Human female meiosis revised: new insights into the mechanisms of chromosome segregation and aneuploidies from advanced genomics and time-lapse imaging

ABSTRACT

BACKGROUND

The unbalanced transmission of chromosomes in human gametes and early preimplantation embryos causes aneuploidy, which is a major cause of infertility and pregnancy failure. A baseline of 20% of human oocytes are estimated to be aneuploid and this increases exponentially from 30 to 35 years, reaching on average 80% by 42 years. As a result, reproductive senescence in human females is predominantly determined by the accelerated decline in genetic quality of oocytes from 30 years of age.

OBJECTIVE AND RATIONALE

Understanding mechanisms of chromosome segregation and aneuploidies in the female germline is a crucial step towards the development of new diagnostic approaches and, possibly, for the development of therapeutic targets and molecules. Here, we have reviewed emerging mechanisms that may drive human aneuploidy, in particular the maternal age effect.

SEARCH METHODS

We conducted a systematic search in PubMed Central of the primary literature from 1990 through 2016 following the PRISMA guidelines, using MeSH terms related to human aneuploidy. For model organism research, we conducted a literature review based on references in human oocytes manuscripts and general reviews related to chromosome segregation in meiosis and mitosis.

OUTCOMES

Advances in genomic and imaging technologies are allowing unprecedented insight into chromosome segregation in human oocytes. This includes the identification of a novel chromosome segregation error, termed reverse segregation, as well as sister kinetochore configurations that were not predicted based on murine models.

WIDER IMPLICATIONS

Elucidation of mechanisms that result in errors in chromosome segregation in meiosis may lead to therapeutic developments that could improve reproductive outcomes by reducing aneuploidy.

Introduction

Human conceptions are afflicted by an extraordinary rate of chromosome errors, and the majority derive from the oocyte (Hassold and Hunt, 2001). In natural conceptions that reach clinical recognition, 35% of human pregnancies are aneuploid. The rate observed in preimplantation embryos is substantially higher, in part because aneuploid embryos have poor developmental potential and are selected against during the peri-implantation stages and throughout foetal life (Capalbo et al., 2014). In natural conception, more than 90% are of meiotic origin and the majority are caused by errors in meiosis I (Hassold and Hunt, 2001; Gabriel et al., 2011). In reproductive aged women, 20–30% of occytes (and up to 70% of oocytes in advanced maternal age (AMA) women) are aneuploid, while just 1–8% of spermatozoa are afflicted (Lu et al., 2012; Wang et al., 2012). In sperm, the incidence of aneuploidy is independent of paternal age (Erickson, 1978; Hassold and Hunt, 2001; Lu et al., 2012; Wang et al., 2012). The analyses of aneuploidy in miscarriages have been invaluable for our appreciation of the serious consequences chromosomal imbalances have for embryonic and foetal development, since a much higher incidence and wider range and representation of chromosomes are detected compared to subsequent developmental stages, including live births (Hassold et al., 1980; Zaragoza et al., 1994).

Since the discovery that aneuploidy is the major cause of congenital disorders (Jacobs and Strong, 1959; Jacobs et al., 1959; Lejeune, Gautier, and Turpin, 1959; Ford et al., 1959a,b), most our knowledge has derived from population-based studies of foetal losses and rare live births. Maternal age is the major factor that influences aneuploidy, giving rise to the characteristic J curve (Erickson, 1978; Hassold and Hunt, 2001; Fig. 1A). However, individual chromosomes follow different age-dependent curves (Nagaoka et al., 2012; Franasiak et al., 2014a, b; Fig. 1B) suggesting that both chromosome-specific as well as general cellular factors conspire to shape the segregation efficiency in human oocytes.

Aberrant recombination patterns on chromosomes that have mis-segregated have also been identified as an important factor, in both male and female gametes (Table I). This is because recombination together with cohesion of sister chromatids establish the unique ‘bivalent’ chromosome structure where homologous partner chromosomes are tethered together, a configuration that is critical for their accurate segregation in meiosis I (Fig. 2A). The remarkable feature is that recombination occurs in foetal oocytes whereas chromosome segregation takes place decades later (Fig. 2A). Since mammalian oocytes are arrested at the G2/M transition (or dictyate stage), this raises the intriguing question of how the bivalent is maintained until the meiotic divisions.

Here, we review recent technological breakthroughs in single-cell genomics and time-lapse imaging of human oocytes, which are revealing unexpected new facets of chromosome segregation. In particular, we discuss the emerging model of Chromosomal Aging, which integrates and relies upon knowledge from human and other mammalian oocytes to explain age-related aspects of female aneuploidy. New insights into the spindle assembly checkpoint (SAC) and spindle formation in human oocytes reveal that the spindle itself may be error-prone. Advances in technology and in vitro biology usher in the potential for novel therapies that could extend the reproductive lifespan of women (Andersen, 2015).

Methods

To evaluate molecular mechanisms that have been implicated in human female meiosis, we carried out a systematic review following the PRISMA guidelines. Pubmed Central was searched for the following MeSH terms: chromosome segregation, human embryo, human oocyte, human germline, meiosis, aneuploidy, meiotic disturbances, meiotic protective factors and aneuploidy, for all full-text articles published in English from 1990 through 2016. Keywords were used in multiple and overlapping combinations to identify those publications strictly relevant to oocyte meiosis. This identified 29 213 publications. Of these, 3855 were reviews and excluded. The remaining abstracts were screened for whether human oocytes or embryos were included. This eliminated 12 953 articles. The remaining 12 405 abstracts were screened further for the inclusion of ‘human aneuploidy’ or ‘human meiosis’. This reduced the number of relevant publications to 295, which were read in full. Reference lists were also cross-checked for additional relevant studies.

For the models for chromosome segregation in meiosis, no systematic review was carried out. We used the references lists from human oocyte studies and relied on reviews from the meiosis field to identify relevant mechanistic studies. The animal studies therefore do not represent a comprehensive review of all available models for aneuploidy.

For illustrative purposes, embryo aneuploidy data from previous publications were retrieved and elaborated. Aneuploidy data were obtained by means of array (a) CGH or quantitiative (q) PCR-based 24-chromosome testing (Capalbo et al., 2013, 2014, 2015, 2016). Whole chromosome aneuploidies were stratified according to female age used as continuous variable (per each year unit) to report chromosome-specific susceptibility to oocyte aging. Data were also stratified at the single chromosome level to investigate monosomy/trisomy ratio observed in embryos. Overall, data from approximately 5000 embryos were included in this elaboration. Only a descriptive examination of peer-reviewed data is reported here for illustrative purposes and without additional inferential analysis performed.

Results

A new genomics era for single cells: recovery of genetic information from all products of meiosis

We have long appreciated the power of being able to detect both the genetic information and chromosomal content in all products of meiosis to infer chromosome segregation patterns and genetic inheritance (Fig. 2A). Specifically, obtaining the recombination patterns allows ‘reconstruction’ of what bivalents might have looked like, prior to chromosome segregation in the meiotic divisions (Fig. 2C). Such analyses, recently been developed for human oocytes and their matched polar bodies, are referred to as MeioMapping (Fig. 2B and C) (Ottolini et al., 2015). Two independent groups sequentially biopsied the polar bodies by either activating the oocyte artificially (Ottolini et al., 2015) or fertilizing the mature oocyte using sperm (Hou et al., 2013; Ottolini et al., 2015). After fertilization, the female pronucleus was extracted (Hou et al., 2013) or a biopsy of the blastocyst was obtained (Ottolini et al., 2015). This is followed by amplification of the DNA, termed whole-genome amplification (WGA; Zhang et al., 1992), which yields sufficient quantity of DNA to create libraries for next-generation sequencing (NGS) or single-nucleotide polymorphism (SNP) arrays (Fig. 3). WGA was initially applied to screen polar bodies for aneuploidies using a content analysis known as comparative genomic hybridization (CGH; Wells et al., 2002). SNP arrays or NGS allow not only content but also genetic information to be discerned (Fig. 3).

Several important observations have been made using MeioMapping because it allows the recovery of all chromosomes involved in the meiotic divisions. Using the pericentromeric SNPs for chromosome fingerprinting, Ottolini and colleagues discovered a new segregation pattern, termed reverse segregation (Fig. 2A). Reverse segregation is characterized by both homologous chromosome segregating their sister chromatids at meiosis I. At meiosis II, the two non-sister chromatids segregate correctly in nearly 80% of cases (Ottolini et al., 2015).

MeioMapping has revealed that although the major aberrant segregation defect is reverse segregation in oocytes from women of AMA (Ottolini et al., 2015), predivision or precocious separation of sister chromatids (PSSC) causes the majority of aneuploidy in activated/fertilized oocytes (Hou et al., 2013; Ottolini et al., 2015). This is because non-sister chromatids segregate to opposite spindle poles with relatively high efficiency at meiosis II during reverse segregation (Ottolini et al., 2015). The molecular mechanisms are unknown. MeioMapping has shown that ~70% of errors occur in meiosis I, but only half of these cause aneuploidy in the oocyte, since ‘corrective’ segregation events at meiosis II may compensate for the initial error (Hou et al., 2013; Ottolini et al., 2015). This is consistent with findings from content analyses of polar bodies and embryos (Verlinsky et al., 2001; Pellestor et al., 2002, 2005; Fragouli et al., 2009; Handyside et al., 2012; Capalbo et al., 2013; Forman et al., 2013). All of the chromosome segregation errors observed in the two studies that followed the chromosomes with fingerprinting also firmly established that the segregation errors are reciprocal, affecting two of the three cells, and therefore arose during meiosis. This is also consistent with findings from chromosomal content analyses (Verlinsky et al., 2001; Pellestor et al., 2002, 2005; Fragouli et al., 2009; Handyside et al., 2012; Capalbo et al., 2013) and is important clinically, since polar bodies can be used to infer aneuploidy in oocytes. This has been the basis for preimplantation genetic testing for aneuploidy (PGT-A; Fig. 4).

Seeing is believing: time-lapse imaging of chromosomes in real time reveals splitting of sister kinetochores at meiosis I

The inference that human chromosomes can split their sister chromatids at meiosis I, during reverse segregation or a single predivision (PSSC) should perhaps not have come as a surprise. The single X chromosome in XO mouse oocytes frequently segregate sister chromatids at meiosis I (LeMaire-Adkins and Hunt, 2000). Development of time-lapse imaging allowed the first visualization of sister chromatid splitting at meiosis I. This came from mouse oocytes deleted for Sycp3 (Kouznetsova et al., 2007), a synaptonemal complex gene that promotes crossing over (Yuan et al., 1998, 2000). In oocytes of Sypc3^-/- females, a few homologue pairs fail to generate crossovers and the chromosomes remain univalent. The univalents segregate both as whole chromosomes or split their sister chromatids at meiosis I (Kouznetsova et al., 2007; Sakakibara et al., 2015).

Direct observation that human chromosomes can split their sister chromatids at meiosis I came when fluorescence time-lapse imaging of human oocytes was achieved. Human immature GV oocytes were injected with mRNA encoding fluorescently labelled kinetochores and spindle proteins, which allowed Zielinska and colleagues to follow the fate of chromosomes during the first meiotic division (Zielinska et al., 2015). The incidence of sister chromatid splitting increased with maternal age. The splitting of sister chromatids in meiosis I provided direct evidence for both the reverse segregation and PSSC patterns that have been inferred from genetic data (Hou et al., 2013; Ottolini et al., 2015) and cytological observations of fixed material (Angell, 1991; Fig. 5B). Notably, human MII oocytes also contain whole chromosome gains and losses (Jagiello et al., 1975; Polani and Jagiello, 1976; Martin et al., 1986; Pellestor et al., 2002), suggesting that PSSC and reverse segregation co-exist with meiosis I non-disjunction. It is currently unclear what factors determine their relative frequencies in oocytes obtained from IVF clinics.

Two different mechanisms appear to be involved in sister chromatid splitting in human oocytes (Zielinska et al., 2015). Univalents are sometimes apparent prior to sister chromatid splitting (Fig. 5C). Univalents or individual sister chromatids have been seen previously in human MI oocytes (Angell, 1995; Garcia-Cruz et al., 2010a, b; Fig. 5B) and their incidence is age-related in both mouse (Henderson and Edwards, 1968) and human (Angell, 1995; Zielinska et al., 2015). Although there is extensive evidence that this is due to age-related deterioration in cohesion in mouse oocytes, the mechanisms for their age-related incidence in human oocytes is still not clear, as discussed below. In other cases, the bivalent still appeared to be intact, but one or both the chromosomes still oriented their sister kinetochores in opposite directions, termed ‘half-inverted’ or ‘fully-inverted’ configurations (Fig. 5D).

Spindle instability and unfused sister kinetochores are permissive of erroneous kinetochore-microtubule attachments

What may cause sister kinetochores to separate precociously? Human meiotic spindles are not only different in their architecture but highly unstable compared to those found in mouse oocytes (Holubcova et al., 2015). Occurring without centrosomes (acentrosomal), spindle formation initiates from chromosomes and the spindles take up to five hours to form after breakdown of the nuclear envelope (germinal vesicle break down). Eight of out ten spindles become multipolar prior to taking on a bipolar configuration. Metaphase I takes an extremely long time, 14–16 h (Angell, 1995; Holubcova et al., 2015) and although the spindle takes on a bipolar configuration prior to the onset of anaphase I and division, the severity of instability is correlated with attachment of sister kinetochores to opposite spindle poles (Fig. 5E). Spindle instability was reported to be age-independent (Holubcova et al., 2015), a finding in contrast with previous studies of fixed oocytes (Battaglia et al., 1996), including oocytes from naturally cycling women.

Spindle instability alone, however, does not explain why sister kinetochores bi-orient. Sister kinetochores are normally co-orientated and sister chromatids are topologically connected by catenation of the sister DNA molecules and protein-mediated cohesion (cohesin complexes; Fig. 5A). In many organisms, sister kinetochores are fused, however, in human oocytes this may not be the case (Patel et al., 2015). Instead, sister kinetochores are often attached to spindle fibres, known as k-fibres, from opposite spindle poles (Patel et al., 2015; Zielinska et al., 2015). The incidence of sister kinetochore splitting and, more generally, the distance between sister kinetochores (termed iKT distances), have been reported to correlate with increased maternal age in both mice (Chiang et al., 2010; Lister et al., 2010) and humans (Duncan et al., 2012; Patel et al., 2015; Zielinska et al., 2015; Lagirand-Cantaloube et al., 2017). The impact on chromosome segregation as determined by the physical distances between sister kinetochores remain somewhat unclear.

The chromosomal aging hypothesis: cohesion loss precedes chromosome segregation errors

As discussed above, the incidence of univalents and sister kinetochore splitting in meiosis I is age-related in both mouse and human oocytes. There is also evidence that this leads to elevated risk of chromosome missegregation in meiosis I or meiosis II (Koehler et al., 2006; Kouznetsova et al., 2007; Chiang et al., 2010; Lister et al., 2010; Merriman et al., 2013; Sakakibara et al., 2015; Zielinska et al., 2015). What is less clear, however, is the origin of univalents and their age-related incidence, which was first noted by Henderson and Edwards (1968) for mouse oocytes. In their seminal study, the authors attempted to reconcile findings from population studies that aneuploidies were often associated with ‘crossover-less’ recombination patterns. They proposed two not mutually exclusive hypotheses: (i) that univalents originated from bivalents that deteriorated during dictyate arrest; or (ii) that univalent incidence increased with maternal age because oocytes entered and exited meiosis (ovulation) in waves, such that foetal oocytes in the early waves had ‘good’ recombination and subsequently good segregation, when females were young butas females aged, oocytes originating from ‘later’ foetal waves would be ovulated, and these had poorer segregation because the recombination was worse. This latter hypothesis is also known as the Production Line hypothesis.

The first part of the Production Line hypothesis predicts that recombination rates should follow a gradient in foetal oocytes, however, this is not the case (Rowsey et al., 2014). The second part, that oocytes with lower recombination rates and therefore compromised segregation potential ovulate later, has not been tested. If anything, recombination rates in children increase as women age (Kong et al., 2004; Coop et al., 2008; Campbell et al., 2015), although this can also be explained by the selection for oocytes with high recombination rates and therefore lower aneuploidy risk (Ottolini et al., 2015). There are several scientific observations that makes it difficult to rule out ovulation timing. Unlike mouse oocytes, human foetal and adult oocytes show tremendous heterogeneity in their recombination rates, in some cases by an order of magnitude (Lenzi et al., 2005; Hou et al., 2013; Ottolini et al., 2015). The incidence of crossover-less foetal oocytes is also extremely high, reported to be nearly 20% in some studies (Tease et al., 2002; Cheng et al., 2009). Thus, preferential ovulation timing could explain the high incidence of univalents in oocytes from women of AMA. Such a model requires that (i) cell-cycle progression over a time scale of decades is linked to recombination status; and (ii) follicular response is dependent upon the cell-cycle status of the oocyte. Although these two requirements are possible, since the meiotic silencing checkpoint determines progression in response to synapsis and recombination (Turner, 2015) and oocyte attrition due to DNA damage halts follicular development (Titus et al., 2013), they remain to be tested or shown to be important for aneuploidy generation in human.

There is ample evidence that cohesion loss may drive the age-dependent univalent formation in mouse. The most conclusive is from the Smc1β^-/- female mouse where recombination levels are narrowly distributed in foetal oocytes, but the incidence of univalents (and aneuploidy) follows an age-dependent gradient (Hodges et al., 2005). SMC1β is particularly intriguing because it is part of the meiotic cohesin complex that mediates cohesion (Fig. 5A) and is haploinsufficient for aneuploidy (Murdoch et al., 2013). In general, because mouse oocytes show highly regulated recombination with virtually no crossover-less oocytes (Hunt and Hassold, 2008), the majority of age-related effects on chromosome structure can be attributed to cohesion loss. Age-dependent cohesion loss accounts for both univalent formation and the loss of sister chromatid cohesion near centromeres. Hence, cohesion deterioration or exhaustion provides a single molecular mechanism for the loss of bivalent maintenance as females age (Jessberger, 2012).

One important prediction of the chromosomal aging model is that if cohesin loss is the cause of aneuploidy, then bivalent deterioration should be correlated to cohesin loss or dysregulation. This appears to be the case in mouse MI oocytes where mSGO2 and mREC8 (Fig. 5A) are preferentially lost from centromeres where the two sister kinetochores are split (Lister et al., 2010). In human oocytes, this prediction is not upheld. Although univalents are observed with increased incidence in oocytes from women of AMA (Henderson and Edwards, 1968; Angell, 1995; Zielinska et al., 2015), univalent structures were not preferentially depleted for cohesin (REC8) compared to bivalent chromosomes in oocytes from women 35 years or younger (Garcia-Cruz, et al., 2010a). Thus, although other studies have reported diminished levels of cohesin proteins (total levels) and mRNAs (Tsutsumi et al., 2014), the chromosomal association of cohesin does not correlate with bivalent structures, at least according to the results of this one study (Garcia-Cruz et al., 2010a). However, it is possible that cohesion maintenance is affected, since only a proportion of cohesin complexes mediate cohesion in mitotic cells (Nishiyama et al., 2010). Having an underlying mechanism for cohesion loss in human oocytes is clearly needed.

The cohesion exhaustion hypothesis is only one of several mechanisms by which meiotic chromosomes may age or be aberrant (age-independent). Defective deacetylation of histone H4 K12 is seen with greater incidence in aged oocytes (human) and has been suggested to account for compaction and chromosome alignment defects, resulting in aneuploidy (van den Berg et al., 2011). Shorter telomeres were preferentially associated with aneuploidy in a study of polar bodies and preimplantation embryos (Treff et al., 2011). A study of preimplantation embryos also reported that aneuploid embryos from women of AMA or suffering from recurrent miscarriage also had, on average, shorter telomeres (Mania et al., 2014). Telomere regulation, however, is challenging to study and changes according to developmental stage (Turner et al., 2010; Turner and Hartshorne, 2013). Telomerase, which extends telomeres, is required for chromosome alignment and spindle stability at metaphase I in mouse (Liu et al., 2002). However, we currently do not know whether this is due to telomeric effects on chromosome segregation or whether the correlation can be explained by the important function of telomeres in recombination and chromosome pairing during the preceding meiotic prophase in foetal oocytes (Shibuya and Watanabe, 2014).

Chromosome-specific maternal age curves

One intriguing feature of chromosomal aging is that it must interact with other factors that cause chromosomes to follow different age-dependent aneuploidy curves (Nagaoka et al., 2012; Franasiak et al., 2014a, b) (Fig. 1B). As illustrated in Fig. 1B, chromosomes 1–9 are relatively shallow, whereas the acrocentric chromosomes and especially 21 and 22 show high rates of missegregation. Chromosomes 15 and 16 are different yet again (Nagaoka et al., 2012).

One major factor that influences chromosome segregation and potentially also sister chromatid cohesion is recombination. The influence of recombination on aneuploidy was first discovered by studying spontaneous miscarriages once chromosome fingerprinting became available (Nagaoka et al., 2012). The positions of recombination events where homologues become tethered appeared perturbed when chromosomes had mis-segregated at meiosis I (Lamb et al., 2005). Some crossovers are ‘too close’ or ‘too far’ away from centromeres and some pairs are crossover-less. Importantly, chromosomes display specific vulnerabilities (Nagaoka et al., 2012). Further studies of recombination in the US National Down Syndrome Project identified several intriguing features, including that Trisomy 21 individuals tended to have a global decrease in recombination events compared to their unaffected siblings (Middlebrooks et al., 2013). Using the approach of mapping recombination across all three products from a single meiosis, global recombination levels were shown to correlate positively with euploidy in both younger women (Hou et al., 2013) and women of AMA (Ottolini et al., 2015). In addition, crossover-less chromosomes are at increased risk of PSSC and reverse segregation, since sister chromatids that did not participate in a recombination event are also at elevated risk of PSSC (Ottolini et al., 2015). Finally, recombination also appears to determine how sister chromatids segregate from each other at meiosis II. Recombinant sister chromatids are more likely to be retained in the oocyte compared to the second polar body in both human (Ottolini et al., 2015) and mouse (Wu et al., 2005). The mechanisms underlying these features are not known.

Pertinent to the discovery that sister chromatids are often separated at meiosis I, recombination near centromeres is associated with aneuploidy in human conceptions (Nagaoka et al., 2012). Whereas terminal crossovers may provide less rigidity of bivalents during tension-mediated attachment to microtubules (Webster and Schuh, 2016), centromeric crossovers may disrupt cohesion or the crossover may become entrapped in the centromeric domain of cohesin that is protected and maintained until meiosis II (Fig. 5A). Studies in fruitflies and budding yeast have also found that recombination near or at centromeres is detrimental for chromosomes (Koehler et al., 1996; Rockmill et al., 2006). Studying the direct link between recombination and sister chromatid dynamics with combined live cell imaging and genomics will likely yield important insight into the impact that centromeric and telomeric crossovers have on chromosome segregation.

Reconstructing the bivalent configuration from MeioMaps: recombination and its link with chromosome segregation

The combined assessment of haplotypes that are determined by recombination also allowed the first direct correlations between recombination and chromosome segregation in single meiosis (Fig. 2). Traditionally, recombination has been studied by assessing MLH1 foci, a cytological marker that correlates in numbers with chiasmata and whose deletion reduces crossover rates or abrogates meiosis in model organisms (Baker et al., 1996; Edelmann et al., 1996; Hunter and Borts, 1997; Tease et al., 2002). Studies from natural pregnancies have firmly established that recombination rates and their position along chromosomes are important in the aetiology of aneuploidy in foetuses and rare live births. Hou et al. (2013) reported that the link between recombination and errors in chromosome segregation is relatively weak in young women, whereas Ottolini et al. (2015) found that recombination frequency explains 18% of the variability in aneuploidy in oocytes obtained from women of AMA. In both cases, a lower genome-wide rate of recombination events lead to higher susceptibility to segregation errors during meiosis.

Epidemiologically, recombination and maternal age are still the strongest factors associated with the risk of aneuploidies. Recombination rates are also linked to increased reproductive success in women, especially as they age (Kong et al., 2004). It is possible that increased recombination events within an oocyte could counteract bivalent deterioration over time thus ensuring normal segregation and a euploid conception. This could explain the elevated recombination rates in children born to ‘older’ mothers in several populations (Kong et al., 2004; Coop et al., 2008; Campbell et al., 2015).

The crossover paradox

If recombination levels are important to ensure chromosome segregation, why does female meiosis generate more ‘vulnerable’ configurations including crossover-less pairs compared to male meiosis, who only have half the number of crossovers (Tease, et al., 2002; Lenzi et al., 2005; Cheng et al., 2009; Ottolini et al., 2015)? Vulnerable crossover configurations such as crossover-less pairs are more than an order of magnitude greater in foetal oocytes than sperm (Cheng et al., 2009). What causes ‘recombination failure’ on chromosome pairs in oocytes compared to adult sperm, where such configurations are virtually never seen? Recent mathematical modelling of crossover distribution in human foetal oocytes suggests that specific steps in the maturation of crossovers may be inefficient in oocytes, but not sperm (Wang et al., 2017). This could explain not only the incidence of crossover-less chromosome pairs but also other vulnerable configurations (e.g. terminal positions) that predispose chromosomes to missegregation decades later, when the adult oocyte ovulates and the meiotic divisions take place.

Do cellular checkpoints contribute to poor genetic quality of human oocytes?

MeioMapping in human oocytes revealed one catastrophic meiosis, where none of the chromosome pairs had recombined and 12 pairs had mis-segregated, leading to gross aneuploidy in the MII oocyte (Ottolini et al., 2015). In this case, a complete failure to initiate meiotic recombination could explain the apparent lack of recombination genome-wide. What is equally interesting is that such oocytes mature suggesting that the meiotic silencing checkpoint is not stringent enough to eliminate defective foetal oocytes from the adult pool (Turner, 2015). The meiotic silencing checkpoint normally eliminates oocytes or spermatocytes that fail to pair (synapse) homologous chromosomes, a process that is highly dependent upon recombination. However, the meiotic silencing checkpoint appears to be noticeably less efficient in female mice than male mice (Cloutier et al., 2015). If the same is true for human oocytes, then inefficient elimination of oocytes with aberrant or vulnerable crossover configurations generated by the highly heterogenous recombination process (Tease et al., 2002; Lenzi et al., 2005) could contribute to the high rate of aneuploidy.

The DNA damage response (DDR) mediates repair as well as cellular arrest or apoptosis in response to endogenous and exogenous damage (Jackson and Bartek, 2009). In meiosis, the DDR also induces recombination via the formation of hundreds of double-strand breaks (Keeney et al., 2014). Most DNA damage repair genes appear to be important for culling oocytes (or spermatocytes) in response to persistent damage (Bolcun-Filas et al., 2014). Mutants that result in damage that does not trigger arrest, such as recombination intermediates (Hwang et al., 2017), could contribute to aneuploidy in affected human oocytes.

The SAC controls the onset of anaphase. In mitotic cells, the SAC monitors correct kinetochore attachment and promotes re-orientation until sister kinetochores are under tension from attachment to opposite spindle poles. Only then is anaphase onset triggered and separase activated (Fig. 5A), allowing sister chromatids to separate. Unattached kinetochores trigger the accumulation of SAC components and formation of the mitotic checkpoint complex (MCC; Cdc20-BubR1, Bub3 and Mad2), which inhibits the activation of the anaphase promoting complex. The SAC is also required to prevent aneuploidy in mouse oocytes (Wassmann et al., 2003; Homer et al., 2005; Niault et al., 2007; McGuinness et al., 2009; Hached et al., 2011; Touati et al., 2015), presumably by ensuring the timely onset of anaphase, which is accelerated when Bub1 or BubR1 are depleted or when only a single copy of MAD2 is present. There is therefore substantial evidence that the SAC is critical for mammalian oocytes, but the question is whether the SAC works the same way in meiosis as in mitosis.

Several lines of evidence suggest that the mechanism that triggers the SAC might be adapted in meiosis. In mitosis, the stretching of sister kinetochores that are under tension from microtubules emanating from opposite spindle poles physically removes Aurora B kinase from its phosphorylation targets, thereby preventing de-stabilization of kinetochore-microtubule interactions, or ‘silencing’ the SAC. Recent observations in mouse oocytes question this model for bivalents in meiosis I. Both Aurora B and Aurora C are localized close to kinetochore-microtubule attachments when these are under tension, suggesting that mouse oocytes do not contain a mechanism for ‘stretching’-dependent phospho-regulation of kinetochore-microtubule attachments (Yoshida et al., 2015). The bivalent tension or stretching model originated from the elegant ‘pricking’ experiments conducted in grasshopper spermatocytes, where bivalents ceased to reorient when a needle that mimicked tension was applied (Nicklas and Koch, 1969). Tension could still be monitored in mammalian oocytes, but the mechanism for kinetochore-microtubule re-orientation would then have to be different compared to mitotic cells.

The second major finding in recent years is that the SAC responds to exogenous DNA damaging agents by arresting cells in MI (Collins et al., 2015; Marangos et al., 2015; Collins and Jones, 2016). The implication is that if the SAC is compromised, chromosomes may divide, despite DNA lesions. The challenge is to determine what endogenous lesions may trigger the SAC and whether such lesions would generate aneuploidy, as opposed to chromosome breakage and cellular arrest. If this is an important function of the SAC in the origin of human aneuploidies, then one would expect to see damaged chromosomes in oocytes.

In addition to non-canonical crosstalk between the DNA damage checkpoint and the SAC, the mitotic surveillance pathway blocks cell-cycle progression in response to prolonged mitosis (M-phase) and the loss of centrosomes (reviewed in Lambrus and Holland, 2017). Intriguingly, DNA damage response genes 53BP1 and USP28 act upstream of p53 to inhibit the proliferation of daughter cells that become aneuploid (Hinchcliffe et al., 2016). We do not know whether this pathway is active in meiosis, but it raises the interesting question of how centrosomes become ‘lost’ in human oocytes and how this is tolerated. One potential caveat is that mouse models targetting 53BP1 are viable, questioning the effect that the mitotic surveillance pathway has in vivo (Fernandez-Capetillo et al., 2002).

Do the cellular checkpoints that are active in murine oocytes contribute towards human aneuploidies? The oocyte-specific facets of cell-cycle control have been proposed to create vulnerability to segregation errors (Nagaoka et al., 2011). There is also evidence that the levels of several checkpoint proteins, such as Mad2, is lowered in aged mouse oocytes. This may lead to decreased error correction, when spindle organization is impaired during meiosis I (Yun et al., 2014). However, many genes show transcriptional dysregulation in aged human oocytes (Grondahl et al., 2010), and improved models will need to correlate several factors including phenotypic expression in haploinsufficiency models in mouse. Variants in both Aurora B and C were recently mapped and functional characterization of the variants, when overexpressed in mouse oocytes, suggested that the different alleles may support different SAC strength (Nguyen and Schindler, 2017; Nguyen et al., 2017). Collectively, such studies suggest that modulation to cell-cycle checkpoints may be important regulators of genome integrity and aneuploidy in the natural population.

Maturation arrest and spindle assembly defects

Although much focus has been on the SAC, recent observations suggest that the spindle itself may cause arrest and, potentially, segregation errors in meiosis. Feng and colleagues investigated the genetic causes of several families showing complete oocyte maturation arrest at the MI stage (Feng et al., 2016). The authors found seven different mutations in the β-tubulin-encoding gene, TUBB8, in women with primary infertility due to meiosis I arrest of their oocytes. The TUBB8 mutations were paternally inherited in five of the families whereas de novo mutations in TUBB8 occurred in the affected women of the other two families. In the same study, the authors not only found that TUBB8 is primarily expressed in human oocytes and embryos, they also determined that the TUBB8 protein is an essential and major component of the oocyte spindle. The introduction of mutant protein, but not wild type TUBB8, into HeLa cells and yeast cells caused altered microtubule dynamics. Mutant TUBB8 protein was also dominant when expressed in mouse and human oocytes, causing maturation arrest, thereby recapitulating the phenotype observed in the women presenting with primary infertility. The study has set a high bar for the functional confirmation that the mutations mapped to specific genes, in this case TUBB8, are the likely cause of pathogenicity.

Subsequent studies of mutations mapped in TUBB8 have revealed further complexities in the role of the meiosis-specific β-tubulin. The phenotypes include embryonic arrest (Chen et al., 2016; Feng et al., 2016). It is currently unknown whether the arrested preimplantation embryos are aneuploid and whether the mature oocytes that do complete the first division have a normal chromosome content. Given the severe spindle defects, it would not be unreasonable to speculate that the MII oocytes could be aneuploid. Assessing chromosome segregation when TUBB8 is perturbed may yield important insights into human aneuploidies.

Influence of genetic factors on human female meiosis

The J curve is a population average. The likelihood of producing an aneuploid oocyte cannot be predicted reliably by maternal age alone. There is a significant variation in aneuploidy rates amongst individuals of the same reproductive age (Franasiak et al., 2014a,b). Relatively young women (<35 years old) appear to have unexpectedly high levels of aneuploid oocytes compared to many women considered to be of AMA (>35 years old). The only genome-wide association study of chromosome errors in human preimplantation embryos identified a quantitative trait locus for mitotic errors, but did not find any quantitative trait loci that were associated with maternal aneuploidies, at least not any that were significantly elevated above the false discovery rate (McCoy et al., 2015). A target sequencing approach identified putative Aurora B and C variants, as discussed above (Nguyen et al., 2017). Given the large volume of data generated from the increased application of aneuploidy testing in ART cycles, further studies may provide insight into this important medical and societal issue in the near future.

Several observations suggest that genetic factors could predispose to both the general baseline and age-related elevation in aneuploidy conceptions. A recent genome-wide screen for new meiotic genes in mouse oocytes revealed hundreds of genes, whose depletion by RNAi affected chromosome segregation (Pfender et al., 2015). This suggests that conducting refined analyses in human oocytes and population-based studies may yet yield new molecular targets. Studies in mice suggest that heterozygosity of SMC1β, a conserved meiosis-specific cohesin subunit, predisposes to aneuploidy (Murdoch et al., 2013). Deletion of both copies of SMC1β predisposes to age-related loss of bivalent structures and therefore to aneuploidy in mouse oocytes (Hodges et al., 2005). The haploinsufficiency studies are important because they suggest dosage sensitivity. This is particularly relevant in human populations where complete deletions (homozygous) of gene activities are relatively rare and usually only found in consanguineous families (O’Driscoll, 2008).

There have been several studies attempting to investigate the link between ovarian aging (premature ovarian insufficiency) and chromosomal aging of human oocytes. Currently, there are no firm links between the two. However, we did find a putative SNP where women with mutations that truncate the product of SYPC3 were at risk of recurrent miscarriage due to aneuploidy in the foetus (Bolor et al., 2009). The findings were interesting because the function of SYCP3 function in murine meiosis is well defined, as it acts as a chromosomal scaf-fold protein that forms a part of the synaptonemal complex and promotes crossing over (Lammers et al., 1994). The association between recurrent miscarriage and mutation in SYCP3, however, was not reproduced by a second study of the same SNP in the same ethnic population (Mizutani et al., 2011) nor in a different population (Hanna et al., 2012).

In summary, there are currently several molecular models from mouse studies that firmly implicate multiple different mechanisms affecting chromosomes and spindles as well as lowered cell-cycle control. However, clinical associations between human aneuploidy and their prognostic value remain to be determined. Once defined, the multifactorial nature of aneuploidy may help the development of genetic biomarkers for increased susceptibility to aneuploidies in women of young reproductive age, so that they may opt for preventive measures, or it may even lead to cures.

Artificial gametes and chromosome therapy in living oocytes and preimplantation embryos

The high incidence of aneuploidy in human oocytes has given rise to extensive prenatal screening programmes as well as genetic testing to select euploid embryos for transfer in IVF clinics (reviewed in Vermeesch et al., 2016). Preimplantation genetic testing (Fig. 4) is only a diagnostic procedure that increases the efficiency of treatment but does not increase the pregnancy chance for a given stimulation cycle. This is important since attrition is still the major factor why couples do not obtain a pregnancy in IVF settings. However, aneuploidy testing does not change the genetic status of the oocytes that a woman produces. Once a woman reaches the point where she no longer produces any chromosomally normal oocytes, aneuploidy screening no longer provides any benefit in terms of pregnancy success. In many countries, this limits the age at which women may undergo an IVF treatment.

How do we improve reproductive success and reduce the effects of maternal aging in the natural population as well as in the clinic? Identification of lifestyle factors that affect natural conception is important. Several factors including smoking, irradiation, oral contraceptives and low socioeconomic status (Christianson et al., 2004; Hunter et al., 2013) have been implicated but their molecular basis has yet to be elucidated. Disentangling the factors that influence aneuploidy may provide us with lifestyle interventions to reduce miscarriage rates and may move the J curve to the right and prevent the early truncation of reproductive lifespan caused by aneuploidy.

One lifestyle intervention may be to target cellular responses that are activated in response to aging, obesity or high lipid diets. Several studies in mice have suggested that targeting the unfolded protein response, which is activated by stress in the endoplasmic reticulum (ER), improves maturation of oocytes as well as developmental outcomes of the offspring. When follicles from obese mice or obese females were exposed in vitro or in vivo, respectively, to salubrinal, this improved maturation rates, conceptions, and health of the offspring (Wu et al., 2010, 2012, 2015). In vivo, salubrinal was only supplied for a very short duration prior to ovulation (Wu et al., 2015), suggesting that anovulation in obese females can be reversed without major interventions. Given that ER stress is implicated in a wide range of cellular pathologies (Martinez et al., 2017), the strength of the in vivo and in vitro data from murine models warrants preclinical assessment of ER inhibitors in improving maturation efficiency of human oocytes.

Another avenue of research is to understand how much of our knowledge from aging models in mice we can translate to humans, and in particular what may be relevant for reproductive aging. There are several progeria models of mouse, including DDR genes (e.g. Murga et al., 2009), that may be interesting to assess given that common variants of several DDR genes were implicated in determining the age of onset of natural menopause (Day et al., 2015). Premature ovarian insufficiency has also been linked to mutations in the DDR genes, including BRCA1 (Titus et al., 2013) and meiosis-specific STAG3 whose protein (Caburet et al., 2014) is a component of meiotic cohesin complexes. Deletion of STAG3 causes persistent DNA damage and failure of chromosome synapsis in meiotic cells (Caburet et al., 2014; Fukuda et al., 2014; Hopkins et al., 2014; Winters et al., 2014), like many other cohesin mutants (McNicoll et al., 2013). However, it is unclear whether the DDR solely dictates maintenance of the follicle pool, or, whether this complex network of more than 450 genes (Pearl et al., 2015) also influences chromosome segregation and therefore, aneuploidy.

Chromosome therapy is a functional correction of aneuploidies in living cells. This has been achieved with two different molecular strategies. Jiang et al. exploited the natural mechanism of XIST, a non-conding RNA that drives the transcriptional silencing of one of the X-chromosomes in females by inducing heterochromatic modifications to the inactive X chromosome (Jiang et al., 2013). The authors transcribed XIST from the DYRK1A locus of chromosome 21 in iPS cells from a Down Syndrome individual. After induction of the ectopically inserted XIST, regions of the targeted chromosome 21 became heterochromatized and transcriptionally silenced, and a ‘chromosome 21 Barr body’ was detectable. The deficit in cell proliferation and differentiation of iPS cells into neural progenitor cells was also reversed by chromosome 21 repression. The heterochromatic silencing was stably maintained after the initial XIST expression induction and the system was compatible with natural female X inactivation, making this a promising strategy for translating dosage compensation to individuals with Down Syndrome.

Another strategy to achieve correct chromosome number is based on the use of a mammalian-specific gene, ZSCAN4 (zinc finger and SCAN domain containing 4) (Akiyama et al., 2015; Amano et al., 2015). We currently only have data from mouse, where it is expressed at the 2-cell stage of preimplantation embryos and is required for genome stability and maintenance of a normal karyotype in mouse ES cells. Using mRNA and Sendai virus vectors encoding human ZSCAN4, trisomy 18 or 21 could be corrected to euploid, apparently without affecting other chromosomes.

One intriguing possibility is whether such new biologics could be developed to enhance the efficiency of ZSCAN4 during oogenesis of GV oocytes in vitro and/or preimplantation development of human embryos.

Alternatives to chromosome therapy include mRNA or use of inhibitors during in vitro maturation or embryo development. Elucidation of putative mechanisms where factors may be rate-limiting for accurate chromosome segregation raises the possibility of supplying the protein or mRNA to the oocyte. Conversely, elevated levels of inhibitory proteins could be treated by use of inhibitors. Such procedures would not alter the genetic make-up of the oocyte, but would facilitate improved genetic integrity of human embryos generated in vitro. Even though, these are only mere perspectives at present, such developments, once available, would counteract the natural mechanism of reproductive aging and increase the cumulative live birth rate per started cycle.

Future Perspectives

Our increasing knowledge of human aneuploidies and their origins together with advances in technologies are allowing unprecedented insight into chromosome errors in human conceptions. The challenge is to translate this knowledge into improved diagnosis and therapies for precision medicine that allows women to make informed choices about their reproductive health. Improving the efficiency of medically assisted reproduction is critical and the future is promising with regards to developing interventions to improve the genetic quality of oocytes, especially as women age reproductively. Improving reproductive aging to match the increased lifespan currently experienced in the human population is important for fertility and health issues associated with the cessation of reproductive functions in women.

One concern that we still need to address is that many of our studies on human gametes are based on oocytes obtained from gonadotrophin-stimulated women. Although there are studies that suggest that gentle stimulation regimes do not affect aneuploidy rates (Plachot, 2001; Labarta et al., 2012), and therefore, presumably, segregation of chromosomes, other studies have found significant effects of gonadotrophin stimulation regime on aneuploidy (Baart et al., 2007; Rubio et al., 2010). It remains to be determined whether the new findings on meiotic segregation are restricted to IVF oocytes or if they are a universal phenomenon. Nevertheless, as our knowledge about meiosis increases in human oocytes from medically assisted reproduction, it is also possible that clinical material that is currently not used for treatment may become an option for fertility treatment, in case mature oocytes are scarce or stimulation is not possible. Basic research on female meiosis may therefore contribute towards further developments in the in vitro maturation treatment of immature oocytes (Coticchio et al., 2015).

Acknowledgements

We thank our colleagues, especially Dr Xavi Aran, for insightful comments and proof-reading of the manuscript. We apologize to our colleagues, whose work we could not cite due to limitations of space.

Authors’ roles

A.C. and E.R.H. conceived the study and wrote the manuscript. D.C. performed manuscript drafting and data preparation. F.M.U. provided critical discussion. L.R. participated in study design and provided critical discussion of the manuscript.

Funding

No specific funding was used for this study.

Conflict of interest

E.R.H. is funded by a Novo Nordisk Young Investigator Award. E.R.H. receives funding from Illumina Ltd. on which the MeioMapping is based. GENERA and GENETYX provide a PGT-A service to their patients. The remaining authors have no conflicts of interest.

References

Author notes