The APC activator fizzy-related-1 (FZR1) is needed for preimplantation mouse embryo development

Most cells, including oocytes and those in embryos, have to regulate their division to ensure faithful segregation of chromosomes. An important player in this process is the anaphase-promoting complex (APC), which ensures degradation of key substrates at specific times in the mitotic cell cycle and during meiosis (Jones, 2011; Pesin and Orr-Weaver, 2008; Peters, 2006). Owing to its essential role in cell division there is much interest in how the activity of the APC is controlled. This is at least in part answered by its ability to bind to one of two protein coactivators, CDC20 and fizzy-related-1 (FZR1; also known as CDH1), whose binding affinity is affected by CDK1 phosphorylation and whose substrate specificities differ (Peters, 2006). APC^CDC20 function is principally restricted to the metaphase–anaphase transition (Izawa and Pines, 2011). In contrast, functions of APC^FZR1 are more varied, due to its ability to be active for a much greater fraction of the cell cycle, when CDK1 is low, and an ability to recognise a much wider range of substrates (Cotto-Rios et al., 2011; Qiao et al., 2010; Takahashi et al., 2012; Tudzarova et al., 2011; Wan et al., 2011).

FZR1 is thought to be absent from the eggs and early embryos of many model organisms (Lorca et al., 1998; Raff et al., 2002; Sigrist and Lehner, 1997). However, mammalian oocytes do contain FZR1 and antisense knockdown approaches have established roles for this protein in both prophase I arrest and maturation during meiosis I (Homer et al., 2009; Marangos et al., 2007; Reis et al., 2006; Reis et al., 2007; Schindler and Schultz, 2009; Yamamuro et al., 2008). Furthermore, APC^FZR1 activity has also been observed in meiosis II following fertilization (Chang et al., 2004). An in vivo knockout approach would prove valuable in order to understand more fully the function of FZR1 in mammalian eggs and embryos, without the ambiguities of antisense knockdown strategies. However, a conventional FZR1 knockout is embryonic lethal (García-Higuera et al., 2008; Li et al., 2008), and without viable pups any examination of its meiotic function is not possible. Furthermore, due to the persistence of maternal stores in knockout embryos, the function of FZR1 at early embryonic developmental timepoints is difficult to determine.

This study set out to investigate the role of maternal FZR1 in the completion of meiosis and in early embryogenesis during preimplantation development using an oocyte-targeted knockout of mice harbouring floxed Fzr1 (Fzr1^fl/fl) alleles and Zona-Pellucida 3 (Zp3)-driven Cre-recombinase. We have previously used this approach to generate oocyte-specific Fzr1 knockout mice (Fzr1^−/−) and confirm that APC^FZR1 plays an important role in maintaining prophase I arrest (Holt et al., 2011). Although such mice contain fewer prophase I oocytes, due to the spontaneous meiotic maturation caused by FZR1 loss, there remains sufficient numbers that stay arrested to allow for examination of its later meiotic and early embryonic functions.

FZR1 was present in germinal vesicle stage (GV) oocytes from control mice containing the floxed Fzr1 allele (Fzr1^fl/fl) but not in Fzr1^−/− oocyte-specific knockouts (Fig. 1A,B). APC^FZR1 activity was not essential for meiosis I because metaphase II (metII) eggs were ovulated by Fzr1^−/− mice (Fig. 1C). The nuclear maturation of such eggs was normal (Fig. 1C), however, the numbers ovulated were about half those of controls (12.6±7.6 versus 24.6±8.7 per female, P = 0.03; Fig. 1D). This reduction is accounted for by the 40% fewer GV oocytes present in knockout mice (Holt et al., 2011).

FZR1 levels dropped by >50% during oocyte maturation in controls (Fig. 1A,B). This is possibly through the activity of a FZR1 degradation pathway (Benmaamar and Pagano, 2005; Listovsky et al., 2004), one of which involves APC^FZR1 mediated autoubiquitylation, consistent with reports of such activity in maturing oocytes (Homer et al., 2009; Reis et al., 2007) as well as a reduction in Fzr1 translation rates observed in maturing oocytes (Chen et al., 2011). Although loss of APC^FZR1 activity during maturation by antisense knockdown was observed to raise aneuploidy rates markedly for in vitro matured oocytes (Reis et al., 2007), there was only a small increase in the percentage of aneuploid eggs from Fzr1^−/− mice, which did not reach a level of statistical significance with the numbers analysed (4%, versus 11%; P = 0.11; Fig. 1E). We conclude therefore that although APC^FZR1 activity has been measured in oocytes both during meiosis I and meiosis II in vitro (Chang et al., 2004; Reis et al., 2007), these activities are not essential for oocyte maturation in vivo.

In embryos collected from Fzr1^fl/fl mice that were mated to wild-type males, we observed a very marked reduction in FZR1 protein following fertilization, such that levels were not easily detected using 50 2-cell embryos or morulae (supplementary material Fig. S1). This is consistent with the transcriptional quiescence of oocytes during meiotic maturation and the degradation of ∼70% maternal mRNA content by the time embryonic genome activation is initiated in 2-cell embryos (Pikó and Clegg, 1982). However, FZR1 levels at the expanded blastocyst stage were comparable to metaphase II eggs (supplementary material Fig. S1) and from this time onwards it is known to have an essential role in trophoblast cell endoreduplication, a process needed for placentation (García-Higuera et al., 2008; Li et al., 2008).

Having established the FZR1 protein profile during preimplantation development we wanted to determine whether maternal stores of FZR1 were essential for fertilization and embryo development. This was important to examine given the already established function that FZR1 plays in oocytes (Holt et al., 2011), and the existence of several maternal effect genes, which have essential roles in embryo development but cannot be rescued by paternal alleles (Ma et al., 2006; Wu et al., 2003; Zheng and Dean, 2009). Therefore, we assessed the fertility of Fzr1^−/− females over a 6-month period by mating to wild-type C57BL6 males.

We observed that the loss of maternal FZR1 during meiosis did not compromise the ability of the oocyte to generate a viable embryo. The cumulative number of pups born to each female and the number of days between litters was the same for control and Fzr1^−/− mice (Fig. 1F,G). Fzr1^−/− females did have smaller litter sizes (6.8±2.6 versus 8.1±2.7 pups, P = 0.02, t-test; Fig. 1H); most likely due to their lower numbers of ovulated metII eggs. All pups appeared healthy and thrived, so we conclude that there is no essential requirement for maternal stores of FZR1 following release from GV arrest in the completion of meiosis, fertilization or embryo development. However, this study design used wild-type males and as such cannot reveal if FZR1 is essential for early embryonic development because of paternal FZR1.

The above experiments show that embryo development is normal in the absence of maternal FZR1 if paternal FZR1 is present. Sperm cells contain FZR1 protein; however, mating of Fzr1^−/− females with Fzr1^−/− males was not possible because sperm-specific Fzr1 knockout blocks spermatogenesis (J.E.H., I.G.-H., S.M., K.T.J. and J. E. McLaughlin, unpublished data). In order to examine embryo development in the absence of any FZR1 we parthenogenetically activated metII eggs from Fzr1^−/− females using strontium-containing medium, which has been used widely in a number of studies (Bos-Mikich et al., 1997; Kline, 1996; Rogers et al., 2006). Embryos were made diploid by the addition of cytochalasin to block cytokinesis (Balakier and Tarkowski, 1976), and the technique employed here gives the same rate of preimplantation development as embryos generated by IVF (Liu et al., 2002). FZR1 protein levels in parthenotes generated from Fzr1^fl/fl females in this manner share the same profile as embryos produced by mating with wild-type males (supplementary material Fig. S1).

We observed no significant difference in the percentage of Fzr1^fl/fl and Fzr1^−/− eggs that were activated in strontium-containing medium (supplementary material Fig. S2A,B). Furthermore, the timing of meiotic exit was not delayed in Fzr1^−/− eggs (supplementary material Fig. S2C). Therefore, although a switch from APC^CDC20 to APC^FZR1 activity can be measured in eggs (Chang et al., 2004), FZR1 is not essential for meiotic exit.

To examine the viability of embryo development in the absence of any maternal or paternal FZR1, parthenotes were further cultured following activation. Interestingly, here we noted a significant delay and early developmental arrest in Fzr1^−/− embryos, with ∼75% becoming arrested at the 2-cell stage (Fig. 2A). These embryos continued to maintain this 2-cell stage arrest, even after 4 days in culture. For those Fzr1^−/− embryos not arrested the mean timings of the first three mitotic divisions were significantly delayed indicating they had difficulties in undergoing cell division (Fig. 2B).

To further analyse the effects of FZR1 loss, we examined the chromatin of the 2-cell arrested embryos. The 5% (n = 6/114) of control Fzr1^fl/fl embryos that remained 2-cell arrested at day 4 were all blocked in interphase with decondensed chromatin contained within a single nucleus of both blastomeres (Fig. 2C). This is consistent with the conventional 2-cell block phenomenon that has previously been observed during mouse embryonic development in vitro (Hamatani et al., 2004; Schultz, 2002; Wang et al., 2004). About 30% (n = 13/41) of the Fzr1^−/− embryos that were 2-cell arrested were similar to controls. However, ∼70% (n = 28/41) showed chromatin defects, suggestive of an inability to maintain genomic integrity. The most common abnormality, affecting 50% (n = 20/41) was the presence of binucleate blastomeres, micronuclei or wholly fragmented chromatin in one or both blastomeres (Fig. 2Di–iii; supplementary material Fig. S3A,B). In 30% of Fzr1^−/− embryos (n = 12/41) at least one of the blastomeres was in mitosis, with arrest characterised in prophase, prometaphase or telophase (Fig. 2Div; supplementary material Fig. S3C).

FZR1 has been implicated in many responses to genomic stress, and can be regarded as a tumour suppressor (Engelbert et al., 2008; García-Higuera et al., 2008; Qiao et al., 2010). Therefore, we decided to examine for γ-H2AX staining in both Fzr1^−/− and Fzr1^fl/fl parthenotes. Ataxia telangiectasia mutated (ATM), a major regulator of DNA damage phosphorylates H2AX, generating the epitope recognised by anti-γ-H2AX antibodies (Lukas et al., 2011). Such γ-H2AX foci were readily present in 2-cell arrested Fzr1^−/− but not Fzr1^fl/fl embryos (Fig. 2E). 80% of Fzr1^−/− embryos, which had become arrested at the 2-cell stage, had γ-H2AX foci but this was less than 15% in Fzr1^fl/fl embryos assessed either at the same developmental timepoint (2-cell stage) or at the same time after egg activation, which was at 4 days when they were morulae (Fig. 2F). This increase in γ-H2AX foci has most likely been generated by breakage of chromatin in the previous mitosis, and is consistent with the fragmented DNA observed in Fzr1^−/− embryonic fibroblasts (García-Higuera et al., 2008).

In order to understand more fully the provenance of the 2-cell Fzr1^−/− embryo arrest and confirm the conclusions drawn above, chromatin was imaged in live 1-cell embryos expressing histone-2B–mCherry cRNA by confocal imaging. In all control Fzr1^fl/fl embryos (n = 13), we observed pronuclear fusion, a process associated with the mixing of the chromosomes (syngamy), followed by passage through mitosis and formation of a 2-cell embryo (Fig. 3A,B). Some embryos possessed lagging chromosomes during anaphase (n = 4/13; Fig. 3C) but these were temporary and they did not result in formation of micronuclei. In contrast to these observations, a third (n = 4/12) of the imaged Fzr1^−/− parthenotes failed to undergo syngamy and following mitotic entry formed two separate spindles. At anaphase, which occurred at the same time for both spindles, the chromatin underwent disjunction forming two nuclei per blastomere in the resulting 2-cell embryo (Fig. 3B). Given that the occurrence of this event is in proportion to the 2-cell arrested binucleate embryos observed previously, we conclude that binucleate 2-cell embryos are generated in Fzr1^−/− parthenotes due to the failure of the two pronuclei in the 1-cell embryo to undergo syngamy during mitotic entry. All the remaining 1-cell Fzr1^−/− parthenotes (n = 8) that did undergo syngamy nonetheless showed lasting defects during mitosis, which included mitotic arrest or formation of micronuclei (Fig. 3C,D).

For the small percentage of Fzr1^−/− embryos not 2-cell arresting, their delayed cleavage during the first three cell cycles could impact on their subsequent compaction, a process normally initiated at the 8-cell stage (Chen et al., 2010; Eckert and Fleming, 2008). Until this time all blastomeres are morphologically alike but at compaction they polarise and this event marks the initiation of blastocyst formation and cellular specialisation into trophectodermal cells and the inner cell mass. We found that compaction was initiated independently of FZR1, at 80 hours post-hCG (P = 0.66, t-test; Fig. 4A). However, because Fzr1^−/− embryos were developing at a slower rate, the actual developmental timepoint of compaction was earlier. Greater than 60% of 4-cell Fzr1^−/− embryos were compacted, but no compaction was ever observed in control Fzr1^fl/fl at this developmental stage (Fig. 4B).

Using E-cadherin as a marker of tight junction formation (Larue et al., 1994; Vestweber et al., 1987), immunostaining was mostly cytoplasmic in non-compacting 4- and 8-cell Fzr1^fl/fl embryos and became far more intense at sites of cell-to-cell contact in compacting morulae. In contrast, cell-to-cell E-cadherin immunostaining was evident in all 4-cell Fzr1^−/− embryos at levels comparable to control morulae (Fig. 4C; supplementary material Fig. S4). Therefore, in the absence of FZR1, although embryonic cleavages are delayed, compaction can still occur and remains independent of the number of blastomeres present. However, despite the ability of Fzr1^−/− embryos to undergo compaction, very few developed beyond the morula state (data not shown).

FZR1 activity has been measured previously during GV arrest (Holt et al., 2011; Marangos et al., 2007; Reis et al., 2007; Schindler and Schultz, 2009; Yamamuro et al., 2008), meiotic maturation (Homer et al., 2009; Reis et al., 2006), and at fertilization (Chang et al., 2004). We conclude that all these processes are not completely dependent on FZR1 because viable pups were born when Fzr1^−/− mothers were bred with wild-type males. This situation of measureable (Hagting et al., 2002; Lindon and Pines, 2004) but non-essential (Engelbert et al., 2008; Floyd et al., 2008; Sigl et al., 2009) APC^FZR1 activity is similar to mitotic exit in somatic cells at anaphase onset.

Previously, Fzr1 loss in oocytes during meiotic maturation using morpholinos was reported to raise aneuploidy rates (Reis et al., 2007). Here, when maturation was allowed to proceed in vivo, FZR1 loss only mildly raised rates. However, a much greater effect is measured in Fzr1^−/− oocytes if a similar methodology used for the antisense knockdown is replicated, by removing the surrounding granulosa cells and culturing oocytes for extended times (Holt et al., 2012). Therefore, it is possible that granulosa cell–oocyte communication affords some protection against aneuploidy resulting from FZR1 loss.

We observed many 2-cell arrested binucleate Fzr1^−/− parthenote embryos. This is attributed to FZR1 depletion rather than cytochalasin addition during the creation of parthenotes because control Fzr1^fl/fl embryos, which did not show any developmental arrest, were also made diploid by the same drug treatment. Additionally the arrest was not due to damage caused to the oocyte by Fzr1 loss during meiosis because if it were then embryonic development would not have been observed in Fzr1^−/− female mice mated with wild-type males. The fact that Fzr1^−/− parthenotes embryos show preimplantation lethality yet embryos generated by the mating of Fzr1^−/− females with wild-type males are viable, we attribute to presence of paternal FZR1 in the latter case. This is our preferred hypothesis based on the fact that FZR1 protein can be measured in sperm cells (J.E.H., I.G.-H., S.M., K.T.J. and E. A. McLaughlin, unpublished data). However despite this, it cannot be ruled out that some aspect of parthenogenetic development is uniquely FZR1 dependent.

Binucleate embryos were created by a failure of the embryo to form a single mitotic spindle during the first division. Normally the two pronuclei come into apposition by a process that is microtubule dependent, and following nuclear envelope breakdown assemble on a single spindle (Schatten et al., 1985). In many of the Fzr1^−/− embryos the pronuclei did not undergo syngamy and instead established separate spindles, which formed with the same orientation and underwent division with similar timings. As such each blastomere in the 2-cell embryo contained two nuclei following mitotic exit. We speculate that FZR1 loss interferes with this microtubule dependent process because we have also observed effects of its loss on the timing of assembly of spindle microtubules during the first meiotic division (Holt et al., 2012).

Early cell division in the preimplantation embryo was FZR1 dependent, even for those embryos that did undergo syngamy. Their division was so slow that compaction occurred at the 4-cell stage, rather than as it should at the late 8-cell stage. However, for both Fzr1^−/− and Fzr1^fl/fl embryos compaction occurred at the same time relative to egg activation, which agrees with previous analysis using DNA replication blocking drugs that concluded this event is timed independently of cell number (Smith and Johnson, 1985; Valdimarsson and Kidder, 1995).

We have performed an immunoblot-based analysis of Fzr1^−/− oocytes to determine how its loss may have affected some known APC^FZR1 substrates involved in cell cycle control or spindle assembly, such as Cdc20, Aurora kinase A, securin, polo-like kinase I, TPX2 and HURP (Holt et al., 2012). Of these Cdc20 showed a 25-fold and cyclin B1 a 5-fold elevation, whereas the other substrates were not increased (Holt et al., 2012). Given Cdc20 overexpression is reported to cause mitotic defects and multinucleated cells in some human cell lines (Mondal et al., 2007), it is possible that the same phenomenon is being observed here in Fzr1^−/− embryos. However, potential FZR1 substrates are numerous so it may not be possible to pinpoint a single defect. For example, FZR1 is implicated not only in mitotic exit and G1 progression but also in DNA repair (Bassermann et al., 2008). This may be quite important for oocytes here because some types of DNA repair are repressed during meiosis and only occur following fertilization (Yuen et al., 2012).

Metazoan embryos of many model organisms have short cell divisions, consisting of alternating S and M phases without gaps, until the time of embryonic genome activation (Dalle Nogare et al., 2009; Gotoh et al., 2011; Lee and Orr-Weaver, 2003; Philpott and Yew, 2008; Zamir et al., 1997). In such systems there is a temporal association of FZR1 expression with the establishment of gap phases at embryonic genome activation (Kramer et al., 2000; Lorca et al., 1998; Sigrist and Lehner, 1997). Early mouse embryos do have gap phases, albeit extremely short, lasting ∼1 hour (Smith and Johnson, 1985). Therefore the essential requirement of FZR1 in mouse embryos may reflect the fact that early mammalian embryonic cell cycles are closer to those of mammalian somatic cells than they are to non-mammalian embryonic divisions.

Here we have found that FZR1 is essential for development at a much earlier timepoint than previously appreciated from knockouts. It is interesting from the present study that Fzr1 loss is so pronounced in mouse embryos. This contrasts with cultured mammalian cell lines, chicken cell lines, embryonic fibroblasts, and yeasts, where although Fzr1 loss can be associated with increased genomic instability, cells are nonetheless mostly viable and continue to proliferate for some time (Engelbert et al., 2008; Floyd et al., 2008; García-Higuera et al., 2008; Li et al., 2008; Schwab et al., 1997; Sudo et al., 2001). The acute sensitivity of embryos to lack of FZR1 suggests an essential role in regulating a key developmental process that remains to be discovered.

Female Fzr1^fl/fl mice (García-Higuera et al., 2008) were mated with ZP3Cre [C57BL/6-Tg(Zp3-cre)93Knw] males. F1 male offspring with Fzr^+/fl/ZP3Cre genotype were then mated with Fzr1^fl/fl females to create oocyte-specific Fzr1 knockout (Fzr1^−/−) and control Fzr1^fl/fl littermates (Holt et al., 2011). Genotyping was described previously (García-Higuera et al., 2008). Female Fzr1^fl/fl or Fzr1^−/− mice, 5- to 6-weeks old, were mated with C57BL6 males of proven fertility. Continuous breeding pairs were maintained and litters were checked for daily.

GV oocytes and metII eggs were collected from 4- to 6-week-old mice as described previously (Chang et al., 2011; Holt et al., 2011). Eggs were activated in Ca²⁺-free KSOM (potassium simplex optimised medium) with 10 mM strontium chloride and 1 µg/ml cytochalasin D for 4 hours (Navarro et al., 2005; Rogers et al., 2006), and then washed in KSOM medium and cultured under paraffin oil in a humidified atmosphere of 5% CO₂ at 37°C.

Histone-2B–mCherry cRNA was made and microinjected into metII eggs as described previously (Lane et al., 2012). Images were captured using an inverted microscope fitted for epifluorescence (Holt et al., 2010; Lane et al., 2010; Madgwick et al., 2006) or using an Olympus FV1000 confocal microscope (Lane et al., 2012). Live imaging of 1-cell embryos was performed at 5-minute intervals. The medium for microinjection and imaging was Hepes-buffered KSOM. Image analysis was performed using ImageJ (NIH, Bethesda, USA) or FV10-ASW2.0 viewer software (Olympus, Japan).

Following the immunoblotting procedure (Holt et al., 2011) PVDF membranes were probed with antibodies to FZR1 (1∶100, Abcam, UK) and glyceraldehyde-3-phosphate dehydrogenase (1∶10,000, Sigma). Horseradish-peroxidise-conjugated secondary antibodies (DAKO, Denmark) were used for detection by ECL.

Eggs/embryos were fixed and permeabilised as described previously (Lane et al., 2010). For aneuploidy analysis eggs were pretreated with monastrol (Lane et al., 2010). Antibodies used were CREST (1∶400, Cortex Biochem, USA), tubulin (1∶200, Invitrogen), E-cadherin (M108, 1∶200, Takara Bio Inc., Japan) and γ-H2AX (1∶100, Abcam, UK). Secondary antibodies were Alexa Fluor 488, Alexa Fluor 555 or Alexa Fluor 633 (Invitrogen), counterstaining with Hoechst (20 µg/ml) or propidium iodide (10 µg/ml). Cells were mounted on glass slides with Citifluor (Citifluor Ltd, UK).

Statistical analyses were performed using GraphPad Prism version 5 (GraphPad Software, CA, USA). Comparison of means was made using either Student’s or Welch’s t-test and all dichotomous data by a Fisher’s exact test.

We thank members of the Jones laboratory for critical reading of the manuscript.