Absence of a robust mitotic timer mechanism in early preimplantation mouse embryos leads to chromosome instability

Preimplantation mammalian embryos frequently comprise a mixture of cells containing the correct chromosome complement (euploid cells) and cells containing an abnormal number of chromosomes (aneuploid cells). These mosaic-aneuploid embryos are common in fertility clinics, with many studies reporting upward of 70% of embryos with some aneuploid cells, indicating that chromosome segregation in the initial mitotic divisions after fertilization is error prone (Hook, 1981; Delhanty et al., 1993; Macklon et al., 2002; Vanneste et al., 2009; van Echten-Arends et al., 2011; McCoy, 2017; Vázquez-Diez and FitzHarris, 2018). Chromosome segregation occurs during mitosis (M phase) and is facilitated by the spindle, a transient structure formed from microtubules that is responsible for gathering, sorting and aligning replicated chromosomes, before dispatching them into two newly forming cells in anaphase, which marks the end of M phase (Compton, 2000; Heald and Khodjakov, 2015; Petry, 2016). However, little is known about how chromosome segregation is regulated in the niche context of early embryogenesis, or why errors that cause mosaicism are so common.

It has long been known that precise regulation of M-phase duration is key to avoiding chromosome segregation errors (Therman et al., 1984). This is exemplified by the spindle assembly checkpoint (SAC), a signaling pathway that inhibits the anaphase promoting complex (APC/C) and thus prevents M-phase completion until all chromosomes are correctly bioriented on the spindle. In most cells, the SAC plays an essential role in the prevention of anaphase and M-phase completion until the very last chromosome is aligned (Lara-Gonzalez et al., 2012; Musacchio, 2015; Lara-Gonzalez et al., 2021); however, in mammalian embryos, the SAC is inefficient at preventing anaphase, and therefore chromosome segregation frequently occurs before all the chromosomes are aligned (Vázquez-Diez et al., 2019). Thus, an insufficiently long M phase as a consequence of SAC weakness likely contributes to chromosome segregation errors in the mammalian embryo.

By contrast, recent work in somatic cells has highlighted the impact of unduly long M phases. First, the sustained tension exerted upon chromosomes by the spindle during a prolonged M phase can cause the cohesion between sister chromatids to be lost and the sister chromatids to separate precociously, a process termed ‘cohesion fatigue’ (Daum et al., 2011; Gorbsky, 2013; Sapkota et al., 2018; Stevens et al., 2011; Worrall et al., 2018). As precociously separated chromosomes activate the SAC, cohesion fatigue in somatic cells causes perpetual M-phase arrest (de Lange et al., 2015; Gorbsky, 2013; Lara-Gonzalez and Taylor, 2012; Stevens et al., 2011). However, the impact of cohesion fatigue in a cell type lacking a strong SAC, such as the mammalian embryo, is unknown. Second, recent experiments in somatic cells have described a ‘mitotic timer’ mechanism, in which cells that experience a slightly prolonged M phase undergo an arrest in the subsequent G1 phase, via a pathway involving USP28, 53BP1, p53 and p21 (Dalton and Yang, 2009; Wong et al., 2015; Fong et al., 2016; Lambrus and Holland, 2017; Lambrus et al., 2016; Meitinger et al., 2016; Uetake and Sluder, 2010). This mechanism detects extensions of M phase in the order of tens of minutes, and is present in cell lines, primary cultures and mid-gestation mouse embryos (Bazzi and Anderson, 2014; Lambrus and Holland, 2017; Lambrus et al., 2016; Uetake and Sluder, 2010). As long M phases suggest stressful mitoses (Dalton and Yang, 2009), the mitotic timer likely serves as a failsafe mechanism preventing the propagation of such cells. Here, we report that a 6 h M phase fails to arrest embryo development, but the same prolongation of M phase can cause considerable cohesion fatigue, and we demonstrate that this could contribute to early embryo aneuploidy.

M phase of the two- to four-cell division lasts for 90-120 min (Vázquez-Diez et al., 2019; Fig. S1D) and is likely more representative of preimplantation development than the one- to two-cell division (Sikora-Polaczek et al., 2006; Reichmann et al., 2018), and thus, we focused on this stage of development. In cultured somatic cells, M phase typically lasts 30-60 min, and an increase of 30 min can cause a subsequent G1 arrest (Uetake and Sluder, 2010). To address whether such a mechanism exists in embryos at the two- to four-cell stage, we first established experimental conditions in which we could arrest cells at metaphase for a defined period of time, followed by a release from M-phase arrest to complete mitosis, and then examined further development of embryos. We employed APCin, a small-molecule APC/C inhibitor that functions by binding to Cdc20 (Sackton et al., 2014). As expected, 100 µM APCin added prior to nuclear envelope breakdown (NEBD) did not prevent mitotic entry, but caused a reversible arrest at metaphase without impacting the durations of other aspects of M phase (Fig. S1). Following washout of the drug, the embryos continued to develop without obvious deleterious impacts upon embryo health. Moreover embryos remained alive during prolonged APCin exposure (Fig. 1A; Fig. S1). To verify that APCin causes M-phase arrest by APC/C inhibition, we imaged GFP-tagged cyclin B, an APC/C substrate, in control and APCin-treated embryos, and confirmed that the degradation of cyclinB-GFP that occurs during anaphase in control embryos was not observed in APCin-treated embryos (Fig. S2). Thus, APCin exposure provides a simple and robust means of prolonging M phase by inhibiting the APC/C in embryos.

Next, to look for evidence of a mitotic timer mechanism, two-cell stage embryos were exposed to 100 μM APCin within 30 min of NEBD for 6, 14 or 24 h, and, following APCin removal, embryo development was subsequently monitored using time-lapse imaging (Fig. 1B-E). Following a 14 h or 24 h treatment, removal of APCin allowed M-phase completion, but subsequently the embryos remained arrested at the four-cell stage (Fig. 1C-E). Strikingly, however, following a 6 h treatment with APCin, the embryos continued to divide, and underwent compaction (to form the morula) and blastocoel formation (to form the blastocyst) (Fig. 1C-E). The resulting blastocysts had comparable cell numbers to control embryos (a mean of 79.93 cells in control compared with 73.37 cells after a 6 h M phase) and a similar mitotic index (Fig. 1F,G). Moreover, the proportions of inner-cell-mass and trophectoderm cells (distinguished using OCT4 antibodies) were similar in control embryos and those that had been arrested for 6 h at M phase (Fig. 1H), suggesting that cell allocation at the morula-blastocyst transition was broadly unaffected. A 6 h M phase that was induced using monastrol, an inhibitor of the ATPase activity of kinesin 5 (Maliga et al., 2002) that destabilizes spindle structure and thus activates the SAC (Kapoor et al., 2000), also failed to enforce an arrest at the four-cell stage (Fig. S3A-D). Thus, a 6 h M phase does not prevent subsequent cell divisions, or development to the blastocyst stage, but cell cycle arrests are elicited by extremely long M phases (>14 h).

In somatic cells, cohesion fatigue can occur when prolonged forces exerted by the spindle upon sister chromatid pairs cause premature separation of sister chromatids during an extended M phase (Daum et al., 2011; Gorbsky, 2013; Sapkota et al., 2018; Stevens et al., 2011). Examination of mitotic spindles in two-cell stage embryos that were arrested in M phase for 2, 6 or 24 h revealed a progressive loss of chromosome alignment (Fig. 2A-D). Centromere labeling revealed that, after 24 h, the sister chromatids of most of the misaligned chromosomes had prematurely separated (Fig. 2D), as opposed to being in coherent sister chromatid pairs. We also noted a progressive extension of the spindle, with a mean spindle length of 26.43 µm in control embryos compared with 45.27 µm in embryos in 24 h M phases (Fig. 2B,C), a commonly observed phenotype in cells that lose sister cohesion, likely reflecting the role of kinetochore microtubules in the regulation of spindle length (Dumont and Mitchison, 2009; Goshima and Scholey, 2010). Next, to quantify the proportions of chromosomes in which sister chromatids had separated, we treated M-phase-arrested cells with monastrol for 2 h immediately prior to fixation in order to collapse the spindle and spatially separate the chromosomes for examination (Fig. 2E), which would otherwise not be possible given that chromosomes in the metaphase plate are crowded. Whereas individual chromatids were almost never seen in control embryos, 5.82% of all sister chromatid pairs had individualized by 6 h and 66.58% by 24 h of APCin treatment (Fig. 2F). We also found that inter-kinetochore distances in coherent sister chromatid pairs increased from 0.58±0.013 µm (indicated as mean±s.e.m.) in controls to 0.73±0.043 µm at 24 h of APCin treatment (P<0.0001), suggesting that loss of sister-pair coherence is preceded by an increase in inter-centromere distance (Fig. 2G), as previously described for somatic cells (Sapkota et al., 2018). Next, we co-treated with APCin and monastrol simultaneously to enforce an M-phase arrest in the absence of a bipolar spindle (Fig. 2H). In 99.44% of cases, chromosomes were still paired after a 24 h M phase under these conditions (Fig. 2I,J), suggesting that, as in somatic cells, spindle forces are necessary for cohesion fatigue. To summarize, a 6 h M phase causes an appreciable level of cohesion fatigue, but the same treatment does not prevent subsequent cell divisions.

We next wondered what effect cohesion fatigue would have upon embryo development. Micronuclei are a widely used marker of chromosomal instability. Whereas in somatic cells, micronuclei are frequently reincorporated into the genome to drive chromothripsis, they remain separate throughout preimplantation development in embryos, and the pattern in which they are inherited at subsequent cell divisions causes aneuploid embryos (Vázquez-Diez et al., 2016). Micronuclei thus serve as a useful surrogate marker of the degree of chromosome segregation error and aneuploidy in embryos (Vázquez-Diez et al., 2016). We found that four-cell stage embryos generated after a 6 h M phase possessed significantly more micronuclei than control embryos (a mean of 0.18 micronuclei per embryo in control compared with 1.88 for embryos that had a 6 h M phase; P<0.001) (Fig. 3A-C). This increased frequency of micronuclei remained evident at the blastocyst stage (3.0 micronuclei in control compared with 5.58 for embryos with a 6 h M phase; P=0.0005) (Fig. 3). This observation is consistent with previous work showing that micronuclei in mouse embryos frequently fail to reincorporate into the principal nucleus in the next cell cycle, and instead remain separate and are repeatedly inherited by only one daughter cell at each subsequent cell division, a series of events that inevitably generates additional aneuploid cells (Vázquez-Diez et al., 2016). We also observed that embryos with a 6 h M phase displayed an appreciable but non-significant increase in the incidence of nuclei with abnormal and concave shapes (Fig. 3D), a phenotype that is associated with aneuploidy (Skinner and Johnson, 2017). Thus, M-phase extension in the two- to four-cell division leads to cohesion fatigue, but cell divisions continue and give rise to blastocysts with increased micronuclei and other nuclear defects.

We next wondered what triggers the interphase arrest in the case of extreme (14 h) M-phase extension, and whether this arrest shares characteristics of the mitotic timer mechanism. First, we conducted live-imaging experiments using PCNA:EGFP as a marker of S phase (Fig. 4A,B), as previously described (Paim and FitzHarris, 2019). We found that 76.67% of the blastomeres that had experienced a 14 h M phase progressed through the subsequent S phase, as evidenced by the transient appearance of clear PCNA:EGFP puncta in the nucleus (Fig. 4A,B). Although this approach was unable to determine the fidelity of S phase and whether replication errors occurred, it nonetheless suggested that the arrest that was incurred after extreme prolongation occurred in G2, in contrast to the G1 arrest that is associated with the mitotic timer in somatic cells.

In somatic cells, the mitotic timer arrest is thought to occur independently of DNA damage, but mitotic prolongation is associated with DNA damage (Lanni and Jacks, 1998; Dalton and Yang, 2009; Orth et al., 2012). Here, we found that extreme (14 h) M-phase prolongation increased the degree of DNA damage in mouse embryos in two ways. First, quantification of γH2AX, a marker of DNA damage, in the principal nucleus following a 14 h M phase indicated significantly more DNA damage in these embryos compared with controls (Fig. 4C,D; Fig. S4). Second, M-phase prolongation was associated with increased levels of micronuclei formation. Micronuclear DNA exhibited extreme γH2AX immunofluorescence, consistent with catastrophic levels of DNA damage (Fig. S4), as previously described (Crasta et al., 2012; Guerrero et al., 2010; Hatch et al., 2013; Vázquez-Diez et al., 2016; Zhang et al., 2015b). Strikingly, when we co-treated embryos with monastrol to collapse the spindle and thereby prevent cohesion fatigue during the 14 h M phase, γH2AX intensity in the principal nucleus was decreased (Fig. 4D; Fig. S4), and micronucleus numbers were reduced (Fig. S4). Thus, the arrest triggered by prolonged M-phase extension occurs in G2, is characterized by DNA damage, and is associated at least in part with loss of sister-chromatid pairing. Next, to identify a link between DNA damage, cohesion fatigue and the four-cell arrest after extreme M-phase prolongation, we exposed two-cell stage embryos to APCin for 14 h, or to APCin with monastrol for 14 h, and subsequently released both groups of embryos into media with or without KU55933, an inhibitor of ATM kinase that plays a role in the DNA damage response (Hickson et al., 2004) (Fig. 4E). Notably, the four-cell stage arrest caused by the 14 h M phase was partially abrogated by monastrol (Fig. 4F,G), the same treatment that prevented cohesion fatigue and reduced DNA damage associated with an extended M phase. Embryo development could also partially be rescued by KU55933-mediated inhibition of the DNA damage response pathway.

In summary, the arrest elicited by the 14 h M phase appears to be distinct from the mitotic timer for three reasons: (1) its lack of sensitivity compared with arrests elicited by tens of minutes of extra M phase in cultured cells, (2) its occurrence in G2, in contrast to the G1 arrest that is normally associated with the mitotic timer and (3) its association with DNA damage, unlike the mitotic timer. We note that, with relatively high variation in γH2AX signals between cells, we cannot exclude the possibility that factors other than DNA damage could contribute to this arrest induced by a prolonged M phase. Whether this arrest represents a physiologically relevant checkpoint in this system remains unclear.

Mammalian embryos are frequently aneuploid. Chromosome mis-segregation in oocytes causes aneuploid eggs and typically homogeneously aneuploid embryos, whereas mis-segregation in post-fertilization mitosis causes embryos that are a ‘mosaic’ of euploid and aneuploid cells. Here, we focused on the source of mitotic aneuploidy, finding that the mitotic timer mechanism, which is acutely sensitive in somatic cells and serves to prevent propagation of cells that underwent an overly long and stressful mitosis, is ineffective in embryos. Given the recent reports showing that the SAC is weak in embryos (Vázquez-Diez et al., 2019), and that the tetraploidy checkpoint, which has been described in some cell types (Andreassen et al., 2001; Ganem et al., 2014; Storchova and Kuffer, 2008), is also absent in embryos (Paim and FitzHarris, 2019), we hypothesize that a lack of quality control in the early embryo likely underpins mosaicism. The recent studies in mammalian embryos echo long-known deficiencies of cell cycle checkpoints in early embryos of other species (Hara et al., 1980; Zhang et al., 2015a; Galli and Morgan, 2016; Gerhold et al., 2018; Chenevert et al., 2020).

Notably, our data illustrate that the simultaneous absence of multiple checkpoints can have severe unexpected consequences. Whereas cohesion fatigue causes M-phase arrest in somatic cells, treatments that cause cohesion fatigue in embryos failed to cause a similar arrest, presumably as a result of the aforementioned SAC weakness, and micronuclei appeared in the subsequent interphase. Subsequently, in the apparent absence of a mitotic timer, these cells could divide further. In somatic cells, micronuclei can reincorporate into the nucleus in the next mitosis (Crasta et al., 2012; Huang et al., 2012; Zhang et al., 2015b); however, in embryos, they are unilaterally inherited at each subsequent cell division, and thereby impart a chromosomal instability upon their progeny for many cell cycles (Vázquez-Diez et al., 2016). Consistent with this, we found that, following a prolonged M phase at the two-cell stage, the micronucleus number per embryo remained high even in blastocysts. Thus, the absence of fully robust SAC and mitotic timer mechanisms means that cohesion fatigue could lead to a cascade of aneuploid cells, and thus constitute an unappreciated contribution to embryo mosaicism.

A major development in assisted reproduction in humans is the introduction of time-lapse microscopy to aid the selection of viable embryos (Aparicio-Ruiz et al., 2018; Chavez et al., 2012; Fishel et al., 2018; Gallego et al., 2019; Kirkegaard et al., 2012; Meseguer et al., 2011), although its effectiveness is uncertain. Whereas embryo timelapse approaches typically consider the timings of cell divisions (i.e. cytokinesis), the timing of NEBD, which could allow the duration of M phase to be calculated, is generally not considered. Based on our observations here, we speculate that, in the future, the integration of M-phase lengths into embryo selection algorithms, perhaps paired with the application of artificial-intelligence-based embryo grading, may improve the ability to select the most viable embryo in the clinic.

Two- to 3-month-old virgin female Crl:CD1 (ICR) mice were purchased from Charles River Laboratories and mated with CD1 male mice. All animals were housed in individually ventilated cages, under controlled light (12 h light, 12 h dark), humidity (40-60%) and temperature (22±2°C), and food and water were available ad libitum. Two-cell embryos were obtained from superovulated mice and were collected in M2 media (Sigma-Aldrich, M7167) and cultured in KSOM media drops (EmbryoMax KSOM; Merck Millipore, MR-020P-5F) covered in mineral oil at 37°C in 5% CO₂. Further details of the animal models can be found in Table S1. The study was completed in accordance with the guidelines of the Canadian Council on Animal Care; experimental procedures were approved by the Comité Institutionnel de Protections des Animaux (CIPA, protocol number: IP18034GFs) of the Centre de Recherche du Centre Hospitalier de l'Université de Montréal (CRCHUM).

For treatments, two-cell stage embryos were washed in pre-equilibrated KSOM supplemented with DMSO (vehicle; Sigma-Aldrich), APCin (100 µM; Tocris Bioscience, 5747) and KU55933 (20 µM; Sigma-Aldrich, SML1109). After the incubation period, embryos were washed through nine 20 µl drops of pre-equilibrated KSOM and then cultured. Exposure to monastrol (200 µM; Sigma-Aldrich, 475879) for 2 h was used to induce monopolar spindle formation prior to fixation for individual kinetochore assessment. New batches of all chemicals were tested for efficacy before experimentation. Further details of the reagents used can be found in Table S1.

Two-cell stage embryos were microinjected with mRNAs in M2 media as described previously (Fitzharris, 2009; FitzHarris et al., 2018). Briefly, we used a PicoPump (World Precision Instruments, Sarasota, FL, USA), intracellular electrometer (Harvard Apparatus) and micromanipulators (Narishige, Amityville, NY, USA) mounted on a Leica DM IL inverted microscope. After microinjection, embryos were left for at least 2 h before live-cell imaging. mRNAs were generated using the mMessage Machine kits T3 (Ambion, AM1348) or T7 (Ambion, AM1344) followed by Poly(A)-Tailling kit (Ambion, AM1350) according to the manufacturer's instructions. The following vectors were linearized and used as a template for in vitro cRNA production: histone 2B (H2B) tagged with red fluorescent protein (H2B:RFP) in pRN4 (gift from Alex McDougall, Observatoire Océanologique de Villefranche-sur-Mer, Villefranche-sur-Mer, France), H2B tagged with enhanced green fluorescent protein (H2B:EGFP) in pRN4, cyclinB1:GFP in pCMX (Addgene 26061) and PCNA:EGFP in pcDNA3.1+poly(A) (gift from Kazuo Yamagata, BOST, Kindai, Japan). Further details of the vectors used can be found in Table S1.

Live-cell brightfield imaging was performed using the Zeiss Axio Observer microscope, equipped with an Axiocam and a 20×/0.8 numerical aperture air objective using brightfield imaging. Embryos were placed in a heated stage top incubator with 5% CO₂ supply at 37°C and images were acquired using the ZEN Blue software. Time-lapse z-stacks of embryos (∼25) were obtained at 10 min intervals for a maximum of 70 h before fixation. Live- and fixed-cell fluorescence imaging was performed using a Leica SP8 confocal microscope fitted with a HyD detector and 20×/0.75 or 63×/1.4 numerical aperture oil objective using a 1.5 μm optical section. Fixed z-stacks with a step size of 2 μm were imaged using LAS-X software. All embryos were placed on a glass-bottomed dish in 2 μl drops of 1% PBS containing bovine serum albumin (BSA) or KSOM media, for fixed or live imaging, respectively, under mineral oil.

For immunofluorescence, embryos were fixed using 2% paraformaldehyde (PFA) in PBS for 20 min, permeabilized in 0.25% Triton X-100 in PBS for 10 min and blocked in 3% BSA in PBS overnight at 4°C. For the CENP-A antibody, a 2 h lambda phosphatase (New England Biolabs, P0753S) pre-treatment at 30°C was performed before staining (Ma and Schultz, 2013). The following primary antibodies were used: human anti-CREST (a gift from Marvin J. Fritzler, Cumming School of Medicine, Calgary, Canada; 1:100), mouse anti-β-tubulin (Sigma-Aldrich, T4026; 1:1000), rabbit anti-CENP-A (New England Biolabs, 2048S; 1:200), rabbit anti-γH2AX (Trevigen, 4418-APC-020; 1:500), mouse anti-OCT4 (Santa Cruz Biotechnology, sc-5279; 1:200). Alexa-labelled secondary antibodies (1:1000) were purchased from Thermo Fisher Scientific. The Alexa Fluor 555 phalloidin-conjugated antibody (Invitrogen, A34055; 1:200) and Hoechst 33342 (Invitrogen, H1399; 1:1000) were used to visualize F-actin and DNA, respectively. Further details of the antibodies can be found in Table S1.

Image analyses were performed using ImageJ/Fiji (Open Source) or Imaris (version 9.6.2, Bitplane) software. Statistical analyses were completed using the Prism software (version 9.0.2, GraphPad Software). Further details of the software used can be found in Table S1. Data were analyzed using one-way ANOVA followed by a post hoc Dunnett's multiple comparison test or using a Kruskal–Wallis non-parametric test followed by a post hoc Dunnett's multiple comparison test. Shapiro–Wilk normality tests were applied where appropriate. The level of significance for all statistical tests was set to P<0.05. All error bars are presented as standard error of the mean (s.e.m.). Similar letters in graphs indicate statistical similarity.

We thank Gaudeline Rémillard-Labrosse for laboratory support and Aurélie Cleret-Buhot for image analysis assistance and access to Imaris software. We also thank Jean-Claude Labbé and Gilles Hickson for critical reading of the manuscript.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200391.