Molecular contribution to embryonic aneuploidy and karyotypic complexity in initial cleavage divisions of mammalian development

Multiple studies across mammalian species, including humans, have established that in vitro-derived embryos suffer from remarkably frequent whole-chromosomal losses and gains, or aneuploidy (Vanneste et al., 2009; Daughtry et al., 2019). Depending on the type and severity of the segregation error, many aneuploid embryos will undergo developmental arrest and/or result in early pregnancy loss if transferred. Estimates of embryonic aneuploidy in vivo are difficult to ascertain, but 50-70% of spontaneous miscarriages following natural conception in women are diagnosed as karyotypically abnormal (Hassold et al., 1980; Schaeffer et al., 2004). Aneuploidy can arise either meiotically during gametogenesis or post-zygotically from the mitotic cleavage divisions of preimplantation development. Although significant effort has been put forth to identify specific contributors to meiotic chromosome mis-segregation, particularly with advanced maternal age (Webster and Schuh, 2017; Schneider and Ellenberg, 2019), much less is known about the molecular mechanisms underlying mitotic aneuploidy generation. This is in spite of findings that mitotic errors are equally or more prevalent than meiotic errors, and arise independently of maternal age or fertility status (Vanneste et al., 2009; Chavez et al., 2012; McCoy et al., 2015a,b). As the first three mitotic divisions are the most error-prone and do not necessarily lead to embryo arrest at the 4- to 8-cell stage, which is when zygotic genome activation (ZGA) largely occurs in most mammals (Braude et al., 1988; Plante et al., 1994), this suggests that maternally inherited factors regulating mitotic chromosome segregation may be lacking or compromised to prevent chromosomally abnormal embryos from continuing in development (Mantikou et al., 2012; Schneider and Ellenberg, 2019; Tšuiko et al., 2019).

Our understanding of the mechanisms mediating mitotic chromosome mis-segregation primarily derives from cancer cells and tumorigenesis, and includes loss or prolonged chromosome cohesion, defective spindle attachments, abnormal centrosome number and relaxed cell cycle checkpoints (Ganem et al., 2009; Soto et al., 2019). Regardless of the mechanism, chromosomes that are mis-segregated during mitosis will become encapsulated into micronuclei and can contribute to aneuploidy in subsequent divisions. However, the cleavage divisions of preimplantation embryo development are fundamentally different from the mitoses experienced by dividing somatic cells. Besides reliance on maternally inherited signaling factors for cell division (Mantikou et al., 2012; Schneider and Ellenberg, 2019; Tšuiko et al., 2019), cleavage-stage embryos produce an increasing number of progressively smaller cells without changing the overall size of the embryo. Moreover, unlike tumors and cancer cells, which tend to express cell cycle checkpoints at high levels (Schvartzman et al., 2010), cleavage-stage human embryos have been shown to underexpress checkpoints and overexpress cell cycle drivers, at least at the RNA level (Kiessling et al., 2010). By monitoring bipolar attachment of spindle microtubules to kinetochores during mitosis, the mitotic checkpoint complex (MCC) prevents activation of the anaphase-promoting complex/cyclosome (APC/C) and delays mitotic progression in the absence of stable bipolar kinetochore-microtubule attachments. This delay, however, is only temporary and cells with an unsatisfied checkpoint will eventually arrest or exit mitosis prematurely. The core components of the MCC are evolutionarily conserved and include CDC20, as well as the serine/threonine kinases, BUB1B, BUB3 and MAD2. Also known as BUBR1, BUB1B is the largest of the MCC proteins and is normally present throughout the cell cycle (Elowe et al., 2010). Without BUB1B/BUBR1, the MCC no longer localizes to unattached kinetochores to prevent incorrect or deficient spindle attachments, resulting in the generation of aneuploid daughter cells (Lampson and Kapoor, 2005). Whether the MCC is functional in the initial mitotic divisions of mammalian preimplantation development is currently unclear and has only been studied in mouse embryos (Wei et al., 2011; Vazquez-Diez et al., 2019), which normally exhibit a low incidence of micronucleation and aneuploidy unless treated with chemicals to inhibit MCC function (Wei et al., 2011; Bolton et al., 2016; Vazquez-Diez et al., 2019; Singla et al., 2020).

Cattle share several key characteristics of preimplantation development with humans, including the timing of the first mitotic divisions, the stage at which the major wave of ZGA occurs and the approximate percentage of embryos that typically reach the blastocyst stage (Alper et al., 2001; Wong et al., 2010; Sugimura et al., 2012). Furthermore, single-nucleotide polymorphism (SNP) genotyping and next generation sequencing (NGS) revealed that the frequency of aneuploidy in cattle is likely similar to humans (Destouni et al., 2016; Hornak et al., 2016; Tšuiko et al., 2017). Destouni et al. also demonstrated that bovine zygotes can segregate parental genomes into different blastomeres during the first cleavage division, but the mechanism by which this occurs has yet to be determined (Destouni et al., 2016). Thus, with the ethical and technical limitations of human embryo research, bovine embryos represent a suitable model for studying the dynamics of micronuclei formation and aneuploidy generation during preimplantation development. In this study, we have used a combination of time-lapse and live-cell fluorescent imaging with single-cell DNA-seq (scDNA-seq) for copy number variation (CNV) analysis to assess mitotic divisions in bovine embryos from the zygote to 12-cell stage and to visualize chromosome segregation in real-time. We also evaluated the lack of MCC function in bovine zygotes to determine whether defective cell cycle checkpoint signaling contributes to aneuploidy during early mammalian embryogenesis.

Because micronuclei formation is an indication of chromosome mis-segregation, we first assessed the prevalence of micronucleation throughout bovine preimplantation development. Although micronuclei-like structures have been detected in bovine embryos previously (Yao et al., 2018), their frequency or whether they were associated with a particular stage of preimplantation development was not determined. To address this, we generated a large number (n=53) of in vitro-produced bovine embryos and fixed them at different stages to evaluate DNA integrity with DAPI and nuclear structure by immunostaining for the nuclear envelope marker LAMIN B1 (LMNB1; Fig. 1A). Immunofluorescent labeling revealed the presence of micronuclei as early as the zygote stage that were distinct from the maternal and paternal pronuclei (Fig. 1B). Several micronuclei, as well as multiple nuclei (multi-nuclei) of similar size, were also detected at the early cleavage stage (Fig. 1C). Overall, 37.7% (n=20/53) of early cleavage-stage embryos exhibited micro-/multi-nucleation in one or more blastomeres, suggesting that their formation before ZGA is conserved between cattle and primates (Chavez et al., 2012; Daughtry et al., 2019). A similar examination of bovine blastocysts also immunostained for the trophoblast marker caudal type homeobox 2 (CDX2) demonstrated that micronuclei often reside in the trophectoderm (TE; Fig. 1D), but can also be contained within the inner cell mass (ICM) of the embryo (Fig. 1E). In order to confirm these findings, we immunostained bovine blastocysts using using an antibody to annexin A2 (ANXA2), a membrane-associated adhesion molecule important for embryo attachment and trophoblast outgrowth (Garrido-Gómez et al., 2012; Wang and Shao, 2020) that is highly expressed in early trophoblasts (Fig. S1A,B), and showed the retention of ANXA2-negative micronuclei in the ICM (Fig. S1C-E). However, it is also possible that these micronuclei were not fully integrated into the ICM and instead were produced from apoptosis, which is known to occur in this compartment of in vitro-derived bovine blastocysts (Gjorret et al., 2003).

To visualize the formation of micro- and multi-nuclei in real-time and determine the fate of these nuclear structures in subsequent divisions, we microinjected bovine zygotes (n=90) with fluorescently labeled modified mRNAs and monitored the first three mitoses by live-cell confocal microscopy (Fig. 1A). Histone H2B and/or LMNB1 were used to visualize DNA and nuclear envelope, respectively, whereas F-actin was able to distinguish blastomeres (Movie 1). Of the microinjected embryos, 18.9% (n=17/90) failed to complete cytokinesis during microscopic evaluation, while 53.3% (n=49/90) exhibited normal bipolar divisions and 27.8% (n=25/90) underwent multipolar divisions from one cell to three cells or more (Fig. 2A). In accordance with our immunostaining findings, 31.1% (n=28/90) of the embryos contained micro- and/or multi-nuclei, and anaphase lagging of chromosomes was detected before their formation in only three of these embryos at the zygote (Fig. 2B) or two-cell (Fig. 2C) stage. Micro- and multi-nucleation was more frequently associated with bipolar divisions (Fig. 2A) and an examination of micronuclei fate demonstrated an equal incidence of unilateral inheritance (Fig. 2D) or re-fusion with the primary nucleus (Fig. 2E), while a smaller percentage appeared to form a chromatin bridge with a neighboring blastomere (Fig. 2F, Fig. S2 and Movie 1). Interestingly, the majority of multipolar embryos (76%; n=19/25) underwent the abnormal division after bypassing the fusion of maternal and paternal pronuclei that is known as syngamy (Fig. 2G), and/or produced daughter cells that did not contain any apparent nuclear structure (Fig. 2H). These results help explain previous findings of blastomeres with uniparental origins and those that completely lacked nuclear DNA when assessed for CNV, respectively (Destouni et al., 2016; Ottolini et al., 2017; Daughtry et al., 2019; Middelkamp et al., 2020).

Given the large aneuploidy range reported (32-85%) and differences in the embryonic stage, proportion of the embryo analyzed or CNV approach used in previous bovine studies (Destouni et al., 2016; Hornak et al., 2016; Tšuiko et al., 2017), we sought to determine the precise frequency of aneuploidy in a large number of bovine embryos (n=38) at the early cleavage stages using high-resolution scDNA-seq (Fig. 1A and Table S1). Each embryo was disassembled into individual blastomeres and all cells were assessed to ensure an accurate representation of the overall embryo, resulting in the analysis of 133 blastomeres from the 2- to 12-cell stage (Fig. 3A). Based on previously described criteria (Daughtry et al., 2019), we classified 25.6% (n=34/133) of the blastomeres as euploid and 35.3% (n=47/133) as aneuploid; 3% (n=4/133) solely contained segmental errors and 17.3% (n=23/133) exhibited chaotic aneuploidy. The remaining cells either failed WGA (10.5%; n=14/133) or were identified as empty due to the detection of only mitochondrial DNA (8.3%; n=11/133). After combining the results from each embryo, we determined that 16% (n=6/38) were entirely euploid, whereas 55% (n=21/38) comprised only aneuploid cells (Fig. 3B). An additional 29% (n=11/38) contained a combination of both euploid and aneuploid blastomeres, known as mosaic. Of the embryos with mosaicism, 18% (n=2/11) had incurred segmental errors only or DNA breaks of 15 Mb in length or larger that did not affect the whole chromosome. Whole-chromosomal losses (58.4%) were more prevalent than whole-chromosomal gains (41.6%), while segmental losses and gains were equally prevalent. The X chromosome was by far the most frequently impacted by whole chromosomal losses and gains, whereas chromosome 5 (human chromosomes 12 and 22), 7 (human chromosomes 5 and 19), 11 (human chromosomes 3 and 9) and 29 (human chromosome 11) were commonly subjected to DNA breakage (Fig. 3C). Although meiotic mis-segregation was identified in 16% (n=6/38) of the embryos (Fig. 3D), mitotic aneuploidy accounted for the majority of errors (66%; n=25/38), with the remaining 18% (n=7/38) exhibiting the karyotypic complexity characteristic of chaotic aneuploidy (Fig. 3E). In addition, most (67%; n=4/6) of the embryos with meiotic errors also experienced mitotic mis-segregation of different chromosomes than those originally affected during meiosis (Fig. 3F). Reciprocal losses and gains, whereby chromosomes lost from one blastomere were found in a sister blastomere, accounted for only 25% (n=7/29) of the mitotic errors (Fig. 3D,F).

Since this chromosome constitution indicated deficient cell cycle checkpoints and there are conflicting reports on whether the MCC is functional at the early cleavage stage in mice (Wei et al., 2011; Vazquez-Diez et al., 2019), our next objective was to determine whether a lack of adequate checkpoints was associated with micro-/multi-nucleation and aneuploidy in bovine embryos (Fig. 1A). Given negligible effects on mouse embryogenesis from knockdown of Mad2, another MCC component (Vazquez-Diez et al., 2019), we focused our attention on BUB1B (also known as BUBR1), the largest of the MCC proteins that is present throughout the cell cycle (Elowe et al., 2010). Two non-overlapping morpholino antisense oligonucleotides (MAOs) were designed to specifically inhibit BUB1B by targeting the ATG translation start site (BUB1B MAO 1) or a sequence upstream within the 5′ UTR (BUB1B MAO 2) and tested before use in embryos (Fig. S3A). Because BUB1B is diffusely expressed in the cytoplasm of cells in interphase and only concentrates at kinetochores in early prophase (Simmons et al., 2019), the visualization of BUB1B expression and localization requires that all cells be undergoing mitosis. However, blastomeres in early cleavage-stage mammalian embryos exhibit asynchronous cell division rather than simultaneously dividing like in other vertebrates (Gilbert, 2000), and the addition of chemicals such as colcemid to arrest cells in metaphase can induce aneuploidy by themselves (Held, 1982; Ligasová and Koberna, 2021). Therefore, we used the Madin-Darby Bovine Kidney (MDBK) epithelial cell line to assess BUB1B expression in non-transfected cells for comparison to those transfected with either a standard control (Std Control) MAO or BUB1B MAO 1 and observed a reduction in BUB1B expression only in the latter by immunofluorescence (Fig. S3B). Quantification of the percentage of MDBK cells with positive BUB1B expression following transfection with different concentrations of each MAO revealed a dose-dependent decrease in BUB1B expression in BUB1B MAO 1 transfected cells (Fig. S3C). This was further assessed by western blot analysis of MDBK cells treated with or without colcemid prior to mitotic shake-off, which takes advantage of cells detaching from the culture substrate and rounding up during mitosis (Ligasová and Koberna, 2021), and showed reduced BUB1B expression following colcemid treatment (Fig. S3D). To determine whether BUB1B knockdown was similarly efficient in embryos, bovine zygotes were microinjected with the Std Control MAO (n=5) or BUB1B MAO 1 (n=5), and BUB1B expression was evaluated by microfluidic quantitative RT-PCR (RT-qPCR). As shown in Fig. S3E, we detected a significant (P=0.007) decrease in BUB1B expression in the BUB1B MAO-injected embryos compared with those injected with the Std Control MAO.

Once we confirmed BUB1B knockdown efficiency, bovine zygotes were microinjected with either Std Control MAO (n=81), BUB1B MAO 1 (n=48) or BUB1B MAO 2 (n=36) and cultured under a time-lapse imaging microscope to monitor development. Each embryo was categorized as having either normal or abnormal divisions for comparison with untreated (non-injected) embryos (n=180). In the BUB1B MAO 1 treatment group, 37.5% (n=18/48) of the zygotes failed to undergo the first cleavage division (Table 1) and a subset (8.3%; n=4/48) of these embryos attempted to divide by forming multiple cleavage furrows (Fig. 4A), but never successfully completed cytokinesis (Movie 2). Of those BUB1B MAO 1 zygotes that divided, more exhibited abnormal cytokinesis (65.4%; n=17/26), including multipolar divisions and/or blastomere asymmetry (Movies 3 and 4, respectively), than normal bipolar divisions (30%; n=9/30). Similar results were obtained following injection with BUB1B MAO 2 (Table 1 and Fig. 4B); despite the phenotypic similarities between the two non-overlapping MAOs, we further assessed BUB1B MAO specificity by conducting embryo rescue experiments with modified BUB1B mRNA that would not be directly targeted by the MAO. BUB1B mRNA with a mutated MAO-binding sequence was microinjected into zygotes, along with BUB1B MAO 1 (n=51), and embryos cultured up to the blastocyst stage (Fig. 4C). Although no embryos formed blastocysts following injection of either the BUB1B MAO 1 or 2, 45% (n=23/51) of the BUB1B MAO 1 and mRNA co-injected embryos underwent cleavage divisions and reached the blastocyst stage (Fig. 4D). This percentage was similar to that obtained from the non-injected embryos and following injection with the Std Control MAO, confirming that the knockdown of BUB1B expression and rescue of BUB1B-induced mitotic defects were specific.

Because none of the BUB1B MAO-injected embryos formed blastocysts and instead arrested at the early cleavage stage, we next examined nuclear structure and CNV in MCC-deficient embryos by immunofluorescence and scDNA-seq, respectively (Fig. 1A). LMNB1 immunostaining revealed both micro- and multi-nuclei in BUB1B MAO 1- and 2-treated embryos that did not attempt division or were unable to complete the first cytokinesis (Fig. 4E, top row). Similar abnormal nuclear structures, as well as empty blastomeres or those with DNA that lacked a nuclear envelope, were also apparent in MCC-deficient embryos that successfully divided (Fig. 4E, bottom row). Disassembly of the embryos into individual cells for assessment of DNA content and CNV analysis demonstrated that, while euploid blastomeres could be obtained following BUB1B MAO injection, MCC deficiency mostly produced blastomeres with chaotic aneuploidy (Fig. 4F). Analogous to some of the non-injected controls (Fig. 3E), a complete loss of certain chromosomes and a gain of up to six copies of other chromosomes were detected, suggesting that the lack of MCC function permits premature mitotic exit, micro-/multi-nucleation and asymmetrical genome distribution upon division.

Based on findings that several BUB1B-deficient embryos did not divide or exhibited abnormal multipolar and/or asymmetric divisions, we sought to determine whether BUB1B knockdown had specific effects on the mitotic machinery or whether it induced a broader dysregulation of mitosis. Therefore, the relative abundance of additional maternal-effect, mitotic, cell cycle, ZGA and cell survival genes was assessed in individual BUB1B MAO 1 versus non-injected and Std Control-injected MAO zygotes (Fig. S3 and Table S2) using microfluidic RT-qPCR. Besides BUB1B (Fig. S3E), the genes encoding several other kinases involved in cytokinesis and chromosome segregation, including aurora kinase B (AURKB), Polo-like kinase 1 (PLK1) and ribosomal protein S6 kinase alpha-5 (RPS6KA5), were all significantly downregulated in BUB1B MAO-injected embryos relative to the controls (Fig. 5A; P≤0.05). Expression of additional genes, including amyloid β precursor protein binding family B member 1 (APBB1), which inhibits cell cycle progression, and those associated with extracellular matrix remodeling [cartilage acidic protein 1 (CRTAC1) and ADAM metallopeptidase with thrombospondin type 1 motif 2 (ADAMTS2)] and the stress response (endoplasmic reticulum lectin 1; ERLEC1) was also significantly decreased in MCC-deficient embryos. In contrast, genes involved in promoting cell cycle progression, such as epithelial cell transforming 2 (ECT2), pogo transposable element derived with ZNF domain (POGZ), centromere protein F (CENPF) and ribosomal protein S6 kinase alpha 4 (RPS6KA4), were significantly upregulated in BUB1B MAO-injected embryos, along with microtubule polymerization (HAUS augmin like complex subunit 6; HAUS6) or orientation (synaptonemal complex protein 3; SYCP3) genes (Fig. 5B; P≤0.05). Because ECT2, POGZ, CENP-F, HAUS6 and SYCP3 have been shown to serve as kinase substrates in dividing somatic cells (Kettenbach et al., 2011; Santamaria et al., 2011; Bibi et al., 2013), this suggests that the inhibition of BUB1B primarily increases the phosphorylation and expression of kinase substrates either directly or indirectly via regulation by other protein kinases, such as AURKB, PLK1 and RPS6KA5.

Aneuploidy is a major cause of embryo arrest, implantation failure and spontaneous miscarriage during mammalian development, and, yet, relatively little is still known about the molecular mechanism(s) mediating mitotic chromosome mis-segregation. Because mitotic errors can be as detrimental as meiotic errors or produce chromosomally mosaic embryos that continue in development and still implant (Magli et al., 2000; Baltaci et al., 2006; Fragouli et al., 2014, 2017; Greco et al., 2015), it is imperative that we identify euploid and mosaic embryos with high implantation potential before they are transferred. Although there are multiple mechanisms that could mediate mitotic chromosome mis-segregation, within the context of preimplantation development, only the MCC has been examined and these studies primarily used mouse embryos treated with chemicals to induce aneuploidy (Wei et al., 2011; Bolton et al., 2016; Vazquez-Diez et al., 2019). In addition, the knockdown of a specific MCC component in mouse zygotes had no effect on the first cleavage divisions when mitotic aneuploidy typically occurs in other mammals (Vazquez-Diez et al., 2019), questioning whether the MCC is functional in early mammalian embryos. Using a combination of live-cell imaging, scDNA-seq, immunofluorescence and genetic manipulation, we visualized mitotic chromosome segregation in real-time from the zygote to the 12-cell stage and assessed the role of the MCC in embryos from an animal model that suffers from a comparable incidence of aneuploidy to that in humans.

Of the cleavage-stage embryos examined by immunostaining or live-cell imaging, over 30% contained micro- or multi-nuclei, and anaphase lagging of chromosomes was detected in a small number of embryos prior to micronuclei formation. When we evaluated other morphological or mitotic characteristics that might indicate how these abnormal nuclear structures formed, we determined that most micronuclei-containing embryos underwent normal bipolar divisions, excluding multipolar cytokinesis as the primary mechanism. However, multipolar divisions were associated with a lack of syngamy and often produced cells that did not contain any apparent nuclear structure (Fig. 6A). Whereas mouse zygotes have been shown to sustain spatial separation of parental genomes by dual-spindle formation (Mayer et al., 2000; Reichmann et al., 2018), embryos from other mammals, including humans, are thought to exhibit syngamy at the zygote stage (Kai et al., 2018; Yao et al., 2018). However, recent studies with both bovine and human embryos showed that the clustering and unification of parental genomes often fails, resulting in micronuclei formation and chromosome mis-segregation (Cavazza et al., 2021), and dual-spindle formation can occur in bovine zygotes despite the presence of paternal centrosomes (Schneider et al., 2021). By avoiding syngamy and then undergoing abnormal cytokinesis, zygotes could differentially segregate entire parental genomes to daughter cells, which has been termed ‘heterogonic division’ (Destouni et al., 2016). Although we were unable to confirm that these embryos underwent heterogonic divisions by SNP analysis due to the lack of maternal DNA input available from all oocyte donors, our findings provide a mechanism for how blastomeres with uniparental origins could arise in both cattle and primates. It may also help explain the potential higher risk of uniparental disomy disorders such as Angelman syndrome and Prader-Willi syndrome in human embryos derived in vitro (Gosden et al., 2003; Hattori et al., 2019). Based on the prevalence of blastomeres with uniparental origins reported in these studies (Destouni et al., 2016; Ottolini et al., 2017; Daughtry et al., 2019; Middelkamp et al., 2020), we suggest that SNP-based strategies should be used to accurately assess embryo ploidy and distinguish the parental source of all DNA, as it may reveal why some seemingly ‘euploid’ embryos fail to implant.

Examination of micronuclei fate in subsequent divisions revealed an equal incidence of unilateral inheritance and fusion back with the primary nucleus, with a smaller percentage of embryos exhibiting a chromatin bridge between blastomeres following micronucleation (Fig. 6B). Because cancer cell micronuclei have been shown to undergo extensive DNA damage upon re-fusion with the primary nucleus (Crasta et al., 2012; Zhang et al., 2015), chromosomal integrity and the effect on developmental outcome will likely differ depending on which of these events occurred. The significance of the chromatin bridging and whether it exacerbates aneuploidy or restores euploidy is currently unknown, but we suspect that this genetic exchange between blastomeres contributes to the high frequency of non-reciprocal mitotic errors observed here and in embryos from other mammals (Vanneste et al., 2009; Chavez et al., 2012; McCoy et al., 2015a; Daughtry et al., 2019). A similar assessment of bovine blastocysts determined that micronuclei can reside in both the placental-derived TE and ICM of the embryo, the latter of which may be more detrimental. It is estimated that 40% of human blastocysts are aneuploid to some extent (Fiorentino et al., 2014; Popovic et al., 2018) and a recent study suggests that 24% of bovine blastocysts are aneuploid, with a high degree of concordance in the incidence of aneuploidy between the TE and ICM lineages (Tutt et al., 2021). However, these percentages are based on the assessment of multiple cells together rather than at the single-cell level, making it difficult to determine the extent of chromosomal mosaicism in an embryo. Using single-cell RNA-seq (scRNA-seq) data to infer chromosome dosage from gene expression differences (Petropoulos et al., 2016; Griffiths et al., 2017), a recent study observed no significant enrichment of aneuploid cells in the TE compared with ICM of human blastocysts, but only 31 ICM cells from 23 embryos was examined (Starostik et al., 2020). Nevertheless, the authors also reported that 31% of the embryos possessed meiotic aneuploidies, while mitotic aneuploidies were detected in 74% of the embryos. This suggests that mitotic errors are more likely to be observed in blastocysts, where they may be more tolerated than meiotic errors, and until aneuploidy is comprehensively examined at the single-cell level in whole blastocysts, the precise distribution of aneuploid cells between the TE and ICM will remain unclear.

Although SNP arrays and NGS have been used previously to assess aneuploidy in cleavage-stage bovine embryos, these studies reported a large aneuploidy range, examined a single stage of development and/or evaluated only part of the embryo (Destouni et al., 2016; Hornak et al., 2016; Tšuiko et al., 2017). After karyotypically reconstructing each embryo and combining the scDNA-seq results, we determined that over half the embryos contained only aneuploid cells. One-third of the remaining embryos were chromosomally mosaic, all of which were primarily the product of non-reciprocal mitotic errors. In those embryos with meiotic errors, most also experienced mitotic mis-segregation of different chromosomes than those originally affected during meiosis, indicating that embryos with meiotic errors are more prone to additional mitotic mis-segregation events that are further propagated during subsequent divisions. The rest of the aneuploid embryos exhibited a compete loss and/or a gain of up to six copies of chromosomes characteristic of chaotic aneuploidy from multipolar divisions (Ottolini et al., 2017; Daughtry et al., 2019). A previous study comparing in vivo ovulated (IVO) versus in vitro matured (IVM) mouse oocytes showed that the aneuploidy frequency in embryos produced with IVO oocytes increased from 3.7% to 9.4% with IVM oocytes (Treff et al., 2016). These percentages were exacerbated by advanced maternal age, increasing from 12.8% in embryos generated with IVO oocytes to 30.8% with IVM oocytes, and suggesting that older oocytes are more susceptible to aneuploidy regardless of whether derived in vivo or in vitro. Because the bovine oocytes used in this study were obtained from young females, we expect that IVM had less impact on the incidence of aneuploidy than if the oocytes were collected from older female donors. Nonetheless, it should be noted that IVM of oocytes can increase aneuploidy by itself, although not likely to the levels observed in this study given the relatively low frequency of meiotic errors.

Because of the discrepancy in whether the MCC is functional in the early cleavage divisions of mouse embryogenesis in previous studies (Wei et al., 2011; Vazquez-Diez et al., 2019), we investigated the consequences of MCC inhibition by targeting BUB1B in bovine zygotes. Following injection, BUB1B MAO embryos either failed to divide even after several attempts or exhibited abnormal divisions that were multipolar and/or asymmetrical (Fig. 6C). Furthermore, immunostaining of the BUB1B MAO-treated embryos that did divide revealed blastomeres with abnormal nuclear structures or completely devoid of nuclear DNA. CNV analysis of blastomeres that contained DNA showed a predominance of chaotic aneuploidy, with a complete loss or excessive number of chromosomal copies as described in some untreated embryos here and previously reported in primate embryos with multipolar divisions (Ottolini et al., 2017; Daughtry et al., 2019). Without BUB1B, we speculate that embryos were unable to obtain proper microtubule-kinetochore attachments prior to the first cytokinesis, resulting in failed MCC and arrest, or underwent premature cell division and chromosome mis-segregation due to MCC dysregulation. Although knockdown of another MCC component, Mad2, in mouse embryos was recently shown to have no effect on blastocyst formation, it did increase the number of micronuclei present at the morula stage (Vazquez-Diez et al., 2019). Both MAD2 and BUB1B bind CDC20 to prevent activation of the APC, but in vitro binding assays demonstrate that BUB1B is 12 times more effective than MAD2 in inhibiting CDC20 (Fang, 2002). In addition, it has been shown in Drosophila that the recruitment of Cdc20 to the kinetochore requires Bub1b and not Mad2 (Li et al., 2010). Based on these findings and the observation that BUB1B differs from the other MCC components in that both underexpression and overexpression result in drastically different phenotypes, including infertility in hypomorphic Bub1b mice (Baker et al., 2004, 2013), we postulate that BUB1B is the foremost checkpoint regulator. Moreover, as BUB1B mRNA is highly expressed in oocytes and has been shown to be a maternal effect gene (Pérez-Mongiovi et al., 2005; Gasca et al., 2007), we suspect that it mainly ensures chromosome fidelity during the error-prone early cleavage divisions, but an embryonic form of BUB1B expressed upon ZGA may also serve a role later in preimplantation development.

Given that many BUB1B-deficient embryos attempted to divide, but did not complete cytokinesis, or exhibited abnormal divisions, we evaluated whether BUB1B abundance affected the expression of other maternal-effect, mitotic, cell cycle, ZGA or cell survival genes. Although some genes involved in extracellular matrix remodeling and the stress response were impacted by BUB1B deficiency, the primary effect of BUB1B knockdown was the downregulation of the protein kinases AURKB, PLK1 and RPS6KA5, and the upregulation of several kinase substrates. Another protein kinase, RPS6KA4, that we have shown is associated with binucleation, mitotic arrest and blastomere lysis in mouse embryos, as well as aneuploidy in human embryos, was significantly upregulated in BUB1B-deficient embryos (Chavez et al., 2014). In somatic cells, BUB1B requires PLK1 to function (Elowe et al., 2007) and together they inhibit kinase activity at kinetochores through positive feedback (Suijkerbuijk et al., 2012). Thus, we suspect that the inhibition of BUB1B increased the phosphorylation and expression of kinase substrates either directly or indirectly through PLK1, or possibly via another polo-like kinase family member. Although we did not evaluate its expression in this study, maternally inherited genotypic variants spanning PLK4 have also been reported to play a role in tripolar divisions and aneuploidy in human embryos (McCoy et al., 2015a, 2018). Whether PLK4 similarly regulates BUB1B function is currently unknown, but BUB1B likely cooperates with a large regulatory network of kinases and their substrates to reinforce MCC function and ensure chromosome fidelity until ZGA. Collectively, our findings confirm that the MCC does indeed maintain proper chromosome segregation in the initial cleavage divisions of mammalian preimplantation development and show that deficiency in BUB1B and/or other maternally derived factors likely contributes to the karyotypic complexity observed in early embryos.

Using a combination of live-cell imaging, scDNA-seq for CNV analysis and genetic manipulation of embryos, we developed an experimental approach to assess mitotic divisions and chromosome segregation throughout bovine preimplantation development (Fig. 1A). First, we fertilized mature oocytes, cultured resultant zygotes under a time-lapse imaging microscope to monitor embryo developmental dynamics, and evaluated DNA integrity and nuclear structure by immunofluorescence up to blastocyst stage (n=53). We confirmed our findings by live-cell confocal microscopy of zygotes microinjected with fluorescently labeled modified mRNAs and visualization of the initial mitotic divisions in real time (n=90). Cleavage-stage embryos between 2 and 12 cells were then disassembled into single blastomeres for comprehensive assessment of meiotic and/or mitotic errors (n=38). Finally, the role of the MCC in aneuploidy generation was determined by microinjecting zygotes with BUB1B MAOs (n=84) or a Std Control MAO (n=81) for comparison with non-injected embryos (n=180) and with embryos co-injected with BUB1B MAO and BUB1B modified mRNA (n=85) using time-lapse monitoring, immunostaining, CNV analysis and/or microfluidic quantitative RT-PCR.

All chemicals were obtained from Sigma-Aldrich or ThermoFisher Scientific unless otherwise stated. Tyrode's albumin lactate pyruvate (TALP) medium with Hepes (TALP-Hepes) was used as washing media and contained 114 mM NaCl, 3.2 mM KCl, 25 mM NaHCO₃, 0.34 mM NaH₂PO₄-H₂O, 10 mM C₃H₅NaO₃, 2 mM CaCl₂-H₂O, 0.5 mM MgCl₂-6H₂O, 10.9 mM Hepes, 0.25 mM sodium pyruvate, 1 μl/ml Phenol Red, 3 mg/ml FAF-BSA and 100 μM gentamicin sulfate. For fertilization, TALP-IVF was used (114 mM NaCl, 3.2 mM KCl, 25 mM NAHCO₃, 0.34 mM NaH₂PO₄-H₂O, 10 mM C₃H₅NaO₃, 2 mM CaCl₂-H₂O, 0.5 mM MgCl₂-6H₂O, 1 μl/ml Phenol Red, 0.25 mM sodium pyruvate, 100 units/ml penicillin, 100 μg/ml streptomycin, 1 μM epinephrine, 0.02 mM penicillamine, 10 μM hypotaurine, 6 mg/ml FAF-BSA and 10 mg/ml heparin).

Cumulus-oocyte complexes (COCs) were retrieved by follicular aspiration of ovaries collected at a commercial abattoir (DeSoto Biosciences). Those COCs with at least three layers of compact cumulus cells and homogeneous cytoplasm were placed in groups of 50 in 2 ml sterile glass vials containing 1 ml of oocyte maturation medium, covered with mineral oil and equilibrated in 5% CO₂. Tubes with COCs were shipped overnight in a portable incubator (Minitube) at 38.5°C. Following 24 h of maturation, COCs were washed three times in TALP-Hepes followed by a final wash in fertilization media, before placement in a four-well dish (Nunc; ThermoFisher Scientific) containing 0.5 ml of fertilization media. Semen from either Racer (014HO07296, Accelerated Genetics) or Colt P-red (7HO10904, Select Sires) was obtained for IVF. Sperm were purified from frozen-thawed straws using a gradient [50% (v/v) and 90% (v/v)] of Isolate (Irvine Scientific), washed twice in fertilization media by centrifugation at 100 g, and diluted to a final concentration of 1 million/ml in the fertilization dish. Fertilization was allowed to commence for 17-19 h at 38.5°C in a humidified atmosphere of 5% CO₂. Zygotes were denuded from the surrounding cumulus cells by vortexing for 4 min in 200 μl of TALP-Hepes with 0.5% (w/v) hyaluronidase (Sigma-Aldrich) and washed in fresh TALP-Hepes.

Denuded zygotes were transferred to custom Eeva 12-well polystyrene dishes (Progyny, formerly Auxogyn) containing 100 µl drops of BO-IVC culture media (IVF Bioscience) under mineral oil (CooperSurgical) and cultured at 38.5°C in a humidified atmosphere of 5% CO₂, 5% O₂ and 90% N₂. Embryos were monitored with an Eeva darkfield 2.2.1 or bimodal (darkfield/brightfield) 2.3.5 time-lapse microscope system (Progyny) housed in a small tri-gas incubator (Panasonic Healthcare) as previously described (Vera-Rodriguez et al., 2015). Images were taken every 5 min with a 0.6 s exposure time. Each image was time stamped with a frame number and all images compiled into an AVI movie using FIJI software version 2.0.0 (Schindelin et al., 2012) for assessment of mitotic divisions by two independent reviewers.

Embryos were washed in PBS with 0.1% BSA and 0.1% Tween-20 (PBST; Calbiochem), and fixed with 4% paraformaldehyde (Alfa Aesar) in PBST for 20 min at room temperature. Once fixed, the embryos were washed with gentle shaking three times for a total of 15 min in PBS-T to remove residual fixative. Embryos were permeabilized in 1% Triton-X (Calbiochem) for 1 h at room temperature and washed in PBST as described above. To block non-specific antibody binding, embryos were transferred to a 7% donkey serum (Jackson ImmunoResearch Laboratories)/PBS-T solution for either 1 h at room temperature or overnight at 4°C. An antibody against LMNB1 (ab16048, Abcam) was diluted 1:1000, while the CDX2 mouse monoclonal antibody (clone CDX2-88, Abcam) and ANXA2 rabbit monoclonal antibody (8235, Cell Signaling Technology) were diluted 1:100 in PBS-T with 1% donkey serum, and embryos stained for 1 h at room temperature or overnight at 4°C. The specificity of the ANXA2 antibody was tested by co-staining a highly-pure day 28 immortalized rhesus placental (iRP) first trimester trophoblast cell line, iRP-D28A (Rosenkrantz et al., 2021), using a mouse monoclonal antibody (clone OV-TL 12/30, Agilent Dako) to the pan-trophoblast marker cytokeratin 7 (KRT7). Primary immunosignals were detected using 488-conjugated donkey anti-rabbit (for LMNB1) or 647-conjugated donkey anti-mouse (for CDX2) Alexa Fluor secondary antibodies (Thermo Fisher Scientific) at a 1:250 dilution with 1% donkey serum in PBS-T at room temperature for 1 h in the dark. Embryos were washed in PBS-T and the DNA stained with either 1 μg/ml DAPI or Hoechst 34580 (Thermo Fisher) for 15 min. Embryos were mounted on slides using Prolong Diamond mounting medium (Invitrogen). Immunofluorescence was initially visualized on a Nikon Eclipse Ti-U fluorescent microscope system and images captured using a Nikon DS-Ri2 color camera and confirmed with a Leica SP5 AOBS spectral confocal system. Z-stacks, 1-5 μm apart, were imaged one fluorophore at a time to avoid spectral overlap between channels. Stacked images and individual channels for each color were combined into composite images using FIJI software version 2.0.0.

Plasmids containing the coding sequence (CDS) for mCitrine-Lifeact (Addgene 54733), which labels filamentous actin (F-actin), mCherry-Histone H2B-C-10 (Addgene 55057) and mCherry-LAMINB1-10 (Addgene 55069) were generated in Dr Michael Davidson's laboratory (National High Magnetic Field Laboratory and Department of Biological Science, Florida State University, USA). Custom primers containing a 5′-T7 promoter sequence were used to amplify each fluorescent tag-mRNA fusion construct as follows: T7_mCitrine_F, CTAGCTTAATACGACTCACTATAGGGCGGTCGCCACCATGGTGA; LifeAct_R, TTACTTGTACAGCTCGTCCATGCCGAGAGTGATCCCGGC; T7_mCherry_F, AATTAATACGACTCACTATAGGGAGAGCCACCATGGTGAGCAA; H2B_R, GCGGCCGCTTTACTTGT; and LAMINB1_R, TCCGGTGGATCCCTACATAA.

PCR amplification was performed with high-fidelity Platinum Taq polymerase (Thermo Fisher) under the following conditions: 94°C for 2 min, followed by 35 cycles of 94°C for 30 s, 70°C for 30 s and 72°C for 3 min. PCR products were purified with the QIAquick PCR Purification kit (Qiagen), then in vitro transcribed using the mMessage Machine T7 Transcription Kit (Invitrogen). Following the synthesis of capped mRNA, the MEGAclear transcription clean up kit (Invitrogen) was used to purify and concentrate the final modified mRNA product.

Bovine zygotes were microinjected with mCitrine-Lifeact and either mCherry-H2B or mCherry-LAMINB1 mRNAs at a concentration of 20 ng/μl each in the presence of Alexa Fluor 488 labeled Dextran (Invitrogen) using a CellTram vario, electronic microinjector and Transferman NK 2 Micromanipulators (Eppendorf). Zygotes that exhibited mCherry fluorescent signal within 4-6 h of microinjection were selected for overnight imaging. Imaging dishes were prepared by placing 20 μl drops of BO-IVC media on glass-bottomed dishes (Matek) and covering with mineral oil. A Zeiss LSM 880 laser-scanning confocal microscope with 10× objective and Fast Airy capabilities was used to capture fluorescent images of embryos for 18-20 h, which encompassed the first three mitotic divisions. Z-stack images were taken every 1.5 µm for a total of 60 slices covering a 90 µm range at 10 min intervals. Each fluorophore was acquired independently to prevent crosstalk and maximize scanning speed. Individual images underwent Airyscan processing using Zeiss software and were compiled into videos with individual embryo labels using FIJI. Assessment of cytoplasmic and nuclear structure in embryos during mitotic divisions was completed by two independent reviewers.

Embryos were disassembled under a stereomicroscope equipped with a heated stage and digital camera (Leica Microsystems) for documentation. The zona pellucida (ZP) was removed from each embryo by a 30 s exposure to warm Acidified Tyrode's Solution (EMD Millipore), followed by 30-60 s in 0.1% (w/v) pronase (Sigma-Aldrich). Once the ZP had been removed, embryos were washed in TALP-Hepes and gently manipulated using a STRIPPER pipettor (Origio), with or without brief exposure to warm 0.05% trypsin-EDTA (Thermo Fisher) as necessary, until all blastomeres were separated. Following disassembly, each blastomere and was washed three times with Ca²⁺ and Mg²⁺-free PBS (Fisher Scientific), collected into individual PCR tubes in 2 μl of PBS and snap frozen on dry ice. Downstream analysis was completed only for embryos where the disassembly process was successful for all blastomeres.

Single bovine blastomeres and unaffected male (GM06034; 60XY) and female (GM06035; 60XX) skin fibroblasts from the Coriell Institute (Camden, NJ, USA) underwent DNA extraction and WGA using the PicoPLEX single-cell WGA Kit (Rubicon Genomics) according to the manufacturer's instructions with slight modifications. Cells were lysed at 75°C for 10 min followed by pre-amplification at 95°C for 2 min and 12 cycles of gradient PCR with PicoPLEX pre-amp enzyme and primer mix. Pre-amplified DNA was further amplified with PicoPLEX amplification enzyme and 48 uniquely indexed Illumina sequencing adapters provided by the kit or custom adapters with indices designed as previously described (Vitak et al., 2017; Daughtry et al., 2019). Adapter PCR amplification consisted of a 95°C hot start for 4 min, four cycles of 95°C for 20 s, 63°C for 25 s and 72°C for 40 s, and seven cycles of 95°C for 20 s and 72°C for 55 s. Libraries were quantified with a Qubit High Sensitivity (HS) DNA assay (Life Technologies). Amplified DNA from each blastomere and fibroblast (50 ng) was pooled and purified with AMPure XP beads (Beckman Coulter). Final library quality assessment was performed on a 2200 TapeStation (Agilent).

Pooled libraries were sequenced on an Illumina NextSeq 500 using a 75-cycle kit with a modified single-end workflow that incorporated 14 dark cycles at the start of the first read prior to the imaged cycles. This step excluded the quasi-random priming sequences that are G-rich and lack a fluorophore for the two-color chemistry used by the NextSeq platform during cluster assignment. A total of 3.5×10⁶ reads/sample were generated. All raw sample reads were demultiplexed and sequencing quality assessed with FastQC (Krueger et al., 2011). Illumina adapters were removed from raw reads with the sequence grooming tool, Cutadapt (Chen et al., 2014), which trimmed 15 bases on the 5′ end and five bases from the 3′ end, resulting in reads of 120 bp on average. Trimmed reads were aligned to the most recent bovine reference genome, BosTau8 (Zimin et al., 2009), using the BWA-MEM option of the Burrows-Wheeler Alignment Tool with default alignment parameters (Salavert Torres et al., 2012). Resulting bam files were filtered to remove alignments with quality scores below 30 (Q<30) as well as alignment duplicates that were likely the result of PCR artifacts with the Samtools suite (Ramirez-Gonzalez et al., 2012). The average number of filtered and uniquely mapped sequencing reads in individual libraries was between 1.9 and 2.2 million.

CNV was determined by comparing a sample (blastomeres) to a known euploid control (fibroblasts) using the integration of two previously developed bioinformatics pipelines (Vitak et al., 2017): variable non-overlapping window circular binary segmentation (VNOWC) and the circular binary segmentation/hidden Markov model (CBS/HMM) intersect termed CHI, which we validated using cells with known aneuploidies (Daughtry et al., 2019). All CNV calls from the two pipelines generated profiles of variable-sized windows that were intersected on a window-by-window basis. Because other low-input sequencing studies have shown that CNV can be reliably assessed at a 15 Mb resolution with 0.5-1× genome coverage (Lee et al., 2013; Zhou et al., 2018), we classified breaks of 15 Mb in length or larger that did not affect the whole chromosome as segmental. Only whole and segmental CNV calls in agreement between the VNOWC and CHI methods at window sizes containing 4000 reads were considered. Chaotic aneuploidy was classified by the loss or gain of greater than four whole and/or broken chromosomes, as previously described (Daughtry et al., 2019). Additional classification of each aneuploidy as meiotic or mitotic in origin was accomplished by determining whether a loss or gain of the same chromosome was detected in all blastomeres (meiotic) or whether reciprocal chromosome losses and gains were observed between blastomeres (mitotic). Non-reciprocal chromosome losses and gains were also classified as mitotic errors and embryos containing blastomeres with both reciprocal and non-reciprocal mitotic errors of the same chromosome were only counted once.

Two non-overlapping MAOs were designed and synthesized by Gene Tools to specifically target bovine BUB1B (Ensembl transcript ID: ENSBTAT00000009521.5). BUB1B MAO 1 (TTTCCTTCTGCATCGCCGCCATC) specifically targeted the ATG start codon of the BUB1B mRNA-coding sequence, while BUB1B MAO 2 (CGATCTGAGGCTCTGAAGAAAGGCC) targeted upstream of MAO 1 in the 5′ UTR of bovine BUB1B. A Std Control MO (CCTCTTACCTCAGTTACAATTTATA) that targets a splice site mutant of the human hemoglobin β-chain (HBB) gene (GenBank accession number AY605051) that is not present in the Bos taurus genome served as a control. Both BUB1B and Std Control MAO were synthesized with a 3′-carboxyfluorescein tag to aid in visualization during cell transfection and embryo manipulation.

Before use in embryos, the BUB1B MAOs were first tested using the MDBK epithelial cell line (Madin and Darby, 1958). MDBK cells were plated on poly-L-lysine treated coverslips and grown to 70% confluency prior to MAO treatment. The cells were incubated with 6 μl/ml Endo-Porter delivery reagent containing DMSO (Gene Tools) and 2, 4 or 8 μM of either BUB1B MAO 1 or Std control MAO and cultured in Eagle's Minimum Essential Medium modified to contain Earle's Balanced Salt Solution, non-essential amino acids, 2 mM L-glutamine, 1 mM sodium pyruvate, 1500 mg/l sodium bicarbonate, 10% (v/v) FBS and antibiotics (50 U penicillin and 50 μg streptomycin) in 5% CO₂ at 37°C. After 36 h, cells were synchronized at metaphase in the presence of 0.03 μg of colcemid (Sigma-Aldrich) for 12 h, and collected for staining at 48 h post MAO treatment. Cells were washed in PBS, followed by a single 20 min fixation using 4% paraformaldehyde (Alfa Aesar) and a permeabilization step with 1% Triton-X (Calbiochem) in PBS. Additional PBS washes were completed prior to blocking with 7% donkey serum (Jackson ImmunoResearch Laboratories) in PBS for either 1 h at room temperature or overnight at 4°C. A primary antibody against BUB1B (ab28193, Abcam) was diluted 1:1000 in PBS with 1% donkey serum, and cells were incubated overnight at 4°C. BUB1B antibody binding was detected using a 568-conjugated donkey anti-rabbit Alexa Fluor secondary antibody (Thermo Fisher) at a 1:250 dilution with 1% donkey serum in PBS at room temperature for 1 h in the dark. Cells were washed in PBS and the DNA stained with 1 μg/ml DAPI for 15 min. The coverslips with adherent cells were then mounted on slides using Prolong Diamond mounting medium (Invitrogen). Immunofluorescence was visualized on a Nikon Eclipse Ti-U fluorescent microscope system and representative fluorescent images captured with a Nikon DS-Ri2 color camera. Using FIJI, background fluorescence was subtracted from the red (BUB1B) channel, followed by individual channels of each color for combination into a composite image. BUB1B immunostaining was visually assessed at each MAO concentration for 100 metaphase cells per treatment group.

To confirm sufficient BUB1B knockdown, MDBK cells were treated with or without colcemid, as described above, before mitotic shake-off (Ligasová and Koberna, 2021) and lysed in RIPA Buffer (Sigma-Aldrich) with a protease inhibitor cocktail (Thermo Fisher). Protein concentrations were calculated by BCA assay (Thermo Fisher) and a total of 20 μg of total cellular protein was loaded per lane. Proteins were separated by SDS-PAGE using 12% polyacrylamide gels (Sigma-Aldrich) and transferred to Immun-Blot PVDF membranes (Bio-Rad). Nonspecific binding was inhibited by blocking the membranes in 5% BSA (Sigma-Aldrich) before immunoblotting. Membranes were incubated overnight at 4°C with the BUB1B primary antibody described above at a 1:1000 dilution, followed by a 1:5000 dilution of a 488-conjugated donkey anti-rabbit Alexa Fluor secondary antibody (Thermo Fisher) for 1 h at room temperature. Incubation with an α-tubulin primary antibody (clone TU-01, Thermo Fisher) diluted 1:1000, followed by a 1:3000 dilution of a 647-conjugated donkey anti-mouse Alexa Fluor secondary antibody (Thermo Fisher) served as a loading control. After each step, the membranes were washed three times with TBST for 10 min. Finally, the blots were visualized using a FluoroChem M fluorescence and chemiluminescence imaging system (Protein Simple), and the images processed in FIJI.

Zygotes underwent cytoplasmic injection with 3′-carboxyfluorescein-labeled MAO at 20 h post fertilization as described above. A concentration of 0.3 mM MAO was used based on previous findings that Std Control MAO at this concentration was the maximum that allowed normal blastocyst formation rates. Following microinjection, embryos were cultured up to the blastocyst stage as described above with or without imaging on the Eeva darkfield 2.2.1 microscope system. Upon developmental arrest, embryos were collected for immunostaining, gene expression analysis or for disassembly into single cells (as described above) for downstream analysis. To further validate MAO specificity, bovine embryos were co-injected with BUB1B-modified mRNA at a concentration of ∼3 nl (75 pg) of mRNA per embryo in addition to BUB1B MAO 1. The BUB1B-coding sequence (CDS) was amplified from the plasmid, pcDNA5-EGFP-AID-BubR1 (Addgene 47330), followed by mutation of the MAO binding site using the Q5 site-directed mutagenesis kit (NEB) according to the manufacturer's instructions. Briefly, custom primers (forward, 5′-aaaaaagagggaGGTGCTCTGAGTGAAGCC-3′; and reverse, 5′-aactgcagccatATGGGATCCAGCTCTGCT-3′) were designed to mutate the region of the BUB1B CDS targeted by the MAO without affecting the amino acid sequence. The uppercase letters indicate the targeting sequence; the lowercase letters are the ‘backbone sequence’ containing 5′ and 3′ UTRs. Exponential amplification of the template plasmid using high-fidelity DNA polymerase was followed by a single step phosphorylation, ligation and DpnI restriction enzyme digestion. NEB 5-α competent cells were transformed with the mutated plasmid, followed by DNA miniprep isolation using QIAprep spin columns (Qiagen). Mutated plasmids were identified by Sanger sequencing performed by the ONPRC Molecular and Cellular Biology Core using a custom designed primer (TTGGTGAATAGCTGGGACTATG). Following identification and isolation, the mutated plasmid served as a template to synthesize a PCR product containing a T7 promoter using Platinum Taq (Invitrogen). Custom primers (forward, CTAGCTTAATACGACTCACTATAGGGAGCGCCACCATGGCTGCAGTTAAAAAAGAG; reverse, CAATCTGTGAGACTTGATTGCCTAGCTCACTGAAAGAGCAAAGCCCCAG) were designed for use with the T7 mMessage mMachine Ultra Kit as described above.

Gene expression was analyzed in non-injected, Std control MAO- or BUB1B MAO-injected embryos using the BioMark Dynamic Array microfluidic system (Fluidigm). All embryos were collected within 36 h of fertilization as described above. Individual embryos were pre-amplified according to the manufacturer's ‘two-step single cell gene expression’ protocol (Fluidigm) using a SuperScript VILO cDNA synthesis kit (Invitrogen), TaqMan PreAmp Master Mix (Applied Biosystems) and gene-specific primers designed to span exons using Primer-BLAST (NCBI). Bovine fibroblasts and no reverse transcription template samples were used as controls. Pre-amplified cDNA was loaded into the sample inlets of a 96×96 dynamic array (DA; Fluidigm) and assayed in triplicate. Genes that exhibited no or very little expression, despite the design and testing of multiple primer sets, were removed from further analysis. A total of 10 reference genes were assayed for use as relative expression controls. Cycle threshold (Ct) values were normalized to the two most stable housekeeping genes (RPL15 and GUSB) using qBase⁺ 3.2 software (Biogazelle). The mean was determined from the calculated normalized relative quantity (CNRQ) values across triplicates±s.e.m. and plotted using Morpheus (https://software.broadinstitute.org/morpheus/).

To determine statistical differences between MAO concentrations in MDBK cells, log-binomial modeling using the Generalized Estimating Equations approach was performed and Tukey's test adjusted P-values reported to adjust for multiple comparisons. For the RT-qPCR results, the Mann–Whitney U-test was used for nonparametric multiple comparisons with Bonferroni correction. An unadjusted P≤0.05 was considered statistically significant.

We gratefully acknowledge Dr Tom Spencer at the University of Missouri-Columbia for the Madin-Darby Bovine Kidney (MDBK) epithelial cells. Drs Crystal Chaw and Stefanie Kaech Petrie in the OHSU Advanced Light Microscopy (ALM) Core were instrumental in developing our imaging protocol, and we very much appreciate all their help and support. We also thank Drs Andrew Adey and Andrew Fields for their assistance and use of the NextSeq500 sequencer. We acknowledge the Oregon National Primate Research Center (ONPRC) Integrated Pathology Core for confocal microscopy (supported by S10RR024585) that operates under the auspices of the ONPRC NIH/OD core grant (P51OD011092).