The Interaction between Cyclin B1 and Cytomegalovirus Protein Kinase pUL97 is Determined by an Active Kinase Domain

Replication of human cytomegalovirus (HCMV) is characterized by a tight virus-host cell interaction. Cyclin-dependent protein kinases (CDKs) are functionally integrated into viral gene expression and protein modification. The HCMV-encoded protein kinase pUL97 acts as a CDK ortholog showing structural and functional similarities. Recently, we reported an interaction between pUL97 kinase with a subset of host cyclins, in particular with cyclin T1. Here, we describe an interaction of pUL97 at an even higher affinity with cyclin B1. As a striking feature, the interaction between pUL97 and cyclin B1 proved to be strictly dependent on pUL97 activity, as interaction could be abrogated by treatment with pUL97 inhibitors or by inserting mutations into the conserved kinase domain or the nonconserved C-terminus of pUL97, both producing loss of activity. Thus, we postulate that the mechanism of pUL97-cyclin B1 interaction is determined by an active pUL97 kinase domain.

also postulated in light of the finding that the HCMV-inhibitory effect of the pUL97 inhibitor maribavir (MBV) was increased when CDK activity was simultaneously suppressed [11]. Although pUL97 is not strictly essential for HCMV replication, the deletion of ORF UL97 or the pharmacological inhibition of pUL97 leads to a drastic reduction in the efficiency of viral replication [28,29]. Its kinase domain includes subdomains (SD) I-XI, which are conserved (aa 337-651) within herpesviral and cellular protein kinases. Notably, the second SD contains an invariant lysine residue (K355), the replacement mutation of which resulted in a complete loss of kinase activity [16,30,31]. As published recently by our group, alternative initiation of translation at codons M1, M74, and M175 results in the expression of three pUL97 isoforms, with partly individual properties in terms of regulation of viral replication and MBV susceptibility [32][33][34]. The formation of pUL97 dimers and oligomers is based on a self-interaction domain, located within amino acids 231-280, that supports strong autophosphorylation mainly occurring at N-terminal residues [35][36][37]. Nuclear translocation of pUL97 is conferred via two classical bipartite nuclear localization sequences (NLS1 and NLS2, amino acids 6-35 and 190-213, respectively) that mediate importin-α binding [32,33]. By phosphorylating a number of cellular and viral substrates, pUL97 regulates various steps during HCMV replication (reviewed in Marschall et al. [3]). In the present study, we refine our previous report on pUL97-cyclin interaction. Specifically, we provide novel data for the high-affinity pUL97-cyclin B1 interaction, with the mode of interaction determined by kinase activity and structural properties of pUL97, and we present the first three-dimensional concept of this previously-unrecognized example of herpesviral kinase-cyclin interaction.

Cell Culture, HCMV Infections and Plasmid Transfection
Human embryonic kidney epithelial cells (293T) were cultivated in Dulbecco's modified Eagle's medium containing 10% FCS and primary human foreskin fibroblasts (HFF) in MEM containing 7.5% FCS. HCMV infection experiments were performed at a multiplicity of infection (MOI) of 0.5 to 1.0 using HCMV strain AD169 or AD169-GFP [38]. Transfection of 293T cells was performed using polyethylenimine reagent (Sigma-Aldrich, St. Louis, MO, USA). For the transfection of a 10 cm dish, 10 µg DNA was mixed with 1 mL PBS and subsequently incubated with 20 µL PEI 2000 in 1 mL PBS for 20 min at room temperature. Thereafter, the solution was mixed with 36 µL PEI 25,000 in 1 mL PBS Viruses 2015, 7 4585 and incubated for 20 min before dropwise addition to the cell layer in a 10 cm dish. After incubation for 4 h at 37˝C, the transfection solution was replaced by fresh medium.

Coimmunoprecipitation (CoIP)
The 293T cells were seeded in 10 cm dishes at a cell number of 5ˆ10 6 and were transfected with expression plasmids coding for full-length pUL97 or mutants of pUL97 using the polyethylenimine transfection technique. Red fluorescent protein (RFP; pDsRed1-N1, BD Clontech) was used as a transfection control. HCMV AD169-infected HFFs were cultivated in cell culture flasks and harvested at 4-5 days post-infection, as indicated. Optionally, inhibitors were added to the culture media one day before harvesting. Plasmid-transfected 293T cells and HCMV-infected HFFs were lysed in 500 mL CoIP buffer (50 mM Tris/HCl [pH 8.0], 150 mM NaCl, 5 mM EDTA, 0.5% NP-40, 1 mM PMSF, 2 µg aprotinin mL´1, 2 µg leupeptin mL´1 and 2 µg pepstatin mL´1) for 20-30 min on ice, centrifuged (14,000 rpm, 4˝C, 10-30 min) and incubated for about 3 h at 4˝C under rotation with antibody-coated protein A sepharose beads (100 µL of a 25 mg/mL solution; Sigma-Aldrich) or Dynabeads Protein A (25 µL per sample; Life Technologies/Novex), respectively. To absorb unspecific protein binding, lysates of 293T cells were incubated for about 1 h at 4˝C with uncoated protein A sepharose beads. The precipitates were washed five times (1 mL each) with CoIP buffer before the samples were subjected to standard Western blot analysis using tag-or protein-specific antibodies for the detection of coimmunoprecipitates and protein expression (ECL staining; New England Biolabs).

Confocal Imaging
HFFs were grown on coverslips and used for infection with HCMV strain AD169. At three days post-infection, cells were fixed with 4% paraformaldehyde solution (10 min, room temperature) and permeabilized by incubation with 0.2% Triton X-100 solution (15 min, 4˝C). Non-specific staining was blocked by incubation with 2 mg/mL human γ-globulin (cohn fraction II, Sigma Aldrich; 40 min, 37˝C).
Indirect immunofluorescence staining was performed by stepwise incubation with primary antibodies for 90 min each at 37˝C, followed by incubation with dye-conjugated secondary antibodies (Alexa, Molecular Probes, Inc., Eugene, OR, USA) for 30 min at 37˝C. Cell samples were mounted with Vectashield Mounting Medium containing DAPI and analyzed using a DMI6000 B microscope and a 63ˆHCX PL APO CS oil immersion objective lens (Leica Microsystems, Wetzlar, Germany). Confocal laser-scanning microscopy was performed with a TCS SP5 microscope (Leica Microsystems). Images were processed using the Meta-Imaging series (Molecular Devices, Sunnyvale, CA, USA) and LAS AF software (version 1.8.2 build 1465; Leica Microsystems).

In Vitro Kinase Assay (IVKA)
In vitro protein phosphorylation was investigated by IVKA analyses using immunoprecipitated proteins from lysates of transiently transfected 293T cells (10 cm dish with approx. 5ˆ10 6 cells per reaction) or HCMV-infected HFFs (T175 flask with approx. 5ˆ10 6 cells per reaction). CoIP of protein complexes for combined CoIP-IVKA reactions was performed according to the CoIP procedure described under 2.3 by cell lysis and immunoprecipitation with antibody-coated protein A beads in CoIP buffer. The final washing step was modified compared to normal CoIP, i.e., by washing twice in 500 µL HNTG buffer (50 mM HEPES [pH 7.5], 150 mM NaCl, 1 mM EDTA, 10% glycerin, and 0.1% Triton X-100) and two times in 500 µL kinase basis buffer (20 mM Tris/Cl [pH 7.5], 0.5 mM MnCl 2 ). Precipitates were incubated for 30 min at 30˝C in 40 µL kinase assay buffer (basis buffer including 1 µM ATP, 1 mM DTT and 2.5 µCi [γ-33 P]-ATP, Hartmann Analytic, Braunschweig, Germany) to start in vitro protein phosphorylation. Optionally, 1 µL of a purified histone mix H1-4 (Roche, Mannheim, Germany) was added as an exogenous substrate protein. In vitro phosphorylation was finally stopped by denaturation at 95˝C for 10 min with 15 µL of 4ˆ-concentrated SDS-PAGE loading buffer. The radiolabeled proteins were subjected to SDS-PAGE and analyzed by autoradiography using a phosphoimager; levels of protein expression and the reliability of specific immunoprecipitation was controlled by Western blot analysis in parallel.

Homology Modeling of pUL97
The crystal structure of human CDK2 (hCDK2) complexed with the quinazoline compound DIN-234325 (PDB code: 2B53) [39] was selected as a template for the homology modeling of pUL97 inhibited by MBV or the quinazoline compound Ax7396. The loop Leu37-Glu40, which was not resolved in PDB entry 2B53, was modeled according to the coordinates of the respective residues from PDB entry 4KD1 [40]. Additionally, two alternative pUL97 models were generated, which were based on the cyclin B-bound conformations of hCDK2 (PDB code: 2JGZ) [41] or hCDK1 (PDB code: 4Y72; [42]). A comparison of the different models will allow an estimate of whether pUL97 may exhibit a similar conformational plasticity as human CDKs. After the sequence alignment of hCDK1, hCDK2 and pUL97, homology models were created with MODELLER 9.14 [43]. Visualization of pUL97 modeling was done with VMD 1.9.1. [44]. Note that a loop formed by the nonconserved amino acids 506-541 of pUL97 is not shown in the model due to a large gap in the sequence alignment.

HCMV Protein Kinase pUL97 Shows a Particularly Strong Interaction with Cyclin B1
Recently, we identified an interaction of the HCMV-encoded protein kinase pUL97 with cyclin T1, and characterized this interaction with data from yeast two-hybrid, CoIP and kinase assays [45]. In particular, the region responsible for cyclin T1 could be narrowed down to amino acids 231-280 of pUL97. In our ongoing analysis (M.M., unpublished data) and in the previous report by Graf et al. [45], we noticed that cyclin interaction of pUL97 was not exclusively restricted to cyclin T1, but also included further cyclin types, such as B1, A, and others, and that these interactions can be detected in HCMV-infected cells. In the present study, we demonstrate an interaction of pUL97 with cyclin B1 that is more prominent than that described for cyclin T1 and that is characterized by several unexpected features. In a CoIP setting, we used a polyclonal anti-cyclin B1 antibody (pAb-cyclin B1) to precipitate endogenous cyclin B1 together with transiently expressed pUL97-Flag ( Figure 1A, lane 3). Recently, we reported the formation of up to three isoforms for pUL97 [32][33][34], which are, likewise, nicely detectable in this setting (lanes 9 and 13), whereby the inactive mutant K355M shows an isoform band pattern at reduced molecular masses compared to wild-type due to the lack of autophosphorylation (lane 10). Interestingly, cyclin B1 interaction could also be detected for the isoforms of pUL97 (lanes 4-5; see mainly the abundantly expressed large and middle isoforms M1 and M74; note that, generally, the quantity of detectable levels of pUL97 isoforms, mutants, as well as wild-type is in part a matter of limited protein stability as reported earlier [34]). Specificity of the CoIP setting was illustrated by the use of three negative controls (lanes 6-8; no antibody, Fc fragment, and pre-immune serum, respectively). The pUL97-cyclin B1 interaction was supported by our confocal imaging analysis with HCMV-infected primary fibroblasts. Imaging data demonstrated a colocalization between pUL97 and cyclin B1 in a distinct speckled pattern concentrated in subnuclear compartments ( Figure 1B, panels 5-9, compared to a mock control in panels 1-4; Figure 1C, panels 1-12) that were identified as early nuclear viral replication centers using the viral DNA polymerase processivity factor pUL44 as a marker for viral replication centers (panels [13][14][15][16][17][18][19][20][21][22][23][24]. Various combinations of antibodies against pUL97 and cyclin B1 were used to confirm this finding and controls were performed with negative antibodies and mock-infected cells ( Figure 1C, panels 25-28; note that in this experiment, at least 50 further mock-infected, cyclin B1-positive cells were detected showing a very similar pattern of cyclin B1 expression). While in mock-infected cells strong cyclin B1 expression was distinctly found in mitotic cells, HCMV-infected cells showed a broad increase in cyclin B1 levels (under the known conditions of HCMV-induced early S phase arrest [11]). A pronounced pUL97-cyclin B1 colocalization was identified in a subnuclear speckled pattern ( Figure 1B, panel 9; Figure 1C, panels 6 and 12), also including other viral proteins accumulating in nuclear replication centers, such as pUL44, which is known to interact with and to be phosphorylated by pUL97 ( [46]; panels 18 and 24; note that cyclin B1 did not entirely colocalize with pUL44, but that fractions of the two proteins were detected in distinct cells in identical subnuclear compartments). A quantitation of the percentage of cells double-positive for pUL97 and cyclin B1 resulted in 17.8˘3.6% of colocalization (when counting exclusively cells with a very prominent signal of colocalization as shown in the inset panels of Figure 1B,C) or, in a total range, 63.2˘7.6% (also including cells comprising less prominent colocalization signals; quantitation was achieved by microscopic counting of n = 7 of ě50 cells, mean values˘SD). This finding is in line with previously published data showing pUL97-cyclin T1 colocalization in nuclear viral replication centers [45].
Viruses 2015, 7 7 quantitation was achieved by microscopic counting of n = 7 of ≥50 cells, mean values ± SD). This finding is in line with previously published data showing pUL97-cyclin T1 colocalization in nuclear viral replication centers [45].

The pUL97-Cyclin B1 Interaction is Dependent on Kinase Activity
A striking feature of our current analysis was the finding that inactive pUL97 did not interact with cyclin B1. When CoIP was performed with the catalytically-inactive version pUL97(K355M)-Flag, only a marginal signal was obtained, which stood in contrast to the strong signal of wild-type pUL97 interaction with cyclin B1 ( Figure 1A, lanes 2-5). This finding was further supported by the use of pUL97 inhibitors in this CoIP setting ( Figure 1A, lanes 11, 12; Figure 2). In the absence of inhibitor, endogenous cyclin B1 interaction was detectable at a particularly high affinity compared to cyclin T1 interaction (Figure 2A, lanes 1 and 8; the specificity of interaction was verified and confirmed by several experiments, using various controls also including pre-immune serum). For both cyclins, interaction could be blocked by treatment with the pUL97 inhibitors Ax7396 [47,48] and MBV [49][50][51]. The block was concentration-dependent and was more prominent for cyclin B1 than for cyclin T1 interaction (Figure 2A, lanes 4-7 and 11-14, respectively). No effect, or only marginal inhibitory effect, on interaction was displayed by the nucleoside analog GCV, which binds as a substrate and is phosphorylated by pUL97 (Figure 2A, lanes 2-3 and 9-10) or by CDK inhibitors R22 [9] and roscovitine [20] ( Figure 2B, lanes 1-4). These data point to a strictly activity-determined mode of cyclin interaction, a concept that is particularly supported by the pronounced loss of cyclin B1 interaction in the case of pUL97 mutation K355M and two chemically distinct pUL97-inhibitory small molecules. It is interesting to see that both types of drugs, i.e., the clinical drug candidate MBV and the experimental compound Ax7396 belonging to the quinazoline class, show similar efficacies in blocking pUL97-cyclin B1 interaction. Our recent findings on the novel quinazoline-type inhibitors of pUL97 Vi7392 and Vi7453 (characterized by anti-HCMV 50% effective concentrations in the low micromolar range, EC 50 ď 4 µM; data not shown) indicate an advantageous effect over drugs used in standard therapy, since they appear to be poor inducers of drug resistance. Previously, we reported that three quinazoline compounds did not show an induction of drug resistance against HCMV in vitro [47,48]. To address the question of drug resistance for the novel quinazolines Vi7392 and Vi74523, HCMV-infected cells were treated with high drug concentrations (10-fold EC 50 ), continuously monitored, and passaged for months. As a striking result, quinazoline-mediated resistance formation was not observed (in contrast to the frequent induction of GCV resistance in control panels). The finding that pUL97 inhibitors of the chemical class quinazolines are non-inducers of pUL97-conferred drug resistance prompted us to hypothesize that quinazolines may differ in their binding mode from other classes of inhibitors. Independent of individual frequencies of drug resistance, these pUL97 inhibitors from different chemical classes had, in common, the ability to block cyclin B1 binding. This suggests that drug binding may be linked with changes in structural conformations of pUL97 as reflected by active and inactive states of the kinase.
Viruses 2015, 7 9 To address the question of drug resistance for the novel quinazolines Vi7392 and Vi74523, HCMVinfected cells were treated with high drug concentrations (10-fold EC50), continuously monitored, and passaged for months. As a striking result, quinazoline-mediated resistance formation was not observed (in contrast to the frequent induction of GCV resistance in control panels). The finding that pUL97 inhibitors of the chemical class quinazolines are non-inducers of pUL97-conferred drug resistance prompted us to hypothesize that quinazolines may differ in their binding mode from other classes of inhibitors. Independent of individual frequencies of drug resistance, these pUL97 inhibitors from different chemical classes had, in common, the ability to block cyclin B1 binding. This suggests that drug binding may be linked with changes in structural conformations of pUL97 as reflected by active and inactive states of the kinase. On day three, cells were lysed and endogenous cyclin B1 or cyclin T1 was immunoprecipitated using polyclonal rabbit antibodies. CoIP and expression control samples were subjected to WB analysis using the antibodies indicated (including rabbit antibody Fc fragment as a specificity control). Note that 3 µM of the pUL97-targeting compounds MBV and Ax7396 showed a strong inhibitory effect on cyclin interaction (i.e., the range of anticytomegaloviral activity exerted by these compounds was determined 1-15 µM), so that this concentration was maintained for further experiments. On day three, cells were lysed and endogenous cyclin B1 or cyclin T1 was immunoprecipitated using polyclonal rabbit antibodies. CoIP and expression control samples were subjected to WB analysis using the antibodies indicated (including rabbit antibody Fc fragment as a specificity control). Note that 3 µM of the pUL97-targeting compounds MBV and Ax7396 showed a strong inhibitory effect on cyclin interaction (i.e., the range of anticytomegaloviral activity exerted by these compounds was determined 1-15 µM), so that this concentration was maintained for further experiments.
For this purpose, drug-inhibited pUL97 was modeled using the structure of an inhibitor-bound CDK2 as a template. Both MBV and Ax7396 ( Figure 3A) fit well into the binding pocket of pUL97 ( Figure 3B), supporting the observation that they inhibit pUL97 activity. Therefore, the lack of cyclin binding may be directly related to the absence of autophosphorylation of inhibited pUL97. However, from the basis of crystal structures of human CDKs it is known that inhibitor binding may not only affect the catalytic activity but also the structural properties of CDKs. One example is the conformation of the activation loop, which is rather flexible in CDK1 and CDK2 and constitutes a part of the cyclin B binding interface [42]. In order to explore whether a similar conformational plasticity may also be feasible for pUL97, additional models were generated using the crystal structures of cyclin B-bound CDK1 and CDK2 as templates. The resulting models suggest that the length of the pUL97 activation loop is sufficient to adopt a conformation similar to that observed in CDK1 and CDK2 ( Figure 3C,D). Based on this finding, it is tempting to speculate that the pUL97 kinase domain may directly interact with cyclin B in a fashion similar to CDK1 or CDK2. However, further experiments are required to verify this hypothesis.

Viruses 2015, 7 10
For this purpose, drug-inhibited pUL97 was modeled using the structure of an inhibitor-bound CDK2 as a template. Both MBV and Ax7396 ( Figure 3A) fit well into the binding pocket of pUL97 ( Figure 3B), supporting the observation that they inhibit pUL97 activity. Therefore, the lack of cyclin binding may be directly related to the absence of autophosphorylation of inhibited pUL97. However, from the basis of crystal structures of human CDKs it is known that inhibitor binding may not only affect the catalytic activity but also the structural properties of CDKs. One example is the conformation of the activation loop, which is rather flexible in CDK1 and CDK2 and constitutes a part of the cyclin B binding interface [42]. In order to explore whether a similar conformational plasticity may also be feasible for pUL97, additional models were generated using the crystal structures of cyclin B-bound CDK1 and CDK2 as templates. The resulting models suggest that the length of the pUL97 activation loop is sufficient to adopt a conformation similar to that observed in CDK1 and CDK2 ( Figure 3C,D). Based on this finding, it is tempting to speculate that the pUL97 kinase domain may directly interact with cyclin B in a fashion similar to CDK1 or CDK2. However, further experiments are required to verify this hypothesis.

Serial Deletions of the pUL97 C-Terminus Abrogate Cyclin B1 Interaction
Our previously published reports implied that an intact C-terminus of pUL97, but not an intact N-terminus, is required for full kinase activity [18,32,34,35]. To prove this suggestion, we generated further C-terminally truncated versions of pUL97 [ORF UL97(1-707) lacking 1, 5, 15, 50 or 112 amino acids] and expressed the proteins in transiently transfected 293T cells. When immunoprecipitated by the use of Flag tag-specific antibodies and analyzed for activity in in vitro kinase assays (IP-IVKA), we detected strong autophosphorylation and histone substrate phosphorylation for full length pUL97 (pUL97-Flag; Figure 4A, lanes 1-2) and the construct lacking only one amino acid (pUL97(1-706)-Flag; Figure 4A, lane 3). In contrast, all other constructs carrying further C-terminal truncations completely lacked activity and, thus, were negative in both autophosphorylation and substrate phosphorylation. (Note that the faint bands for histones in lanes 4-8 correlate with background phosphorylation levels shown in lane 9).
Next, we addressed the question whether these catalytically inactive C-terminally truncated mutants of pUL97 were likewise negative for cyclin B1 interaction. In a CoIP setting (similar to the one shown in Figure 2), we performed an immunoprecipitation of endogenous cyclin B1 and searched for those versions of pUL97 that were positive in CoIP. As a striking result, we exclusively obtained positive signals for full-length pUL97(1-707) and pUL97(1-706) ( Figure 4B, lanes 1-2), whereas all further truncated versions were negative (lanes 3-6). CoIP controls with an Fc fragment (incapable of recognizing pUL97-Flag) or pAb-HA (indicating self-interaction between pUL97-Flag and pUL97-HA; it should be stressed that this interaction might be indirect, thus possibly bridged by a number of cellular proteins) served to monitor the specificity of reactions ( Figure 4B, lanes 7-8). This finding argues for an important role of the C-terminus in pUL97, not only for the folding of an intact catalytic center of the kinase domain but also for cyclin B1 recruitment. Moreover, it was also interesting to see that pUL97-pUL97 interaction (an activity basically mapped to the N-terminal region 231-280; [35]) showed an identical dependence on the intact C-terminus. Here again, those truncations removing more than one amino acid from the C-terminus abrogated pUL97-pUL97 interaction ( Figure 4C). Thus, we postulate that the C-terminal region of pUL97, although not contained within the classic kinase domain (defined by conserved subdomains I-XI; [16]), has an overall importance for catalytic activity and protein interactions. Truncations of the C-terminus may lead to structural alterations that are incompatible with normal pUL97 functionality, including phosphorylation activity, cyclin B1 interaction and pUL97-pUL97 interaction. In this regard, it might be relevant that all manipulations described in Figures 1, 2 and 4 (i.e., replacement mutation K355M, treatment with kinase inhibitors and C-terminal truncation, respectively) have one consequence in common, namely to prevent pUL97 autophosphorylation. For this reason, it is tempting to speculate that pUL97 autophosphorylation might be a trigger that induces so far uncharacterized structural changes in order to establish a pUL97 conformation that is competent for cyclin B1 binding. Figure 4. Importance of the C-terminus of pUL97 for kinase activity, cyclin B1 binding and pUL97-pUL97 interaction. Full-length pUL97 and C-terminally truncated versions lacking 1, 5, 15, 50 or 112 amino acids were transiently transfected into 293T cells (concentrations of plasmid DNA were adjusted between 10 µg and 20 µg to achieve homogeneous expression levels). Three days post-transfection, cells were lysed and endogenous cyclin B1 (B) or cotransfected pUL97-HA (C) was immunoprecipitated. Coimmunoprecipitated proteins (pAb-cyclin B1, positive control mAb-HA, negative controls rabbit Fc fragment and pre-immune serum not shown) were analyzed via WB analysis (B,C). For measuring in vitro kinase activity of C-terminally truncated versions of pUL97, an IVKA was performed with the immunoprecipitates (A). In reactions 2-9, a standard substrate of pUL97, i.e., purified histones H1-H4 (comigrating in one band as resulting from the chosen SDS-PAGE conditions), was added exogenously together with [γ-33 P]-ATP. Phosphorylated proteins were separated by SDS-PAGE/WB and analyzed on a phosphoimager plate. . Importance of the C-terminus of pUL97 for kinase activity, cyclin B1 binding and pUL97-pUL97 interaction. Full-length pUL97 and C-terminally truncated versions lacking 1, 5, 15, 50 or 112 amino acids were transiently transfected into 293T cells (concentrations of plasmid DNA were adjusted between 10 µg and 20 µg to achieve homogeneous expression levels). Three days post-transfection, cells were lysed and endogenous cyclin B1 (B) or cotransfected pUL97-HA (C) was immunoprecipitated. Coimmunoprecipitated proteins (pAb-cyclin B1, positive control mAb-HA, negative controls rabbit Fc fragment and pre-immune serum not shown) were analyzed via WB analysis (B,C). For measuring in vitro kinase activity of C-terminally truncated versions of pUL97, an IVKA was performed with the immunoprecipitates (A). In reactions 2-9, a standard substrate of pUL97, i.e., purified histones H1-H4 (comigrating in one band as resulting from the chosen SDS-PAGE conditions), was added exogenously together with [γ-33 P]-ATP. Phosphorylated proteins were separated by SDS-PAGE/WB and analyzed on a phosphoimager plate.

The Cyclin B1-pUL97 Complex Precipitated from HCMV-Infected Cells Comprises Substrate-Binding and Substrate-Phosphorylating Activity
The specific phosphorylation of proteins present in pUL97-cyclin B1 coimmunoprecipitates was investigated by the use of HCMV-infected primary fibroblasts. For this purpose a combined CoIP-IVKA was performed to demonstrate the activity and substrate specificity of pUL97 contained in the cyclin complex (the identity of proteins was confirmed by WB restaining of the identical WB membranes; Figure 5, panels CoIP control). Autophosphorylation of pUL97 and phosphorylation of one of the most abundant viral substrate proteins, pp65 [52], was detected in the control sample derived from direct pUL97 immunoprecipitation ( Figure 5, lane 4). Both activities, i.e., auto-and substrate phosphorylation of pUL97, albeit to a lower intensity, were similarly contained in the complex derived from cyclin B1 CoIP (lane 2). An upregulation of distinct CDKs/cyclins during HCMV replication had been described by various reports [11] and this finding was confirmed for cyclin B1 in our experiment, with an increased level in HCMV-infected cells compared to uninfected cells ( Figure 5, lanes 2-5 versus lane 1, expression control). The pUL97 specificity of the reaction was indicated by suppression with the pUL97-directed inhibitor MBV (lane 3) and the reliability of the setting was monitored by negative controls (lanes 1 and 5). Of note, we did not detect substantial amounts of cyclin B1 phosphorylation by pUL97 (or any cross-phosphorylation by putatively coimmunoprecipitated CDK activity). This lack of detectable amounts of pUL97-specific in vitro phosphorylation of cyclin B1 was also confirmed by equivalent settings using other cyclins as CoIP targets (i.e., using antibodies against cyclins T1, A or H; data not shown). This finding was surprising, although it should be mentioned that the IVKA reactions were performed in an assay buffer previously optimized for pUL97 that only allows some residual activity of CDKs or other possibly associated protein kinases. In principle, we favor a model of cyclin-associated higher order protein complexes that might include more than one kinase activity (possibly additional CDK activity), which, however, includes activity regulation of the kinases and possibly changes in substrate specificity (see below, Figure 5). At this point, it appears also possible that pUL97-cyclin interactions within such complexes might at least to some extent be bridged by a third protein. Importantly, the strong phosphorylation signal obtained for pp65 indicated that at least one viral substrate protein is closely associated with the pUL97-cyclin B1 complex in HCMV-infected cells. Further investigation of the nature of such multiprotein complexes is presently being performed by the use of a recently-established, sensitive, tandem mass spectrometry approach [53].

Discussion
Interactions between herpesviral protein kinases and host cell-derived cyclins have not been well-characterized to date. The first exception was given by the previously identified interaction between HCMV pUL97 and cyclin T1 [45]. This interaction, although not detectable in high abundance but rather at limited levels of signals, was demonstrated for endogenous cyclin T1, both upon pUL97 overexpression and HCMV productive infection. In addition, an interaction with several types of Combined coimmunoprecipitation-in vitro kinase assay (CoIP-IVKA) demonstrating the pUL97-specific phosphorylation of proteins in pUL97-cyclin B1 coimmunoprecipitates derived from HCMV-infected cells. HFFs were infected with HCMV AD169-GFP (approx. MOI 1.0). Four days post-infection, 3 µM of the pUL97 inhibitor was added to the medium to address the question of pUL97-specific phosphorylation. On day five, GFP expression was controlled by microscopic inspection before cell lysates were collected. Cyclin B1 (lanes 1-3) or pUL97 (lane 4) was immunoprecipitated using polyclonal antibodies as indicated; nonspecific precipitation was excluded by performing control IPs with Fc fragment (lane 5) and Flag-specific antibodies (not shown). CoIP samples were subjected to an IVKA reaction and phosphorylated proteins were detected by SDS-PAGE/WB and phosphoimager exposure. A restaining of the radioactive blot (CoIP control) and staining of additional expression control WBs indicated the reliability of protein precipitation and expression levels, respectively.

Discussion
Interactions between herpesviral protein kinases and host cell-derived cyclins have not been well-characterized to date. The first exception was given by the previously identified interaction between HCMV pUL97 and cyclin T1 [45]. This interaction, although not detectable in high abundance but rather at limited levels of signals, was demonstrated for endogenous cyclin T1, both upon pUL97 overexpression and HCMV productive infection. In addition, an interaction with several types of cyclins was also found, i.e., besides cyclin T1 also B1 and A. This finding was confirmed by the data of the present study, stressing the point that pUL97 interaction with cyclin B1 occurs at a level of higher affinity than with cyclin T1. In the initial report by Graf et al. [45], the pUL97 region responsible for cyclin T1 interaction was mapped to amino acids 231 to 280, located in the poorly-structured N-terminal portion of pUL97. This cyclin T1-binding region overlaps with the self-interaction domain [35]. The finding already pointed to a possible link between cyclin interaction and pUL97-pUL97 interaction. It was speculated that the formation of pUL97 dimers and oligomers may be bridged through cyclin interaction. This concept has been further substantiated by the important finding of the present report that pUL97-cyclin B1 interaction is strictly dependent on kinase activity. Three points of evidence support this notion, i.e., cyclin interaction was experimentally abrogated by (i) treatment with pUL97 inhibitors; (ii) replacement of an essential lysine of the ATP-binding site (K355M); and (iii) deletion of C-terminal parts of pUL97 also required for catalysis. This indicates that pUL97-cylin interaction is limited by structural properties of an active kinase domain. It is currently unknown whether the binding of a small molecule, such as the applied quinazoline compounds or drug candidate MBV, is able to sterically block the access of the kinase domain to cyclin interaction or whether a block of cyclin interaction is caused by a binding-incompetent conformation of pUL97. However, the finding that three different experimental situations (all associated with an inactive state of pUL97) led to an abrogation of cyclin interaction strongly favors the latter possibility. This is also supported by our bioinformatical investigations providing the first theoretical concept for conformational changes within the pUL97 kinase domain that may directly determine cyclin interaction.
Since our current state of investigations does not suggest a dependency of pUL97 activity on cyclin binding, we postulate a refined model based on the following considerations of the pUL97 structure-activity relation: The biochemical state of de novo-synthesized pUL97 might allow for the correct folding of an initial conformation of the kinase domain characterized by a basic level of activity. In a next step, pUL97 might undergo massive autophosphorylation (possibly also including additional transphosphorylation by cellular kinases such as CDKs; reviewed by Marschall et al. [3,54]). The state of autophosphorylation is considered a putative trigger for a conformational change that might render the kinase domain accessible to cyclin interaction. This conformational change might also provide full intrinsic activity to the catalytic cleft within the kinase domain. Thus, in a cyclin-associated form, pUL97 might be fully competent for substrate interaction and downstream phosphorylation activity. More biochemical data will be necessary to validate this concept and, in particular, our ongoing research based on mass spectrometry analysis of protein complexes and the detailed investigation of drug-kinase interaction may provide important information in the near future.

Author Contributions
M.S. performed main parts of the experimental work and contributed to writing the manuscript; E.S. and H.S. performed bioinformatical analyses and contributed to writing the manuscript; C.H., R.W., T.B. and T.L. performed additional experimental work and provided scientific and methodological support; C.H. contributed to the editorial work during manuscript revision; M.M. designed the study, evaluated data and wrote main parts of the manuscript.