Cyclin E Constrains Cdk5 Activity to Regulate Synaptic Plasticity and Memory Formation

doi:10.1016/j.devcel.2011.08.009

Dev Cell. Author manuscript; available in PMC 2012 October 18.

Published in final edited form as:

Dev Cell. 2011 October 18; 21(4): 655–668.

Published online 2011 September 22. doi: 10.1016/j.devcel.2011.08.009

PMCID: PMC3199337

NIHMSID: NIHMS328000

Cyclin E Constrains Cdk5 Activity to Regulate Synaptic Plasticity and Memory Formation

Junko Odajima,^1,⁵ Zachary P. Wills,^2,^5,⁶ Yasmine M. Ndassa,³ Miho Terunuma,⁴ Karla Kretschmannova,⁴ Tarek Z. Deeb,⁴ Yan Geng,¹ Sylwia Gawrzak,¹ Isabel M. Quadros,⁴ Jennifer Newman,⁴ Manjusri Das,¹ Marie E. Jecrois,¹ Qunyan Yu,¹ Na Li,¹ Frederic Bienvenu,^1,⁷ Stephen J. Moss,⁴ Michael E. Greenberg,² Jarrod A. Marto,³^* and Piotr Sicinski¹^*

¹Department of Cancer Biology, Dana Farber Cancer Institute, and Department of Genetics Harvard Medical School, Boston, MA 02115, USA

²Department of Neurobiology, Dana Farber Cancer Institute, and Department of Biological Chemistry and Molecular Pharmacology Harvard Medical School, Boston, MA 02115, USA

³Department of Cancer Biology, Dana Farber Cancer Institute, and Department of Biological Chemistry and Molecular Pharmacology Harvard Medical School, Boston, MA 02115, USA

⁴Center for Neuroscience Research, Department of Neuroscience, Tufts University School of Medicine, Boston, MA 02111, USA

⁵These authors contributed equally to this work

⁶Present address: Department of Neurobiology, University of Pittsburgh, Pittsburg, PA 15213, USA

⁷Present address: Institute of Functional Genomics, UMR 5203 CNRS – U 661 INSERM – Universite de Montpellier, 34094 Montpellier, France

^*Correspondence: Email: jarrod_marto/at/dfci.harvard.edu (J.A.M.)

^*Email: peter_sicinski/at/dfci.harvard.edu (P.S.)

The publisher's final edited version of this article is available at Dev Cell

See other articles in PMC that cite the published article.

Publisher's Disclaimer

SUMMARY

Cyclin E is a component of the core cell cycle machinery, and it drives cell proliferation by regulating entry and progression of cells through the DNA synthesis phase. Cyclin E expression is normally restricted to proliferating cells. However, high levels of cyclin E are expressed in the adult brain. The function of cyclin E in quiescent, postmitotic nervous system remains unknown. Here we use a combination of in vivo quantitative proteomics and analyses of cyclin E knockout mice to demonstrate that in terminally differentiated neurons cyclin E forms complexes with Cdk5 and controls the formation of synapses by restraining Cdk5 activation. Ablation of cyclin E led to a decreased number of synapses, reduced number and volume of dendritic spines, and resulted in impaired synaptic plasticity and memory formation in cyclin E-deficient animals. These results reveal a cell cycle-independent role for a core cell cycle protein, cyclin E, in synapse function and memory.

INTRODUCTION

The mammalian core cell cycle machinery is composed of cyclins and their associated cyclin-dependent kinases (Cdks). Cyclin-Cdk complexes phosphorylate cellular proteins, thereby driving cell cycle progression (Malumbres and Barbacid, 2009).

Cyclins are induced in a coordinated fashion to enable cell proliferation. Thus, in response to growth factor stimulation D-type cyclins (D1, D2 and D3) are upregulated and subsequently bind and activate Cdk4 and Cdk6. Cyclin D-Cdk4/6 kinase phosphorylates the retinoblastoma protein, pRB, leading to the release of E2F transcription factors and to transcriptional induction of the E-type cyclins (cyclins E1 and E2) (Sherr and Roberts, 2004). The two E-cyclins share significant amino acid identity, are co-expressed in all proliferating cell types, and appear to have entirely overlapping functions (Geng et al., 2001; Gudas et al., 1999; Koff et al., 1991; Lauper et al., 1998; Lew et al., 1991; Zariwala et al., 1998).

Cyclin E affects cell proliferation by multiple mechanisms. Once induced in late G1 phase, E-cyclins bind and activate Cdk2, Cdk1 and further phosphorylate pRB. In addition, cyclin E-Cdk holoenzyme phosphorylates proteins involved in initiation of DNA replication, proteins governing centrosome duplication, histone biosynthesis and cell cycle progression. All these functions contribute to the well-established role for cyclin E in cell proliferation (Hwang and Clurman, 2005).

Consistent with their growth-promoting functions, overexpression of cyclins E1 and E2 is seen in a substantial fraction of human cancers, including mammary carcinomas, lung, endometrial, gastric, colorectal and ovarian cancers, as well as sarcomas, lymphomas and leukemias. In several cancer types overexpression of cyclin E was shown to confer poor prognosis (Hwang and Clurman, 2005).

In the past, we and others studied the functions of cyclin E in development by generating cyclin E-knockout mice. Cyclin E deficient (E1^−/−E2^−/−) mice died early during gestation due to placental abnormalities (Geng et al., 2003; Parisi et al., 2003). In addition, cyclin E was required for normal heart development and for cell cycle re-entry of embryonic fibroblasts (Geng et al., 2003). All these findings were consistent with the role for cyclin E in cell proliferation.

Expression of cyclin E is usually limited to proliferating cells. Consequently, quiescent organs of adult mice express very little or no cyclin E protein. An exception to this rule is provided by the observations that cyclin E is expressed at high levels in the brains of adult mice (Geng et al., 2001; Ikeda et al., 2010; Miyajima et al., 1995). However, the function of cyclin E in non-proliferating, terminally differentiated brain cells remained unknown.

In this study, we provide evidence that in terminally differentiated neurons cyclin E regulates formation of synapses, by inhibiting Cdk5, an essential regulator of neuronal differentiation. These findings reveal a very unexpected function of a core cell cycle protein in postmitotic neurons, and may have implications for our understanding of neurological disorders, such as Alzheimer disease, in which Cdk5 disregulation has been implicated as a causative factor (Cruz and Tsai, 2004).

RESULTS

Cyclin E Expression in Adult Brain

We started our analyses by verifying that brains of adult mice express high levels of cyclin E. This was in contrast to other adult organs composed of non-proliferating cells, which expressed very little cyclin E (Figure 1A). We observed that the levels of cyclin E in brains of adult mice were similar to those seen in embryonic brains, the latter containing high proportion of proliferating neuronal progenitors (Figure 1A). Developmental analysis revealed that in brains, cyclin E levels peak during embryonic development, decline at birth, and then increase again during postnatal life when neuronal differentiation takes place (Figure 1B). In adult brains, cyclin E is expressed in most anatomical regions, including cerebral cortex and hippocampus (Figure 1C). Co-immunostaining of adult brain sections for cyclin E and neuronal-specific marker NeuN revealed that cyclin E is expressed in terminally differentiated neurons, but not in glial cells (Figures 1D, S1A and S1B). Subcellular fractionation of adult brain cortices demonstrated that cyclin E is present predominantly in the cytoplasm, and to a lesser extent in the nucleus (data not shown). Consistent with these findings, immunofluorescence studies of cultured primary neurons revealed that cyclin E is localized in the cell soma, and enriched in dendrites as well as in axons (Figures 1E and S1C, and data not shown). These observations are in contrast to dividing cells, where cyclin E is largely nuclear (Hwang and Clurman, 2005; Ohtsubo et al., 1995), suggesting that in postmitotic neurons cyclin E may function in a manner distinct from its role in the cell cycle.

Figure 1

Cyclin E Expression in Postmitotic Neurons

Quantitative Proteomic Analyses of Adult Mouse Brain

To gain molecular insights into cyclin E function in the brains of adult animals, we generated a knock-in mouse strain expressing Flag- and hemagglutinin (HA)-tandemly tagged version of cyclin E1. We reasoned that these mice would allow us to use sequential immunoaffinity purifications with anti-Flag and -HA antibodies, followed by quantitative mass spectrometry, to compare the set of cyclin E1-interacting proteins in adult versus in embryonic brains.

To generate tagged knock-in mice, DNA sequences encoding Flag- and HA-tags were inserted into the amino-terminus of the endogenous cyclin E1 locus through gene targeting in embryonic stem cells (Figure 2A). Once homozygous mice were obtained, we verified that in the tissues of knock-in mice the expression pattern and function of the tagged cyclin E1 faithfully mirrored those of the endogenous cyclin E1 in wild-type animals (Figures S2A–S2D, and data not shown).

Figure 2

In Vivo Quantitative Proteomic Analyses

We then purified cyclin E1-associated protein complexes from adult and embryonic brains using sequential immunoaffinity purifications with anti-Flag and -HA antibodies, and performed protein identification using high-sensitivity nanoflow liquid chromatography tandem mass spectrometry (LC-MS/MS) (Figure S2E and Table S1). To obtain a quantitative measure of the differences between the sets of cyclin E1-interacting proteins in adult versus embryonic brains, we used iTRAQ reagents (Ross et al., 2004) in a quantitative mass spectrometry approach. Briefly, proteins present in purification products from adult and embryonic brains were separately labeled with distinct iTRAQ stable isotope reagents. Subsequently, the samples were mixed and subjected to quantitative LC-MS/MS analysis.

We found that known cell cycle partners of cyclin E1, such as Cdk1 and Cdk2 were enriched among cyclin E1-containing complexes in embryonic brains (Figures 2B, 2C and Table S2), consistent with cell cycle-promoting function of cyclin E in proliferating cells. Surprisingly, in adult brains we found that Cdk5, a Cdk-related protein that regulates neuronal differentiation (Dhavan and Tsai, 2001), was strongly enriched in cyclin E1-containing complexes (Figures 2B, 2C and Table S2). We verified abundant association of cyclin E1 with Cdk5 in adult, but not in embryonic brains by immunoprecipitation – western blotting (Figures 3A and 3B). We estimated using immunodepletion analyses that virtually all cyclin E1 molecules are bound to Cdk5 in adult brains (Figure 3C).

Figure 3

Analyses of Cyclin E-Cdk5 Complex

Consistent with these findings, developmental analyses revealed that embryonic brains, which contain rapidly proliferating neuronal progenitors, express cyclin E and Cdk2, but very little Cdk5. Upon birth, the expression of Cdk2 is extinguished, while the levels of Cdk5 become upregulated, concomitant with induction of cyclin E (Figure 1B).

In addition to Cdk5, p27^Kip1, an inhibitor of cyclin-dependent kinases, was enriched among cyclin E1-binding proteins in adult brain (Figures 2B, ,2C2C and Table S2), suggesting a possibility of a ternary cyclin E1-Cdk5-p27^Kip1 complex. We determined by sequential immunoprecipitations that cyclin E1, Cdk5 and p27^Kip1 indeed form a ternary complex in adult mouse brain (Figure S3A).

Cyclin E-Cdk5 Complex Is Catalytically Inactive and Is Induced During Synaptogenesis

Cdk5 is expressed in non-proliferating, postmitotic neurons in the nervous tissue, where it binds and becomes catalytically activated by its regulatory partners, p35 and p39 (Lew et al., 1994; Tsai et al., 1994; Tsai et al., 1993). These p35/p39-Cdk5 complexes control many aspects of neuronal differentiation by phosphorylating several neuronal substrates (Dhavan and Tsai, 2001). It was previously shown, based on the crystal structure of p25- (a fragment of p35) Cdk5 complex, that p35/p39 activates Cdk5 via a mechanism that is distinct from the one used by cyclins to activate Cdk2 (Tarricone et al., 2001). This, together with the presence of a Cdk-inhibitor p27^Kip1 in ternary complexes with cyclin E1-Cdk5 (Figure S3A) suggested that Cdk5 would be catalytically inactive in these complexes. We tested this prediction by in vitro kinase assays, and found that cyclin E1-Cdk5 was unable to phosphorylate several known Cdk5 substrates (Figures 3D and S3B). These results suggested that cyclin E may inhibit Cdk5 activity (i.e. prevent Cdk5 activation) by recruiting it into catalytically inactive complexes, and by sequestering it away from p35/p39, the only known Cdk5 activator. To test this hypothesis, we overexpressed increasing amounts of cyclin E1 along with constant levels of Cdk5 and p35 in heterologous cells. Consistent with our hypothesis, these analyses revealed that cyclin E1 can compete with p35 for Cdk5 binding (Figures 3F and 3G).

In order to further investigate the function of cyclin E–Cdk5 in neurons, we isolated mouse embryonic hippocampal neurons and cultured them in vitro. We found that expression of cell cycle proteins, such as cyclin A and Cdk2 declined as cells matured. In contrast, cyclin E and Cdk5 became concomitantly induced in the course of neuronal differentiation (Figure 3E). Furthermore, the induction of cyclin E and Cdk5 in hippocampal neurons coincided with the induction of synaptic proteins and with the onset of synaptogenesis (Figure 3E, and data not shown). We noted a similar correlation in vivo, during postnatal brain development (Figure 1B). Since Cdk5 was shown to play an important role in synapse formation and synaptic plasticity (Dhavan and Tsai, 2001), our findings suggested the possibility that cyclin E might influence Cdk5 function in these processes.

Decreased Numbers of Synapses in Cyclin E-deficient Neurons

To analyze the function of cyclin E in synapse physiology, we generated a cyclin E1^flox/flox mouse strain (Figure S4A). We next interbred cyclin E1^flox/flox and cyclin E2^−/− animals, and obtained cyclin E1^flox/floxE2^−/− mice (Figure S4B).

We started our analyses by dissecting hippocampi from these mice and by establishing in vitro cultures of dissociated hippocampal neurons. Ten days after plating, postmitotic neurons were transfected with GFP and Cre. Subsequently, synapse formation was assessed by staining the cultures for the presynaptic marker Synapsin I and excitatory postsynaptic marker PSD95, and the number of co-localized Synapsin I and PSD95 puncta on transfected neurons was quantified. Strikingly, acute deletion of cyclin E in postmitotic neurons resulted in a significant decrease of PSD95/Synapsin I puncta (Figures 4A, 4B and S4D), indicating a reduced number of excitatory synapses. We extended these observations by measuring spontaneous miniature excitatory postsynaptic currents (mEPSC) in cultured cyclin E-deficient neurons. mEPSCs gauge the strength and number of synaptic inputs onto a recorded neuron. We found that an acute ablation of cyclin E led to a significant decrease of mEPSC amplitudes and frequency (Figures 4C, 4D, and S4E), which is indicative of reduced synapse number and reduced synaptic transmission. Collectively these results revealed a significant decrease in the number and function of excitatory synapses following an acute ablation of cyclin E in postmitotic neurons.

Figure 4

Reduced Synapse Numbers in Cyclin E-null Neurons

Decreased Numbers of Dendritic Spines in Cyclin E-null Neurons

To further substantiate these findings, we analyzed the impact of cyclin E ablation within the context of an intact neuronal circuit. To this end, we prepared hippocampal slices from cyclin E1^flox/floxE2^−/− mice. We biolistically transfected the slices with Cre and GFP, and quantified the number of dendritic spines, which represent postsynaptic sites of excitatory synapses in the brain, in GFP-positive CA1 hippocampal pyramidal neurons. Consistent with our previous findings, we observed reduced numbers of dendritic spines upon acute cyclin E shutdown in all dendritic regions of CA1 hippocampal neurons (apical, basal and distal) (Figures 4E and 4F).

Collectively, our analyses of cyclin E-deficient neurons and hippocampal slices indicated that cyclin E promotes the establishment of dendritic spines and the formation of functional synapses. We next investigated a possible mechanism by which cyclin E might affect these processes.

Mechanistic Analyses of Cyclin E-Cdk5 Synaptic Function

Based on our observations that cyclin E-Cdk5 complexes are catalytically inactive (Figure 3D), and that cyclin E can potentially compete with p35 for Cdk5 binding (Figures 3F and 3G), we hypothesized that cyclin E serves to sequester Cdk5 away from its activators p35 and p39, into inactive cyclin E-Cdk5 complexes. We further hypothesized that ablation of cyclin E and the loss of inhibitory cyclin E-Cdk5 complexes results in hyperactivation of Cdk5, leading to the observed reduction in synapse number. This hypothesis was consistent with several reports that Cdk5 can play an inhibitory role in synapse formation and function (Chergui et al., 2004; Kim et al., 2006; Kim and Ryan, 2010; Morabito et al., 2004).

We tested this model first by overexpressing Cdk5 in in vitro cultured control neurons. We found that Cdk5 overexpression indeed phenocopied cyclin E ablation, namely it reduced the numbers of excitatory synapses (Figures 4A and 4B). Moreover, we found that knock-down of Cdk5 (Figure S1D), or expression of dominant-negative Cdk5 mutant (Nikolic et al., 1996) reversed the synaptic deficits in cyclin E-null neurons (Figures 4A and 4B). Lastly, we observed that knock-down of Cdk5 in organotypic hippocampal slices restored normal numbers of dendritic spines in cyclin E-deficient neurons (Figure 4G). All these observations were consistent with the notion that cyclin E constrains Cdk5 activity and that hyperactivation of Cdk5 leads to synaptic phenotypes in cyclin E-null neurons.

Molecular Analyses of Brains from Conditional Cyclin E Knockout Mice

A direct test of the model proposed above would be to determine the phosphorylation status of synaptic Cdk5 substrates on Cdk5-specific residues in the brains of cyclin E-null mice. To accomplish this, we interbred cyclin E1^flox/floxE2^−/− mice with Nestin-Cre animals, which constitutively express Cre recombinase in the nervous system (Graus-Porta et al., 2001), and verified efficient deletion of cyclin E in the brains of cyclin E knockout – Nestin-Cre mice (Figures S4B and S4C). Since in Nestin-Cre mice, expression of Cre is turned on in proliferating embryonic neuronal progenitors (Graus-Porta et al., 2001), ablation of cyclin E in embryogenesis might have affected brain development. However, we established that ablation of cyclin E did not affect proliferation rates throughout embryonic brain development (Figure S5A). Also, the number of nestin-positive neural stem/progenitor cells was unperturbed in cyclin E-null fetal brains (Figure S5B). Adult cyclin E-deficient brains displayed normal anatomical structures (Figures S5C–S5E) and normal staining with a wide panel of neuronal and glial cell markers such as Ctip2, Cux1, GABA, calbindin, parvalbumin, Olig2 and GFAP (Figures S1A, S1B, and data not shown). Moreover, normal numbers of proliferating progenitor cells were seen in the subventricular zone of adult cyclin E-deficient brains (data not shown). Hence, ablation of cyclin E in the nervous system did not adversely affect proliferation rates and development of the brain, as we have previously shown to be the case with most other organs in cyclin E-deficient embryos (Geng et al., 2003).

We then dissected hippocampi and cerebral cortices from adult cyclin E knockout – Nestin-Cre mice and analyzed the phosphorylation status of Wave1 (Kim et al., 2006) and Synapsin I (Matsubara et al., 1996), two known synaptic Cdk5 targets, using antibodies directed against Cdk5-specific phosphoresidues on these proteins (Ser553 on Synapsin I, Ser310 and Ser441 on Wave1). Consistent with our hypothesis of Cdk5 hyperactivation, we observed increased phosphorylation of these residues in cyclin E-deficient hippocampi as well as in cortices (Figures 5A and 5B). In addition, increased phosphorylation of Cdk5 substrates was also seen in in vitro cultured hippocampal neurons from cyclin E knockout – Nestin-Cre mice (data not shown). Importantly, acute ablation of cyclin E in in vitro cultured postmitotic neurons led to increased phosphorylation of Wave1, while knock-down of Cdk5 reversed this effect (Figures 5C and 5D), confirming that hyperactivation of Cdk5 is also seen upon acute deletion of cyclin E. Phosphorylation of Wave1 on Ser310 and Ser441 by Cdk5 was shown to inhibit Wave1 activity (Kim et al., 2006). Wave1 promotes dendritic spine maturation, and its inhibition leads to a decrease in the number of mature dendritic spines. Moreover, Wave1 activity is essential for hippocampus-dependent learning and memory (Kim et al., 2006; Soderling et al., 2003).

Figure 5

Molecular Analyses of Brains from Adult Conditional Cyclin E Knockout Mice

In addition to this direct inhibition of synaptic proteins, Cdk5 can also negatively regulate the function of glutamate N-methyl-D-aspartic acid (NMDA) receptors (Chergui et al., 2004). Specifically, Cdk5 was shown to inhibit phosphorylation of the NR1 subunit of NMDA receptors, by inhibiting cAMP-dependent protein kinase (PKA), which phosphorylates NR1 on Ser897 (Chergui et al., 2004). Importantly, PKA-mediated phosphorylation of NR1 is required for synaptic and cell surface localization of this NMDA receptor subunit, and therefore for the receptor function (Bird et al., 2005; Crump et al., 2001). Hence, hyperactivation of Cdk5 is predicted to reduce NR1 phosphorylation on Ser897, and consequently to reduce its incorporation into membrane-bound NMDA receptors. Indeed, we observed reduced phosphorylation of NR1 subunit on Ser897, and reduced surface expression of NR1 in cyclin E-null hippocampi (Figures 5E–5H), consistent with increased Cdk5 activity in cyclin E knockout brains, and with the inhibitory role of active Cdk5 on NMDA receptor assembly (Chergui et al., 2004).

Synaptic Defects in Brains of Cyclin E Knockout Mice

We next investigated the brains of conditional cyclin E knockout - Nestin-Cre animals for morphologic signs of impaired synapse structure and function. First, we analyzed dendritic spines in hippocampal CA1 pyramidal neurons by Golgi impregnation. Our analyses revealed that the spine length, width and estimated spine volume were substantially reduced in adult cyclin E-null brains (Figures 6A, 6B, S6A and S6B). Consistent with the above, analyses of transmission electron microscopic images of CA1 stratum radiatum demonstrated a reduced length of postsynaptic densities in cyclin E-null brains (Figures 6C–6E). Hence, ablation of cyclin E in the brains of conditional cyclin E knockout animals led to both molecular (reduced incorporation of NMDA receptor subunit) and ultrastructural synaptic abnormalities.

Figure 6

Synaptic and Dendritic Spine Deficiencies in Conditional Cyclin E Knockout Mice

To substantiate these observations at a functional level, we analyzed spontaneous excitatory postsynaptic currents (sEPSC), which gauge synaptic activity, using acute hippocampal slices. We found that the frequency of sEPSC in CA1 pyramidal neurons was greatly reduced in cyclin E knockout brains (Figures 6F, 6G and S6C–S6E). To evaluate changes in the activity of NMDA receptors, we isolated NMDA component of EPSC as described in Experimental Procedures. The NMDA component made up 41.5 ± 6.9% (SEM) in control mice, and it was reduced to 27.5 ± 6.5% (SEM) in cyclin E knockouts (Figures S6F and S6G). These observations reveal reduced synaptic transmission, and reduced NMDA-dependent currents in the brains of cyclin E knockout animals.

Defective Synaptic Plasticity and Memory Deficits in Cyclin E Knockout Mice

The ultrastructural, molecular and electrophysiological synaptic abnormalities seen in cyclin E knockout brains suggested that cyclin E-deficient mice might display impaired learning and memory, as these processes critically depend on proper synaptic function. Of note, Cdk5 knockout mice were shown to display enhanced memory formation, consistent with inhibitory role of Cdk5 in synaptic plasticity (Hawasli et al., 2007). Therefore, we hypothesized that cyclin E knockout mice, which showed signs of Cdk5 hyperactivation, might display impaired memory.

Long-term potentiation (LTP) represents an electrophysiological correlate of learning and memory (Cooke and Bliss, 2006). Therefore, we first analyzed LTP in CA1 hippocampal neurons of conditional cyclin E knockout – Nestin-Cre animals, following theta-burst stimulation of the Schaffer collateral inputs. We observed reduced LTP in cyclin E knockouts, revealing an impaired synaptic plasticity upon cyclin E loss (Figures 7A, S7A–S7C). Importantly, LTP in CA1 pyramidal neurons is NMDA receptor-dependent (Tsien et al., 1996), hence its reduction is in agreement with the reduced cell surface expression of an essential NMDA receptor subunit, NR1, found in cyclin E-null brains (Figures 5E–5H).

Figure 7

Impaired Synaptic Plasticity and Memory in Conditional Cyclin E Knockout Mice

To extend these observations to an in vivo setting, we turned to the Morris water-maze task. This test is also critically dependent on hippocampal CA1 NMDA receptors function (Tsien et al., 1996), and it measures spatial learning and memory. Importantly, expression of the Nestin-Cre transgene was shown to have no effect by itself on performance of mice in this test (Raivich et al., 2004). Conditional cyclin E knockout and control mice were trained to learn the position of a water-submerged platform (Figure S7D), and then the platform was removed to assess the retention of memory for spatial configuration. Cyclin E knockout mice showed impaired performance in this test (Figure 7B), revealing spatial memory deficits. We next tested hippocampal- and NMDA receptor-dependent (Bernasconi-Guastalla et al., 1994; Nakazawa et al., 2003) cognitive flexibility by moving the platform to the opposite quadrant, and re-training the animals to find the platform in a new location (reversal learning). Cyclin E knockout mice also showed a significant impairment in spatial reversal learning (Figures 7C, S7E and S7F). Collectively, all these observations revealed that cyclin E loss results in impaired synaptic plasticity, decreased memory formation and impaired cognitive flexibility, phenotypes opposite to the ones observed upon Cdk5 ablation. Importantly, the overall behavioral pattern was not perturbed in cyclin E-deficient mice, as evidenced by normal performance of these animals in a battery of control locomotor, sensorimotor and behavioral tests (Figures S7G–S7K).

DISCUSSION

Cyclin E belongs to the core cell cycle machinery, which has been conserved from yeast to humans. The function of this machinery is to drive cell cycle progression. Indeed, many phosphorylation substrates of yeast and mammalian cyclin-Cdk enzymes represent proteins involved in cell division (Hwang and Clurman, 2005; Sherr and Roberts, 2004; Ubersax et al., 2003). Consistent with this notion, analyses of knockout mice lacking particular cyclins or Cdks revealed distinct proliferative deficits in specific compartments (reviewed in Sherr and Roberts, 2004). The relatively narrow, tissue-specific phenotypes of most cyclin- or Cdk-deficient strains underscored the overlapping, redundant functions of individual cyclins and Cdks in driving cell proliferation.

In clear distinction to these proliferative phenotypes, the work presented here revealed very unexpectedly that mammalian cyclin E, a cancer-related protein that normally serves to drive cell division, plays a cell cycle-independent and rate-limiting function in terminally differentiated neurons by controlling synapse formation and function. We found that acute ablation of cyclin E in postmitotic neurons led to a reduced number of synapses and dendritic spines, and resulted in decreased synaptic transmission. To ablate cyclin E in vivo, we used conditional cyclin E knockout - Nestin-Cre mice. In these animals, deletion of cyclin E occurred during embryonic development, which might have impacted our analyses of adult cyclin E-deficient brains. However, we verified that ablation of cyclin E did not affect proliferation rates of embryonic brains and did not disturb brain development, as was the case in most other compartments of cyclin E-null mice (Geng et al., 2003). Moreover analyses of brain-specific cyclin E knockout mice revealed very similar defects to the ones observed upon acute cyclin E shutdown in postmitotic neurons. In fact, the phenotypes seen in conditional cyclin E knockout – Nestin Cre mice were milder than those encountered upon acute cyclin E shutdown, a phenomenon seen in other knockouts and ascribed to developmental compensatory mechanisms (Lin et al., 2008).

We propose that cyclin E affects synapse function and memory formation by sequestering Cdk5 into catalytically inactive complexes, thereby affecting the phosphorylation status and function of synaptic proteins (Figure 6H). Hence, cyclin E acts in this setting as a Cdk antagonist, in distinction to the well-described function of cyclins as Cdk activators. Consistent with this model, overexpression of Cdk5 phenocopied acute cyclin E loss, while knock-down of Cdk5 restored normal numbers of synapses and dendritic spines in cyclin E-null neurons. Moreover, we observed hyperphosphorylation of synaptic Cdk5 substrates on Cdk5-specific sites upon acute cyclin E shutdown in vitro, and in the brains of cyclin E knockout animals. Hyperactivation of Cdk5 is also expected to indirectly affect other events relevant for learning and memory, such as NR1 subunit localization shown here, but also phosphorylation of CREB on Ser133 by PKA (Sindreu et al., 2007). These abnormalities likely collectively contribute to decreased synaptic function and to memory deficits, observed in cyclin E-knockout mice. It should be noted that in addition to its inhibitory function, Cdk5 was also shown to play positive roles in synaptogenesis and synapse function, for example by phosphorylating the CASK protein and by regulating the interaction between CASK and liprin-α (Samuels et al., 2007). It remains to be seen how these steps proceed in the absence of cyclin E.

Further studies will be needed to determine whether ablation of cyclin E results in decreased formation of new synapses and dendritic spines, reduced maintenance of the existing ones, or both. Importantly, active learning has been associated both with formation and maintenance of new dendritic spines in the brains (Xu et al., 2009; Yang et al., 2009). Given the influence of cyclin E in this process, it will be interesting to determine whether the levels of cyclin E in the brain change locally during learning and memory formation. A related question is what pathways control the abundance of cyclin E in terminally differentiated brain cells. In proliferating cells, the initial induction of E-type cyclins is driven at the transcriptional level by E2F transcription factors, while proteolytic degradation of cyclin E in late S phase is brought about by SCF^Fbw7 ubiquitin ligase (Hwang and Clurman, 2005). In postmitotic cells the E2Fs are thought to be held in an inactive state by the pRB and pRB-like “pocket” proteins, therefore it is likely that the proteolytic machinery would be responsible for modulating cyclin E levels.

Of note, the ability of cyclin E to interact with Cdk5 in neurons was also reported by Miyajima et al. (1995), however our current work reveals the functional relationship between these two proteins. To the best of our knowledge, this is the first demonstration of a function for a cyclin protein in the formation of neuronal synaptic circuits and memories.

Cdk5 is considered as an essential regulator of neuronal differentiation, and it acts in a complex with its non-cyclin activators p35 or p39. Cdk5-p35/p39 complexes phosphorylate several neuronal proteins, and control a wide range of functions, including synapse formation and plasticity, neuronal migration, axon guidance, membrane transport and neurite outgrowth (Dhavan and Tsai, 2001). It remains to be seen whether cyclin E affects all these Cdk5 functions, or it only serves to control Cdk5 in its synaptic role.

Another question raised by this work is whether abnormal levels or activity of cyclin E underlie human learning disabilities, cognitive disorders and other pathological conditions. Indeed, cyclin E is a major target of PARK2, an E3 ubiquitin-ligase, mutations in which are the most common cause of early-onset Parkinson's disease. (Staropoli et al, 2003; Lüking et al., 2000). Moreover, Cdk5 has been implicated in several human neurological and neurodegenerative diseases, such as Alzheimer's disease (Cruz and Tsai, 2004; Dhavan and Tsai, 2001), making cyclin E a potential player in these conditions. Interestingly, recent work indicates that reduction of Cdk5 levels can decrease pathological changes and reduce neuronal loss in an Alzheimer's disease mouse model (Piedrahita et al., 2010). These observations suggest an exciting possibility that cyclin E, a protein we demonstrate to bind and inhibit Cdk5, might offer a therapeutic tool for ameliorating this devastating disease.

EXPERIMENTAL PROCEDURES

Generation of Tagged Cyclin E1 Knock-in Mice

A targeting vector to knock-in a Flag-HA tag at the N-terminus of the cyclin E1 gene was constructed by replacing a start codon in the second exon with DNA sequences encoding an initiation methionine and Flag-HA tag. The targeting vector was electroporated into embryonal stem (ES) cells and homozygous cyclin E1^Ntag/Ntag animals were obtained using standard procedures (Geng et al., 2003).

Generation of Conditional Cyclin E Knockout Mice

Conditional cyclin E1 gene-targeting construct was assembled by inserting loxP sites into ScaI sites in the introns 4 and 11 of the murine cyclin E1 gene. The gene-targeting construct was electroporated into ES and homozygous cyclin E1^flox/flox mice were obtained. Cyclin E1^flox/flox mice were interbred with cyclin E2^−/− mice (Geng et al., 2003). Nestin-Cre animals (Graus-Porta et al., 2001) were from the Jackson Laboratory.

Purification of Cyclin E1-containing Complexes from Embryonic and Adult Brains and Quantitative Proteomic Analyses

For a single purification, we used pooled 30 brains dissected from 1-month old mice, or 20 heads dissected from E14.5 embryos. Tissues were homogenized, cyclin E1 and its associated proteins were immunoprecipitated using anti-Flag M2 agarose (Sigma), eluted twice with Flag peptides (Sigma), then immunoprecipitated again with anti-HA antibody (12CA5 ascites fluid, Covance) coupled to protein A sepharose beads (Amersham). Complexes were eluted from HA beads with 0.1 M glycine (pH 2.5). For LC-MS/MS, purified protein complexes containing at least 200–300 ng of cyclin E1 were used for a single run. LC-MS/MS, data analysis, iTRAQ labeling and quantitative mass spectrometry are described in Extended Experimental Procedures.

Protein Extraction, Immunoblot Analyses and In Vitro Kinase Assays

These were performed using standard procedures. For the list of antibodies, please see Extended Experimental Procedures. For kinase assays, cell lysates were immunoprecipitated with HA beads and incubated at 30°C for 30 min in reaction buffer with 10 μCi [γ³²P]-ATP containing one of the following substrates: pRB (0.5 μg; Santa Cruz), Histone H1 (1 μg; Roche) or GST-Synapsin I (1 μg).

Immunostaining of Brain Sections

Mice were cardiac perfused with PBS and then with paraformaldehyde (4% in PBS). Brains were extracted, postfixed in paraformaldehyde, embedded in tissue freezing medium (Triangle Biomedical Sciences) and frozen. Coronal sections (12 μm) were cut, stained as described in Extended Experimental Procedures, mounted with Fluoromount-G (SouthernBiotech), and analyzed on a fluorescent (Nikon E600) or confocal microscope (Leica SP5X).

Hippocampal Neuron Cultures

Dissociated hippocampal neurons were prepared from 16- to 17-day-old mouse embryos. Neurons were maintained in Neurobasal Medium (Invitrogen), supplemented with 2% B27 (Invitrogen), penicillin (100 U/ml), streptomycin (100 mg/ml), and 2 mM glutamine. Neurons were transfected using the Lipofection method (Invitrogen) according to the manufacturer's suggestions. For immunocytochemistry, neurons were fixed in 4% paraformaldehyde/4% sucrose in PBS, incubated with the indicated antibodies (see: Extended Experimental Procedures) and imaged using a laser scanning Zeiss Pascal microscope.

Organotypic Slice Culture

Transverse hippocampal slices of 350 μm were prepared from 5- to 7-days-old mice and cultured. Slice cultures were transfected using a Helios Gene Gun (Biorad) at day 7 of culture (day in vitro, DIV). For each bullet, up to 3 different plasmids comprising a total of 50 μg of DNA were coated onto 12.5 mg of 1.6 μm gold particles, according to the manufacturer's protocol (Biorad). Slices were fixed at 13 DIV in 2.5% paraformaldehyde and 4% sucrose and processed for immunofluorescence as described above.

Confocal Image Analysis and Quantification

For dissociated neurons, images were obtained using a Zeiss Pascal confocal microscope. On average, a z stack of 5–7 optical sections (0.5 μm/section) was taken for each neuron imaged. Images were collected from 10 to 15 neurons per coverslip, with two coverslips for each condition per experiment. Synapse density, puncta size and intensity were measured using Metamorph software (Universal Imaging Corporation). Statistical significance was assessed using one-way ANOVA with a Bonferroni Post Test (Prism-GraphPad Software). For dendritic spine analysis, a z series projection of proximal apical secondary dendrites was made using 10–50 stacks (0.3 μm/stack), each averaged four times. Statistical significance was assessed using a one-way ANOVA with a Bonferroni Post Test (Prism-GraphPad Software).

Electrophysiological Analyses of In Vitro Neuronal Cultures

Primary hippocampal neurons were nucleofected with GFP or GFP and Cre after dissection. To record currents from those cells, the whole-cell configuration of the patch-clamp technique was used. Cells were voltage-clamped at an electrode potential of −60 mV. The analysis of mEPSC amplitude and frequency were performed with MiniAnalysis software (Synaptosoft). Currents were recorded using an Axopatch 200B amplifier, low-pass filtered at 2 kHz, digitized at 10 kHz using a Digidata 1320A interface, and acquired using pCLAMP10 software (Molecular Devices) on to the hard drive of a PC for off-line analysis.

Surface Labeling of Brain Slices

Hippocampal slices were prepared with a microslicer (Leica VT100S) and placed into biotinylation reagent solution of 1mg/ml NHS-SS-biotin (Pierce). Slices were homogenized and mixed with 40 μl of Streptavidin beads (Pierce). The beads were collected; the bound proteins were analyzed by SDS-PAGE and immunoblotting.

Golgi Impregnation and Analyses of Dendritic Spines

Brains dissected from adult (2–3 months-old) male conditional cyclin E knockout - Nestin-Cre and control littermates were stained using FD Rapid GolgiStain Kit (FD Neuro Technologies). Coronal sections (100 μm) were cut and analyzed using a conventional upright microscope under transmitted light. Using Neurolucida software (MBF Bioscience), CA1 pyramidal neurons were reconstructed and spine densities and sizes were quantified. Both spine length and width were measured using Metamorph software. Statistical significance was assessed using a multifactorial ANOVA with a Bonferroni Post Test (Prism-GraphPad Software).

Electron Microscopy

Animals were perfused intracardially with 2.5% glutaraldehyde, 2% paraformaldehyde. Tissues were postfixed with 1% osmium tetroxide (OsO4)/1.5% potassium ferrocyanide (KFeCN6) and embedded in TAAB Epon. Ultrathin sections were stained with lead citrate and examined in a JEOL 1200EX or a TecnaiG² Spirit BioTWIN transmission electron microscope. Images were recorded with an AMT 2k CCD camera. PSD length was measured using Metamorph software. Statistical significance was assessed using a multi-factorial ANOVA with a Bonferroni Post Test (Prism-GraphPad Software).

Electrophysiological Analyses of Acute Hippocampal Slices

Transversal hippocampal brain slices were prepared from 2–3 months old male conditional cyclin E knockout - Nestin-Cre and control mice. LTP was induced by theta-burst stimulation of Schaffer collaterals repeated five times at 30 s intervals. Whole cell patch-clamp recordings from CA1 neurons were made at holding potential of −70 mV, low-pass filtered at 2 kHz and sampled at rate of 10 kHz (Molecular Devices). Off-line analysis of excitatory synaptic currents (EPSC) was made using Minianalysis (Synaptosoft) and pClamp (Molecular Devices) software. To isolate NMDA component of EPSC recordings were made in magnesium (Mg)-free solution prior and during application of NMDA receptor antagonist, APV (20 μM).

Behavioral Analyses

These were performed on adult 2–3 month old male conditional cyclin E knockout - Nestin-Cre and control littermates, as described in Extended Experimental Procedures.

Supplementary Material

Click here to view.^{(5.4M, doc)}

Click here to view.^{(1.1M, xls)}

Click here to view.^{(62K, xls)}

ACKNOWLEDGEMENTS

We thank Drs. J-S Guan, M. Morabito, G. Corfas, Q. Ma, L-H. Tsai, H. Gardner, M. Ericsson, J. Wang, L. Richey, Y. Haydon and J.C. Dodart for help and advice, Y. Kim and P. Greengard for anti-phospho-S310 and -S441 and anti-total Wave1 antibodies, R. Betarbet for anti-phospho-Ser897 NR1 antibody, T. Ryan and P. De Camilli for Synapsin I cDNA. This work was supported by NIH grant R01CA108950 (to P.S.).

Footnotes

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Extended Experimental Procedures, seven figures and two tables.

REFERENCES

Bernasconi-Guastalla S, Wolfer DP, Lipp HP. Hippocampal mossy fibers and swimming navigation in mice: correlations with size and left-right asymmetries. Hippocampus. 1994;4:53–63. [PubMed]
Bird GC, Lash LL, Han JS, Zou X, Willis WD, Neugebauer V. Protein kinase A-dependent enhanced NMDA receptor function in pain-related synaptic plasticity in rat amygdala neurones. J Physiol. 2005;564:907–921. [PubMed]
Chergui K, Svenningsson P, Greengard P. Cyclin-dependent kinase 5 regulates dopaminergic and glutamatergic transmission in the striatum. Proc Natl Acad Sci USA. 2004;101:2191–2196. [PubMed]
Cooke SF, Bliss TV. Plasticity in the human central nervous system. Brain. 2006;129:1659–1673. [PubMed]
Crump FT, Dillman KS, Craig AM. cAMP-dependent protein kinase mediates activity-regulated synaptic targeting of NMDA receptors. J Neurosci. 2001;21:5079–5088. [PubMed]
Cruz JC, Tsai LH. Cdk5 deregulation in the pathogenesis of Alzheimer's disease. Trends Mol Med. 2004;10:452–458. [PubMed]
Dhavan R, Tsai LH. A decade of CDK5. Nat Rev Mol Cell Biol. 2001;2:749–759. [PubMed]
Geng Y, Yu Q, Sicinska E, Das M, Schneider JE, Bhattacharya S, Rideout WM, Bronson RT, Gardner H, Sicinski P. Cyclin E ablation in the mouse. Cell. 2003;114:431–443. [PubMed]
Geng Y, Yu Q, Whoriskey W, Dick F, Tsai KY, et al. Expression of cyclins E1 and E2 during mouse development and in neoplasia. Proc Natl Acad Sci USA. 2001;98:13138–13143. [PubMed]
Graus-Porta D, Blaess S, Senften M, Littlewood-Evans A, Damsky C, Huang Z, Orban P, Klein R, Schittny JC, Muller U. Beta1-class integrins regulate the development of laminae and folia in the cerebral and cerebellar cortex. Neuron. 2001;31:367–379. [PubMed]
Gudas JM, Payton M, Thukral S, Chen E, Bass M, Robinson MO, Coats S. Cyclin E2, a novel G1 cyclin that binds Cdk2 and is aberrantly expressed in human cancers. Mol Cell Biol. 1999;19:612–622. [PMC free article] [PubMed]
Hawasli AH, Benavides DR, Nguyen C, Kansy JW, Hayashi K, Chambon P, Greengard P, Powell CM, Cooper DC, Bibb JA. Cyclin-dependent kinase 5 governs learning and synaptic plasticity via control of NMDAR degradation. Nat Neurosci. 2007;10:880–886. [PubMed]
Hwang HC, Clurman BE. Cyclin E in normal and neoplastic cell cycles. Oncogene. 2005;24:2776–2786. [PubMed]
Ikeda Y, Matsunaga Y, Takiguchi M, Ikeda MA. Expression of cyclin E in postmitotic neurons during development and in the adult mouse brain. Gene Expr Patterns. 2011;11:64–71. [PubMed]
Kim Y, Sung JY, Ceglia I, Lee KW, Ahn JH, et al. Phosphorylation of WAVE1 regulates actin polymerization and dendritic spine morphology. Nature. 2006;442:814–817. [PubMed]
Kim SH, Ryan TA. CDK5 serves as a major control point in neurotransmitter release. Neuron. 2010;67:797–809. [PMC free article] [PubMed]
Koff A, Cross F, Fisher A, Schumacher J, Leguellec K, Philippe M, Roberts JM. Human cyclin E, a new cyclin that interacts with two members of the CDC2 gene family. Cell. 1991;66:1217–1228. [PubMed]
Lauper N, Beck AR, Cariou S, Richman L, Hofmann K, Reith W, Slingerland JM, Amati B. Cyclin E2: a novel CDK2 partner in the late G1 and S phases of the mammalian cell cycle. Oncogene. 1998;17:2637–2643. [PubMed]
Lew DJ, Dulic V, Reed SI. Isolation of three novel human cyclins by rescue of G1 cyclin (Cln) function in yeast. Cell. 1991;66:1197–1206. [PubMed]
Lew J, Huang QQ, Qi Z, Winkfein RJ, Aebersold R, Hunt T, Wang JH. A brain-specific activator of cyclin-dependent kinase 5. Nature. 1994;371:423–426. [PubMed]
Lin Y, Bloodgood BL, Hauser JL, Lapan AD, Koon AC, et al. Activity-dependent regulation of inhibitory synapse development by Npas4. Nature. 2008;455:1198–1204. [PMC free article] [PubMed]
Lucking C, Durr A, Bonifati V, Vaughan J, De Michele G, et al. Association between early-onset Parkinson's disease and mutations in the parkin gene. N. Engl. J. Med. 2000;342:1560–1567. [PubMed]
Malumbres M, Barbacid M. Cell cycle, CDKs and cancer: a changing paradigm. Nat Rev Cancer. 2009;9:153–166. [PubMed]
Matsubara M, Kusubata M, Ishiguro K, Uchida T, Titani K, Taniguchi H. Site-specific phosphorylation of synapsin I by mitogen-activated protein kinase and Cdk5 and its effects on physiological functions. J Biol Chem. 1996;271:21108–21113. [PubMed]
Miyajima M, Nornes HO, Neuman T. Cyclin E is expressed in neurons and forms complexes with cdk5. Neuroreport. 1995;6:1130–1132. [PubMed]
Morabito MA, Sheng M, Tsai LH. Cyclin-dependent kinase 5 phosphorylates the N-terminal domain of the postsynaptic density protein PSD-95 in neurons. J Neurosci. 2004;24:865–876. [PubMed]
Nakazawa K, Sun LD, Quirk MC, Rondi-Reig L, Wilson MA, Tonegawa S. Hippocampal CA3 NMDA receptors are crucial for memory acquisition of one-time experience. Neuron. 2003;38:305–315. [PubMed]
Nikolic M, Dudek H, Kwon YT, Ramos YF, Tsai LH. The cdk5/p35 kinase is essential for neurite outgrowth during neuronal differentiation. Genes Dev. 1996;10:816–825. [PubMed]
Ohtsubo M, Theodoras AM, Schumacher J, Roberts JM, Pagano M. Human cyclin E, a nuclear protein essential for the G1-to-S phase transition. Mol Cell Biol. 1995;15:2612–2624. [PMC free article] [PubMed]
Parisi T, Beck AR, Rougier N, McNeil T, Lucian L, Werb Z, Amati B. Cyclins E1 and E2 are required for endoreplication in placental trophoblast giant cells. Embo J. 2003;22:4794–4803. [PubMed]
Piedrahita D, Hernandez I, Lopez-Tobon A, Fedorov D, Obara B, et al. Silencing of CDK5 reduces neurofibrillary tangles in transgenic alzheimer's mice. J Neurosci. 2010;30:13966–13976. [PMC free article] [PubMed]
Raivich G, Bohatschek M, De Costa C, Iwata O, Galiano M, et al. The AP-1 transcription factor c-jun is required for efficient axonal regeneration. Neuron. 2004;43:57–67. [PubMed]
Ross PL, Huang YN, Marchese JN, Williamson B, Parker K, et al. Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol Cell Proteomics. 2004;3:1154–1169. [PubMed]
Samuels BA, Hsueh YP, Shu T, Liang H, Tseng HC, Hong CJ, Su SC, Volker J, Neve RL, Yue DT, Tsai LH. Cdk5 promotes synaptogenesis by regulating the subcellular distribution of the MAGUK family member CASK. Neuron. 2007;56:823–837. [PMC free article] [PubMed]
Sherr CJ, Roberts JM. Living with or without cyclins and cyclin-dependent kinases. Genes Dev. 2004;18:2699–2711. [PubMed]
Sindreu CB, Scheiner ZS, Storm DR. Ca2+ -stimulated adenylyl cyclases regulate ERK-dependent activation of MSK1 during fear conditioning. Neuron. 2007;53:79–89. [PMC free article] [PubMed]
Soderling SH, Langeberg LK, Soderling JA, Davee SM, Simerly R, Raber J, Scott JD. Loss of WAVE-1 causes sensorimotor retardation and reduced learning and memory in mice. Proc Natl Acad Sci USA. 2003;100:1723–1728. [PubMed]
Staropoli J, McDermott C, Martinat C, Schulman B, Demireva E, Abeliovich A. Parkin is a component of an SCF-like ubiquitin ligase complex and protects postmitotic neurons from kainate excitotoxicity. Neuron. 2003;37:735–749. [PubMed]
Tarricone C, Dhavan R, Peng J, Areces LB, Tsai LH, Musacchio A. Structure and regulation of the CDK5-p25(nck5a) complex. Mol Cell. 2001;8:657–669. [PubMed]
Tsai LH, Delalle I, Caviness VS, Jr., Chae T, Harlow E. p35 is a neural-specific regulatory subunit of cyclin-dependent kinase 5. Nature. 1994;371:419–423. [PubMed]
Tsai LH, Takahashi T, Caviness VS, Jr., Harlow E. Activity and expression pattern of cyclin-dependent kinase 5 in the embryonic mouse nervous system. Development. 1993;119:1029–1040. [PubMed]
Tsien JZ, Huerta PT, Tonegawa S. The essential role of hippocampal CA1 NMDA receptor-dependent synaptic plasticity in spatial memory. Cell. 1996;87:1327–1338. [PubMed]
Ubersax JA, Woodbury EL, Quang PN, Paraz M, Blethrow JD, Shah K, Shokat KM, Morgan DO. Targets of the cyclin-dependent kinase Cdk1. Nature. 2003;425:859–864. [PubMed]
Xu T, Yu X, Perlik AJ, Tobin WF, Zweig JA, Tennant K, Jones T, Zuo Y. Rapid formation and selective stabilization of synapses for enduring motor memories. Nature. 2009;462:915–919. [PMC free article] [PubMed]
Yang G, Pan F, Gan WB. Stably maintained dendritic spines are associated with lifelong memories. Nature. 2009;462:920–924. [PubMed]
Zariwala M, Liu J, Xiong Y. Cyclin E2, a novel human G1 cyclin and activating partner of CDK2 and CDK3, is induced by viral oncoproteins. Oncogene. 1998;17:2787–2798. [PubMed]