Nuclear removal during terminal lens fiber cell differentiation requires CDK1 activity: appropriating mitosis-related nuclear disassembly

The ocular lens and cornea are the only clear tissues in the body. Opacification of the normally clear lens (called a cataract), afflicts 80% of the elderly population and remains the most common cause of blindness worldwide, afflicting over 19 million people (WHO, 1998; http://www.who.int/whr/1998/en/). As such, cataracts contribute significantly to the $139 billion spent annually in the USA alone due to compromised vision (http://preventblindness.org/significant-increase-costs-vision-related-diseases).

Mammalian lenses develop from a surface ectoderm-derived vesicle. The anterior cells of the vesicle differentiate into lens epithelial cells, while the cells that make up the posterior half of the vesicle differentiate into primary fiber cells. A single layer of epithelial cells lines the anterior hemisphere of the lens. Although, initially, all lens epithelial cells proliferate, only a band of epithelial cells slightly anterior to the lens equator undergoes cell division in the mature lens. Posterior to this zone, the epithelial cells begin to differentiate to form secondary fiber cells. These fibers elongate and eventually form the bulk of lens tissue. The lens continues to grow throughout life, such that the original fibers, or oldest cells, occupy the lens center with progressively younger fiber cells found closer to the lens surface. Proper differentiation of fiber cells also involves formation of an organelle-free zone (OFZ) comprising fiber cells from which light-scattering intracellular organelles, including the cell nucleus, are removed (reviewed by Wride, 2011). Failure to form the OFZ results in a cataractous lens. Although denucleation was observed more than a century ago (Rabl, 1899), the molecular mechanisms leading to lens fiber cell denucleation remain poorly understood.

The lysosomal nuclease DNase IIβ (DLAD) is essential for breaking down lens fiber cell DNA and establishing an OFZ and clear lens (Nishimoto et al., 2003; De Maria and Bassnett, 2007; Nakahara et al., 2007). Lysosomal/cytoplasmic DLAD gains access to and destroys chromatin DNA upon disassembly of the nuclear membrane (Bassnett and Mataic, 1997). This is reminiscent of mitosis. In proliferating cells, mitosis requires cyclin-dependent kinase 1 (CDK1) in conjunction with cyclins A or B to phosphorylate nuclear membrane lamins to destabilize the nuclear envelope (reviewed by Nigg, 1993). Activated CDK1 also aids in regulating mitotic chromatin (reviewed by Nigg, 1993; Kotak et al., 2013; Orthwein et al., 2014; Zheng et al., 2014). By contrast, post-mitotic cells rarely exhibit CDK1 and cyclin A/B expression (King et al., 1994; Tommasi and Pfeifer, 1995). However, post-mitotic lens fiber cells contain both CDK1 and cyclin B protein (He et al., 1998), suggesting that these proteins might initiate nuclear envelope disassembly to provide access for DLAD during terminal differentiation.

The CDK inhibitors p57^KIP2 and p27^KIP1 also regulate CDK1 and the G1/S transition-regulating kinase CDK4 (Sherr and Roberts, 1999). Increased synthesis of p57^KIP2 and p27 ^KIP1 characterizes the withdrawal from the cell cycle and the initiation of lens fiber cell differentiation (Zhang et al., 1997,, 1998; Lovicu and McAvoy, 1999; Nagahama et al., 2001; Reza et al., 2007). As differentiation progresses, fiber cells continue to elaborate crystallins: the major fiber cell gene products.

Several lines of evidence suggest that a ubiquitin proteasome system participates in nuclear breakdown. First, poly-ubiquitylated conjugates increase in equatorial epithelial cell nuclei just prior to fiber cell differentiation, and localize to differentiating fiber cell nuclei (Shang et al., 1999). Second, zebrafish containing a mutation in the 26S proteasome gene Psmd6 experience abnormal retention of fiber cell nuclei, as well as a number of cell cycle alterations in the lens epithelium (Imai et al., 2010). Proteosomal degradation of cyclins A and B during mitosis inactivates CDK1, facilitating reformation of the nuclear membrane in daughter cells subsequent to karyokinesis. Cyclins A and B re-accumulate during the G2 phase of the cell cycle to activate CDK1 in preparation for the next mitosis. However, transgenic mice expressing a mutated ubiquitin (K6W-Ubiquitin) in the lens fiber cells accumulated p27^KIP1, decreased phosphorylation of nuclear lamins A and C, retained nuclei within the usual OFZ, delayed synthesis of crystallins, and were cataractous (Caceres et al., 2010). Together, these data suggested that CDK1 is prominent in directing lens fiber cell denucleation.

Here, we tested the hypothesis that, as in mitotic cells, the disassembly of the nuclear envelope in terminally differentiating fiber cells requires CDK1. We suggest that, in contrast with cycling cells, where CDK activators and inhibitors are cyclically regulated, in lens fibers there is a unidirectional pathway in which high levels of p57^KIP2 and p27^KIP1 keep CDK1 inactive in immature fiber cells. However, prior to the formation of the OFZ, diminishing levels of these CDK inhibitors lead to CDK1 activation, lamin phosphorylation, disassembly of the nuclear membrane, entry of DLAD, reorganization of chromatin and destruction of the nucleus. Deletion of Cdk1 from the lens lineage facilitated the testing of this hypothesis.

Although previous studies have documented the presence of CDK1 protein in lens fiber cells (He et al., 1998), the subcellular localization of CDK1 remained unknown. As expected, the lens epithelium expressed abundant CDK1 and much of the enzyme appeared to be localized to nuclei in epithelial and outer cortical fiber cells (Fig. 1, zones 1, 2). In secondary lens fiber cells, the overall level of CDK1 expression declined as development advanced (compare right with left side, lower panels). Although CDK1 was obvious in both the cytoplasm and nuclei of elongating cells (Fig. 1, zone 2), it remained most concentrated in the nuclei of the deeper fiber cells (Fig. 1, zones 3, 4).

Transgenic mice homozygous for the loxP-flanked (floxed) allele of Cdk1 and hemizygous for the MLR10 Cre transgene (MLR10; Cdk1^L/L) were created to remove CDK1 from the lens. Cre expression in MLR10 transgenic mice initiates at E10.5 and this transgene can effectively delete loxP-flanked alleles in both lens epithelial cells and lens fiber cells (Zhao et al., 2004). In the MLR10; Cdk1^L/L lenses, the overall expression of CDK1 protein became mosaic by E15.5 (supplementary material Fig. S1D-F) with few CDK1-positive epithelial cells remaining by E17.5 (Fig. 2B,D, white arrows). By comparison with expression in lens, MLR10; Cdk1^L/L retinas displayed no alterations in CDK1 expression relative to control littermates (Fig. 2E, supplementary material Fig. S1F), indicating that the Cre transgene properly targeted the lens without affecting other tissues within the optic cup. CDK1 protein persisted in postnatal epithelial cells and fiber cells from both control lenses. Western blots corroborated the diminution of CDK1 at E18.5 in MLR10; Cdk1^L/L lenses (Fig. 2F). The remaining protein indicates that some epithelial cells escape Cre-mediated deletion or, alternatively, represents persistent CDK1 protein produced from transcripts that existed prior to the deletion of Cdk1.

To isolate Cdk1 deletion to lens fiber cells, MLR39 transgenic mice were bred to Cdk1^L/L animals to generate MLR39; Cdk1^L/L mice. Cre expression in MLR39 mice initiates at embryonic day 12.5 (E12.5) and, within the lens, remains exclusively in the fiber cell compartment (Zhao et al., 2004). Immunohistochemical and western blot comparisons between the lenses from MLR39; Cdk1^L/L mice and those of Cre negative control littermates (Cdk1^L/L) failed to reveal significant diminution of CDK1 in fiber cells from the P0 MLR39; Cdk1^L/L mice (supplementary material Fig. S1A-C,G). In addition, MLR39; Cdk1^L/L lenses remained clear and appeared histologically identical to control lenses (data not shown). The persistence of CDK1 protein in the lens fibers indicated that the MLR39 transgene failed to delete the Cdk1 gene early enough to significantly reduce CDK1 protein in the fiber cell compartment. Therefore, all subsequent analyses employed MLR10; Cdk1^L/L mice.

The gross morphology of MLR10; Cdk1^L/L lenses appeared similar to control lenses at E12.5, prior to fiber cell denucleation, (Fig. 3A,B). However, by E15.5, the nuclei in the differentiating secondary fibers at the bow region of MLR10; Cdk1^L/L lenses appeared 152% larger by cross-sectional area, relative to those of control littermates (Fig. 3, compare the size of nuclei in the dashed ovals in D with those in C).

Mouse fiber cells normally begin losing their nuclei at approximately E16-18 (Kuwabara and Imaizumi, 1974). Consistent with this observation, lenses with intact Cdk1 excluded nuclei from central fibers from E17.5 onwards (Fig. 3E, dashed oval, E′, zone 6). By contrast, E17.5 MLR10; Cdk1^L/L lenses retained nuclei in the center of the lens, resulting in a failure to form an OFZ (Fig. 3F,F′, see persistent nuclei in zone 6). Bassnett and others documented descriptive criteria for nuclei during the lens fiber cell differentiation process. We used these criteria to distinguish the nuclei in E17.5 MLR10; Cdk1^L/L and control lenses (Bassnett, 2009). The youngest secondary fiber cells, near the lens equator in zone 1, possessed densely packed elongated oval nuclei in control lenses (Fig. 3E,E′, zone 1). In zone 1 of the MLR10; Cdk1^L/L lens, the fiber cell nuclei were also oval, but fewer in number and their average cross-sectional area was 172% larger than those of control lenses (Fig. 3, compare zone 1 in F′ with E′). From E17.5 onwards, distinctly fewer cells appeared to undergo secondary fiber cell differentiation in the MLR10; Cdk1^L/L lenses.

In zone 2, the nuclei in the fibers of control lenses were more sparse and less elongated than in zone 1 (Fig. 3E, compares zone 1 and 2). CDK1-deficient lenses also contained relatively fewer nuclei in zone 2 but they appeared considerably larger than the comparable region in the control lens (Fig. 3F′,E′, compare zones 2) and virtually unchanged in size from zone 1 (Fig. 3F′, compare zones 1 and 2). In zone 3, the control fiber nuclei assumed a smaller, more-rounded shape (compared to zones 1 and 2) and also stained more darkly with Hematoxylin (suggestive of nuclear condensation). The CDK1-deficient fibers of zone 3 contained a mixture of large, elongated, more-rounded nuclei that appeared similar, in size and shape, to the zone 3 nuclei of control fiber cells (Fig. 3E′,F′, compare zones 3). Sparsely packed spherical nuclei, similar in size to those in zone 3, persisted in the control lenses in zones 4 and 5 (Fig. 3E′, zones 4, 5). The MLR10; Cdk1^L/L fiber nuclei appeared fewer in number with more variable size in zones 4 and 5 (compared with the control) and this pattern persisted into zone 6 (Fig. 3F′). By contrast, zone 6 of the control lens was devoid of nuclei (Fig. 3E′, zone 6).

The entry of DLAD into the nuclear compartment requires the disassembly of the nuclear membrane, which normally occurs as fiber cells approach the OFZ. Analogous to mitotic events, we postulated that the nuclear membrane might remain intact in fiber cells lacking CDK1, thus preventing the entry of DLAD. CDK activity is controlled by cyclin-dependent kinase inhibitors. As lens epithelial cells begin to differentiate, they express p27^KIP1 and p57^KIP2, and both of these CDK inhibitors cooperate in the initiation of cell cycle withdrawal during embryonic fiber cell differentiation (Zhang et al., 1998; Lovicu and McAvoy, 1999; Nagahama et al., 2001; Kase et al., 2005). In keeping with these reports, both control and MLR10; Cdk1^L/L lens epithelial cells express p57^KIP2 prior to reaching the lens equator; however, expression in both genotypes abruptly ends as the cells begin to detach from the lens capsule (Fig. 4, compare bracketed region in A and D). By comparison, the expression of p27^KIP1 extends further into the fiber cell mass than p57^KIP2 in control lenses (Fig. 4B, bracketed region), and becomes undetectable just prior to nuclear degradation in the OFZ. Additional coordinated events that occur prior to the formation of the OFZ in control lenses include the nuclear concentration of CDK1 (Fig. 1, zones 3, 4), the adoption of a spherical nuclear morphology (Fig. 3E′, zones 4, 5), the disappearance of p27^KIP1 (Fig. 4B), and the phosphorylation of nuclear lamins A and C (Fig. 4C). In CDK1-deficient lenses, p57^KIP2 and p27^KIP1 are also observed in the outer regions, but p27^KIP1 expression is below detection limits in the inner fiber cell mass (Fig. 4, compare E with B). In addition, CDK1-deficient nuclei fail to exhibit phosphorylation of Lamin A/C (Fig. 4, compare F with C).

Concurrent with the onset of phosphorylation of lamin A/C in control lenses, DLAD moves from the periphery of the nucleus to within the nucleus (Fig. 4H). As indicated in the top inset panel, there is little DLAD evident in early nucleated fiber cells. By contrast, as indicated in the bottom inset panel, uniform DLAD staining is observed throughout the nuclei just prior to the formation of the OFZ. The entrance of DLAD in control lenses anticipates the DNA degradation as shown by abundant TUNEL-positive foci in the developing OFZ (Fig. 4I, yellow staining, arrows). Although DLAD generates 3′-phosphoryl/5′-hydroxyl ends following endonucleic cleavage of DNA (Shiokawa and Tanuma, 1999), endogenous phosphatases rapidly convert the 3′-PO₄⁻ ends to 3′-OH ends that can be labeled by the TUNEL assay (De Maria and Bassnett, 2007). By contrast, in MLR10; Cdk1^L/L lenses, DLAD accumulated around the periphery of central fiber cell nuclei, rather than entering the nucleus, as seen in the control lenses (Fig. 4, compare the lower inset of K with the lower inset of H), and fewer fiber cell nuclei demonstrate degradation of DNA (Fig. 4, compare yellow nuclei in L to I).

The nuclear mitotic apparatus protein (NuMA) is mechanistically involved in chromosome segregation that precedes nuclear disassembly and mitosis (Gribbon et al., 2002; Abad et al., 2007; Kotak and Gonczy, 2014). During the metaphase-anaphase transition, CDK1-induced phosphorylation on threonine 2055 (T2055) results in movement of NuMA from the cell membrane to the spindle poles, resulting in chromosome segregation (Kotak et al., 2013). Western analysis indicated the presence of pNuMA in control lenses (Fig. 5C). Immunohistochemical analyses confirmed the presence of pNuMA in control, Cdk1^L/L, lenses (Fig. 5A1-3). In these lenses, pNuMA is observed distributed or as multiple foci throughout the entire nuclei of epithelial cells and early differentiating fiber cells (Fig. 5A1). Consistent with a role for pNuMA in organizing chromatin, fewer prominent pNuMA puncta are observed in more differentiated fibers (Fig. 5A2) of the control lens and this is echoed by the punctate pattern of chromatin staining (Fig. 3E1-3 and Fig. 5A1-3; supplementary material Fig. S2B, specifically #2-4). Strikingly, however, rather than coalescing to a few or to two prominent foci as they do in mitosis, in Cdk1^L/L control mouse lenses they appear to coalesce to a single large focus in the denucleating cells (Fig. 5A3). By comparison, pNuMA was present at considerably lower levels in western blots in MLR10; Cdk1^L/L lenses and it was below the limit of immunofluorescent detection in the CDK1-deficient lens fiber cells (Fig. 5B1-3). Some pNuMA was observed in a few epithelial cells in these lenses (Fig. 5B). Consistent with an absence of pNuMA foci in the MLR10; Cdk1^L/L lens, the nuclei remain larger and the chromatin remains heterogeneously spread or less focused throughout the nucleus (Fig. 3F1 and Fig. 5B1-3; supplementary material Fig. S2C,D). Interestingly, total NuMA was also decreased in the CDK1-deficient lens (supplementary material Fig. S4).

Despite retaining fiber cell nuclei, CDK1-deficient lenses removed mitochondria from central fiber cells, as indicated by diminished Tom20 (an outer mitochondrial membrane protein) immunoreactivity in fiber cells (Fig. 6, asterisks in A,D). Some mitochondria remained in outer fibers but disappeared in the central fiber cells in both genotypes (Fig. 6, white arrows in A,D). Immunostaining for the endoplasmic reticulum marker protein disulfide isomerase, PDI, also appeared similar between the MLR10;Cdk1^L/L and the control mice (Fig. 6C,F), and no endoplasmic reticulum is indicated in mature central fiber cells (Fig. 6, inside dotted border C,F). This contrasts with the prolonged expression of PDI in lenses from dominant-negative Ncoa6 transgenic mice that also exhibit retention of nuclei in lens fiber cells (Wang et al., 2010). Thus, although mitochondria (Fig. 6A,D) and endoplasmic reticulum (Fig. 6C,F) were destroyed in both genotypes, only MLR10; Cdk1^L/L lenses retained nuclei in central fiber cells (Fig. 6E, yellow arrowheads), consistent with the hypothesis that CDK1 deficiency specifically inhibits denucleation rather than causing a generalized inhibition of organelle destruction.

Although the number and density of lens epithelial cell nuclei were similar in the control and MLR10; Cdk1^L/L lenses at E12.5 (Fig. 7E), by E15.5 there were fewer total epithelial cell nuclei, and fewer epithelial cell nuclei per unit area, in MLR10; Cdk1^L/L lenses than in control lenses (Fig. 7, compare B with A,D and C,E). There was, however, no increase in epithelial cell apoptosis in CDK1-deficient lenses at E15.5 or E17.5 (Fig. 4I,L; supplementary material Fig. S3B-D). Therefore, decreased cell survival fails to explain the loss of epithelial cell population density in the MLR10; Cdk1^L/L lens.

As CDK1 plays an essential role in cell cycle regulation, assays for BrdU incorporation (an S-phase indicator) and phosphorylated histone H3 (pHH3) (a marker of late G2 phase, immediately prior to mitosis) were used to assess the cell cycle in MLR10; Cdk1^L/L lenses. MLR10; Cdk1^L/L lenses exhibited an S-phase index (proportion of BrdU positive nuclei) similar to that of control lenses at both E12.5 and E15.5 (Fig. 8, compare B with A, solid bars in I). However, by E17.5, epithelial cells in the MLR10; Cdk1^L/L lens displayed a significantly higher S-phase index, despite the decreased cell density of the epithelial cell layer. Likewise, starting at E15.5 and continuing to at least E17.5, a significantly higher percentage of lens epithelial cells in MLR10; Cdk1^L/L mice exhibit pHH3 immunoreactivity than in age-matched control mice (Fig. 8, compare F with E, H with G, lightly shaded bars in I). As MLR10; Cdk1^L/L lens epithelial cells exhibited both BrdU incorporation and pHH3 expression, the cell cycle appeared to be active in these epithelial cells. However, the widespread expression of pHH3 in MLR10; Cdk1^L/L lens epithelial cells suggested that although the mutant lens nuclei prepare to leave G2, they may not actually enter mitosis. Consistent with this hypothesis, we observe a 265% increase in nuclear cross-sectional area in comparison with control lens epithelial cell nuclei (Fig. 8, compare K with J,L). The reduced number of lens epithelial cells coupled with the increased nuclear size suggests that CDK1-deficient lens epithelial cells bypass mitosis and simply undergo endoreduplication of their DNA during the cell cycle. This also explains the reduced number of, and larger, nuclei in the differentiating secondary fiber cells of MLR10; Cdk1^L/L lenses.

In mammals, destruction or extrusion of nuclei occurs as a normal event during differentiation only in erythrocytes, keratinocytes and lens fiber cells. Of these, only lens fiber cells destroy nuclei within the cell. Despite many investigations over the past century, the molecular mechanism(s) by which fiber cell denucleation occurs has remained a mystery (Vrensen et al., 1991, 2004; He et al., 1998, 2010; Ivanov et al., 2005; Xie et al., 2007; Rivera et al., 2009; Wang et al., 2010; Gupta et al., 2011; Ma et al., 2011; Jarrin et al., 2012; Rodrigues et al., 2013). Preliminary work suggested that the lens ‘appropriates’ normal mitotic mechanisms in order to accomplish denucleation: specifically, that nuclear membrane disassembly occurs after phosphorylation of nuclear lamins and that stabilization of the CDK1 inhibitor p27^KIP1 delays denucleation (Caceres et al., 2010). Furthermore, the persistence of CDK1 and its activator cyclin B and entry of DLAD, while p27^KIP1 levels decline (He et al., 1998) (Figs 1 and 4), as well as observations of delayed denucleation in DLAD^−/− mice (Nishimoto et al., 2003), all suggested that CDK1 directs fiber cell denucleation. In proof of principle experiments, here we show that the phosphorylation of lamin, entry of DLAD in to the nuclear compartment and denucleation per se require CDK1, thus elucidating upstream events leading towards lens fiber denucleation. Focalization of pNuMA appears to be a consequence of CDK1 activity.

As germline deletion of Cdk1 results in pre-implantation lethality (Santamaria et al., 2007), the MLR10 CRE transgene was exploited to remove Cdk1 specifically from the lens (Fig. 2). Whereas normal fiber cells exhibited phosphorylation of the known CDK1 substrate lamin A/C in the region just prior to fiber cell denucleation, it remained unphosphorylated in CDK1-deficient fiber cells. Furthermore, in the absence of CDK1, DLAD remained outside the fiber cell nuclei (Fig. 4K, lower inset) and these cells failed to denucleate (Fig. 3D,F,F′). Taken together, these findings indicate that fiber denucleation requires CDK1 activity. The specificity of the requirement of CDK1 for removal of nuclei per se is implied by observations that breakdown of the mitochondria and endoplasmic reticulum occurred on schedule in MLR10; Cdk1^L/L lens fibers (Fig. 6).

NuMA is a CDK1-dependent regulator of mitosis. Although elucidation of regulation of its function remains in progress, it is clear that, during mitosis, CDK1-dependent phosphorylation causes NuMA to concentrate at spindle poles and to induce a redistribution of dynein that results in chromosome segregation and eventually the division of the nucleus (Lelievre et al., 1998; Abad et al., 2007; Kotak et al., 2013; Zheng et al., 2014). Data in Fig. 5 indicate that NuMA is also involved in the chromatin organization that precedes lens fiber denucleation (Fig. 5A1-3; supplementary material Fig. S2). Interestingly, whereas in mitotic cells two foci are formed, in denucleating lens cells only a single large pNuMA focus is observed in these wild-type mice. By contrast, NuMA T2055 remained largely unphophorylated in CDK1-deficient fiber cells, and chromatin was less consistently organized (Fig. 5B1-3; supplementary material Fig. S2). Together, the data suggest that lens fiber cells appropriate from normal mitosis the crucial function of CDK1-driven phosphorylation of lamin and NuMA in order to direct lens cell denucleation and development. Whereas in mitosis CDK1 drives the coordinated disassembly of the nuclear membrane and organization of chromatin to allow for formation of daughter cells, in lens fibers chromatin and nuclei are destroyed. In addition to directing phosphorylation and localization of NuMA, CDK1 seems to regulate the levels of the native protein and this may be related to the levels of the pNuMA. The relationship between CDK1 and NuMA is consistent with a feedback mechanism. Elucidation of additional steps in regulation of these events during lens development will be the topic of future investigations.

Interestingly, MLR10; Cdk1^L/L lenses contained fewer nuclei in both the epithelium (Fig. 7) and fiber cells (Figs 3 and 4) than controls, despite exhibiting retention of fiber cell nuclei in the putative OFZ. The reduction in total fiber cells and nuclei in the MLR10; Cdk1^L/L lens appears to result from a reduction in epithelial cells required to fuel continued secondary fiber cell differentiation (Figs 2 and 3). Furthermore, the loss of lens epithelial cells in CDK1-deficient lenses occurred without detectable increases in apoptosis, suggesting that CDK1 is dispensable for lens cell survival (Fig. 4; supplementary material Fig. S3). In addition, in MLR10; Cdk1^L/L lenses, there was no extensive migration of the lens posterior epithelium, as observed in lens fibers in which DNA damage repair is compromised (Wang et al., 2010). If it is assumed that apoptosis is a consequence of DNA damage, these data suggest CDK1-driven entry of DLAD into the nucleus, and the initial disassembly of DNA, operate upstream of the requirements for retention of DNA integrity.

Another remarkable feature of MLR10; Cdk1^L/L lenses is the disparate size of primary versus secondary fiber cell nuclei (Fig. 3D,F versus 3C,E). Nuclei within inner fiber cells of both genotypes (including nuclei of what should be the OFZ in MLR10; Cdk1^L/L) are of similar dimensions. (Fig. 3, compare zones 4-6 in F′ with zones 4 and 5 in E′). However, nuclei in outer fibers of CDK1-deficient lenses, though fewer in number, appear distinctly larger (Fig. 3, compare zones 1 and 2 in F′ with zones 1 and 2 in E′; Fig. 4, compare the p57^KIP2-positive nuclei in D with those in A). We posit that this difference in the size of primary versus secondary fiber cell nuclei in MLR10; Cdk1^L/L lenses results from a difference in the number of genomic duplications experienced by the two different populations of precursor cells, as well as poorer organization due to limited NuMA, as noted above. As the deletion of the floxed Cdk1 allele commences at E10.5, the older, primary fiber cells would largely have been in the process of withdrawing from the cell cycle before the knock down of CDK1 was taking effect. By contrast, the future secondary fiber cells would still be epithelial cells at E10.5 and would likely go through one or more rounds of DNA synthesis before withdrawing from the cell cycle. The large secondary fiber cell nuclei in MLR10; Cdk1^L/L lenses precisely match the phenotype expected if lens epithelial cells underwent endoreduplication of DNA without mitosis in the absence of CDK1 prior to differentiation. This is supported by the higher proliferation index of CDK1-deficient lens epithelial cells (Fig. 8). Furthermore, the higher proportion of MLR10; Cdk1^L/L lens epithelial cells in S-phase or G2 phase and that are enlarged (Figs 7 and 8) is consistent with previous studies documenting the requirement of CDK1 for nuclear disassembly in mitosis and meiosis during development (Adhikari et al., 2012). Likewise, Cdk1 null pre-implantation mouse embryos that reach the blastocyst stage and mouse embryonic fibroblasts conditionally deleted for Cdk1 exhibit a reduced number of (abnormally large) nuclei (Santamaria et al., 2007; Diril et al., 2012).

In conclusion, this discovery of a requirement for CDK1 activity for a terminal differentiation pathway, including removal of nuclei and establishing an OFZ, expands the known functions for this protein beyond those for mitosis and meiosis. In the lens, CDK1 deficiency fails to induce apoptosis or prevent the onset of secondary fiber cell differentiation. The fundamental process of nuclear disassembly apparently requires lamin phosphorylation by CDK1 and includes NuMA-related chromatin organization. The finding that these processes can occur independently of cell division implies that CDK1 may play important roles in other aspects of nuclear function. There are several disease-related laminopathies, including Emery-Dreifuss muscular dystrophy (EDMD), dilated cardiomyopathy (DCM), limb-girdle muscular dystrophy and Hutchinson-Gilford progeria syndrome, most of which profoundly affect non-proliferating cells (reviewed by Ho and Lammerding, 2012). This work suggests that CDK1, and perhaps other regulators of nuclear structure during mitosis, may play an unappreciated role in terminally differentiated cells.

Mice were used in accordance with the ARVO statement for the Use of Animals in Ophthalmic and Visual Research. Transgenic mice expressing CRE in the lens fiber cells (MLR39) and in the entire lens (MLR10) have been described previously (Zhao et al., 2004). The Cdk1^L/L mice were generated using a conditional targeting vector assembled on a pBluescript II KS(+) backbone (Stratagene). The targeted region of Cdk1 contained exon 3 originally amplified with Cdk1 primers containing engineered restriction sites (underlined) (forward, CGG GGT ACC TAG ATA GCT AGG GAA TCC GCA CCT GA; reverse, GCG TCC GGA GGC AGC TAC CAG AGG TGC TAA GTA AG) with flanking LoxP sites. The Cdk1 5′ arm contained exon 2, whereas the 3′ arm contained exons 4 and 5. The neomycin gene flanked by FRT sites was inserted into intron 2 as a selectable marker. Transfection of linearized pBluescript, screening of targeted TC-1 embryonic stem (ES) cells and injection of blastocysts to produce chimeric males were performed as described previously (Dix et al., 1996). Agouti male chimera offspring were mated with C57BL/6NCrl females and then backcrossed onto the 129S genetic background. Experimental mice were maintained on a mixed genetic background segregating for alleles originating in FVB/N, 129S and C57BL/6NCrl strains.

The gestational age of experimental embryos was determined by vaginal plug detection, set at embryonic day 0.5 (E0.5). One hour prior to embryo collection, pregnant dams were administered (0.1 mg/g body weight) 5-bromo-2′-deoxyuridine (BrdU) dissolved in phosphate-buffered saline (PBS) at a concentration of 100 mg/ml. For paraffin wax-embedded sections, embryos were collected and fixed in 10% neutral buffered formalin (NBF). Standard protocols were used to process and embed tissues in paraffin wax before sectioning at 5 µm. For frozen sections, lenses were fixed in 4% neutral buffered paraformaldehyde for 90 min at 4°C, embedded in OCT, frozen and sectioned at 10 µm. Cryosections were permeablized in 0.05% Triton X-100/PBS for 2 min, blocked in 5% donkey serum and 5% BSA in PBS for 30 min at room temperature before being incubated with DLAD, phosphorylated lamin A/C or PDI antibodies. The primary antibody for DLAD was generated as previously described (Nakahara et al., 2007) and used at a dilution of 1:500 overnight at 4°C. The secondary antibody used for visualization of DLAD was conjugated goat anti-hamster (Jackson ImmunoResearch, 127-035-160; 1:250). Detection of DLAD was via DAB peroxidase substrate kit (Vector Labs, SK4100). Images were collected using an Olympus light microscope BX51. Primary antibodies for phosphorylated NuMA (at threonine 2055) have been described previously (Kotak et al., 2013). Primary antibodies for phosphorylated lamin AC, BrdU, p57^KIP2 and CDK1 (ab58528, ab6326, ab4058, and ab7953, respectively) were obtained from Abcam. The primary antibody for p27^KIP1 (BD610241) was obtained from BD Biosciences. Primary antibodies for Tom20 (sc-11415) and total NuMA (sc-48773) were obtained from Santa Cruz Biotechnology, whereas antibodies for phosphorylated histone H3 (Ser10) and PDI were obtained from Millipore (16-189) and Sigma-Aldrich (P7122), respectively. All primary and secondary antibodies were used at a 1:100 dilution, with the exception of DLAD (noted above), total NuMA and phosphorylated histone H3, which were used at a 1:50 dilution. Primary antibodies were detected using secondary antibodies attached to fluorescent probes (Alexa Fluor 488 goat anti-rabbit IgG, Alexa Fluor 546 goat anti-rat IgG, FITC for donkey anti-rabbit IgG, 711-095-152 and Cy3 for donkey anti-mouse IgG). Sections were counterstained with DAPI (Vector Labs, H-1200). Photomicrographs were captured on a Zeiss 710 Laser Scanning Confocal System at the Center for Advanced Microscopy and Imaging at Miami University. Standard Hematoxylin and Eosin-stained sections were used to analyze the structure of the lens, and images were captured using a Nikon TI-80 microscope.

Quantifying indirect immunofluorescent labeling on tissue sections has been previously described (Garcia et al., 2011; Plageman et al., 2011; Madakashira et al., 2012). All immunofluorescent assays were photographed with identical exposure times. ImageJ software was used to measure the signal intensity of the pixels (RGB) and given areas. ImageJ software allowed for the selection of the plot to be measured. The BrdU and TUNEL index represented the ratio of the lens cell nuclei positive for the mentioned markers over the total DAPI stained nuclei in the ocular lens.

Lenses from MLR10; Cdk1^L/L and Cdk1^L/L mice at embryonic day 17.5 were immediately fixed in 10% NBF for 1 h. After fixation, the lenses were washed in PBS and stained with DAPI. The lenses were then placed in between two coverslips with a drop of PBS and a series of images were collected at varying depths using the Zeiss 710 Laser Scanning Confocal System, and finally reconstructed into a three-dimensional image. The cross-sectional area of each individual nucleus was determined using IMAGEPRO software at the Center for Advanced Microscopy and Imaging at Miami University.

Cdk1^L/L, MLR39; Cdk1^L/L and MLR10; Cdk1^L/L lenses were taken at birth. The epithelial cell layer and fiber cell mass were physically separated in Cdk1^L/L and MLR39; Cdk1^L/L lenses, and homogenized in RIPA buffer [50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS] with phosphatase and protease inhibitors (Pierce, catalog number 78440). The protein concentration was determined by BCA assay (Pierce, catalog number 23227). Protein lysates were separated on a 10% SDS-polyacrylamide gels and transferred to PVDF membranes (Millipore, catalog number IPVH10100), blocked with 5% non-fat dry milk for 1 h at room temperature and incubated overnight at 4°C with antibodies to CDK1 (1:2000, Abcam, ab7953) or pNuMA (1:1000; Kotak et al., 2013). After incubation with HRP-conjugated secondary antibody (1:1000; Cell Signaling Technology, 7074) for 2 h, the proteins were analyzed on X-ray films following the addition of the chemiluminescent substrate Lumiglo (Cell Signaling Technology, 7003).

We thank Dr Pierre Gönczy for the gift of the pNUMA antibody; Devin G. Bruney for technical assistance; Molly M. Schleicher, Sheldon Rowan and Adam S. LeFever for critical review of the manuscript; and Ales Cvekl for guidance.