitously, depending for instance on how the randomly oriented axis of the first zygotic cleavage (Louvet-Vallee et al. 2005, Zernicka-Goetz 2005) apportions some spatially restricted developmental information laid in the ooplasm. This does not exclude cases where both blastomeres are not totipotent and the embryo is, therefore, unviable. This can be due to the failure of both blastomeres reflecting errors of chromosome segregation at first zygotic mitosis (most aneuploidies preclude development to term; Rabinowitz et al. 2012, Babariya et al. 2017) or the lack of developmental competence of some oocytes, i.e. not all of them developed after fertilization because of a deficiency in maternal factors (Christians 2017). Furthermore, cases where both blastomeres are not totipotent can also be consistent with unequal segregation of determinants or ingredients that should be inherited in a defined amount by the sister two-cell blastomeres to support their totipotency. Although we need to exercise caution about these determinants, because segregation of proven ones has yet to be documented in mammalian embryos and histochemical diagnostic methods have limitations (Littwin & Denker 2011), we are not totally clueless. As mentioned earlier, totipotency of a blastomere in mice, for example, could correlate with a number of ≥4 epiblast cells formed by the blastomere progeny at the time of implantation (Morris et al. 2012) and, therefore, rely on some determinants regulating the accumulation of a blastomere’s progeny cells in the epiblast (e.g. NANOG). This leads to the second conclusion of this review: The molecular signature of totipotency is obviously complex, yet the search for it might be facilitated if it is oriented toward the determination of the epiblast lineage (Fig. 6).

Based on the findings reviewed above, we suggest a revised model of mammalian totipotency discontinuity, where mosaicism and imperfect regulation (imperfect because it is unable to offset the initial difference) are sequentially involved. Only the zygote is totipotent with certainty in the mammalian life cycle. Its organization is mosaic, so that the first cleavage axis divides still elusive determinants between the two-cell stage blastomeres, which can – but do not have to be – both totipotent. Depending on their composition, blastomeres evolve in regulating their developmental program to successfully achieve the formation of a proper blastomere (primary regulation) and then a complete organism. Thus, our model does not run counter to the vast body of data attesting the regulative ability of the mammalian embryo, but it rather advocates against their rigid interpretation that discounts a preexisting constraint of maternal origin playing a role in totipotency before regulation could gain momentum.

Note added in proof

While this paper was under revision it was reported that two long noncoding RNAs, namely LincGET (Wang et al. 2018) and Neat1 (Hupalowska et al. 2018), are asymmetrically expressed between sister blastomeres of mouse embryos at the 2-cell stage. These lncRNAs interact physically with CARM1 and promote blastomere contribution to the inner cell mass, thereby shedding light onto the elusive link between blastomere potency and gene expression in 2-cell stage.

Declaration of interest

M B serves on the Editorial Board of Molecular Human Reproduction and of the International Journal of Developmental Biology. The other authors have nothing to disclose.

Funding

This work did not receive specific funding, but the authors were supported by the Max Planck Institute for Molecular Biomedicine (M B, E C), Deutsche Forschungsgemeinschaft (M B, G F), and Sorbonne Université and Centre national de la Recherche scientifique (E S C) during its preparation.