Advertisement

Modulation of p27Kip1 levels by the cyclin encoded by Kaposi's sarcoma‐associated herpesvirus

David J. Mann, Emma S. Child, Charles Swanton, Heike Laman, Nic Jones

Author Affiliations

  1. David J. Mann*,1,2,
  2. Emma S. Child1,
  3. Charles Swanton2,
  4. Heike Laman2 and
  5. Nic Jones2
  1. 1 Department of Biology and Biochemistry, Brunel University, Uxbridge, Middlesex, UB8 3PX, UK
  2. 2 Imperial Cancer Research Fund, 44 Lincoln's Inn Fields, London, WC2A 3PX, UK
  1. *Corresponding author. E-mail: David.Mann{at}brunel.ac.uk

Abstract

DNA tumour viruses have evolved a number of mechanisms by which they deregulate normal cellular growth control. We have recently described the properties of a cyclin encoded by human herpesvirus 8 (also known as Kaposi's sarcoma‐associated herpesvirus) which is able to resist the actions of p16Ink4a, p21Cip1 and p27Kip1 cdk inhibitors. Here we investigate the mechanism involved in the subversion of a G1 blockade imposed by overexpression of p27Kip1. We demonstrate that binding of K cyclin to cdk6 expands the substrate repertoire of this cdk to include a number of substrates phosphorylated by cyclin–cdk2 complexes but not cyclin D1–cdk6. Included amongst these substrates is p27Kip1 which is phosphorylated on Thr187. Expression of K cyclin in mammalian cells leads to p27Kip1 downregulation, this being consistent with previous studies indicating that phosphorylation of p27Kip1 on Thr187 triggers its downregulation. K cyclin expression is not able to prevent a G1 arrest imposed by p27Kip1 in which Thr187 is mutated to non‐phosphorylatable Ala. These results imply that K cyclin is able to bypass a p27Kip1‐imposed G1 arrest by facilitating phosphorylation and downregulation of p27Kip1 to enable activation of endogenous cyclin–cdk2 complexes. The extension of the substrate repertoire of cdk6 by K cyclin is likely to contribute to the deregulation of cellular growth by this herpesvirus‐encoded cyclin.

Introduction

The critical period in cell‐cycle control occurs late in the first growth phase (G1), when cells must assess whether to irrevocably commit themselves to divide or to enter a quiescent state. This transition is determined by the balance between positive and negative factors: the D‐type cyclins (D1, D2 and D3) complexed to cyclin‐dependent kinases (cdk) 4 or 6 and cyclin E–cdk2 promote cell cycling whilst cdk inhibitors such as p16Ink4a and p27Kip1 retard cell division (reviewed in Sherr and Roberts, 1995; Weinberg, 1995; Bartek et al., 1996). The central role of these molecules in the regulation of proliferation is emphasized by the frequency of anomalies in their expression and activities in human cancer (reviewed in Hall and Peters, 1996).

Elevations in the levels of p27Kip1 are implicated in the maintenance of quiescence in a number of cell types (Kato et al., 1994; Nourse et al., 1994; Coats et al., 1996; Rivard et al., 1996). Although mutations in the p27Kip1 gene are rare (Kawamata et al., 1995; Pientenpol et al., 1995; Ponce‐Castaneda et al., 1995), it is becoming increasingly clear that aberrant levels of p27Kip1 can have a significant bearing on prognosis in a number of human cancers. For example, low levels of p27Kip1 correlate with poor rates of survival for breast and colorectal carcinoma patients (Catzavelos et al., 1997; Porter et al., 1997) and aggressive colorectal carcinomas often exhibit elevated levels of p27Kip1 degradation (Loda et al., 1997). It is evident, therefore, that the cellular mechanisms which regulate the levels of this cdk inhibitor must be stringently controlled. This has been demonstrated experimentally in T lymphocytes and serum‐deprived fibroblasts where elevated p27Kip1 levels appear to be the primary determinant in the maintenance of the quiescent state (Nourse et al., 1994; Coats et al., 1996; Rivard et al., 1996).

One apparent mechanism by which p27Kip1 levels can be modulated is through ubiquitin‐mediated proteolysis (Pagano et al., 1995). Regulation of ubiquitin‐mediated proteolysis is often achieved by phosphorylation of the target protein making it more susceptible to degradation (for example, see Chen et al., 1995; Lanker et al., 1996). This may also be the case with p27Kip1 since its downregulation is enhanced through cyclin E–cdk2‐mediated phosphorylation on Thr187 (Muller et al., 1997; Sheaff et al., 1997; Vlach et al., 1997), although a causal link to ubiquitin‐mediated degradation has not been demonstrated.

Much of our understanding of growth control in eukaryotes stems from the study of its dysregulation by viral proteins. For example, the ability of the simian virus 40 (SV40) T antigen, adenovirus E1A and human papilloma Virus E7 to displace proteins sequestered by the retinoblastoma protein (pRb) led to the realization of the central role of pRb in preventing S phase entry (reviewed in Ewen, 1994). We have recently described a novel mechanism by which viral proteins can subvert normal growth control: cyclins encoded by certain γ‐herpesviruses can confer resistance to their associated catalytic cdk subunit against the cdk inhibitor proteins (Swanton et al., 1997). One such cyclin, K cyclin, is encoded by human herpesvirus 8 (HHV8), also known as Kaposi's sarcoma‐associated herpesvirus. This human tumour virus is strongly implicated as a causative agent for Kaposi's sarcoma (Schulz et al., 1998) and a number of lymphoproliferative disorders including body cavity‐based lymphomas (Cesarman et al., 1995). K cyclin forms an active complex with cdk6 (Chang et al., 1996). The resistance of this cyclin–cdk complex to inhibition by p16Ink4a, p21Cip1 and p27Kip1 enables efficient subversion of G1 arrests imposed by elevated levels of these cdk inhibitors (Swanton et al., 1997).

Here we describe experiments designed to address the mechanism by which K cyclin overcomes a p27Kip1‐imposed growth arrest. We show that the K cyclin–cdk6 complex can phosphorylate p27Kip1 on Thr187, thus triggering the degradation of this cdk inhibitor by the proteasome. The phosphorylation of p27Kip1 results from the ability of K cyclin to extend the range of protein targets that can be phosphorylated by cdk6. The combination of resistance to inhibition and modulation of cdk substrate specificity by K cyclin is likely to be important for HHV8‐induced deregulation of normal growth control.

Results

Overexpression of p27Kip1 in mammalian cells leads to arrest in G1 due to the action of this cdk inhibitor on G1‐specific cyclin–cdk complexes (Kato et al., 1994; Polyak et al., 1994; Toyoshima and Hunter, 1994). We have recently demonstrated that K cyclin can efficiently bind to and activate cdk6 and that K cyclin overexpression can bypass a p27Kip1‐imposed G1 arrest (Swanton et al., 1997). A p27Kip1‐imposed G1 blockade is achieved by inhibition of not only cdk4/6‐containing complexes, but also inhibition of cyclin E–cdk2. Thus, we sought to address the mechanism by which this effect was achieved with a view to further understanding the role of K cyclin in subversion of G1/S control. A number of mechanisms could explain the ability of K cyclin overexpression to circumvent p27Kip1‐mediated G1 arrest: (i) K cyclin may complex with cdk2 (as well as with cdk6) to functionally complement cyclin E and form active p27Kip1‐resistant holoenzymes with cdk2; (ii) K cyclin–cdk6 complexes may exhibit an extended substrate repertoire such that they can functionally substitute for cyclin–cdk2 complexes; (iii) K cyclin expression may facilitate activation of the endogenous cyclin–cdk2 complexes by, for example, promoting the phosphorylation and subsequent downregulation of p27Kip1.

K cyclin extends the substrate repertoire of cdk6

To test the substrate range of specific holoenzymes, we co‐infected Sf9 cells with recombinant baculovirus directing the expression of a cyclin and a cdk. Lysates from these cells were immunoprecipitated through the cdk component and the immune complexes were tested for their ability to phosphorylate various substrates. Figure 1A demonstrates that all cyclin–cdk combinations tested generated immunoprecipitable kinase activity against the C‐terminus of pRb synthesized in bacteria as a glutathione S‐transferase (GST) fusion protein. Immunoprecipitations from control lysates did not yield detectable kinase activity demonstrating the specificity of the assay (Figure 1A). Histone H1 was also included in these assays and, in confirmation of previous observations (Chang et al., 1996), K cyclin–cdk6 complexes could efficiently phosphorylate this substrate to significantly greater levels than observed with cyclin D1–cdk6. Indeed, with similar levels of pRb kinase activity, K cyclin–cdk6 was as efficient as cyclin E–cdk2 in phosphorylating histone H1.

Figure 1.

Characterization of the substrate specificity of K cyclin–cdk6 complexes. (A) Lysates of Sf9 cells co‐infected with recombinant baculovirus directing expression of either K cyclin and cdk6 or cyclin E and cdk2 were immunoprecipitated through the cdk subunit and assayed for their ability to phosphorylate pRb or histone H1. (B) Lysates of Sf9 cells co‐infected with recombinant baculovirus directing expression of cdk6 and either cyclin D1 or K cyclin were immunoprecipitated through the cdk subunit. Immunoprecipitates were normalized for pRb kinase activity (upper panel). Equivalent amounts of cyclin D1–cdk6 and K cyclin–cdk6 pRb kinase activity were assayed for their ability to phosphorylate the cdk2 substrates p27Kip1, Id‐2 and cdc25a.

We next tested the activity of these kinases against a panel of cdk2 substrates (cdc25a, Id‐2 and p27Kip1; Hoffmann et al., 1994; Hara et al., 1997; Muller et al., 1997; Sheaff et al., 1997; Vlach et al., 1997). Baculovirus‐infected Sf9 cell lysates were immunoprecipitated through cdk6 and the kinase activity of each immune complex was assessed against pRb (Figure 1B). Then, using equivalent amounts of pRb kinase activity, each immunoprecipitate was tested for its ability to phosphorylate p27Kip1, Id‐2 and cdc25a. As shown in Figure 1B, K cyclin–cdk6 was able to phosphorylate each of these cdk2 substrates whereas cyclin D1–cdk6 was a very inefficient kinase against these proteins. Thus, K cyclin–cdk6 displays an extended substrate repertoire when compared with cyclin D1–cdk6 and more closely resembles cyclin E–cdk2.

K cyclin can activate cdk2

We next tested the ability of K cyclin to bind to and activate cdk2. Cdks were in vitro transcribed and translated in the presence of 35S methionine and mixed with an unlabelled in vitro transcribed and translated cyclin. After 30 min incubation, the samples were immunoprecipitated through the cyclin and the products resolved by SDS–PAGE (Figure 2A). Autoradiography revealed that K cyclin complexed efficiently with cdk6 and more weakly with cdk2 and cdk3. A similar pattern of interaction was seen with cyclin D1. In contrast to the strong binding of cyclin D1 to cdk4, interaction of K cyclin with cdk4 was weak although co‐infection of insect cells with K‐cyclin‐ and cdk4‐expressing viruses does result in an active cdk complex (C.Swanton and N.Jones, unpublished observations).

Figure 2.

K cyclin forms active complexes with cdk2. (A) cdk subunits were synthesized by in vitro transcription/translation in the presence of [35S]methionine and mixed with either Flag‐tagged cyclin D1 or Flag‐tagged K cyclin also synthesized by coupled‐transcription/translation but with unlabelled methionine. After 30 min incubation, samples were immunoprecipitated through the epitope tag of the cyclin and products resolved by SDS–PAGE. Gels were dried and subjected to autoradiography to reveal the presence or absence of interacting labelled cdk subunits in the cyclin immunoprecipitate. (B) Lysates of Sf9 cells co‐infected with recombinant baculovirus directing expression of either K cyclin and cdk2 or cyclin E and cdk2 were assayed for their ability to phosphorylate pRb in the absence or presence of the cdk inhibitors p21Cip1 (10, 100 or 1000 ng) or p27Kip1 (7.5, 75 or 750 ng). Reaction products were resolved by SDS–PAGE and visualized by autoradiography. (C) U2OS cells were transiently transfected with 10 μg of either empty pcDNA3 vector (control) or vector containing Flag‐tagged K cyclin or cyclin D1. Lysates were prepared and subjected to immunoprecipitation with anti‐flag antibody and the immunoprecipitated proteins subjected to immunoblot analysis with the indicated antibodies.

To assess whether K cyclin–cdk2 complexes were active kinases, Sf9 cells were co‐infected with recombinant baculovirus directing expression of cdk2 and either cyclin E or K cyclin. Lysates from these cells were tested for their ability to phosphorylate the C‐terminus of pRb. Both cyclin–cdk2 complexes readily phosphorylated pRb (Figure 2B). Lysates from Sf9 cells infected with either a cyclin‐ or a cdk‐expressing baculovirus did not yield significant pRb kinase activity (data not shown). We next tested the sensitivities of these activities to the cdk inhibitors p21Cip1 and p27Kip1. p21Cip1 and p27Kip1 produced in bacteria were included in the kinase assays at increasing concentrations. Cyclin E–cdk2 activity was rapidly abolished by addition of either inhibitor, whereas K cyclin–cdk2 activity was at least 10‐fold less sensitive to p21Cip1 and p27Kip1 than cyclin E–cdk2, with activity only fully abolished at the highest concentration of inhibitor tested. The K cyclin–cdk2 complexes were, however, more sensitive to p21Cip1 and p27Kip1 than K cyclin–cdk6 complexes (Swanton et al., 1997). Thus, K cyclin can bind to and activate cdk2 to generate a holoenzyme that is partially resistant to the inhibitory action of p21Cip1 and p27Kip1. K cyclin–cdk2 complexes, like K cyclin–cdk6 complexes, fail to bind stably to either p21Cip1 or p27Kip1 (results not shown; Swanton et al., 1997).

The results described above showed that K cyclin could associate with and activate different cdks in vitro. We determined whether this was also the case in vivo by transfecting U2OS cells with an expression vector containing a flag‐epitope tagged version of K cyclin. The cyclin was subsequently immunoprecipitated and associated proteins analyzed by Western blotting. Cells transfected with epitope‐tagged cyclin D1 were similarly assayed. As shown in Figure 2C, K cyclin was found to be associated with cdk6, cdk4 and cdk2. In contrast, cyclin D1 associated efficiently with cdk4, poorly with cdk6 and not at all with cdk2. Thus, the in vivo data corroborated the pattern of cdk interaction seen with K cyclin in vitro. These results are at variance with the results of Godden‐Kent et al. (1997) who demonstrated efficient interaction with cdk6 only. It is not clear why the two studies differ in this regard but may result from the use of different cell lines. The immunoprecipitated K cyclin complexes were active in phosphorylation of pRb and were insensitive to inhibition with p27Kip1 (data not shown).

Phosphorylation and downregulation of p27Kip1 by K cyclin expression

Downregulation of p27Kip1 has been reported to be triggered by phosphorylation of the cdk inhibitor on Thr187 by cyclin E–cdk2 complexes (Muller et al., 1997; Sheaff et al., 1997; Vlach et al., 1997). Given that K cyclin–cdk6 complexes can phosphorylate p27Kip1 (Figure 1B) we investigated whether this kinase can phosphorylate p27Kip1 on this specific threonine. Sf9 cells were co‐infected with baculoviruses directing the expression of cdk6 and K cyclin or cdk2 and cyclin E. Three days post‐infection, cells were lysed and extracts used to phosphorylate the C‐terminus of p27Kip1 (amino acids 91–198 fused to GST) in the presence of [γ‐32P]ATP. Reaction products were resolved by SDS–PAGE, transferred to nylon membrane and digested with trypsin. Peptides were resolved by two‐dimensional peptide mapping and phosphorylated products detected by autoradiography. GST–p27Kip1(91–198) phosphorylated by either cyclin E–cdk2 or K cyclin–cdk6 generated maps with a single major phosphopeptide (Figure 3A and B, respectively). Mixing equal amounts of radioactivity from each tryptic digestion gave rise to a single phosphopeptide spot after two‐dimensional mapping (Figure 3C), illustrating that these two cyclin–cdk holoenzymes phosphorylated GST–p27Kip1(91–198) on peptides with identical electrophoretic and chromatographic properties. Phosphoamino acid analysis of this phosphopeptide from samples labelled with either K cyclin–cdk6 or cyclin E–cdk2 demonstrated that it contained only phosphothreonine (Figure 3G).

Figure 3.Figure 3.
Figure 3.

Analysis of site of phosphorylation of p27Kip1 by K cyclin–cdk6. A C‐terminal fragment of p27Kip1 synthesized as a GST fusion protein in bacteria was bound to glutathione–Sepharose beads and phosphorylated by either K cyclin–cdk6 (A) or cyclin E–cdk2 (B) produced in Sf9 cells via infection with recombinant baculovirus. Sepharose‐bound C‐terminal p27Kip1 was washed with PBS plus 0.1% Tween 20 and resolved by SDS–PAGE. Radioactive bands were excised and subjected to tryptic digestion prior to two‐dimensional phosphopeptide analysis. Mixing equal counts per minute from C‐terminal p27Kip1 phosphorylated by either K cyclin–cdk6 or cyclin E–cdk2 prior to phosphopeptide analysis yielded a single radioactive spot (C). Phosphorylation of full length, wild‐type (D) or T187A mutant (E) p27Kip1 fused to GST by K cyclin–cdk6 generated a single major phosphopeptide. Mixing equal counts per minute from K cyclin–cdk6 phosphorylated wild‐type and T187A mutant p27Kip1 generated two distinct phosphopeptides (F). Radioactive phosphopeptides from maps shown in (A, B, D and E) were recovered from cellulose plates and hydrolysed prior to phosphoamino acid analysis (G).

We next performed similar experiments using full‐length p27Kip1 fused to GST together with a mutant version of full‐length p27Kip1 in which Thr187 is replaced with Ala [GST–p27Kip1(T187A)]. Two‐dimensional phosphopeptide mapping of the full‐length wild‐type p27Kip1 fusion protein treated with K cyclin–cdk6 generated a single major phosphopeptide (Figure 3D) which was indistinguishable by mixing experiments with that from phosphorylated C‐terminal p27Kip1 (not shown) and contained only phosphothreonine (Figure 3G). Surprisingly, GSTp27Kip1(T187A) also generated a single major phosphopeptide after phosphorylation with K cyclin–cdk6 (Figure 3E). However, this phosphopeptide did not comigrate with that from K cyclin–cdk6‐treated wild‐type p27Kip1 (Figure 3F) and contained only phosphoserine (Figure 3G). Taken together, these results demonstrate that K cyclin–cdk6, like cyclin E–cdk2, can phosphorylate p27Kip1 on Thr187. In addition, upon mutation of Thr187 to Ala, there is an increase in susceptibility of a Ser residue to phosphorylation by the K cyclin–cdk6 holoenzyme.

We next examined the effects of K cyclin expression on p27Kip1 in terms of both phosphorylation status and protein abundance in mammalian cells. For these experiments, we used a cell line derived from NIH 3T3 cells in which K cyclin expression is under negative control of the bacterial lacI gene product and thus can be induced by the addition of isopropyl‐β‐d‐galactopyranoside (IPTG) (NIH 3T3‐K cells, Swanton et al., 1997). Cells were quiesced by culture for 72 h in serum‐depleted medium. K cyclin expression was then induced in the absence of serum stimulation and p27Kip1 status analysed during the subsequent 24 h. K cyclin expression was clearly detectable by 8 h after addition of IPTG (Figure 4A). This induction correlates with an increase in the proportion of p27Kip1 present in a slower migrating form; this slower mobility form of p27Kip1 could be converted to the faster migrating form by treatment with calf intestinal alkaline phosphatase (not shown) as previously described (Muller et al., 1997). Thus, this retarded gel mobility clearly reflects the increased phosphorylation status of p27Kip1. The site of phosphorylation responsible for this mobility shift is Ser10 of p27Kip1 rather than Thr187 (D.J.Mann and N.Jones, unpublished observations); in the in vitro assays using GST–p27Kip1 fusions as substrates, Ser10 is phosphorylated very inefficiently, probably due to the proximity of this amino acid residue to the fusion site. Within 12 h of induction of K cyclin expression, all of the detectable p27Kip1 had the mobility of the slower migrating phospho‐form and the abundance of p27Kip1 had decreased by >50% (Figure 4A). Parallel analysis of NIH 3T3‐K cells in the absence of K cyclin expression indicated no change in the phosphorylation status or abundance of p27Kip1 (results not shown). In addition, we measured the kinase activity associated with both K cyclin and cyclin E by immunoprecipitation kinase assays. K cyclin‐associated kinase activity (in anti‐HA epitope tag immunoprecipitations) was low in uninduced cells and increased in parallel to increasing levels of K cyclin protein (Figure 4A and B). Cyclin E‐associated kinase activity was undetectable at early time points but increased as p27Kip1 levels declined (Figure 4B). Thus, the phosphorylation and downregulation of p27Kip1 correlated precisely with the increase in K cyclin‐associated kinase activity. The lack of cyclin E‐associated kinase activity at early time points indicates that p27Kip1 phosphorylation and downregulation are unlikely to be due to the cyclin E–cdk2 holoenzyme. In addition, preliminary estimates place the half life of p27Kip1 in the presence of K cyclin at ∼80 min compared with 270 min when K cyclin expression is repressed (D.J.Mann, unpublished results). Taken together, these data are consistent with a model in which K cyclin‐dependent phosphorylation of p27Kip1 on Thr187 triggers the subsequent downregulation of the cdk inhibitor leading to enhanced cyclin E‐associated kinase activity.

Figure 4.

K cyclin expression leads to p27Kip1 hyperphosphorylation and downregulation. NIH 3T3‐K cells were quiesced by maintenance in 0.2% serum for 72 h. Cultures were then washed with fresh Dulbecco's modified Eagle's medium (DMEM) plus 0.2% serum and incubated in the presence of 5 mM IPTG for the time indicated. Equal amounts of protein from extracts of these cells were immunoblotted with the indicated antibodies. The p27Kip1 blot was reprobed sequentially with anti‐HA antibody to detect K cyclin and then with anti‐α‐tubulin as a loading control (A). Equal amounts of protein from extracts of quiescent NIH 3T3‐K cells treated for the indicated time with 5 mM IPTG were immunoprecipitated with the antibodies to HA epitope tag of K cyclin (solid bars) or to cyclin E (open bars) and assayed for kinase activity towards pRb (B).

To assess whether p27Kip1 downregulation was a specific event in cells expressing K cyclin, subconfluent growing cultures of NIH 3T3‐K cells were maintained in the absence or presence of IPTG for 48 h. Cells in which K cyclin expression was induced contained significantly lower levels (3‐ to 4‐fold) of p27Kip1 than parallel cultures of NIH 3T3‐K cells in the absence of K cyclin expression (Figure 5A). Control cultures of parental NIH 3T3 cells contained invariant levels of p27Kip1 irrespective of the absence or presence of IPTG (results not shown). Compared with serum‐deprived quiescent cultures, asynchronous dividing cultures have significantly reduced levels of p27Kip1. Thus, K cyclin expression caused the further downregulation of this already depressed basal level of cellular p27Kip1. Induction of K cyclin expression did not significantly affect the levels of other proteins involved in cell‐cycle control (cyclins D1, D3 or E or cdks 4 or 6, see Figure 5A), demonstrating the specificity of the K cyclin‐dependent downregulation of p27Kip1. Downregulation of cyclins D1 and E has, like p27Kip1, been demonstrated to involve ubiquitin‐mediated proteolysis (Clurman et al., 1996; Won and Reed, 1996; Diehl et al., 1997). Indeed, the downregulation of p27Kip1 by K cyclin induction could be blocked by LLnL, an inhibitor of cysteine proteases and the proteasome but not by the cysteine protease inhibitor E64 (Figure 5B). These observations indicate that the phosphorylation and downregulation of p27Kip1 by K cyclin expression is a specific event involving the proteasome and not a general property of K cyclin on elevating such cellular degradation.

Figure 5.

Specificity of K cyclin‐dependent p27Kip1 downregulation and involvement of the proteasome in this process. NIH 3T3‐K cells were maintained in DMEM plus 10% serum in the absence or presence of 5 mM IPTG for 48 h prior to cell lysis. Equal amounts of protein from extracts of these cells were immunoblotted with the indicated antibodies (A). To investigate whether the downregulation of p27Kip1 upon K cyclin expression was sensitive to inhibitors of the proteasome, NIH 3T3‐K cells were maintained in DMEM plus 10% serum in the absence or presence of 5 mM IPTG for 48 h. LLnL (which inhibits cysteine proteases and the proteasome) or E64 (which inhibits non‐proteasomal cysteine proteases) were then added to 50 μM final concentration and the cells harvested 4 h later. Equal amounts of protein from extracts of these cells were immunoblotted with anti‐p27Kip1 antisera (B).

K cyclin‐mediated downregulation of p27Kip1 requires cdk6

The data presented thus far indicate that K cyclin expression is associated with phosphorylation and downregulation of p27Kip1. We have also demonstrated that K cyclin can bind to and activate not only cdk6 but also cdk2, albeit with less efficiency in vitro (Figure 2). In order to dissect the role of these two distinct cdk subunits in the downregulation of p27Kip1, we utilized dominant negative versions of each cdk in which a key catalytic Asp residue (D145 in cdk2 and D163 in cdk6) is mutated to Asn. These mutant cdks are catalytically inactive but can still bind to their cognate cyclins and thus can be used as dominant negative proteins (van den Heuval and Harlow, 1993). We transfected NIH 3T3‐K cells with plasmids directing the expression of green fluorescent protein (GFP) and dominant negative versions of either cdk2 or cdk6. IPTG or vehicle was added to the cells 48 h after transfection and 14 h later the cells were trypsinized. Fluorescence‐activated cell sorting was then used to separate the GFP‐positive transfected cells from the bulk untransfected cell population. Equal numbers of GFP‐positive cells expressing either dominant negative cdk2 or dominant negative cdk6 in the absence or presence of K cyclin were lysed and subjected to immunoblot analysis to determine the levels of p27Kip1 (Figure 6). In the presence of K cyclin expression, dominant negative cdk6 but not dominant negative cdk2 effectively blocked the K cyclin‐mediated downregulation of p27Kip1. These data are consistent with a model in which K cyclin–cdk6 but not K cyclin–cdk2 is responsible for p27Kip1 downregulation. In the absence of K cyclin expression, cells expressing dominant negative cdk2 contained approximately twice as much p27Kip1 as cells expressing dominant negative cdk6, indicating that p27Kip1 levels are responsive to inhibition of cyclin–cdk2 complexes, as demonstrated previously (Sheaff et al., 1997; Vlach et al., 1997).

Figure 6.

K cyclin–cdk6 is responsible for p27Kip1 downregulation. NIH 3T3‐K cells were transfected with plasmids directing expression of GFP and dominant negative versions of either cdk2 or cdk6. Forty‐eight hours after transfection, IPTG (5 mM) or vehicle was added to the cells and 14 h later the cells were trypsinized and collected. Fluorescence‐activated cell sorting was then used to separate the GFP‐positive transfected cells from the bulk untransfected cell population. Cells were lysed and protein from 1×105 cells was resolved by SDS–PAGE and subjected to immunoblotting with the antibody indicated. Dominant negative cdk6 runs with a retarded mobility due to the presence of an epitope tag.

K cyclin‐dependent escape from p27Kip1‐mediated G1 arrest relies on Thr187 of p27Kip1

We next addressed the importance of p27Kip1 phosphorylation and downregulation to the ability of K cyclin to circumvent the p27Kip1‐imposed G1 arrest by utilizing the Thr187 to Ala (T187A) mutant version of p27Kip1. This mutant was tested for its ability to exert a G1 arrest. Transient transfection of wild‐type and T187A mutant p27Kip1 into U2OS cells demonstrated that both forms of p27Kip1 efficiently arrested cells in G1 phase (Figure 7); both proteins were expressed at similar levels as judged by immunoblotting, although the T187A mutant p27Kip1 was ∼25% more abundant than wild‐type p27Kip1 (results not shown). We next repeated these experiments but cotransfected K cyclin and again examined the increase in the G1 cell population. As shown in Figure 7, K cyclin expression enabled cells to efficiently bypass a G1 arrest imposed by wild‐type p27Kip1. However, overexpression of T187A mutant p27Kip1 caused a G1 arrest which was resistant to K cyclin expression. These results demonstrate that the release from a p27Kip1‐mediated G1 blockade by K cyclin expression is dependent on the presence of Thr187. Taken together with the other evidence presented, these data imply that the ability of K cyclin to overcome a p27Kip1‐mediated G1 arrest is linked to the ability of K cyclin–cdk6 complex to phosphorylate p27Kip1 and thereby facilitate its downregulation.

Figure 7.

K cyclin expression cannot prevent G1 arrest imposed by mutant p27Kip1. U2OS cells were transfected with plasmids directing the expression of the cell surface marker CD8, either wild‐type or T187A mutant p27Kip1 and either pcDNA3 (solid bars) or pcDNA3‐K cyclin (white bars). Seventy‐two hours after transfection cells were harvested and the cell‐cycle distribution of the CD8 positive cells determined. Data are represented as the percentage change in the number of cells in the G1 phase of the cell cycle relative to the G1 population of cells transfected with cell surface marker only. The bars represent the mean of duplicate transfections from a single experiment and are representative of three separate experiments each performed in duplicate. At least 15 000 cells were gated for each sample. Untransfected cells displayed the following cell‐cycle distribution: G1, 47%; S, 21%; G2/M, 32%.

Discussion

By restricting cyclin–cdk activity, the cdk inhibitors provide a powerful means of preventing cell proliferation. Overexpression of p16Ink4a or p27Kip1 in mammalian cells leads to arrest in G1 due to the action of these cdk inhibitors on G1‐specific cyclin–cdk complexes, p16Ink4a inhibiting the cdks associated with the D‐type cyclins (that is, cdk4 and cdk6) and p27Kip1 inhibiting both cyclin D‐ and cyclin E‐dependent activities (Serrano et al., 1993; Kato et al., 1994; Polyak et al., 1994; Toyoshima and Hunter, 1994; Lukas et al., 1995).

We have recently demonstrated that the cyclin encoded by ORF 72 of HHV8 (K cyclin) is able to circumvent G1 arrests imposed by both the p16Ink4a and p27Kip1 cdk inhibitors and allow S phase entry in the presence of these growth repressors (Swanton et al., 1997). K cyclin‐mediated suppression of the G1 arrest imposed by overexpression of p16Ink4a can be readily explained by the preference of the viral cyclin for binding to and activation of the endogenous cdk6 (Chang et al., 1996), the K cyclin–cdk6 complex being resistant to p16Ink4a‐mediated inhibition (Swanton et al., 1997). Thus, K cyclin can functionally substitute for the D‐type cyclins in activating cdk6. However, the G1 arrest imposed by p27Kip1 is enforced by blocking the action of both the D‐type cyclin–cdks and cyclin E–cdk2 complexes. Therefore, the expression of K cyclin must overcome the requirement for both cyclin–cdk complexes in promoting the G1→S transition. In this report we have explored the mechanism of this phenomenon. We show that it is due to an expanded repertoire of substrates that can be phosphorylated by K cyclin–cdk6 when compared with cyclin D1–cdk6. One of these substrates is the p27Kip1 inhibitor itself, which upon phosphorylation is downregulated. By promoting such downregulation, K cyclin overcomes a p27Kip1‐imposed G1 arrest.

Previous studies have demonstrated that both K cyclin and the structurally similar V cyclin (encoded by the related herpesvirus saimiri; Jung et al., 1994), when complexed to cdk6, generated efficient pRb kinases (Jung et al., 1994; Chang et al., 1996; Swanton et al., 1997). When we compared the preference of K cyclin–cdk6 with that of cyclin D1–cdk6 for phosphorylation of pRb or histone H1, the viral cyclin‐directed kinase showed little discrimination between these two standard substrates whereas cyclin D1–cdk6 showed a strong preference for pRb (Figure 1). This result suggested that the virally encoded cyclin may broaden the substrate specificity of the associated cdk6 subunit. We tested the ability of K cyclin–cdk6 complexes to phosphorylate three cdk2 substrates: the cdk inhibitor p27Kip1 (Muller et al., 1997; Sheaff et al., 1997; Vlach et al., 1997), the basic helix–loop–helix transcription factor Id‐2 (Hara et al., 1997) and the dual specificity phosphatase cdc25a (Hoffmann et al., 1994). In all three cases, efficient phosphorylation of the cdk2 substrate was observed by K cyclin–cdk6 but not, or very weakly, by cyclin D1–cdk6 (Figure 1). These results strongly support the idea that K cyclin binding can extend the substrate range of cdk6 to include at least a subset of cdk2 substrates (although our results indicate that K cyclin–cdk6 cannot phosphorylate all cdk2 substrates in vivo, see below). The mechanism of this extended substrate repertoire is unknown at present. The parameters that dictate cdk substrate specificity are poorly defined so that analysis of this property of the viral cyclins should prove to be informative. We are currently trying to distinguish between two possibilities: (i) that cyclin–substrate interactions dictate cdk specificity or (ii) that cyclin‐imposed constraints on cdk structure determine cdk specificity. This second model is attractive in that it may account for the more efficient activation of cdk6 by K cyclin (when compared with cyclin D1) and also the ability of K cyclin–cdk6 complexes to resist inhibition by the p16Ink4a cdk inhibitor (Swanton et al., 1997).

The identification of p27Kip1 and cdc25a as substrates for K cyclin–cdk6 suggests the possibility that their phosphorylation plays an important role in K cyclin‐mediated stimulation of S phase entry. Phosphorylation of cdc25a has been shown to stimulate its phosphatase activity in vitro (Hoffmann et al., 1994), implying that cdc25a has greater potential to activate endogenous cdks by catalysing the removal of inhibitory threonine and tyrosine phosphorylations. Phosphorylation of the cdk inhibitor p27Kip1 on Thr187 leads to the downregulation of p27Kip1 protein (Muller et al., 1997; Sheaff et al., 1997; Vlach et al., 1997). We have provided evidence that K cyclin–cdk6 can phosphorylate p27Kip1 on this threonine residue in vitro (Figure 3) and in vivo (Figure 6) and that K cyclin induction in NIH 3T3 fibroblasts leads to a reduction in the endogenous p27Kip1 levels through proteasome‐mediated degradation (Figure 4). The importance of this downregulation is clearly demonstrated by the inability of K cyclin to overcome a G1 arrest imposed by p27Kip1 when Thr187 is changed to Ala (Figure 7).

The resistance of the G1 blockade imposed by the Thr187→Ala substitution mutant of p27Kip1 to K cyclin expression indicates that K cyclin is not sufficient to enforce DNA synthesis when p27Kip1 levels remain elevated. These data imply that S phase entry requires the recruitment of cellular factors that are sensitive to elevated p27Kip1 levels. One likely candidate for this role is cyclin E. Our data are compatible with a model in which the viral cyclin is able to fulfil only part of the role of cyclin E–cdk2 in vivo and that by reducing the levels of p27Kip1, K cyclin expression can lead to the generation of an environment in which cyclin E–cdk2 activity is favoured enabling co‐operation between the viral cyclin and the endogenous cyclin E to facilitate S phase entry.

The ability to override normal growth control and force cells to cycle is a common characteristic of DNA tumour viruses. This property is essential to viral propagation since it ensures the appropriate cellular environment for viral replication. To achieve S phase entry, both cyclin D‐ and cyclin E‐dependent activities must be stimulated or their actions mimicked. DNA tumour viruses have evolved systems by which to accomplish this feat. In the case of HHV8, we have shown that this is likely to derive from the production of cyclin–cdk complexes which are resistant to inhibition and which display broader target specificity. Other DNA tumour viruses achieve similar ends by different means. For example, adenovirus 5 overcomes G1 arrest largely through the action of the E1A oncoprotein which displaces cellular factors from pRb sequestration (Bagchi et al., 1990; Bandara and La Thangue, 1991; Chellappan et al., 1991), thus negating the requirement for cyclin D–cdk activity (Lukas et al., 1995), and binds to and inactivates p27Kip1 thereby freeing cdk2‐containing complexes (Mal et al., 1996).

K cyclin not only has the capacity to interact with cdk6 (Chang et al., 1996) but also weakly interacts with cdk2, cdk3 and cdk4 (this report). In each case, the resulting complexes are able to phosphorylate pRb in vitro (this study; C.Swanton and N.Jones, unpublished observations). These observations highlight important differences between K cyclin and D type cyclins, especially cyclin D1, with which K cyclin shows greatest similarity. Cyclin D1 fails to activate any cdk other than cdk4 or cdk6 (C.Swanton and N.Jones, unpublished observations). At present it is not clear whether the ability of K cyclin to form active complexes with these other kinase partners is physiologically relevant, although both cdk2 and cdk3 appear to have essential roles in the G1→S transition since overexpression of dominant negative mutants of each leads to G1 arrest (van den Heuval and Harlow, 1993; Hofmann and Livingston, 1996). However, of all of the binary combinations of K cyclin and cdk subunits, the K cyclin binds to cdk6 most efficiently and the resulting complex appears to be the most active and exhibits the greatest resistance to cdk inhibitors. It is likely, therefore, that activation of cdk6 is the critical function of K cyclin. This conclusion is substantiated by the observation that cdk6 is an abundant cdk in lymphocytes (Meyerson and Harlow, 1994), the target cell type of human herpesvirus 8. Given that lymphocytes are largely maintained in G0/G1 by high levels of p27Kip1 (Nourse et al., 1994; Fero et al., 1996; Kiyokawa et al., 1996; Luo et al., 1996; Nakayama et al., 1996), the properties of K cyclin which we describe are likely to be of fundamental importance to viral‐mediated deregulation of normal growth control.

Materials and methods

Plasmids and antibodies

The following plasmids have been described previously: CD8, pJ7Ω‐p27Kip1 (Mann and Jones, 1996), pRSET‐p21Cip1, pRSET‐p27Kip1, pRSET‐K cyclin (Swanton et al., 1997), all pKS‐cdk constructs and pKS‐cyclin D1 (Parry et al., 1995). Epitope‐tagged dominant negative cdk2 and cdk6 (van den Heuval and Harlow, 1993) were subcloned into pEFcx (a gift from Dr R.Treisman, Imperial Cancer Research Fund, UK).

K cyclin and cyclin D1 were epitope tagged at their N‐termini with two FLAG epitopes by ligating a double‐stranded oligonucleotide into the unique NcoI site of both cyclins (this restriction site includes the initiating methionine codon) in the pRSET constructs (Swanton et al., 1997). The oligonucleotide sequences were 5′‐CATGGACTACAAGGACGACGACGACAAGGACTACAAGGACGACGACGACAAGGC‐3′ (coding) and 5′‐CATGGCCTTGTCGTCGTCGTCCTTGTAGCCCTTGTCGTCGTCGTCCTTGTAGTC‐3′. The tagged cyclins were transferred to pvl1393 (Invitrogen) via BamHI–EcoRI digestion. Baculovirus directing the expression of cyclin E was a gift from C.Sherr, cdk2 from D.Morgan and cdk6 from M.Meyerson. For expression in mammalian cells, the fragments containing the tagged cyclins were transferred into the pcDNA3 vector (Invitrogen).

GST–cdc25a and GST–Id‐2 were gifts from H.Okayama and E.Hara, respectively. GST–p27Kip1 was created by subcloning the p27Kip1 coding sequence from pJ7Ω (Mann and Jones, 1996) into pGEX‐KG. To generate the p27Kip1 C‐terminus as a GST fusion, PCR was performed using GST–p27Kip1 as template and oligonucleotides 5′‐GCGCCCATGGAGCCCCCGCGGCCCCCCAAAGGTGCCTGC‐3′ and 5′‐GCGCCTCGAGTTACGTTTGACGTCTTCTGAGGCCAGGCTTC‐3′. The resulting fragment was digested with NcoI and XhoI and ligated into pGEX‐KG cut with the same enzymes. For transfection experiments, p27Kip1 was subcloned into pcDNA3 (Invitrogen) from pRSET‐p27Kip1 using BamHI and EcoRI. The Thr187→Ala substitution mutant was created by PCR using pcDNA3‐p27Kip1 as template and oligonucleotides and 5′‐GCGCGGATCCATGTCAAACGTGCGAGTGTC‐3′ and 5′‐GCGCGAATTCTACGTTTGACGTCTTCTGAGGCCAGGCTTCTTGGGCGCCTGC‐3′. The PCR product was digested with BamHI and EcoRI and subcloned into pcDNA3 cut with the same enzymes. The DNA sequence of all PCR products was verified.

The following antibodies were used: anti‐HA (12CA5, Boehringer Mannheim, 1538 816), anti‐p27Kip1 (Santa Cruz, sc‐528 and sc‐527), anti‐cyclin E (Santa Cruz, sc‐481), anti‐cyclin D3 (Santa Cruz, sc‐182), anti‐cyclin D1 (a gift from Dr G.Peters, Imperial Cancer Research Fund), anti‐cdk4 (SantaCruz, sc‐260), anti‐cdk6 (Santa Cruz, CB02), anti‐FLAG (Sigma, M2) and anti‐α‐tubulin (TAT‐1, Imperial Cancer Research Fund antibody service).

Cells and baculovirus

NIH 3T3‐K cells were isolated, cultured and quiesced as described (Swanton et al., 1997). K cyclin expression was induced as follows: cells were washed twice with phosphate‐buffered saline (PBS) and re‐fed with DMEM containing either 0.2% or 10% fetal bovine serum (as appropriate) and 5 mM IPTG. When necessary, LLnL or E64 (Sigma) were added directly to the culture medium at 50 μM final concentration. Culture of U2OS cells, transfection, analysis of the cell‐cycle distribution of transfected cells and fluorescence‐activated cell sorting have been described (Mann and Jones, 1996). Recombinant baculovirus were produced using the BaculoGold system (PharMingen). Co‐infection of Sf9 cells with recombinant baculoviruses has been described (Kato et al., 1993).

In vitro binding assays and immunoblotting

In vitro binding assays were performed as described (Swanton et al., 1997), cDNAs being transcribed and translated using the TNT system (Promega).

For immunoblotting, cells were washed twice with PBS, lysed in situ in RIPA buffer (150 mM NaCl, 1% NP‐40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris pH 8.0) and collected by scraping. After normalizing for protein content, samples were boiled in 1× SDS–PAGE sample buffer and subjected to SDS–PAGE using 10% gels to resolve cyclins and cdks and 12.5% gels for p27Kip1 analysis. Following electrophoresis, proteins were transferred to nitrocellulose membranes (Schleicher and Schuell). Individual proteins were detected using specific antisera, the immunoreactive bands being visualized by using appropriate horseradish peroxidase‐conjugated secondary antibody and subsequent detection by enhanced chemiluminescence (Amersham).

In vitro kinase assays, phosphopeptide mapping and phosphoamino acid analysis

To generate extracts for use in in vitro kinase assays, Sf9 cells infected with appropriate recombinant baculoviruses were lysed as described (Kato et al., 1993; Parry et al., 1995). Extracts were either used directly or immunoprecipitated with an appropriate antibody. Immunoprecipitation kinase assays were performed as described (Matsushime et al., 1994; Swanton et al., 1997) using [γ‐32P]ATP (ICN), GST–pRb and/or histone H1 (Boehringer Mannheim). Inhibition experiments were performed as described (Swanton et al., 1997) using p21Cip1 and p27Kip1 isolated from bacteria harbouring the appropriate pRSET plasmid (kindly provided by M.Hall and G.Peters). GST fusion proteins were isolated from bacteria and bound to glutathione–Sepharose according to the manufacturer's instruction (Pharmacia).

For phosphopeptide mapping, GST‐bound products from in vitro kinase assays (Parry et al., 1995) were washed with PBS containing 0.1% Tween 20, resolved by SDS–PAGE and transferred to PVDF membrane (Du Pont). Radioactive bands were identified by autoradiography, excised, digested with trypsin (Sigma) and peptides resolved in two dimensions on cellulose thin layer chromatography plates (Kodak) as described (Boyle et al., 1991). Peptide resolution was by electrophoresis at pH 1.9 in the first dimension followed by chromatography using 3:10:12:15 acetic acid:pyridine:water:butan‐1‐ol in the second dimension. Phosphoamino acid analysis was performed as described (Neufeld et al., 1989).

Acknowledgements

We thank M.Hall, E.Hara, E.Harlow, M.Meyerson, D.Morgan, H.Okayama, G.Peters, C.Sherr and R.Treisman for gifts of reagents, P.Parker for help with the phosphoamino acid analysis, D.Davies and A.Rae for FACS analysis and H.Land for critical reading of the manuscript. This work was funded by the Imperial Cancer Research Fund (N.J.) and the Cancer Research Campaign ([CRC] grant SP2414/0101) and Royal Society (D.J.M.).

References