Aurora B and C kinases regulate chromosome desynapsis and segregation during mouse and human spermatogenesis

The precise control of chromosome dynamics during meiosis is critical to ensure fertility and prevent aneuploidy. Meiosis involves chromosome replication followed by two rounds of chromosome segregation (meiosis I and II), resulting in the formation of up to four haploid gametes. Meiosis I differs from mitosis because homologous chromosomes segregate, whereas sister chromatids remain associated until meiosis II. Crossover recombination events between homologous chromosomes are essential to mediate their bidirectional segregation during meiosis I (Hunter, 2015; Jordan, 2006). To mediate interactions between homologous chromosomes and facilitate completion of homologous recombination, meiotic cells assemble a proteinaceous scaffold termed the synaptonemal complex (SC). The SC not only acts as a bridge between homologs, but it mediates protein interactions and signaling pathways required for meiotic progression (Rog et al., 2017).

The SC is a zipper-like tripartite protein complex comprising two lateral elements (LEs) and a central region. LE components include meiosis-specific axial proteins (SYCP2 and SYCP3) and cohesin complexes, which collectively form a core between each pair of sister chromatids. The LEs of a pair of homologous chromosomes are bridged together by a series of transverse filaments (SYCP1) and central region proteins (SYCE1–SYCE3, TEX12 and sororin) (Cahoon and Hawley, 2016; Gao and Colaiácovo, 2018; Gómez et al., 2016; Jordan et al., 2017). Synapsis and crossover recombination between homologous chromosomes is completed by the pachytene sub-stage of meiotic prophase (Cole et al., 2012; Gao and Colaiácovo, 2018). Upon completion of homologous recombination, spermatocytes are licensed to progress through prophase to the metaphase I (G2/MI) transition (Hochwagen and Amon, 2006). The SC is disassembled in a coordinated manner. Initially, transverse filament and central region proteins are removed during diplonema, allowing homologs to begin disengaging (Cahoon and Hawley, 2016). During diakinesis, LE proteins and cohesins are depleted from the axis but are retained at the kinetochore region (Ishiguro et al., 2011). Final stages of SC disassembly are coordinated with other aspects of the transition to prometaphase/metaphase I, such as chromatin condensation and formation of bivalents (Clemons et al., 2013).

Treatment of mammalian spermatocytes with okadaic acid (OA), a PP1 and PP2 phosphatase inhibitor, stimulates spermatocytes to undergo a number of hallmarks of the G2/MI transition, including SC disassembly and formation of condensed bivalents (Wiltshire et al., 1995). Based on this finding, it was proposed that cell cycle kinases play an important role in regulating SC disassembly. In mice, inhibition of cyclin-dependent kinases (CDKs) prevents LE disassembly (Sun et al., 2010). In addition, polo-like kinase 1 (PLK1) directly phosphorylates SYCP1, TEX12 and SYCE1 proteins, and inhibition of these modifications during an OA-induced prophase exit prevents SC disassembly (Jordan et al., 2012).

Aurora kinases (AURKs) are implicated in regulating SC dynamics during meiosis. In budding yeast there is a single AURK, Ipl1, which promotes efficiency of SC disassembly and integrates chromosome restructuring events with cell cycle progression (Jordan et al., 2009; Newnham et al., 2013). Mammalian spermatocytes, however, express three AURK paralogs (AURKA, AURKB and AURKC) (Tang et al., 2006). Despite a high degree of sequence similarity between the kinase domains, the three mammalian AURKs display unique functions and localizations. AURKA localizes to centrosomes and spindle microtubules, and plays important roles in ensuring bipolar spindle formation (Sugimoto et al., 2002). Both AURKB and AURKC (hereafter denoted collectively as AURKB/C) can function as the catalytic subunit of the chromosome passenger complex (CPC), which is composed of the scaffold inner centromere protein (INCENP), and two regulatory subunits survivin and borealin (Slattery et al., 2008). The CPC localizes to pericentromeric heterochromatin and along chromosome arms beginning at diplonema, and concentrates at kinetochores during diakinesis (Parra et al., 2003). Localization of the CPC in mouse spermatocytes during the G2/MI transition suggests AURKB/C are at the right place at the right time to regulate SC dynamics (Tang et al., 2006). Here, we have used AURKA and AURKB/C inhibitors, mouse mutant models and human spermatocytes to further investigate the roles of these AURK paralogs in regulating SC disassembly.

Following SC disassembly and chromatin condensation, bivalents must become bioriented at metaphase I to ensure accurate segregation of homologous chromosomes. In mitotic cells the CPC ensures chromosome biorientation by destabilizing kinetochore–microtubule attachment errors and activating the spindle assembly checkpoint (Carmena et al., 2012). During mouse spermatogenesis, AURKB, AURKC and other CPC components are enriched at metaphase I kinetochores and are retained there during both meiotic divisions (Parra et al., 2003; Tang et al., 2006). Using mutant alleles of Aurkb and Aurkc, we tested the hypothesis that the CPC is required for biorientation and accurate segregation of chromosomes during spermatogenesis in mice. Furthermore, we examine whether mutation of Aurkb and Aurkc affects spindle midzone and midbody formation, as CPC components also localize to these cell compartments during anaphase and telophase, respectively, during meiosis I and II (Parra et al., 2003; Tang et al., 2006).

Mouse spermatocytes were induced to undergo the G2/MI transition via treatment with OA during a short-term culture. Juvenile mice at 18 days post-partum (dpp), undergoing the semi-synchronous first wave of spermatogenesis, were used to obtain an enriched pool of pachytene-stage spermatocytes. To assess LE disassembly, spermatocytes were monitored by immunolabeling SYCP3 on chromatin spreads following 5 h of culture (Fig. 1A,B). Without addition of OA, the G2/MI transition does not occur in cultured spermatocytes. In contrast, cells treated with OA progressed to diakinesis with LEs disassembled from the chromosome axis, remaining only at kinetochores (Fig. 1A,B). As previously reported, OA treatment also resulted in the loss of centromere pairing between homologs (Fig. S1A) (de Almeida et al., 2019).

To study the role of AURKs on SC disassembly, we used small-molecule inhibitors with various affinities for the three AURK paralogs. Specifically, the pan-AURK inhibitor ZM447439 (ZM) (Ditchfield et al., 2003), the AURKA inhibitor MLN8054 (MLN) (Manfredi et al., 2007) and the AURKB/C inhibitor AZD1152 (AZD) (Yang et al., 2007) were used. Treatment of pachytene-stage spermatocytes with AURK inhibitors alone for 5 h did not induce meiotic progression, SC disassembly or loss of centromeric pairing (Fig. S1A–C). Treatment of spermatocytes with OA in combination with ZM significantly impaired LE disassembly, but did not affect the removal of proteins within the central region of the SC (Fig. 1C,D; Fig. S1D,E), which complements results from a previous study (Sun and Handel, 2008). The same defect in LE disassembly was observed during OA plus AZD treatment, while no defect in desynapsis was observed under OA plus MLN conditions (Fig. 1C,D). In addition to SYCP3, the meiotic-specific cohesin component REC8 was also retained along the axis under OA plus AZD and OA plus ZM treatment (Fig. S1F,G). We then assessed total levels of LE proteins by western blotting and found that both SYCP2 and SYCP3 were modestly higher after OA plus AZD treatment compared to OA treatment at the 5 h time point (Fig. S2A–C). Together, these findings suggest that AURKB/C activity regulates LE disassembly, and that AURKA is not required for this event during the transition from diplonema to diakinesis.

We then extended our analysis of AURK requirements during SC disassembly in mouse to human. Human spermatocytes were isolated via STA-PUT density sedimentation and SC dynamics were monitored by SYCP3 immunolabeling of chromatin spread preparations (Fig. 1E). SC assembly and synapsis occurred normally in human spermatocytes obtained from each donor (Fig. 1E; Fig. S2D). Treatment of human pachytene-stage enriched spermatocytes with OA induced the G2/MI transition (Fig. 1F,G). Human spermatocytes treated with OA and ZM or AZD, but not with MLN, displayed defective LE disassembly, demonstrating that AURKB/C activity is required for LE disassembly in human spermatocytes. Thus, these data indicate that a requirement of AURKB/C for mediating LE disassembly in spermatocytes is conserved between mouse and human.

To further study the roles of AURKB and AURKC during mammalian spermatogenesis, we used germ-cell-specific Aurkb conditional knockout (cKO) and Aurkc knockout (KO) mice (Fernández-Miranda et al., 2011; Kimmins et al., 2007). We first tested the Stra8-Cre transgene to conditionally mutate Aurkb in the male germline. Stra8-Cre is expressed in spermatogonia through preleptotene stage spermatocytes (Sadate-Ngatchou et al., 2008). However, cKO of Aurkb using Stra8-Cre resulted in reduced testis size and weight because of depletion of germ cells prior to meiotic entry (Fig. S3), thereby precluding analysis of spermatocytes. Instead, optimal conditional deletion of Aurkb was driven by Spo11-Cre, which is expressed in spermatocytes shortly after meiotic entry (Lyndaker et al., 2013) and efficiently excised the floxed Aurkb gene (Fig. S4A,B). Therefore, we used the Spo11-Cre transgene for the remainder of our study.

Hematoxylin and eosin (H&E)-stained testis sections of control, Aurkb cKO, and Aurkc KO mice showed no histological aberrancies compared to control mice (Fig. 2A). We analyzed SC dynamics by assessing LE (SYCP3, REC8 and HORMAD1) and central region (sororin) components of the SC, during meiotic progression and did not observe any defects in Aurkb cKO and Aurkc KO mice (Fig. 2B–D; Fig. S4C,D). These results demonstrate that the critical axis restructuring events that occur during the G2/MI transition are not affected by the absence of either AURKB or AURKC.

We then assessed the subcellular localization of all three AURKs (Fig. 3A–C). Neither AURKB kinetochore localization in Aurkc KO nor AURKC localization in Aurkb cKO spermatocytes was altered. Similarly, the expected localization of AURKA to the centrosome and spindle was observed in Aurkb cKO and Aurkc KO mutant spermatocytes. Protection of centromeric sister chromatid cohesion by shugoshin-2 (SGOL2), a known AURKB substrate, during the first meiotic division is critical to ensure sister chromosome mono-orientation and prevent aneuploidy (Llano et al., 2008). In both Aurkb cKO and Aurkc KO spermatocytes, SGOL2 protein was loaded during the first meiotic division, similar to what is seen in controls (Fig. 3D). We also determined that the localization of REC8 was unchanged in Aurkb cKO and Aurkc KO mice relative to control mice (Fig. 3E). Quantification of spindle morphology and chromosome alignment revealed no defects at metaphase I. Spermatocytes from both Aurkb and Aurkc mutant mice progressed to form bipolar spindles with aligned chromosomes at the metaphase plate, suggesting spindle assembly checkpoint function was unperturbed (Fig. 3F,G). Taken together, our results demonstrate that deletion of either AURKB or AURKC does not disrupt spermatogenesis and that these two kinases functionally compensate for one another within the testis.

Because AURKB and AURKC appear to have redundant functions in spermatogenesis (Figs 2 and 3; Fig. S4A–D), we assessed spermatocytes from mice lacking both kinases. Histological analyses of testis sections obtained from adult Aurkb/c double knockout (dKO) mice showed severe disruption of meiotic progression, with accumulation of primary spermatocytes with condensed chromosomes, and an increase of spermatocytes undergoing chromosome segregation (Fig. 4A,B). Assessment of adult testis sections with TUNEL staining showed a significant population of Aurkb/c dKO primary spermatocytes undergoing apoptosis; most of these cells harbored condensed chromosomes including those at metaphase and anaphase stages (Fig. 4C–E). Aurkb/c dKO adult mice had a 1.4-fold reduction in testis weight and a 1.8-fold reduction in sperm density compared to control mice (Fig. 4F,G). Fertility tests showed Aurkb cKO and Aurkc KO mice were fertile and had similar litter sizes compared to controls when mated to wild-type females. In contrast, Aurkb/c dKO male mice produced only 4.4±0.3 (mean±s.e.m.) progeny per litter compared to 7.9±0.5 from control males (Fig. 4H). In sum, absence of both AURKB and AURKC in mouse spermatocytes results in aberrant meiotic progression and reduced fertility.

Owing to the accumulation of late-meiosis I stage spermatocytes observed in adult Aurkb/c dKO testis histology (Fig. 4A–E), we next assessed juvenile mice aged to 33 dpp. Normally, at this age the synchronous first wave of spermatogenesis would have completed meiosis I and II to produce an enriched population of haploid round spermatids (Ahmed and de Rooij, 2009). Immunolabeled cryosections of seminiferous tubules from Aurkb/c dKO mice at 33 dpp revealed a 3.4-fold increase in tubules that contained a buildup of diplotene-stage spermatocytes with linear stretches of SYCP3 and a condensed γH2AX (the phosphorylated form of H2AX)-rich XY sex body, which lacked linear SYCP1 signal (Fig. 5A–C). In addition, Aurkb/c dKO spermatocytes lacked AURKB signal confirming efficient Cre-mediated Aurkb deletion (Fig. S4E). The seminiferous tubules from Aurkb/c dKO mice were also reduced in size by 38%, further suggesting a defect in meiotic progression at 33 dpp (Fig. 5D). To visualize condensed chromatin, we utilized DAPI staining and immunolabeling against the testis-specific linker histone variant H1T, which is incorporated from late pachynema to the round spermatid stage (Drabent et al., 1996). Assessment of chromatin spread preparations demonstrated that Aurkb/c dKO spermatocytes retained linear stretches of SYCP3 on fully condensed, H1T-positive bivalent chromosomes (Fig. 5E). This observation contrasts with control spermatocytes where SYCP3 is only present at kinetochores of condensed bivalents (Fig. 5E). Despite the LE disassembly defect, some Aurkb/c dKO spermatocytes progress to metaphase I. However, axial stretches of SYCP3 were detected in 55±3.2% (mean±s.e.m.) of Aurkb/c dKO metaphase I stage spermatocytes (Fig. 5F,G). These results demonstrate that LE disassembly is perturbed in Aurkb/c dKO spermatocytes.

Pachytene and diplotene-stage spermatocytes isolated from Aurkb/c dKO mice at 20 dpp were treated with OA to determine whether completion of the G2/MI transition could be chemically induced outside the context of the seminiferous tubule. Although an accumulation of diplotene-stage spermatocytes was observed at 33 dpp, at 20 dpp no significant difference in the prophase distribution of spermatocytes was observed between control and Aurkb/c dKO mice (Fig. S4F). Following OA treatment, Aurkb/c dKO spermatocytes failed to efficiently disassemble SYCP3 (Fig. 5H,I). These observations complement the phenotype observed when wild-type spermatocytes were treated with OA and the AURKB/C inhibitor, AZD (Fig. 1). In contrast, Aurkb cKO and Aurkc KO spermatocytes did not display this defect and progressed to diakinesis following OA treatment in a similar manner to control spermatocytes (Fig. 5H,I). Therefore, the enzymatic activity of AURKB and AURKC kinases are required for the efficient removal of LE stretches and the timely completion of desynapsis.

Despite the accumulation of diplotene-stage spermatocytes and abnormal LE disassembly (Fig. 5), some Aurkb/c dKO spermatocytes reached metaphase I and underwent chromosome segregation. However, only 42.5±3.3% (mean±s.e.m.) Aurkb/c dKO spermatocytes were scored to have fully aligned bivalents at metaphase I compared to 83.2±3.1% of control spermatocytes (Fig. 6A,B). The misalignment was not due to mislocalization of AURKA (Fig. 6C) or the kinetochore proteins crucial for sister chromatid mono-orientation during meiosis I, SGOL2 and MEIKIN (Fig. 6D,E; Fig. S5A) (Kim et al., 2015; Llano et al., 2008), or due to premature loss of REC8 cohesins (Fig. 6F; Fig. S5B,C). We also assessed the localization of PLK1, another cell cycle kinase that regulates accurate chromosome segregation (Petronczki et al., 2008). Aurkb/c dKO and control spermatocytes had the same PLK1 localization pattern, where PLK1 localized to kinetochores and centrosomes (Fig. 6G). We assessed a spindle assembly checkpoint (SAC) protein, MAD2 (also known as MAD2L1), which normally localizes to kinetochores during prometaphase, and remains there until ubiquitous bipolar microtubule-kinetochore attachment satisfies the SAC (Lara-Gonzalez et al., 2012). During the prometaphase-to-metaphase I transition, control and Aurkb/c dKO spermatocytes containing misaligned chromosomes accumulate MAD2 foci on all misaligned kinetochores (Fig. 6H–I). When bipolar kinetochore attachment was achieved in control spermatocytes, MAD2 was no longer present. In contrast, the majority of Aurkb/c dKO spermatocytes had persistent misaligned chromosomes that continued to harbor MAD2 foci at their kinetochores (Fig. 6H–J). These results demonstrate that AURKB and AURKC functions are required for the efficient kinetochore–microtubule attachments prior to chromosome segregation in spermatocytes. In addition, the SAC is still active in the Aurkb/c dKO spermatocytes, which is likely responsible for the accumulation of M-phase spermatocytes observed (Fig. 4A,B). Nevertheless, some Aurkb/c dKO spermatocytes eventually progressed through meiosis, likely due to exhaustion of the SAC, and chromosomes segregated, often erroneously.

Given the increased incidence of misaligned chromosomes in Aurkb/c dKO spermatocytes, we next assessed whether these metaphase defects resulted in chromosome missegregation. Indeed, absence of AURKB and AURKC resulted in 52.7±4% (mean±s.e.m.) of spermatocytes containing chromosome missegregation errors during meiosis I and II, compared with only 12.2±2.5% in control spermatocytes (Fig. 7A–C). Furthermore, we observed DNA bridges during anaphase, which may be indicative of an inability to resolve DNA catenenes. It was recently shown that Aurora B is required for metaphase arrest upon topoisomerase II α (TOP2A) inhibition (Pandey et al., 2020). Therefore, we assessed the localization of TOP2A during meiosis I. As previously reported (Gómez et al., 2013), TOP2A primarily localized to kinetochores in primary spermatocytes at metaphase I (Fig. 7D). The localization of TOP2A was unchanged in Aurkb/c dKO spermatocytes compared to controls.

AURKB, AURKC and other CPC components localize to the spindle midzone during anaphase I (Parra et al., 2003; Tang et al., 2006). Therefore, we assessed the localization of a known AURKB substrate, kinesin family member KIF4 (here referring to both KIF4A and KIF4B), which is a plus-end-directed motor that ensures a robust anaphase spindle midzone and prevents over-elongation of the spindle during chromosome segregation (Hu et al., 2011). We determined that initial localization of KIF4 during anaphase I was not altered in the Aurkb/c dKO compared to control (Fig. 7E). In contrast, KIF4 levels were diminished 1.5-fold at the central spindle at cytokinesis in Aurkb/c dKO spermatocytes (Fig. 7E,F). Aberrant KIF4 localization and midzone formation results in the formation of DNA bridges between newly dividing cells (Mazumdar et al., 2004). The defects observed in Aurkb/c dKO spermatocytes align well with a failure of KIF4 localization and function, as we observed DNA bridges during meiosis I and II (Fig. 7A,B).

Despite the inefficient disassembly of the LE, chromosome misalignment, SAC fatigue, chromosome missegregation, and DNA bridges during anaphase (Figs 4–7), some Aurkb/c dKO spermatocytes develop into round spermatids, and even elongated sperm that are capable of fertilization (Fig. 4A,F,G). Therefore, we also characterized post-meiotic cells in control and Aurkb/c dKO mouse testes. We observed a population of round spermatids containing centromeric SYCP3 signal without the formation of a mature acrosome structure, and a later population lacking SYCP3 signal and a fully mature acrosome (Fig. 7G). In Aurkb/c dKO spermatocytes, the population of early SYCP3 positive round spermatids was significantly increased compared to control mice (mean±s.e.m. of 59.4 ±3.3% and 14.8 ±1.6%, respectively) (Fig. 7H). This suggests that a large proportion of round spermatids that form in the Aurkb/c dKO are not capable of further differentiation to develop into mature sperm, which may be a consequence of aneuploidy and genome damage. Collectively, these results demonstrate that AURKB and AURKC overlap in function and are important to ensure accurate chromosome segregation and genome integrity during mammalian spermatogenesis.

The changes in chromosome structure that occur during meiotic prophase revolve around the SC, a highly ordered proteinaceous scaffold that forms between homologous chromosomes. Formation of the SC is integral to meiosis as it ensures accurate pairing and recombination between homologous chromosomes. Following completion of recombination and formation of crossovers, the SC is disassembled, and homologous chromosomes undergo condensation to become bivalents, which are capable of biorientation and segregation during the first meiotic division. For budding yeast, worms and mouse it has been well documented that cell cycle kinases are important for regulating SC disassembly (Cahoon and Hawley, 2016; Gao and Colaiácovo, 2018). By culturing human pachytene stage spermatocytes in the presence of the phosphatase inhibitor OA, we demonstrated that the requirement of AURKB and AURKC activity during SC disassembly is conserved.

In budding yeast, the sole Aurora kinase, Ipl1, is important to coordinate SC disassembly during meiosis. Ipl1 blocks cell cycle progression during early meiotic prophase by suppressing S-phase CDK activity (Newnham et al., 2013). In the context of depleted Ipl1, spindle pole body maturation and separation, as well as cell cycle progression are decoupled and cells erroneously progress to metaphase with linear stretches of SC components still present at the axis (Jordan et al., 2009; Newnham et al., 2013). In mammals, three Aurora kinases are expressed during spermatogenesis, AURKA, AURKB and AURKC (Nigg, 2001). By using small-molecule inhibitors we determined that AURKB and AURKC are required for disassembly of the SC lateral elements in mouse and human spermatocytes. This specificity is an advancement of previous work that showed that pan-inhibition of AURKs in mouse spermatocytes inhibited lateral element disassembly (Sun and Handel, 2008). We used mouse mutants to delineate the functions of AURKB and AURKC and determined that AURKB and AURKC can compensate for one another with regards to SC disassembly and chromosome segregation during meiosis. In the absence of both AURKB and AURKC, however, we demonstrate that SC disassembly is delayed and this defect is resistant to OA treatment. As AURKB and AURKC are likely to have multiple functions during meiosis, including H3 phosphorylation, SC disassembly and chromosome segregation, it is difficult to say how much the Aurkb/c dKO SC disassembly defect contributes to the chromosome segregation defects and spermatocyte apoptosis observed. Future work in understanding whether SC components are direct targets for AURKB/C will help delineate this further. Phosphorylation of SC components during the SC disassembly has previously been reported (Fukuda et al., 2012; Jordan et al., 2012). Interestingly, both SYCP2 and SYCP3 contain putative AURK phosphorylation motifs, and phosphopeptides containing these motifs have been identified in large-scale phosphoproteomic studies from mouse testes (Huttlin et al., 2010). Because AURKB and AURKC both localize to the chromosome arms during SC disassembly, they are present at the right place at the right time to modify SC components. In future work, identification of AURKB/C targets during spermatogenesis will improve our understanding of the meiotic functions Aurora kinases play during gametogenesis.

Depletion of Ipl1 during meiosis in budding yeast revealed that Ipl1 is required for biorientation of homologs during meiosis I and sister chromatids during meiosis II (Meyer et al., 2013; Monje-Casas et al., 2007). As with mitosis (Pereira et al., 2001; Tanaka et al., 2002), it was proposed that Ipl1 is required to promote microtubule attachment turnover until all homologs and sister chromatids are properly attached to microtubules and correctly oriented on the metaphase I and II spindles. Similar to meiotic depletion of Ipl1 in budding yeast (Monje-Casas et al., 2007; Jordan et al., 2009), Aurkb/c dKO spermatocytes undergo chromosome missegregation following a delay in the metaphase I to anaphase I transition. We observed that in the absence of AURKB and AURKC, the SAC protein MAD2 continued to localize to the kinetochores of misaligned metaphase chromosomes, resulting in an accumulation of metaphase I spermatocytes. Several studies of mitosis have suggested that AURKB plays an important role in destabilizing kinetochore–microtubule attachments at sister kinetochores that fail to form amphitellic attachment to microtubules (Biggins and Murray, 2001; Ditchfield et al., 2003; Pinsky et al., 2006). In these instances, AURKB destabilizes the kinetochore–microtubule attachments, subsequently allowing for the chromatids to achieve bipolar attachment. Our observations align well with these studies in that, without AURKB and AURKC, kinetochore attachment errors are unable to be corrected. Nevertheless, although we continue to observe MAD2 at misaligned kinetochores, homologous chromosomes eventually segregate in some Aurkb/c dKO spermatocytes. This may be due to SAC fatigue, which results in the chromosome missegregation we observe in Aurkb/c dKO spermatocytes. On the other hand, the Aurkb/c dKO males are sub-fertile, suggesting that other kinases are capable of eventual correction of these misaligned kinetochores, which warrants further investigation. Indeed, studies have shown that AURKA is capable of correcting kinetochore–microtubule attachments when bivalents migrate to a spindle pole (Chmátal et al., 2015; Ye et al., 2015). We also observe PLK1 at kinetochores in Aurkb/c dKO spermatocytes (Fig. 7G), and, together with MPS1, PLK1 has been shown to regulate the SAC when bound to BUB1 (Dou et al., 2019; Ikeda and Tanaka, 2017; O'Connor et al., 2016; von Schubert et al., 2015). Nevertheless, the anaphase promoting complex/cyclosome (APC/C) is activated and the absence of AURKB and AURKC results in chromosome segregation errors.

In addition to chromosome missegregation, we also observed chromosome bridges in Aurkb/c dKO spermatocytes, which may be due to merotelic attachments. However, we do not see defective localization of MEIKIN, which is crucial for mono-orientation during meiosis I (Kim et al., 2015). Alternatively, the failure to resolve chromosome entanglements in the Aurkb/c dKO spermatocytes may be due to aberrant TOP2A topoisomerase function (Spence et al., 2007). TOP2A is required for AURKB activation at the kinetochore, and relocalization to the spindle midzone (Coelho et al., 2008; Pandey et al., 2020). Although TOP2A has not been identified as an AURKB/C substrate, and we do not see mislocalization of TOP2A in the absence of AURKB and AURKC, it is possible that TOP2A function is influenced by AURKB/C. Alternatively, the reduction in KIF4 levels seen at the spindle midzone and midbody during anaphase and telophase in the Aurkb/c dKO spermatocytes could be the cause of the chromosome bridges observed. Aberrant KIF4 localization and midzone formation results in the formation of DNA bridges between newly dividing cells (Mazumdar et al., 2004).

Prior to our study, it was known that AURKA was expressed in spermatogonia, spermatocytes, spermatids and spermatozoa (Johnson et al., 2018). However, the specific subcellular localization in spermatocytes had not been assessed in detail. We showed that AURKA is enriched at the centrosomes and spindle microtubules during the first meiotic division. We propose that AURKA is required for centrosome biogenesis during mammalian spermatogenesis. However, creation of an Aurka mutant mouse to assess the meiotic roles of AURKA has not been reported and will be an important matter for future work. The function of AURKA may be aligned with the spindle pole body biogenesis functions discovered for Ipl1 during meiosis in budding yeast, which was shown to be important to prevent premature separation of spindle pole bodies during meiosis (Newnham et al., 2013; Shirk et al., 2011).

AURKA appears to predominantly function in the cytoplasm during centrosome maturation, and AURKB and AURKC are required for coordinating SC disassembly, chromosome compaction, and chromosome segregation during spermatogenesis. We show that AURKA cannot compensate for the absence of AURKB and AURKC in spermatocytes. In contrast, AURKA supports meiosis in the absence of AURKB and AURKC in mouse oocytes by localizing to the kinetochores to become a component of the CPC (Nguyen et al., 2018). Furthermore, deletion of Aurkb or Aurkc in oocytes results in abnormal meiosis (Nguyen et al., 2018; Schindler et al., 2012). These sexual dimorphisms may be attributed to the spatiotemporal differences between spermatogenesis and oogenesis. For instance, bipolar spindles are formed by separation of two centrosomes in mammalian spermatocytes, whereas multiple fragmented acentriolar microtubule organizing centers amalgamate to form a bipolar spindle in mammalian oocytes (Clift and Schuh, 2015; So et al., 2019; Solc et al., 2015). The primary function of AURKA is to promote bipolar spindle assembly (Barr and Gergely, 2007; Sardon et al., 2008), and the differences in microtubule organization during male compared to female gametogenesis may explain why AURKA is unable to compensate for the absence of AURKB and AURKC.

Additionally, in contrast to spermatocytes, SYCP3 does not localize to the kinetochores during meiosis I in oocytes. Therefore, another possibility is that SYCP3 prevents localization of AURKA to the kinetochore in spermatocytes.

Thus far, conditional mutation of Aurkb in oocytes has been directed to occur during the dictyate arrest following SC disassembly (Nguyen et al., 2018). Therefore, the role of AURKs during SC disassembly during oogenesis has not been assessed and should be an avenue of future research.

Expression of a kinase inactive AURKB mutant led to impaired spermatogenesis with multinucleated spermatocytes (Kimmins et al., 2007). Similarly, a mutation in human AURKC that generates a kinase inactive truncation results in polyploid sperm (Dieterich et al., 2007). Thus, mutation of the kinase domain in AURKB and AURKC had a greater impact on spermatogenesis than the cKO and KO approaches used here. Both AURKB and AURKC can bind to INCENP and function as the catalytic subunits of the CPC; however, this binding cannot occur simultaneously (Sasai et al., 2016). Mutant forms of either AURKB or AURKC may deplete the pool of active CPC available to a developing spermatocyte. Therefore, mutations that affect the catalytic function of AURKB or AURKC could display a dominant-negative effect, as has been observed in oocytes (Balboula and Schindler, 2014; Fellmeth et al., 2016; Nguyen et al., 2017), and interfere with CPC function.

Our data demonstrate that AURKB and AURKC can functionally compensate for one another during spermatogenesis. AURKB and AURKC ensure coordination of chromosome restructuring events during the G2/MI transition during spermatogenesis, and they maintain an ordered progression through meiosis I and II that is critical for accurate chromosome segregation (Fig. 8). Future work must be directed toward determining Aurora kinase substrates at each stage of gametogenesis. For spermatogenesis, in vivo synchronization of mutant and control mice using the retinoic acid inhibitor WIN 18,446, coupled with phosphopeptide enrichment and quantitative mass spectrometry would be an ideal approach to identify key changes in the phosphoproteome during meiotic progression (Hogarth et al., 2013).

Mice were bred by the investigators at Rutgers University (Piscataway, NJ) and Johns Hopkins University (Baltimore, MD) under standard conditions in accordance with the National Institutes of Health and U.S. Department of Agriculture criteria, and protocols for their care and use were approved by the Institutional Animal Care and Use Committees (IACUC) of Rutgers University and Johns Hopkins University.

Studies involving deidentified donor testes tissues have been reviewed and designated by Johns Hopkins University Bloomberg School of Public Health IRB as ‘not human subjects research’ (IRB No: 00006700).

To deplete AURKB levels in developing spermatocytes, mice harboring a conditional knockout allele of Aurkb (STOCK-Aurkb^tm2.1Mama) were used, and have been described previously (Fernández-Miranda et al., 2011). Conditional mutation was achieved by the addition of a hemizygous Cre recombinase transgene under the control of meiosis specific promoters. In this study, both the promoter for Stra8 [B6.FVB-Tg(Stra8-iCre)1Reb/LguJ], and the promoter for Spo11 [Tg(Spo11-cre)1Rsw] were assessed. Aurkc KO mice (B6;129S5-Aurkc^tm1Lex) were generated by Lexicon Pharmaceuticals and described previously (Kimmins et al., 2007).

Primers used during this study are described in Table S1. PCR conditions were 90°C for 2 min; 30 cycles of 90°C for 20 s; 58°C for 30 s; and 72°C for 1 min.

For genotyping of Aurkc KO mice, the copy number of Neo was quantified by real-time PCR per the manufacturer's protocol. Briefly, tails were digested in 400 μl of lysis buffer [125 mM NaCl, 40 mM Tris-HCl pH 7.5, 50 mM EDTA, pH 8, 1% (v/v) sarkosyl, 5 mM DTT and 50 μM spermidine] with 6 μl of Proteinase K (Sigma #P4850) for 2 h at 65°C. After dilution of 1:30 in water, the lysates were boiled for 5 min to denature Proteinase K. Then, 2 µl diluted DNA were added to each reaction. Primers to detect Neo (F, 5′-CTCCTGCCGAGAAAGTATCCA-3′; R, 5′-GGTCGAATGGGCAGGTAG-3′) were used at a final concentration of 300 nM and primers to detect Csk (for sample normalization; F, 5′-CTGGCCATCCGGTACAGAAT-3′; R, 5′-TGCAGAAGGGAAGGTCTTGCT-3′) were used at a final concentration of 100 nM. The TAMRA-quenched Neo probe (ABI) was conjugated to 6-fluorescein amidite and used at a final concentration of 100 nM and the TAMRA-quenched Csk probe was conjugated to the fluorophore VIC and used at a final concentration of 100 nM. The comparative Ct method was used to calculate the Neo copy number.

Deidentified human organ donor-derived human testes utilized in this study were obtained from three donors who were 43, 23, and 18 years old, respectively.

Mixed mouse germ cell populations were isolated as described previously (Bellvé, 1993; La Salle et al., 2009). Mid-prophase enriched spermatocytes were isolated from 18 dpp mice, undergoing the semi-synchronous first wave of spermatogenesis.

Human mixed germ cell populations were liberated from testis material following two enzymatic digestions, as described previously (Liu et al., 2015; Yao et al., 2017). During the first digestion, seminiferous tubules were isolated by incubation with 2 mg/ml collagenase, and 1 μg/μl DNase I for 15 min in an oscillating (100 rpm) water bath at 34°C. To release germ cells, the seminiferous tubules were then treated with 3 mg/ml collagenase, 2.5 mg/ml hyaluronidase, 2 mg/ml trypsin and 1 μg/μl DNase I for 13 min in an oscillating (100 rpm) water bath at 34°C. Following the final enzymatic treatment, seminiferous tubules were mechanically disrupted using a transfer pipette for 3 min on ice. The germ cell mixture was then centrifuged for 7 min at 500 g, resuspended in 25 ml of 0.5% BSA in KRB, and filtered through a 70 μm nitex mesh to create a single cell suspension.

Enriched primary spermatocytes were isolated as using STA-PUT gravity sedimentation as previously described with minor adjustments (La Salle et al., 2009). A density gradient was created by flowing 550 ml of 4% BSA in KRB and 550 ml of 2% BSA in KRB into the 25 ml of cell suspension in 0.5% BSA in KRB. Cells were sedimented for 3 h prior to elution and fractionation into 12×75 mm glass culture tubes. Aliquots from each fraction were assessed to determine the purity of isolated primary spermatocytes, as identified from cell shape and size. Fractions containing abundant (80% pure) primary spermatocytes were pooled, counted, and centrifuged at 500 g to resuspend them at a cell concentration of 2.5×10⁶ cells/ml.

Both mouse and human spermatocytes were cultured at 32°C in 5% CO₂ in HEPES (25 mM)-buffered MEMα culture medium (Sigma) supplemented with 25 mM NaHCO₃, 5% fetal bovine serum (Atlanta Biologicals), 10 mM sodium lactate, 59 μg/ml penicillin and 100 μg/ml streptomycin. Spermatocytes were stimulated to undergo the G2/MI transition by a 4 μM okadaic acid (OA) (Sigma) treatment for 5 h. To assess the role of Aurora kinases on SC disassembly during an OA-induced G2/MI transition, spermatocytes were treated with the small-molecule inhibitors MLN8054 (Selleck Chemicals), AZD1152 (Sigma), and ZM447439 (Selleck Chemicals) at a concentration of 5 μM.

Mouse and human chromatin spread preparations were performed as previously described (de Vries et al., 2012; Jordan et al., 2012). Table S2 describes the primary antibodies and their dilutions used in this study. Secondary antibodies conjugated to Alexa Fluor 488, 568, or 633 against human, rabbit and mouse IgG (Life Technologies) were used at a 1:500 dilution. Chromatin spreads, and tubule squash preparations were mounted in Vectashield plus DAPI medium (Vector Laboratories).

To quantify prophase I substage distributions in Figs 1, 2, and 4, chromatin spreads were made from at least three biological replicates. In addition, the chromatin spreads were performed in duplicate. During analysis the mean percentage of each substage was determined after counting 100–200 cells per technical replicate.

For histological assessment, mouse testis tissue was fixed in bouins fixative (Ricca Chemical Company) prior to paraffin embedding. Serial sections (5-μm thick) were mounted onto slides and stained with H&E. For cryo-sectioning, testis tissue was embedded in O.C.T. compound (Fisher) and frozen on dry ice. Serial sections (5-μm thick) were mounted onto slides and immunolabeled with primary and secondary antibodies as described above.

Mouse tubule squash preparations were performed as previously described (Wellard et al., 2018). Full Z-stack captured images were utilized to manually identify spindle morphology and chromosome alignment.

Epididymal sperm counts were performed as previously described (Wang, 2002).

Protein was extracted from germ cells using RIPA buffer (Santa Cruz Biotechnology) containing 1× protease inhibitor cocktail (Roche). Protein concentration was calculated using a BCA protein assay kit (Pierce), and 20 μg of protein extract was loaded per lane of a 7.5%, 12% or 4–15% gradient SDS-PAGE gels (Bio-Rad). To detect proteins >100 kDa, a 7.5% gel was used, and proteins <100 kDa were run on a 12% gel. Protein isolated from STA-PUT-enriched pachytene spermatocytes was loaded onto the 4–15% gradient gel. Following protein separation, proteins were transferred to PVDF membranes using a Trans-Blot Turbo Transfer System (Bio-Rad). Primary antibodies and dilution used are presented in Table S2. For detection of primary antibodies, goat anti-mouse and goat anti-rabbit IgG horseradish peroxidase-conjugated antibodies (Invitrogen) were used as secondary antibodies. Antibody signal was detected via treatment with Bio-Rad ECL western blotting substrate and captured using a Syngene XR5 system.

Nuclear spread and tubule squash images were captured using a Zeiss CellObserver Z1 linked to an ORCA-Flash 4.0 CMOS camera (Hamamatsu), and histology images were captured using a Zeiss AxioImager A2 with an AxioCam ERc 5 s (Zeiss) camera. Images were analyzed with the Zeiss ZEN 2012 blue edition image software and Photoshop (Adobe) was used to prepare figure images.

Student's t-tests, as indicated in figure legends, were used to evaluate the differences between groups using GraphPad Prism software.

The authors thank the Washington Regional Transplant Community for their assistance in obtaining deidentified human testis donations for research, and Marcos Malumbres (CNIO) for providing Aurkb^tm1c mice. Additionally, the authors thank the following researchers for generously providing antibodies for use in this study; Tang K. Tang (Aurora C), José Luis Barbero (SGOL2), Yoshi Watanabe (MEIKIN), Mary Ann Handel (H1T), and Susannah Rankin (Sororin). We would also like to thank Edward Culbertson, Anita Ramachandran, Tianlu Ma, Christopher Shults, Alexandra Nguyen, Amanda Gentilello, and Suzanne Quartuccio for their assistance with mouse genotyping and mouse characterization, and Marina Pryzhkova for her assistance with testis cryosectioning.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.248831.reviewer-comments.pdf