(Received for publication, December 2, 1996, and in revised form, December 24, 1996)
From the Departments of Anatomy and ¶ Medical
Biochemistry, The University of Calgary, Calgary, Alberta, Canada and
the
Department of Biochemistry, The Hong Kong University of
Science and Technology, Hong Kong
The tumor suppressor retinoblastoma protein (RB) plays a central role in cellular growth regulation, differentiation, and apoptosis. Phosphorylation of RB results in a consequent loss of its ability to inhibit cell cycle progression. However, how RB phosphorylation might be regulated in apoptotic or postmitotic cells, such as neurons, remains unclear. Here we report that neuronal Cdc2-like kinase (Nclk), composed of Cdk5 and a neuronal Cdk5 activator (p25nck5a), can bind and phosphorylate RB. Since RB has been shown recently to associate with D-type G1 cyclins and viral oncoproteins through a common peptide sequence motif of LXCXE, Nclk binding may be mediated by a related sequence motif (LXCXXE) found in p25nck5a. We demonstrate (i) in vitro binding of bacterially expressed p25nck5a to a GST-RB fusion protein, (ii) coprecipitation of GST-RB and reconstituted Cdk5·p25nck5a, and (iii) phosphorylation of GST-RB by bacterially expressed Cdk5·p25nck5a kinase and by Cdk5·p25nck5a kinase purified from bovine brain. Finally, we show that immunoprecipitation of RB from embryonic mouse brain homogenate results in the coprecipitation of Cdk5 and that Cdk5 kinase activity is maximal during late embryonic development, a period when programmed cell death of developing neurons is greatest. Taken together, these results suggest that Nclk can bind to and phosphorylate RB in vitro and in vivo. We infer that Nclk may play an important role in regulating the activity of RB in the brain, including perhaps in apoptosing neurons.
The retinoblastoma gene product, RB,1 initially identified as a tumor suppressor, is now also believed to play a key role in cellular processes such as cellular growth regulation, apoptosis, and differentiation. The growth regulatory function of RB was first demonstrated when loss of RB correlated with the development of retinoblastoma and several other human tumors (for review, see Refs. 1 and 2). Reintroduction of a functional RB gene into RB-deficient tumor cells results in reduced cell growth in culture and in nude mice (3, 4). Interestingly, unscheduled inactivation of RB by viral oncoproteins (HPV E7 and SV40 T-antigen) in transgenic mice results in massive apoptosis in differentiating and differentiated neurons and retinal cells (5-8). In addition, mouse embryos homozygous for mutated RB do not survive past day 16 of gestation as a result of aberrant terminal differentiation of erythrocytes and neurons as well as massive cell death in the central nervous system (9-11). These results suggest that RB has an essential role in the differentiation and apoptosis of specific-cell types. However, the biochemical mechanisms underlying these various aspects of RB activity are still unclear.
Structural and functional studies reveal that RB is composed of several distinct protein-binding domains (A/B pocket, large A/B pocket, and C pocket) that, when hypophosphorylated, can bind and inhibit the activity of a variety of cellular proteins required for cell cycle progression (for review, see Ref. 2). Viral oncoproteins can displace RB-binding proteins and drive normally resting cells into S phase (Refs. 12-16; for review, see Ref. 17). The ability of RB to bind and inactivate the regulatory proteins is reduced upon its phosphorylation at various sites. During progression through the mammalian cell cycle, RB becomes hyperphosphorylated and inactivated just prior to the G1/S phase transition as well as during S and G2 (for review, see Ref. 18). During later stages of mitosis RB is progressively dephosphorylated and becomes hypophosphorylated in early G1 phase of the cell cycle. In growth arrested, senescent, or terminally differentiated cells, RB remains largely hypophosphorylated (18, 19) and therefore inhibitory to mitotic progression. The kinases responsible for RB phosphorylation in vivo include members of the cyclin-dependent kinase (Cdk) family, including Cdk2 in association with cyclin E and A as well as Cdk4 and Cdk6 in association with D-type cyclins (18).
Neurons express Cdk5, which is a variant member of the Cdk family. Mammalian neuronal Cdc2-like kinase (Nclk) is composed of Cdk5 and a 25-kDa regulatory subunit (20, 21). The regulatory subunit is an amino-terminal truncated derivative of a 35-kDa protein expressed specifically in neurons of the central nervous system and these and a related protein are designated as neuronal Cdk5 activators (p25nck5a, p35nck5a, and p39nck5ai) (22, 23). Recent data suggest that Nclk has a role in neurite outgrowth and in regulating neurocytoskeletal dynamics (24-26). However, the precise function of Nclk and its kinase targets have yet to be clearly defined. RB is present in neurons where Nclk contributes the major Cdc2-like kinase activity (27-31), and we have therefore explored the possibility that RB may be a potential substrate for Nclk. We find that purified Nclk from bovine brain and reconstituted bacterially expressed Nclk phosphorylate RB in vitro. In addition, the activator protein of Nclk (p25nck5a) binds directly to RB, presumably via a LXCXE-related sequence motif (LXCXXE) in p25nck5a. Finally, Cdk5 coimmunoprecipitates with RB from embryonic mouse brain homogenate, and its kinase activity is maximal during late embryonic development (when RB phosphorylation and neuronal apoptosis are greatest), thus suggesting that the interaction between Nclk and RB may be functionally important in vivo.
Isopropyl--D-thiogalactopyranoside
was purchased from Life Technologies, Inc. [
-32P]ATP
(4500 Ci/mmol) was obtained from ICN. Nitrocellulose membrane was
obtained from Millipore. P81 phosphocellulose was purchased from
Whatman. All other chemicals were purchased from Sigma.
Histone
H1 kinase activity of the reconstituted enzymes was measured by
incubating the reaction mixture, containing 20 mM MOPS, pH
7.4, 30 mM MgCl2, 100 µM
[-32P]ATP (1000 cpm/pmol), and 100 µM
histone H1 peptide P9KTPKKAKKL18, at 30 °C
for 20 min as described previously (22).
[
-32P]phosphate incorporation into the substrate
peptide was quantitated by liquid scintillation using a Beckman LKB
1215 scintillation counter. RB phosphorylating activity was determined
using the expressed RB peptides (0.1 µg/µl) as substrates under the
same conditions. Samples were analyzed by SDS-PAGE followed by
autoradiography.
GST-fusion proteins were purified as described (32).
pGEX-2T recombinant plasmids expressing p70N-RB (N-terminal
portion of human RB, residues 1-379), p45C-RB (C-terminal
portion of human RB, residues 773-928), and p82L-A/B-RB
(large A/B pocket of human RB, residues 379-928) were kindly provided
by Drs. Harlow and Kaelin and were transformed into Escherichia coli strain DH5. The host cells were cultured in 4 liters of LB
medium containing 100 µg/ml ampicillin to
A600 nm = ~0.7 at 37 °C. The expression of
GST-fusion proteins was induced with 0.4 mM
isopropyl-
-D-thiogalactopyranoside at room temperature overnight. The cells were harvested by centrifugation at 5000 × g for 15 min, washed with 1 × phosphate-buffered
saline, suspended in 30 ml of lysis buffer (50 mM Tris-HCl,
pH 7.4, 2 mM EDTA, 1 mM DTT, 0.25 mM phenylmethylsulfonyl fluoride, 1 µg/ml leupeptin, 1 µg/ml aprotinin, and 1 µg/ml antipain) and lysed three times using
a French press (Aminco) with the cell pressurized to 16,000 p.s.i. The
resulting lysate was centrifuged at 18,000 × g for 20 min. The supernatant was incubated for 1 h at 4 °C with 4 ml of
glutathione-agarose (Sigma) previously equilibrated with the same
buffer. After loading the slurry onto a 10-ml column, the column was
washed with 10 bed volumes of 1 × phosphate-buffered saline
supplemented with 0.25 M KCl, 0.1% Tween 20, 1 mM DTT, 0.25 mM phenylmethylsulfonyl fluoride,
1 µg/ml leupeptin, 1 µg/ml aprotinin and 1 µg/ml antipain,
followed by 5 bed volumes of 1 × phosphate-buffered saline
supplemented with 1 mM DTT. The expressed GST-fusion
proteins were eluted with 5 mM reduced glutathione in 50 mM Tris-HCl, pH 8.0, 1 mM DTT.
0.3 µg of GST-p25nck5a and GST-Cdk5 were incubated with 1 unit of thrombin at room temperature for 40 min, and the digested GST portions were removed by precipitating with glutathione-agarose beads. GST-removed supernatant fractions containing either p25nck5a alone (0.1 µg) or Cdk5·p25nck5a (0.1 µg/each) were mixed with 1 µg of GST-p82L-A/B-RB and incubated at room temperature for 40 min. Complex formation between GST-p82L-A/B-RB and p25nck5a (or Cdk5·p25nck5a) was assessed by measuring the ability of p25nck5a or Cdk5·p25nck5a to bind the glutathione-agarose beads. The presence of p25nck5a and Cdk5·p25nck5a was determined by Western blot analysis with p25nck5a- and Cdk5-specific antibodies as described below.
SDS-PAGE, Immunoblots, and AntibodiesSDS-PAGE was performed by the method of Laemmli (33) in 10 or 12.5% vertical slab gels. For Western blot analyses, samples were analyzed by SDS-PAGE followed by transfer to nitrocellulose membranes (Millipore). The blots were probed with primary antibodies (1/500 dilution of a rabbit polyclonal p25nck5a-specific antibody (22) or 1/2000 dilution of a rabbit monoclonal Cdk5-specific (DC-17) antibody (Santa Cruz Biotechnology)) followed by incubation with a 1/2500 dilution of secondary antibodies (either goat anti-rabbit IgG or goat anti-mouse IgG conjugated with horseradish peroxidase (Pierce)) and developed by ECL reagents as specified by the manufacturer (Amersham Life Science, Inc.).
ImmunoprecipitationEmbryonic and adult mice brains (E10, E12, E14, E16, E18, E20, and adult) were isolated as described previously by Drago et al. (34) and homogenized in buffer containing 20 mM HEPES, pH 7.4, 1 mM EDTA, 1 mM DTT, 0.25 mM phenylmethylsulfonyl fluoride, 1 µg/ml leupeptin, 1 µg/ml aprotinin, and 1 µg/ml antipain. 150 µg of the homogenates were precleared with 20 µl of protein G Plus-agarose beads (Santa Cruz Biotechnology) and incubated with 20 µl (2 µg) of RB-specific antibody (IF8; Santa Cruz Biotechnology) for 2 h on a rotating platform at 4 °C. The immunocomplexes were absorbed to protein G Plus-agarose beads and washed extensively with 1 × Tris-buffered saline with 0.1% Tween 20. The presence of RB and Cdk5 in the immunoprecipitated samples was revealed by immunoblotting with a RB-specific (C-15) antibody and a Cdk5-specific (DC-17) antibody from Santa Cruz Biotechnology.
Protein Concentration DeterminationProtein concentration was determined by Bradford microassay (35).
Protein kinases capable of phosphorylating retinoblastoma protein have proven to be members of the cyclin-dependent kinase (Cdk) family. Cdc2, Cdk2, Cdk4, and Cdk6 phosphorylate RB in vitro at a subset of sites that are known to be phosphorylated in vivo, and both Cdc2 and Cdk2 have been found in association with RB in vivo (36-41). The cyclin component(s) of these RB kinases have yet to be definitively established, although transfection of cyclins A, D, and E can stimulate RB phosphorylation in vivo (42, 43). In postmitotic neuronal cells, the major Cdk is Cdk5. It has been shown that cyclin D1 can bind to Cdk5, but the complex displays no demonstrable kinase activity, possibly due to the presence in these complexes of the Cdk inhibitor, p21CIP1 (44). Recently, a highly active neuronal Cdc2-like kinase (Nclk) that is composed of Cdk5 and a 25-kDa regulatory subunit (p25nck5a) has been purified from postmitotic neurons of the central nervous system (20, 21). Since Nclk can phosphorylate cytoskeletal and membrane-associated proteins (neurofilament proteins, Tau, and synapsin-I; 25, 26, 45) and its expression pattern and activity correlate with terminal differentiation of neurons (46), it has been suggested that Nclk has a role in regulating neurocytoskeletal dynamics and/or neuronal differentiation (21, 24, 47). Another prominent feature of late embryonic brain development is the apoptosis of many neurons that apparently have failed to form proper connections. Because Cdk5 kinase activity increases during brain development while Cdc2, Cdk2, and Cdk4 activities decrease (47, 48), and because phosphorylation of RB is involved with both cell cycle progression (12-19) and apoptosis (7, 8, 49, 50) and because loss of RB activity can lead to enhanced neuronal death (5-11), we asked whether RB might be a substrate for Nclk.
We first established by several means that RB could serve as a
substrate for Nclk in vitro. Nclk purified from bovine brain and known to exhibit high histone H1 kinase activity (21) was examined
for its RB-phosphorylating activity using various truncated forms of
human GST-RB proteins: GST-p70N-RB, GST-p45C-RB
and GST-p82L-A/B-RB. As shown in Fig.
1A, each RB peptide was capable of being
phosphorylated by Nclk. To demonstrate that the RB-phosphorylating
activity of bovine brain Nclk results from Cdk5·p25nck5a but
not from Cdk5·cyclin D1 complexes, we performed similar assays using
bacterially expressed and reconstituted GST-Cdk5·GST-p25nck5a
as well as GST-Cdk5·GST-cyclin D1. We previously had shown that reconstitution of GST-Cdk5 and GST-p25nck5a resulted in an
active H1 kinase (Ref. 51 and Fig. 1C), and when tested with
RB we found that reconstituted Nclk was also capable of efficiently
phosphorylating all three RB substrates (Fig. 1B). In
contrast, GST-Cdk5 complexed with GST-cyclin D1 showed negligible
histone H1 kinase activity (Fig. 1C) nor was it capable of
phosphorylating RB (data not shown). The addition of purified
Cdk-activating kinase from bovine brain (which in separate tests
activated Cdk2·cyclin A kinase activity 20-fold; Ref. 52) did not
significantly increase Cdk5·cyclin D1 histone H1 kinase activity.
However, when increasing concentrations of cyclin D1 were added to the
mixture of Cdk5 and p25nck5a, a substantial decrease in histone
H1 kinase activity was observed. These results confirm that excess
cyclin D1 can interact with Cdk5, disrupting its interaction with
p25nck5a and resulting in a kinase-inactive complex (even in
the absence of added p21CIP1; Ref. 44). Our results therefore
suggest that in vivo purified or in vitro
reconstituted Cdk5·p25nck5a can phosphorylate RB and that
Cdk5·cyclin D1 complexes are either inactive or they may target
presently unidentified substrates.
Because Nclk evidently can phosphorylate nonoverlapping NH2-terminal and COOH-terminal RB peptide fragments in vitro (Fig. 1), we considered the distribution within the intact protein of residues that might serve as targets for the enzyme. Of the 16 Ser/Thr-Pro motifs that are potential Cdk phosphorylation sites in RB, at least eight (Ser-249, Thr-252, Thr-373, Ser-780, Ser-807, Thr-821, and Thr-826) have been shown to be phosphorylated in vivo (36, 53-56), three of which are in GST-p70N-RB, four in GST-p82L-A/B-RB and three in GST-p45C-RB. RB is therefore a potentially valid substrate of both in vitro reconstituted and purified in vivo Nclk, although it remains to be determined which sites are phosphorylated in vivo by this enzyme and whether their phosphorylation alters RB function. Because recent studies suggest that Ser-780 is preferentially phosphorylated by Cdk4·cyclin D1, but not by Cdk2·cyclin E/A (56), it is likely that Nclk may also display specific RB phosphorylation patterns in vivo.
Nclk Forms a Stable Complex with RB both in Vitro and in VivoThe in vitro phosphorylation of RB by Nclk suggests that Nclk may interact with RB in vivo. Because LXCXE and the related LXCXXE sequence motifs in D-type cyclins and several viral oncoproteins (such as adenovirus E1A protein, SV-40 T-antigen, and human papilloma virus-16 E7 protein) mediate the formation of complexes with RB (Refs. 42 and 57; for review, see Ref. 58), we searched for LXCXE and LXCXXE motifs in Nck5a. We found the LXCXXE sequence motif (L152RCLGE157) in Nck5a, as well as related motifs (L187RCLGD192) in p39nck5ai and (L372AMGTD377) in p67, another Cdk5 activator found in rat spinal cord (59). Thus, the presence of these related motifs in all the Cdk5-activating proteins raises the intriguing possibility of functional convergence for binding to RB.
Because p25nck5a contains a potential
LXCXXE binding site for RB, we next determined
whether it can bind directly to RB in vitro. GST-p82L-A/B-RB was incubated with GST-free
p25nck5a and the ability of the p25nck5a to
coprecipitate following the addition of glutathione-agarose beads was
assessed. As demonstrated in Fig. 2 (lane 4),
Western blot analysis with Nck5a-specific antibody showed that a
significant portion of p25nck5a precipitated with
glutathione-agarose beads due to its interaction with
GST-p82L-A/B-RB (but not with GST alone; lane
6). As a further control, GST-free p25nck5a by itself did
not precipitate with the beads under the same conditions (lane
2) but remained in the supernatant fraction (lane 3).
Therefore, our results suggest that p25nck5a can bind directly
to the expressed RB fragment containing the large A/B pocket. In
contrast, similar in vitro binding assays with GST-free Cdk5
revealed no significant coprecipitation with GST-p82L-A/B-RB (data not shown). Similar binding
experiments were conducted with p25nck5a and
GST-p70N-RB and GST-p45C-RB, revealing that
p25nck5a also associates with the other fragments of RB
(consistent with its ability to phosphorylate them), although the
efficiency of its binding appears somewhat lower (data not shown).
We further investigated whether the Cdk5·p25nck5a complex can
bind stably to RB and whether this interaction facilitates RB
phosphorylation. To do so, we assessed the coprecipitation of
Cdk5·p25nck5a with GST-p82L-A/B-RB. We first
removed the GST portion from both GST-Cdk5 and GST-p25nck5a by
thrombin digestion, followed by clearance of residual GST-containing fragments by precipitation with glutathione-agarose beads. The remaining Cdk5 and p25nck5a polypeptides were combined,
incubated with GST-p82L-A/B-RB, and reprecipitated with
glutathione-agarose beads. As shown in Fig.
3A, Western blot analyses with Cdk5- and
Nck5a-specific antibodies indicated that the addition of the
glutathione-agarose beads to a mixture of GST-p82L-A/B-RB
and reconstituted Cdk5·p25nck5a results in precipitation of
both Cdk5 and p25nck5a, whereas precipitation was not observed
in the absence of GST-p82L-A/B-RB. In addition, when the
precipitated sample was incubated with [-32P]ATP and
Mg2+, phosphorylation of GST-p82L-A/B-RB was
observed (Fig. 3B), consistent with the formation of a
kinase-active complex between Cdk5·p25nck5a and RB.
Previous work has demonstrated the expression of RB in a variety of
tissues, including brain, during development (27-29). To determine
whether the binding between Nclk and RB that we detect in
vitro might also exist in vivo, we immunoprecipitated
RB-containing complexes from embryonic mouse brain and tested for the
coprecipitation of Nclk. As shown in Fig. 4, RB-specific
antibody precipitated RB from crude homogenates of embryonic mouse
brain (E18) and coprecipitated Cdk5, whereas nonspecific mouse IgG
precipitated neither (the comigration of p25nck5a with antibody
light chains precluded its specific detection in this assay). Our
results therefore strongly suggest the existence of Nclk·RB complexes
in vivo.
Nclk Phosphorylation of RB Is Maximal during Late Fetal Development of the Brain
To assess the possible significance of the binding
and phosphorylation of RB by Nclk, we measured Nclk kinase activity
during development using RB as a substrate (Fig. 5).
Brain tissues were isolated from adult mice and embryos at 10-20 days
post coitum. Nclk immunoprecipitated from these tissue
homogenates was then immunoblotted with Cdk5-specific antibody (Fig.
5A), revealing a gradual increase in Cdk5 abundance as
described previously (46). The same preparations were then incubated
with GST-p82L-A/B-RB, [-32P]ATP, and
Mg2+ and phosphorylation of RB peptide assessed by
autoradiography (Fig. 5B). After correction for changes in
the abundance of Cdk5, we found that Nclk kinase specific activity
increased rapidly at 16 days of development, was maximal by 18 days,
and declined to lower levels in adulthood (Fig. 5C). Similar
results were obtained when histone H1 peptide was used as a substrate
for Nclk (data not shown), and we estimate that maximal Nclk kinase
specific activity at E18 was 5-fold greater than the level observed in adult brain. The down-regulation of Nclk activities for E20 and adult
that we observe is clearly not due to a diminished abundance of Cdk5,
which is essentially constant over this time period. Instead, it
appears most likely that the decreased kinase specific activity arises
either from a decrease in the abundance of an activator protein or the
enhancement of an inhibitory pathway. These possibilities are currently
under investigation.
It is striking that peak Nclk activity is observed at a time when neurogenesis in the mouse brain is already largely complete and apoptosis of nerve cells is most prominent. Approximately half of all neurons generated during development of the mammalian nervous system die (60) and in the cerebral cortex of the fetal mouse brain neuronal apoptosis is maximal between E14 and E18 (61), just as we see for Nclk activation. Previous work has already documented differential RB phosphorylation during embryonic mouse brain development (29), although the precise significance of this observation has been unclear. Because the inactivation of RB by viral oncoproteins can induce apoptosis in neurons (5-8) and mice homozygous for defective alleles of RB show depleted neuronal populations (9-11), it is intriguing to speculate that highly active Nclk in late fetal development may be responsible for the hyperphosphorylation and inactivation of RB as a prelude to apoptosis in those neurons that are destined to die. It is also possible that inappropriate Nclk phosphorylation of target proteins is an important feature of changes arising during neurodegenerative diseases, where neuronal death and Nclk activation are both observed (21, 46, 47).2
In summary, our results suggest that Nclk can bind to and phosphorylate RB in vitro, that it also associates with RB in vivo, and that these interactions are mediated by a direct association between p25nck5a and RB. Furthermore, Nclk kinase activity is maximal during late fetal development of the brain, a period when neuronal apoptosis is prominent. These relationships, and the mechanisms by which Nclk kinase activity are developmentally regulated, are currently under investigation.
We thank Drs. Harlow and Kaelin for providing pGEX-2T recombinant plasmids expressing p70N-RB, p45C-RB and p82L-A/B-RB. We also extend our gratitude to William Durham and Markéta Gogela-Spehar for their helpful assistance in the final stage of completion of this manuscript.