Genetic substitution of Cdk1 by Cdk2 leads to embryonic lethality and loss of meiotic function of Cdk2

Cyclin-dependent kinases (Cdks) together with cyclins constitute the central components of the cell cycle machinery(Morgan, 1997). In eukaryotic cells, several Cdk-cyclin complexes, including Cdk2-cyclin E, Cdk1-cyclin B,Cdk4-cyclin D, Cdk6-cyclin D and Cdk2-cyclin A, drive cell cycle progression,and it was believed that their functions are confined to specific stages of the cell cycle (Morgan, 1997). For example, Cdk4 and Cdk6 are thought to be involved in early G1, whereas Cdk2 is essential to complete G1 and initiate S-phase. Cdk4 and Cdk6 form active complexes with D-type cyclins to initiate the cell division cycle by phosphorylating the Retinoblastoma protein (Rb). Thereafter, in late G1 phase,the activation of Cdk2 by cyclin E further phosphorylates Rb and drives cells through the G1-S restriction point. Later, Cdk2 complexes with cyclin A and is essential for S-phase progression (for a review, see Mittnacht, 1998; Weinberg, 1995). Recently, it was reported that Cdk2 can also interact with the mitotic cyclin, cyclin B,but the exact function of this complex is not known(Aleem et al., 2005). In addition, it has been reported that Cdk2 is present predominantly in the nucleus throughout the cell cycle (Moore et al., 1999; Pines and Hunter, 1991; Satyanarayana et al., 2008). Cdk1 (Cdc2a - Mouse Genome Informatics), in association with cyclin B, is essential to control entry into and exit from mitosis; Cdk1 is present mainly in the cytoplasm and translocates to the nucleus only during mitosis after complexing with cyclin B(Dunphy et al., 1988; Izumi and Maller, 1993; Pan et al., 1993; Riabowol et al., 1989).

In contrast to mammalian cells, in budding yeast a single Cdk, the transcriptional product of the CDC28 gene, regulates diverse cell cycle transitions by associating with multiple stage-specific cyclins(Nasmyth, 1993; Reed et al., 1982). On the basis of the concepts derived from the yeast cell cycle, it was hypothesized that the functions of the multiple Cdks in eukaryotic cells are redundant and one or two Cdks might be sufficient to drive cells through the different phases of the cell cycle. In support of this, recent studies have demonstrated that Cdk2, Cdk4 and Cdk6 single-knockout mice are viable, do not show any severe phenotypes and display minor defects in cell cycle properties, indicating functional redundancy between the different Cdks(Berthet et al., 2003; Malumbres et al., 2004; Ortega et al., 2003). Notably,Cdk1, which was originally identified as an essential mitosis-promoting kinase, can compensate for the loss of Cdk2 by complexing with cyclin E to drive cells through the G1-S transition, even though Cdk1 is only ∼65%identical to Cdk2 (Aleem et al.,2005). In addition, a recent study has demonstrated that Cdk1 alone is sufficient to drive the eukaryotic cell cycle in early embryogenesis and in mouse embryonic fibroblasts (MEFs)(Santamaria et al., 2007). However, at the whole-organism level, the compensation of Cdk2 function by Cdk1 appears to be only partial, as Cdk2 knockout males and females are sterile, displaying dysfunctional and atrophic testes and ovaries(Berthet et al., 2003; Ortega et al., 2003). This indicated that Cdk2 is essential for meiosis and that Cdk1 cannot functionally compensate for the loss of Cdk2. In this context, it is of interest to explore whether there are any possible ways in which Cdk2 might compensate for the loss of Cdk1. Deletion of Cdk1 or a gene-trap mutation in the Cdk1 gene leads to early embryonic lethality (our unpublished results) (Santamaria et al.,2007), indicating that Cdk1 is essential for the survival of mice. This implies that Cdk2 cannot compensate for the loss of Cdk1 when expressed from its own locus. The inability of Cdk2 to take over the function of Cdk1 could be attributed to: (1) intrinsic differences between the Cdk1 and Cdk2 proteins, such as substrate specificity or interaction with binding partners;(2) differences in the timing of expression of Cdk1 and Cdk2 during the different phases of cell cycle; and/or (3) differences in their sub-cellular localization. It is of interest to explore whether Cdk2 acquires some of the properties of Cdk1 when Cdk2 is expressed directly from the Cdk1locus in vivo, and whether it would be able to compensate for the loss of Cdk1. This hypothesis is derived from recent findings that genetic replacement of cyclin D1 by cyclin E can rescue the phenotypes of cyclin D1 knockout mice(Geng et al., 1999). Similarly, it has been shown that cyclin D2 rescues the loss of cyclin D1 when expressed from the D1 locus (Carthon et al., 2005). Furthermore, H-Ras (Hras1) can substitute for K-Ras(Kras) and supports normal embryonic development(Potenza et al., 2005). These studies provide evidence that the timing of expression and the genetic locus play important roles in determining the functions of a protein. By genetically replacing Cdk1 with Cdk2, it is possible to study whether Cdk2 can rescue the loss of Cdk1 in vivo. At the same time, it is of interest to determine whether Cdk2 can retain its own functions when expressed from the Cdk1 locus. In this context, it is also important to determine how efficiently Cdk2 performs its own mitotic cell cycle and meiotic functions in germ cells when expressed from the Cdk1 locus, as Cdk1 cannot functionally rescue the meiotic functions of Cdk2 in Cdk2^-/- mice.

To better understand the importance of genomic location and timing of Cdk2 expression and the possible compensation for loss of Cdk1 by Cdk2, we generated a mouse in which a Cdk2 cDNA was knocked into the Cdk1 locus (Cdk1^Cdk2KI). We found that substitution of both copies of Cdk1 by Cdk2 leads to early embryonic lethality, similar to deletion of Cdk1, even though the knockin Cdk2 is expressed from the Cdk1 locus. In addition, in order to study the consequences of Cdk2 expression from the Cdk1 locus on the function of Cdk2, we generated Cdk2^-/-Cdk1^+/Cdk2KI mice, in which one copy of Cdk2 and one copy of Cdk1 are expressed from the Cdk1 locus with a deletion of the Cdk2 gene in the original Cdk2 locus. From this study, we found that both male and female Cdk2^-/-Cdk1^+/Cdk2KI mice are sterile, similar to Cdk2^-/- mice, even though they express the Cdk2 protein from the Cdk1 locus in testis.

To construct the Cdk1^Cdk2KI targeting vector, we purchased a 129/Sv mouse genomic BAC clone harboring the genomic Cdk1 locus(Resgene, pBeloBACII clone J21 plate 305; PKB576). Using BAC recombineering technology (Lee et al., 2001),we first inserted an FRT site into intron 2 of Cdk1. Then, a cassette harboring [Cdk1 homology arm 5′-Cdk2cDNA-IRES-β-galactosidase-FRT-loxP-PGK-EM7-neomycin-poly(A)-FRT-loxP Cdk1 homology arm 3′] was inserted in place of exon 3 of the Cdk1 locus. The insertion site maintains all Cdk1 exon 3 splicing sequences and results in a transcript including Cdk1 exon 1(5′ UTR), Cdk1 exon 2 (including the ATG start codon plus 11 amino acids) and the Cdk2^Δ12-HA cDNA. We designed this construct to induce Cdk2^12AACdk1 expression under potential regulatory sequences including the Cdk1 5′ UTR,promoter, intron 1 and intron 2. Moreover, over the first 12 amino acids, Cdk1 and Cdk2 are very similar [they differ by four amino acids (in red in Fig. 1A)], suggesting that this region would not affect Cdk2 properties (see Fig. S1 in the supplementary material). The Cdk1^Cdk2KI locus was then retrieved into pBluescriptLight-HSVTK (Liu et al.,2003) and, after recombination, a 22 kb fragment of the Cdk1^Cdk2KI locus was selected by ampicillin and kanamycin resistance. After electroporation and selection with G418 and gancyclovir,three independent embryonic stem cell clones were identified which had the Cdk1^Cdk2KI locus correctly targeted. Positive clones were screened by β-galactosidase expression, Southern blot and PCR. The following primers were used: 5′-ACCATGTATATGTTAGATCGTAG-3′(PKO553), 5′-TCGCTTTCAAGTCTGATCTTCT-3′ (PKO554) and 5′-CGATATTAGGGTGATTAAGTTCC-3′ (PKO043). Wild-type clones yield a band of 300 bp, whereas the mutant clone produces a band of 450 bp. Germline transmission was obtained from two clones and these were used to generate chimeric animals.

Mice were housed under standard conditions and were maintained on a 12-hour light/dark cycle. Mice were fed a standard chow diet containing 6% crude fat and were treated in compliance with the National Institutes of Health guidelines for animal care and use.

Twelve- to fifteen-week-old Cdk2^+/+, Cdk2^-/-, Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI male mice were used and all animals were operated upon under sterile conditions between 9 am and 12 pm, as described previously(Satyanarayana et al., 2003). Mice were anesthetized by intraperitoneal injection of avertin and were subjected to partial (70%) hepatectomy (PH). After 2 hours of BrdU labeling,mice were sacrificed at 24 (n=3), 48 (n=4) and 72(n=4) hours after PH. For BrdU pulse labeling, 10 μl/g body weight of labeling reagent (10:1, 5-bromo-2-deoxyuridine:5-fluro-20-deoxyuridine;Cell Proliferation Kit RPN20, Amersham) was administered intraperitoneally 2 hours before sacrifice. After euthanizing the mice, portions of the liver lobes were fixed separately for BrdU, Hematoxylin and Eosin andβ-galactosidase staining.

MEFs were prepared as described previously(Berthet et al., 2003) from E13.5 Cdk2^+/+, Cdk2^-/-,Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/- Cdk1^+/Cdk2KI embryos. Serum starvation and stimulation experiments were done as described previously(Satyanarayana et al.,2008).

Proliferation of Cdk2^+/+, Cdk2^-/-,Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/- Cdk1^+/Cdk2KI MEFs in response to serum starvation (0.1% FBS) and stimulation (10% FBS) was analyzed in 96-well plates as described previously (Satyanarayana et al., 2008).

Serum-starved (DMEM medium with 0.1% FBS, 96 hours) and serum-stimulated(DMEM medium with 10% FBS) cells from 100-mm culture dishes were transferred on to coverslips in 12-well plates at a density of 1×10⁵cells per well and probed with specific antibodies at 0, 6, 12 and 24 hours after stimulation. Immunocytochemical staining was conducted as described(Satyanarayana et al., 2008). Primary antibodies against Cdk2 and HA-tag (to detect knockin Cdk2-HA)(Berthet et al., 2003) were used at 1:200 dilution. At each time point, the staining pattern was analyzed in several low-power fields (63×) and the images were captured with a confocal laser-scanning microscope (LSM510, Zeiss).

Slides were rehydrated and a microwave antigen-retrieval step was performed for 13 minutes in 10 mM sodium citrate (pH 6.0) containing 0.05% Tween 20. The sections were then treated with 3% hydrogen peroxide for 10 minutes. Blocking was carried out using 2.5% horse serum, 1% BSA in PBS for 30 minutes. Slides were incubated at room temperature for 1 hour with the following primary antibodies: Cdk2 (1:1000), Cdk1 (1:200), HA-Cdk2 (1:5000)(Berthet et al., 2003) and Cdk2(Abcam). Antibody detection was achieved using the anti-rabbit ImmPRESS Reagent Kit (Vector Labs) according to the manufacturer's protocol. Slides were counterstained with Mayer's Hematoxylin, mounted with Permount mounting media, and coverslips were applied.

BrdU immunohistochemical staining on formalin (Sigma, HT50-1-128)-fixed 5-μm liver sections was performed as described(Satyanarayana et al., 2008). An Axioplan2 imaging microscope (Zeiss) was used to photograph and analyze the BrdU staining pattern (and likewise for H&E, β-galactosidase and apoptotic staining). At least 3000 nuclei were counted per slide and the percentage of BrdU-positive nuclei calculated.

Frozen sections of liver, testes, ovaries and embryos were warmed to room temperature for ∼20 minutes. The tissue sections were fixed in acetone for 10 minutes and then air dried. Slides were rinsed with distilled water (2 minutes), incubated in Hematoxylin (Richard-Allan Scientific, 7231) for 3 minutes, and then washed with distilled water twice for 2 minutes. The slides were treated with clarifier (Richard-Allan Scientific, 7402) for 2 minutes,followed by a brief wash with distilled water. After immersing the slides in Bluing Reagent (Richard-Allan Scientific, 7301) for 1 minute, they were washed with water (2 minutes), incubated in 95% ethanol for 1 minute, and then with Eosin Y (Richard-Allan Scientific, 7111) for 20 seconds. Then, the slides were incubated in 100% ethanol (three times, 1 minute each), followed by xylene(three times, 1 minute each).

The tissues or embryos were fixed in formaldehyde/glutaraldehyde fixative[192.6 ml PBS, 5.4 ml 37% formaldehyde, 1.6 ml 25% glutaraldehyde, 0.4 ml IGEPAL (NP40 substitute; Sigma, I3021)]. Sections (10 μm) were prepared from these fixed samples. The slides were rinsed three times for 5 minutes each in PBS. After placing the slides in a humid chamber, theβ-galactosidase staining solution [17.1 ml PBS, 0.5 ml X-gal (40 mg/ml in DMSO), 0.5 ml 100 mM K₃Fe(CN)₆ (Fluka, 60299), 0.5 ml 100 mM K₄Fe(CN)₆ (Fluka, 60279), 40 μl 1 M MgCl₂] was applied directly onto the tissue sections and incubated overnight at 31°C in the dark. After washing, the slides were counterstained in a 0.1% Neutral Red solution for 30 seconds. The sections were dehydrated in 100% ethanol and then incubated three times in xylene, 1 minute each.

Testes were fixed in 10% formalin (NBF; Sigma, HT50-1-128). Apoptotic staining followed the manufacturer's protocol (Chemicon, S7100).

Whole-cell lysates from passage-three MEFs were prepared as described(Berthet et al., 2003). For western blotting, 50 μg of protein was separated on 12.5% polyacrylamide gels (Bio-Rad), transferred onto Immobilon-P transfer membranes (Millipore,IPVH00010) using semi-dry blotting, and probed with the following primary antibodies: Cdk2, Cdk1, Cdk4, HA-Cdk2, cyclin B1 as described previously(Berthet et al., 2003), cyclin E1 (gift of Bruno Amati, European Institute of Oncology, Milan, Italy), cyclin D1 (Neomarkers, RB-010p), p27 (Zymed, 71-9600) and actin (Santa Cruz, C0306). All antibodies were used at 1:1000. For kinase assays (Cdk2 and HA-Cdk2), 250μg of protein from cell lysates and 7 μl of anti-Cdk2 antibody-coupled agarose A beads [as described by Berthet et al.(Berthet et al., 2003)] or HA-antibody-coupled agarose A beads (Roche, 11815016001) were used and the kinase assays performed as described previously(Aleem et al., 2005). For co-immunoprecipitation assays (HA-Cdk2/cyclin E1, HA-Cdk2/cyclin A2), 400μg of protein from cell lysates and 7 μl of HA-coupled agarose A beads were used.

Recently, we reported that Cdk1 can bind to cyclin E and take over the functions of Cdk2 in its absence to drive cells through the G1-S transition(Aleem et al., 2005). To explore the reverse situation, i.e. whether Cdk2 can functionally substitute for Cdk1, we knocked the Cdk2 cDNA into the Cdk1 locus(Fig. 1A). The knockin construct was electroporated into embryonic stem (ES) cells. The homologous recombination event in heterozygous ES cells was identified by lacZreporter gene expression (Fig. 1B), PCR (Fig. 1C)and Southern blot (data not shown). Heterozygous ES cells were injected into mouse blastocysts to generate chimeras. The chimeric mice were backcrossed to produce Cdk1^+/Cdk2KI heterozygous mice. At this point,mouse embryonic fibroblasts (MEFs) were prepared from E13.5 embryos of Cdk1^+/Cdk2KI mice and the expression and kinase activity of knockin HA-tagged Cdk2 was analyzed. This analysis indicated that knockin HA-Cdk2 was expressed and displayed kinase activity similar to that of endogenous Cdk2 (Fig. 1D). Heterozygous mice were then intercrossed to generate homozygous Cdk1^{Cdk2KI/Cdk2KI} mice. Out of 258 mice analyzed, 88 (34%)were Cdk1^+/+ and 170 (66%) were Cdk1^+/Cdk2KI, but no Cdk1^{Cdk2KI/Cdk2KI}mice were obtained. In addition, a total of 143 embryos were analyzed at E9.5(n=23), E10.5 (n=19), E12.5 (n=23), E13.5(n=35), E14.5 (n=16), E16.5 (n=13) and E18.5(n=14). All embryos analyzed were either Cdk1^+/+or Cdk1^+/Cdk2KI and none displayed the Cdk1^{Cdk2KI/Cdk2KI} genotype. Also, 69 blastocyts were recovered from the uteri of four Cdk1^+/Cdk2KIsuper-ovulated females and from six Cdk1^+/Cdk2KI naturally bred females with Cdk1^+/Cdk2KI male mice. None of these blastocysts displayed any abnormal phenotypes, such as fragmentation,shrinkage or degeneration, but all blastocysts genotyped were either Cdk1^+/+ or Cdk1^+/Cdk2KI and none was homozygous for Cdk1^Cdk2KI. However, when we crossed Cdk1^+/Cdk2KI male or female mice with C57BL6 wild-type mice, both wild-type and Cdk1^+/Cdk2KI litters were obtained at the expected frequency. This indicates that the Cdk1^Cdk2KI germ cells are viable and functional, but that the fertilized homozygous embryos are unable to reach the blastocyst stage. Our analysis suggests that the substitution of Cdk1 by Cdk2 leads to early embryonic lethality before E3.5, comparable to the effect of deleting the Cdk1 gene or a gene-trap mutation in the Cdk1 locus (our unpublished results) (Santamaria et al.,2007). To test whether the loss of p53 (Trp53 - Mouse Genome Informatics) could rescue the phenotype of Cdk1^{Cdk2KI/Cdk2KI} mice, Cdk1^+/Cdk2KImice were crossed with p53^-/- mice. We did not obtain any mice, embryos or blastocysts with the p53^-/-Cdk1^{Cdk2KI/Cdk2KI} genotype (data not shown). This indicates that loss of p53 does not rescue the phenotype of Cdk1^{Cdk2KI/Cdk2KI} mice and that at least one copy of Cdk1 is essential for survival.

To identify the consequences for the function of Cdk2 when expressed from the Cdk1 locus, we crossed Cdk1^+/Cdk2KI mice with Cdk2^+/- mice. From such crosses we obtained Cdk2^+/+Cdk1^+/Cdk2KI mice, in which two copies of Cdk2 are expressed from the original Cdk2locus and one copy of knockin Cdk2 from the Cdk1 locus, as well as Cdk2^-/- Cdk1^+/Cdk2KI mice, in which only one copy of knockin Cdk2 is expressed from the Cdk1locus and Cdk2 expression from the endogenous locus is abolished. Littermate Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/- Cdk1^+/Cdk2KI mice did not display any morphological differences. Similarly, no significant differences were observed between Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/- Cdk1^+/Cdk2KI mice in the histopathology of any of the tissues analyzed (data not shown). Surprisingly, male and female Cdk2^-/- Cdk1^+/Cdk2KI mice were sterile, similar to Cdk2^-/- mice(Berthet et al., 2003; Ortega et al., 2003), even though they express a functional copy of Cdk2 from the Cdk1locus. The testes and ovaries of adult Cdk2^-/-Cdk1^+/Cdk2KI mice were atrophic and were only about half the size of those of Cdk2^+/+Cdk1^+/Cdk2KImice (Fig. 2A,B). To determine the cause of the sterility, a histological analysis was conducted on post-natal day 90 (P90) ovaries of Cdk2^-/-Cdk1^+/Cdk2KI and Cdk2^+/+Cdk1^+/Cdk2KI mice. In Cdk2^+/+Cdk1^+/Cdk2KI ovaries, the development of oocytes was normal and the correct follicle stages were observed(Fig. 2I,K). By contrast, in Cdk2^-/- Cdk1^+/Cdk2KI ovaries, no follicles were observed and the oocytes failed to develop in the atrophic ovaries, similar to what is observed in Cdk2^-/- females(Fig. 2J,L)(Berthet et al., 2003; Ortega et al., 2003). A histological analysis was also conducted on sexually immature and mature testes of Cdk2^-/- Cdk1^+/Cdk2KI mice. This revealed that in the testes of adult Cdk2^-/-Cdk1^+/Cdk2KI mice (P90), the size of the seminiferous tubules was much smaller than normal, resulting from a substantial depletion of spermatocytes (Fig. 2C-F),similar to that reported previously for Cdk2^-/- mice(Berthet et al., 2003; Ortega et al., 2003). The numbers of spermatogonia and Sertoli cells were not affected in testes of Cdk2^-/- Cdk1^+/Cdk2KI mice(Fig. 2F). In addition, we took advantage of the lacZ reporter in the Cdk2 knockin construct and analyzed the expression of knockin Cdk2 from the Cdk1 locus byβ-galactosidase staining in testes of Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI mice. We found that lacZ was abundantly expressed in the testes of Cdk2^+/+Cdk1^+/Cdk2KI mice and extensive β-galactosidase staining was observed in the spermatocytes(Fig. 2G). By contrast, in Cdk2^-/- Cdk1^+/Cdk2KI mice, no expression was detected in the seminiferous tubules owing to the substantial depletion of spermatocytes (Fig. 2H), but faint β-galactosidase staining was observed in the testes(Fig. 2H).

In addition to adult testes, we also performed a histological analysis of testes from P10 and P20 Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI mice. Hematoxylin and Eosin staining of P10 testes revealed no significant differences between Cdk2^-/-Cdk1^+/Cdk2KI and Cdk2^+/+Cdk1^+/Cdk2KI mice, which were similar to Cdk2^-/- and Cdk2^+/+ mice,respectively, as reported previously (Fig. 3Aa,b,Ba,b) (Ortega et al.,2003). We observed a similar expression pattern of knockin Cdk2(β-galactosidase staining) in the P10 testes of Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/- Cdk1^+/Cdk2KI mice(Fig. 3Ac,Bc). Nevertheless, we detected a marked increase in apoptosis of primary spermatocytes in P10 testes of Cdk2^-/- Cdk1^+/Cdk2KI as compared with Cdk2^+/+Cdk1^+/Cdk2KI mice(Fig. 3Ad,Bd). In contrast to P10 testes, visible defects were observed in P20 testes of Cdk2^-/- Cdk1^+/Cdk2KI as compared with Cdk2^+/+Cdk1^+/Cdk2KI mice(Fig. 3Ca,b,Da,b). At this stage in development, the first wave of germ cells is completing the second meiotic division and developing into round spermatids. Earlier stages of spermatogenesis can also be detected in tubules of P20 mice. P20 Cdk2^-/- Cdk1^+/Cdk2KI testes were ∼20-30%smaller than Cdk2^+/+Cdk1^+/Cdk2KItestes (data not shown). Histological analysis revealed the absence of round spermatids in P20 Cdk2^-/- Cdk1^+/Cdk2KI testes,similar to Cdk2^-/- testes(Fig. 3Da,b). In addition, we found extensive germ cell apoptosis in P20 Cdk2^-/-Cdk1^+/Cdk2KI as compared with Cdk2^+/+Cdk1^+/Cdk2KI testes(Fig. 3Cd,Dd). In accordance with this germ cell apoptosis and depletion of spermatocytes, diminished expression of knockin Cdk2 (β-galactosidase) was detected in P20 Cdk2^-/- Cdk1^+/Cdk2KI(Fig. 3Dc) as compared with Cdk2^+/+Cdk1^+/Cdk2KI(Fig. 3Cc) testes.

To determine why knockin Cdk2 fails to participate in and complete meiosis,we analyzed its expression and localization further in adult testes. A thorough analysis of 3-month-old adult testes revealed that Cdk2^-/- Cdk1^+/Cdk2KI spermatocytes were able to reach further stages of meiosis than Cdk2^-/-spermatocytes. In Cdk2^-/- Cdk1^+/Cdk2KI testes,there was a greater number of spermatocytes reaching the pachytene stage of meiosis (arrows in Fig. 4Da, Fig. 2F) than in Cdk2^-/- testes, where cells with pachytene morphology and condensed sex body were rarely observed (arrows in Fig. 4Ca, Fig. 2D). A possible explanation for the lack of full rescue is that Cdk2 was not being expressed at the appropriate time from the Cdk1 locus. In order to explore this possibility, we analyzed the expression of Cdk1, Cdk2 and knockin Cdk2 in wild-type and mutant testes. The localization of Cdk1 in wild-type testes was cytoplasmic in spermatogonia but was decreased and possibly absent in leptotene and zygotene cells (arrows, Fig. 4Aa,Ba). A resumption of Cdk1 expression was evident in the nuclei and cytoplasm of pachytene spermatocytes, but Cdk1 was not detectable in round spermatids. In Cdk2^-/- and Cdk2^-/-Cdk1^+/Cdk2KI testes, Cdk1 was expressed in nearly all cells present except Sertoli cells (Fig. 4Ca,Da). Wild-type testes showed weak cytoplasmic staining of Cdk2 in spermatogonia and faint-to-no staining in leptotene and zygotene cells(arrows, Fig. 4Ab). By contrast, pachytene spermatocytes displayed solid Cdk2 staining in the nucleus. Additionally, some round spermatids were strongly positive for Cdk2. In contrast to Cdk2^+/+ testes, robust staining of Cdk2 was detected in all cell types in Cdk2^+/+Cdk1^+/Cdk2KI testes. This could be due to the fact that the endogenous and knockin Cdk2 were expressed at different stages of meiosis and at least one of the two forms was present in the cells(Fig. 4Bb). When we used HA antibodies to detect knockin Cdk2, the expression was limited to only a few cells (Fig. 4Bc). We did not detect any Cdk2 staining in Cdk2^-/- testes, as expected(Fig. 4Cb). Surprisingly, the knockin Cdk2 was present in nearly all remaining cells in testes of Cdk2^-/- Cdk1^+/Cdk2KI mice(Fig. 4Dc). This expression was confirmed by HA antibody staining (Fig. 4Bc,Dc); neither wild-type nor Cdk2^-/- mice stained for HA (Fig. 4Ac,Cc).

Previous studies in wild-type mouse cells reported that Cdk2 is predominantly present in the nucleus, irrespective of the cell cycle stage(Moore et al., 1999; Satyanarayana et al., 2008). By contrast, Cdk1 is present primarily in the cytoplasm and translocates to the nucleus only during mitosis after nuclear breakdown(Bailly et al., 1989; Bailly et al., 1992). To identify the consequences of genetic replacement for the translocational properties of Cdk2 we monitored, by immunofluorescence, the localization of endogenous Cdk2 and knockin Cdk2 (expressed from the Cdk1 locus) in serum-starved and serum-stimulated Cdk2^+/+Cdk1^+/Cdk2KI MEFs at different stages of the cell cycle(Fig. 5Aa-Bh). Our analysis revealed that endogenous Cdk2, as well as HA-tagged knockin Cdk2 expressed from the Cdk1 locus, were predominantly present in the nucleus in serum-starved cells, although we also found a certain amount of staining in the cytoplasm for both endogenous and knockin Cdk2(Fig. 5Ae,Be). Between 6 and 24 hours after serum stimulation, we detected the endogenous or knockin Cdk2 primarily localized in the nucleus (Fig. 5Af-Ah,Bf-Bh). Similarly, when we monitored the localization of knockin Cdk2 in the absence of endogenous Cdk2 in Cdk2^-/-Cdk1^+/Cdk2KI MEFs, we did not observe any difference in the localization pattern of knockin Cdk2, as it was present predominantly in the nucleus even though it was expressed from the Cdk1 locus (data not shown).

To identify whether cells expressing three copies of Cdk2 had any proliferative advantage over wild-type or Cdk2^-/- MEFs, we measured the proliferation rate of Cdk2^+/+Cdk1^+/Cdk2KI, Cdk2^+/+ and Cdk2^-/- MEFs. Our analysis indicated that there was no significant difference in the proliferation rate of Cdk2^+/+Cdk1^+/Cdk2KI as compared with Cdk2^+/+ MEFs, even though they express an extra copy of Cdk2 from the Cdk1 locus(Fig. 5C). Similarly, we did not observe any significant difference in the proliferation rate of Cdk2^-/- Cdk1^+/Cdk2KI MEFs as compared with Cdk2^+/+ MEFs or those of the other two genotypes(Fig. 5C). Co-immunoprecipitation assays revealed that the knockin HA-tagged Cdk2 was able to form a complex with cyclin E1 (Fig. 5D, eleventh panel from the top) and cyclin A2(Fig. 5D, twelfth panel),similar to endogenous Cdk2 as described previously(Elledge et al., 1992; Sheaff et al., 1997). In addition, when we determined the expression pattern of some of the Cdks and cyclins that play a role in the G1, S and G2 phases of the cell cycle, we did not observe any significant differences in their expression levels between Cdk2^+/+, Cdk2^-/-, Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI genotypes(Fig. 5D), with the exception of an increase in cyclin E expression when Cdk2KI was present(Fig. 5D, lanes 3 and 4). This indicates that the expression of Cdk2 from the Cdk1 locus did not affect its cell cycle functions, and that the presence of an extra copy of Cdk2, or the loss of one copy of Cdk1, does not have any impact on the cell cycle.

To identify the consequences of loss of Cdk2 on the transcriptional activation of Cdk1, we used Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI mice, taking advantage of the lacZreporter gene in our target vector to follow the transcriptional activation of Cdk1 in vivo by β-galactosidase staining. We employed the well-established liver regeneration after partial hepatectomy (PH) model system (Fausto, 2000; Kountouras et al., 2001) to study the transcriptional activation of Cdk1 during in vivo cell cycle initiation and progression in the presence or absence of Cdk2(Fig. 6Aa-Bd). In response to PH, S-phase initiation occurs at 24 hours, peaks at 48 hours, and the first round of replication is completed at ∼72 hours(Fausto, 2000; Satyanarayana et al., 2004). In the presence of Cdk2 (Cdk2^+/+Cdk1^+/Cdk2KI), initiation of Cdk1 transcription occurred 24 hours after PH, as revealed by weak β-galactosidase staining in the regenerating liver (Fig. 6Ab). At 48 hours after PH, β-galactosidase staining was stronger than at 24 hours and ∼50% of the cells stained forβ-galactosidase (Fig. 6Ac). At 72 hours after PH, more than 90% of the cells displayedβ-galactosidase staining and the staining pattern was even stronger than at earlier time points (Fig. 6Ad). This analysis suggests that the increase in the staining pattern was due not only to increased transcriptional activation, but also to the accumulation of more protein. In contrast to the transcriptional activation of Cdk1 in Cdk2^+/+Cdk1^+/Cdk2KI mice, Cdk2^-/-Cdk1^+/Cdk2KI mice displayed premature transcriptional activation of Cdk1 as revealed by a robust β-galactosidase staining pattern at 24 hours after PH, when ∼40% of the cells already stained for β-galactosidase (Fig. 6Bb). At later time points (48 hours after PH), the staining pattern appeared more intense (∼70% of cellsβ-galactosidase-positive) than in Cdk2^+/+Cdk1^+/Cdk2KI mice (Fig. 6Bc). However, at 72 hours, the β-galactosidase staining pattern was similar in Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI mice (Fig. 6Bd).

In addition to monitoring the transcriptional activation of Cdk1by β-galactosidase staining, we also monitored the initiation and progression of the cell cycle 24 to 72 hours after PH in Cdk2^+/+, Cdk2^-/-, Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI mice. We observed that S-phase was slightly delayed in Cdk2^-/- as compared with Cdk2^+/+ mice, especially at 24 hours(Fig. 6Cb,Db,E), as reported previously (Satyanarayana et al.,2008). The initiation and peak of S-phase were not altered, but the percentage of BrdU-positive cells was decreased at 24 (and 48) hours after PH in Cdk2^-/- regenerating livers as compared with Cdk2^+/+ livers (Fig. 6Cb,E). In contrast to Cdk2^-/- mice, Cdk2^-/- Cdk1^+/Cdk2KI mice did not display any difference in the regenerative response as compared with Cdk2^+/+ mice, and the percentage of BrdU-positive cells between 24 and 72 hours after PH was similar to that of Cdk2^+/+ or Cdk2^+/+Cdk1^+/Cdk2KI mice (Fig. 6Ca,c,d,Da,c,d,E). This indicates that the knockin Cdk2 expressed from the Cdk1 locus is able to mimic the cell cycle function of endogenous Cdk2. In addition, it appears that the presence of an extra copy of Cdk2 in Cdk2^+/+Cdk1^+/Cdk2KImice did not confer any proliferative advantage, and one copy of Cdk1was sufficient for normal liver regeneration after PH. The differential transcriptional activation of Cdk1 during different stages of liver regeneration prompted us to explore the transcriptional activation of Cdk1 during embryogenesis and in adult tissues.

The presence of the lacZ reporter gene in the targeting vector allowed us to explore the transcriptional activation of Cdk1 during different stages of embryogenesis. Similarly, we monitored the activity status of Cdk1 in most of the adult tissues of Cdk2^+/+Cdk1^+/Cdk2KI mice. Cdk1 transcriptional activation was observed in most of the organs of the embryo between E14.5 and E20.5 (Fig. 7C′-F″). During earlier stages of embryogenesis (E12.5 and E13.5), a faint β-galactosidase staining was detected predominantly in the abdominal region, including the liver, kidney, different components of the digestive system, in the heart and lungs(Fig. 7A′-B″). By contrast, no β-galactosidase staining was detected in E9.5 embryos (data not shown). As a result of the ubiquitous expression of Cdk1 between E14.5 and E20.5, no significant difference in the transcriptional activation of Cdk1 between Cdk2^+/+Cdk1^+/Cdk2KI (in the presence of Cdk2) and Cdk2^-/- Cdk1^+/Cdk2KI (in the absence of Cdk2)embryos was detected (data not shown). Owing to the design of the knockin construct with an IRES/β-gal cassette, it is possible that the transcriptional activity of the Cdk1 locus was underestimated by a factor of up to 10, which would explain the low staining in early embryos.

Furthermore, when we analyzed the transcriptional activation of Cdk1 by lacZ expression in several adult tissues in Cdk2^+/+Cdk1^+/Cdk2KI mice, we did not observe expression of Cdk1 (β-galactosidase staining) in most of the tissues, including brain, heart, liver, lung, kidney and skin(Fig. 8A-E). In the case of the thymus, β-galactosidase staining was mainly observed in the medulla(Fig. 8F′,F″). In spleen, β-galactosidase staining was detected in the hematogenous red pulp (Fig. 8G,G″). In contrast to these other organs, robust expression of Cdk1 was observed in testis: spermatids, spermatocytes and Sertoli cells were solidly stained forβ-galactosidase (Fig. 8H′,H″). This observation is in accordance with previous reports that Cdk1 is widely expressed in germ cells (see Ravnik and Wolgemuth, 1999)(see Fig. 4Aa,Ba,Ca,Da). When we analyzed the expression level of endogenous Cdk2 and HA-tagged knockin Cdk2 in different tissues of adult Cdk2^+/+Cdk1^+/Cdk2KI mice, expression was absent in most of the adult tissues, except for spleen, testes and thymus(Fig. 8I).

It has been hypothesized that the presence of multiple Cdks in mammalian cells poses additional levels of control during cell cycle initiation and progression and that certain Cdk/cyclin combinations perform tissue-specific functions (Aleem and Kaldis,2006; Pagano and Jackson,2004). Furthermore, it was suggested that the availability of multiple Cdks in mammalian cells offers compensatory mechanisms in the absence of one or more Cdks. In support of this hypothesis, a single deletion of Cdk2, Cdk4 or Cdk6 does not affect the survival of mice(Berthet et al., 2003; Malumbres et al., 2004; Ortega et al., 2003). In addition, it was shown that double knockout of Cdk4 and Cdk6, or of Cdk2 and Cdk4, leads to embryonic lethality (Berthet et al.,2006; Malumbres et al.,2004; Santamaria et al.,2007). These studies imply that the loss of one or two Cdks is compensated partially or completely by other Cdk-cyclin complexes. Notably,Cdk1 compensates for the loss of Cdk2 by complexing with cyclin E(Aleem et al., 2005). Nevertheless, compensation of Cdk2 by Cdk1 appears to be only partial as Cdk2 knockout males and females are sterile(Berthet et al., 2003; Ortega et al., 2003). This indicates that Cdk1 cannot fulfil the meiotic functions of Cdk2. To date,whether any of the Cdks can substitute for the functions of Cdk1 has not been explored. The fact that Cdk1, although a mitotic kinase, is able to perform the functions of the S-phase kinase Cdk2, raises the possibility that Cdk2 might substitute for the loss of Cdk1. This hypothesis is strengthened further by the observation that Cdk2 can bind to cyclin B1(Aleem et al., 2005).

Contrary to this hypothesis, it has been reported that the deletion of Cdk1 leads to early embryonic lethality, with embryos dying before E3.5 (Santamaria et al., 2007)(our unpublished results). This indicates that none of the Cdks can compensate for the loss of Cdk1 in terms of lethality. We hypothesized that if the timing of transcriptional activation and the genomic location of Cdk2 match those of Cdk1, Cdk2 might acquire some of the properties of Cdk1 and thereby compensate for the loss of Cdk1. However, even when Cdk2 was expressed from the Cdk1 locus, we did not observe any rescue of the lethality. We found that genetic replacement of Cdk1 by Cdk2 leads to early embryonic lethality, similar to Cdk1 deletion, and embryos die before E3.5. This indicates that Cdk1 is essential for the initial divisions that lead to the formation of the blastocyst. In addition, deletion of p53 in the knockin background did not rescue the phenotypes caused by the substitution of Cdk1 by Cdk2. From this genetic replacement study, we were only able to obtain heterozygous knockin mice (Cdk1^+/Cdk2KI), in which one copy of Cdk2 is expressed from the Cdk1 locus and the other allele encodes wild-type Cdk1. Our work indicates that at least one copy of Cdk1 is essential for the survival of mice and that Cdk2 cannot substitute for Cdk1 function, even when expressed from the Cdk1 locus. Among the possible reasons for the failed rescue is that the localization of Cdk2KI differs from that of Cdk1, although differences in substrate specificity cannot be excluded either.

Moreover, when we crossed Cdk1^+/Cdk2KI with Cdk2^+/- mice and analyzed Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI mice, we found that both male and female Cdk2^-/- Cdk1^+/Cdk2KI mice were sterile, similar to Cdk2^-/- mice, even though they express Cdk2 from the Cdk1 locus. When we analyzed the transcriptional activation of the Cdk1 locus and the presence of knockin Cdk2 in testes by lacZ expression (β-galactosidase staining), HA western blotting and HA immunohistochemistry, we found abundant expression of knockin Cdk2 in the testes of Cdk2^+/+Cdk1^+/Cdk2KImice. However, in adult Cdk2^-/- Cdk1^+/Cdk2KImice, testes and ovaries were atrophic, similar to Cdk2^-/-mice. In Cdk2^-/- Cdk1^+/Cdk2KI mice, we observed that the seminiferous tubules were smaller than normal with a substantial depletion of germ cells, similar to Cdk2^-/- mice. As a result of the substantial depletion of germ cells, the expression of knockin Cdk2 was diminished in the seminiferous tubules. Furthermore, knockin Cdk2 was expressed similarly in germ cells of Cdk2^+/+Cdk1^+/Cdk2KI and Cdk2^-/-Cdk1^+/Cdk2KI P10 testes when the germ cells were not developed beyond tetraploid primary spermatocytes. By contrast, atrophic testes lacked round spermatids and displayed extensive apoptosis of germ cells in Cdk2^-/- Cdk1^+/Cdk2KI mice at P20, indicating that although knockin Cdk2 was expressed, it was unable to complete the pachytene stage and the cells instead underwent apoptosis.

We observed that Cdk2^-/- spermatocytes arrested and accumulated mostly prior to pachytene. This arrest appears to be incomplete,as we observed occasional cells with pachytene morphology. We believe that this arrest can be overcome by knockin Cdk2, as we see more cells with pachytene morphology in Cdk2^-/- Cdk1^+/Cdk2KIthan in Cdk2^-/- mice. It appears, however, that this rescue is only partial, as these spermatocytes arrest later in pachytene. Given that the subcellular localization of knockin Cdk2 appears to reflect that of the endogenous Cdk2, we conclude that the timing of expression of Cdk2 is crucial for its meiotic function(s). These results suggest the existence of a certain time window for the requirement of Cdk2. When Cdk2 is not expressed at that particular time point, the cells fail to complete meiosis even though Cdk2 is expressed subsequently, as indicated by the continuous HA and Cdk2 staining in the tubules of the Cdk2^-/-Cdk1^+/Cdk2KI mice after pachytene. Our results indicate that the genetic relocation of Cdk2 to the Cdk1 locus abolished Cdk2 meiotic function and as a result Cdk2^-/-Cdk1^+/Cdk2KI mice are sterile, similar to Cdk2^-/- mice. This indicates that the genetic locus and timing of Cdk2 expression determine the meiotic functions of Cdk2.

When we analyzed the subcellular localization of knockin Cdk2, we found that it was predominantly localized in the nucleus irrespective of the cell cycle stage, similar to endogenous Cdk2(Moore et al., 1999). Although expressed from the Cdk1 locus, knockin Cdk2 retains its subcellular localization. This indicates that the genomic locus does not play a significant role in determining the translocational property of a protein, at least in the case of Cdk2. Similarly, we found that knockin Cdk2 is able to form a complex with cyclin E1 and cyclin A2 and displays kinase activity similar to endogenous Cdk2. This excludes the possibility that the presence of the HA tag affected the properties, and thereby meiotic function, of knockin Cdk2. In addition, when we analyzed the proliferation rate of Cdk2^-/- Cdk1^+/Cdk2KI MEFs, we did not observe any significant difference to Cdk2^+/+Cdk1^+/Cdk2KI MEFs. This indicates that the knockin Cdk2 is able to perform its function in the mitotic cell cycle and form a complex with cyclin E1. In addition, analysis of cell cycle initiation and progression in vivo revealed that there was no significant difference between Cdk2^-/- Cdk1^+/Cdk2KI and Cdk2^+/+Cdk1^+/Cdk2KI mice, indicating that knockin Cdk2 was able to rescue the slight S-phase delay originally identified in Cdk2^-/- mice during liver regeneration. Furthermore, we analyzed the transcriptional activation of the Cdk1locus by lacZ reporter gene expression using liver regeneration as an in vivo cell cycle model. This analysis revealed that Cdk1transcriptional activation occurred earlier in the absence of Cdk2, suggesting that premature activation of Cdk1 is essential in the absence of Cdk2 in order to promote the G1-S transition. This observation is in accordance with our recent finding that Cdk1, as judged by protein level, is induced at an earlier time point in the absence of Cdk2[(Satyanarayana et al., 2008),see Fig. 5C therein]. It appears that premature transcriptional and translational activation of Cdk1 are essential in the absence of Cdk2 to drive cells through the G1-S transition by binding to cyclin E. In this context, it will be interesting to determine which molecular mechanisms are responsible for coordinating the transcriptional activation of Cdk1 and Cdk2. When we analyzed the transcriptional activation of Cdk1 in adult tissues by lacZ expression, we did not observeβ-galactosidase staining in most of the tissues. This might be due to the fact that most of the adult organs are quiescent and mitotically inactive. By contrast, we found solid transcriptional activation of Cdk1 during different stages (E14.5 to E20.5) of embryogenesis. Our results suggest that Cdk1 is essential for the differentiation and development of various organs during embryogenesis.

The present study indicates that Cdk1 is essential for the survival of mice. Genetic substitution of Cdk1 by Cdk2 leads to early embryonic lethality. This indicates that Cdk2 cannot substitute for the loss of Cdk1, even when the timing of transcription and genetic location of Cdk2 match those of Cdk1. Most interestingly, Cdk2 loses its meiotic function when expressed from the Cdk1 locus, even though it is able to perform its mitotic cell cycle functions by complexing with cyclin E1 and cyclin A2. In addition, an increase in the transcriptional activation of Cdk1during late embryogenesis (E14.5 to E20.5) indicated that Cdk1 is not only essential for early embryogenesis, but might also be essential in the latter stages of embryogenesis.

We thank Mary Beth Hilton, Matt McCollum and Angie Smith for animal care;Eileen Southon and Susan Reid for technical help generating the knockin mice;Donna Butcher, Roberta Smith and the Pathology/Histotechnology Laboratory,LASP, NCI-Frederick, for tissue sectioning and staining; Scott Lawrence and Richard Frederickson for helping with embryo photographs; members of the Kaldis laboratory and David Page for discussions and support. This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.