Serine 13 of the human cytomegalovirus viral cyclin-dependent kinase UL97 is required for regulatory protein 14-3-3 binding and UL97 stability

The human cytomegalovirus (HCMV) UL97 protein is a conserved herpesvirus protein kinase (CHPK) and a viral cyclin-dependent kinase (v-CDK). However, mechanisms regulating its activity in the context of infection are unknown. Here, we identified several cellular regulatory 14-3-3 proteins as UL97-interacting partners that promote UL97 stability. Humans are known to encode seven isoforms of 14-3-3 proteins (β, ε, η, γ, σ, θ, and ζ) that bind phosphoserines or phosphothreonines to impact protein structure, stability, activity, and localization. Our proteomic analysis of UL97 identified 49 interacting partners, including 14-3-3 isoforms β, η, and γ. Furthermore, coimmunoprecipitation with Western blotting assays demonstrated that UL97 interaction with 14-3-3 isoforms β, ε, η, γ, and θ occurs in a kinase activity-dependent manner. Using mutational analysis, we determined the serine residue at amino acid 13 of UL97 is crucial for 14-3-3 interaction. We demonstrate UL97 S13A (serine to alanine substitution at residue 13) retains kinase activity but the mutant protein accumulated at lower levels than WT UL97. Finally, we show both laboratory (AD169) and clinical (TB40/E) strains of HCMV encoding UL97 S13A replicated with WT kinetics in fibroblasts but showed decreased UL97 accumulation. Taken together, we conclude that 14-3-3 proteins interact with and stabilize UL97 during HCMV infection.

Mechanisms that regulate the activity of cellular CDKs are well understood and include the phosphorylation of specific CDK residues, the interaction with cyclins and CDK inhibitors (CKIs), and associations with chaperone complexes (14). Despite the functional similarities between UL97 and cellular CDKs, our understanding of UL97 regulatory mechanisms is limited. UL97 is phosphorylated by CDK9/cyclin T1, but the significance of this event is unknown (15). Furthermore, UL97 associates with the cellular cyclins A2, B1, H, and T1, but whether these associations affect kinase activity or substrate specificity has not been established (15)(16)(17). While the CDK inhibitor p21 does not efficiently inhibit UL97-mediated phosphorylation compared with cellular CDKs (1), other CKIs have not been tested for their ability to regulate UL97 kinase activity. Similarly, if and how the other v-CDKs (EBV BGLF4; HHV-6a, HHV-6b, and HHV-7, U69; KSHV ORF36) (2) are regulated is not well established.
We show here that UL97 binds to five different 14-3-3 proteins (β, ε, η, γ, and θ) dependent upon the Ser-13 residue of UL97, an autophosphorylation site found in its unique N-terminal extension. Interaction with 14-3-3 proteins increases UL97 stability but did not substantially contribute to viral fitness in the in vitro productive replication assays commonly employed to study HCMV. Our work identifies a unique function for the N terminus of UL97 and highlights 14-3-3 proteins as additional cellular regulators of HCMV.
We found the residues that control 14-3-3 isoform-specific binding to GluN2C also control isoform-specific binding to UL97. For example, while the σ isoform fails to interact with UL97, a triple mutant converting the three key σ residues listed previously to the corresponding residues of the other 14-3-3 isoforms (A147S/H180Y/I191C) was able to bind UL97 (Fig. 2B). Furthermore, 14-3-3 isoforms β, ε, η, γ, and θ with these amino acids mutated to the σ isoform-specific residues (e.g., for the β isoform: S147A/Y180H/C191I) showed reduced interaction with UL97 (Fig. 2C). While these three residues dictate isoform-specific binding, they cannot do so exclusively, as the ζ isoform shares the binding promoting residues ( Fig. 2A) but still fails to bind UL97. Interestingly, α helix 6 of the ζ isoform initiates with a nonpolar amino acid (alanine-134) while all the other isoforms contain a polar, noncharged residue at this position (cysteine, serine, or threonine) ( Fig. 2A) (35). A mutant form of the ζ isoform with the nonpolar alanine replaced with a polar, noncharged cysteine (A134C) showed improved binding (Fig. 2D), whereas a reciprocal θ isoform mutant (C134A) showed reduced binding (Fig. 2E). In total, we conclude that UL97 utilizes isoform-specific residues in the sixth, seventh, and eighth α helices of 14-3-3 proteins for binding specificity.
14-3-3 proteins often bind to phosphorylated Ser/Thr residues on their clients (23). We found that a kinase dead (KD) allele of UL97 (K355Q) (catalytic lysine 355 converted to glutamine; Fig. 4A) accumulated to lower levels and showed reduced or eliminated binding to all five 14-3-3 proteins tested (Fig. 4B). The lack of autophosphorylation of UL97-KD can be observed by its enhanced electrophoretic mobility as compared to WT UL97. Dose-response experiments (Fig. 4C) and their quantitation (Fig. 4D) showed that UL97-KD, when accumulated to similar levels as UL97-WT, failed to associate with or less efficiently associated with all 14-3-3 proteins tested (Fig. 4, C and D). While incapable of autophosphorylation, UL97 N-terminal fragments still associate with 14-3-3 proteins, although far less efficiently than the full-length protein (Fig. 3), perhaps because they are inefficiently phosphorylated by cellular kinases (15). We conclude that the kinase activity of UL97, perhaps for autophosphorylation, is required for efficient 14-3-3 binding. All nine previously identified (19) autophosphorylation sites ( Table 2 and Fig. 4A) are within the newly identified 14-3-3 binding domain of UL97 (Fig. 3). Furthermore, we found 10 potential 14-3-3 binding sites in UL97 (Table 2) predicted by at least one of three independent algorithms (36) but only two of which were autophosphorylation sites (Ser-11 and Ser-13). In total, there are 17 serine or threonine residues in UL97 that could potentially mediate 14-3-3 binding (Table 2 and Fig. 4A). Of these, only Ser-13 is contained within the previously mapped 14-3-3 binding domain, autophosphorylated, and predicted by all three algorithms as a potential 14-3-3 binding site (Table 2).
Serine 13 is not required for the kinase activity of transfected UL97 14-3-3 binding can promote protein activity (39). While UL97-S13A showed some ability to autophosphorylate implying it is an active kinase, we wanted to test the ability of this mutant protein to phosphorylate a known UL97 substrate. The retinoblastoma (Rb) tumor suppressor is a well-established substrate for UL97 (1,2,4), whose phosphorylation can be detected both by the decreased electrophoretic mobility of the protein on SDS-PAGE, as well as by immunoblotting with a phosphospecific antibody (e.g., Rb-P-Thr-826). Rb is not phosphorylated when transfected into Saos-2 cells (Fig. 7) unless a kinase that phosphorylates Rb is cotransfected (40)(41)(42). UL97 phosphorylates Rb during Saos-2 transfections in the absence of overexpressed 14-3-3 proteins (1, 2, 4). However, in an attempt to maximize any differences in UL97-mediated Rb phosphorylation between UL97 and UL97-S13A downstream of differential 14-3-3 binding, we conducted new Saos-2 assays in the presence of overexpressed 14-3-3 proteins. When WT UL97 was coexpressed with each of the 14-3-3 proteins to which it binds (β, ε, η, γ, or θ), coexpressed Rb proteins were phosphorylated (Fig. 7), as expected. To gauge the ability of UL97-S13A to phosphorylate Rb, we transfected slightly larger amounts of plasmid to achieve the steady state protein level of WT UL97, as UL97-S13A is less stable (Fig. 5). Similar to WT, during coexpression with the 14-3-3 proteins, UL97-S13A resulted in Rb phosphorylation (Fig. 7). We conclude that UL97-S13A is an active kinase during transient transfections. 14-3-3 proteins stabilize HCMV UL97  n.s. * Figure 6. 14-3-3 antagonist Difopein decreases the stability of WT UL97. A, U-2 OS cells were transfected with an expression vector for WT or S13A mutant UL97 (0.4 μg) and the indicated Flag-tagged 14-3-3 isoform (0.6 μg) together with the indicated amounts (μg) of an expression plasmid for EYFP fused to Difopein. Plasmid amounts in transfections were equalized with EYFP only expression plasmids. After 48 h, cells were harvested and lysates were subjected to Western blotting with the indicated antibodies. B, the transfection with either no (white bar) or 0.3 μg (gray bar) of an expression plasmid for EYFP fused to Difopein shown in panel (A) were performed in biological triplicates. The protein level of HA-tagged UL97 was quantified and normalized to GAPDH. Values are presented relative to the value in the absence of EYFP fused to Difopein for each transfection of WT or S13A-UL97 (set at 100%). Error bars denote SDs. Statistical analysis utilized a two-tailed unpaired Student's t test. *p < 0.05; **p < 0.01; n.s., not significant (p > 0.05). Differing amounts of expression vector for WT UL97 or the S13A mutant of UL97 were used. Lysates were analyzed by Western blotting with the indicated antibodies. A representative image from one of three independent biological replicates is shown.

Serine 13 is not required for the kinase activity of UL97 during HCMV infection
Because transfected UL97-S13A is an active kinase, we predicted the mutant protein expressed during infection would be active as well. Indeed, the levels of Rb phosphorylated at threonine-826 were similar during HCMV infection with either rBAD-HA97 or rBAD-HA97-S13A at either high (Fig. 11A) or low (Fig. 11B) multiplicity of infections (MOIs). Neither UL97 nor UL97-S13A phosphorylated the cellular CDK-specific serine-249 or threonine-252 sites (1), as expected (Fig. 11, A and B). Furthermore, the Rb-suppressed E2F-1 gene product accumulated to similar levels, indicating that Rb was equivalently inactivated by both UL97 and UL97-S13A during infection (Fig. 11). We conclude that UL97-S13A is an active kinase during HCMV infection.

Serine 13 contributes to the protein stability of UL97 during HCMV infection
Because transfected UL97-S13A is less stable than WT, we predicted the mutant protein would also be less stable during HCMV infection. Indeed, UL97 levels were higher during both high (Fig. 11, A and C) and low (Fig. 11, B and D) MOI infections of rBAD-HA97 compared to rBAD-HA97-S13A. We conclude serine-13 promotes UL97 stability during HCMV infection.
Finally, we asked if the inability of UL97-S13A to bind 14-3-3 proteins impacts the ability of UL97 to render infected cells sensitive to the antiviral prodrug ganciclovir (GCV) that it phosphorylates and activates (47,48) or impacts the sensitivity of UL97 itself to maribavir (MBV), a small molecule inhibitor of UL97 kinase activity (49) recently approved by the FDA for patients with post-transplant CMV infections that do not respond to available antivirals (50,51). We found that in both AD169 and TB40/E strains, viruses expressing UL97 or UL97-S13A showed similar sensitivity to each drug (Fig. 13, A-D), although AD169 and TB40/E expressing UL97-S13A showed subtle resistance to MBV (Fig. 13, B and D). We conclude that 14-3-3 binding by UL97 minimally affects maribavir sensitivity but not ganciclovir sensitivity. In total, we conclude that serine 13 of UL97 mediates binding to 14-3-3 proteins and promotes the steady state accumulation of the v-CDK, UL97.

Discussion
Here, we identify and present the novel interaction between HCMV UL97 and cellular 14-3-3 isoforms β, ε, η, γ, and θ. Three 14-3-3 proteins (β, η, and γ) were found in both biological replicates of our UL97 interactome screen. Two additional 14-3-3 proteins (ε and θ) were also found in both biological replicates but also at low level in one of the two controls. However, they showed a >5-fold enrichment over controls (see Supporting information). Similarly, two previously identified UL97-binding partners (HSP90AA1 and TUBA1B) also showed a >5-fold enrichment over controls, even though they were found at low levels in one of the two controls (see Supporting information). The interactions between UL97 and 14-3-3 proteins depend upon both the kinase activity of UL97 as well as the autophosphorylation site at residue serine-13.
The association of 14-3-3 proteins with the N terminus of UL97 enhanced UL97 protein stability. A stability determinant in the N terminus of UL97 matches well with the observation  Figure 8. Phosphorylation of UL97 serine 13 increases its interaction with 14-3-3 proteins. U-2 OS cells were transfected with an expression vector for either an HA-tagged S13A mutant or a kinase-dead derivative of UL97 (0.7 μg) together with expression vectors for the indicated Flag-tagged 14-3-3 isoform (0.6 μg) and expression vectors for either a V5-tagged S13A mutant or kinase-dead derivative of UL97 (0.7 μg). After 48 h, lysates (Input) were harvested, subjected to immunoprecipitation with anti-HA antibody (IP: anti-HA), and analyzed by Western blot with the indicated antibodies. A representative image from one of two or three independent biological replicates is shown 14-3-3 proteins stabilize HCMV UL97 that the full-length protein is the most abundant UL97 isoform in HCMV-infected cells (22). Indeed, the unique long N-terminal region of UL97 is predicted to be structurally disordered (18) and such intrinsically disordered proteins are susceptible to proteolysis (52,53). The lack of an N-terminal extension in gammaherpesvirus v-CDKs implies that they may not bind 14-3-3 proteins, and indeed our preliminary efforts to isolate complexes between 14-3-3 proteins and BGLF4, the EBV v-CDK, have failed. 14-3-3 proteins regulate protein stability in multiple ways, including blocking proteolysis by sterically preventing client polyubiquitination (28), relocalizing clients away from degradation machinery (54), or can even act as chaperones to help proteins achieve or maintain more stable structures (55). How the 14-3-3 proteins stabilize UL97 remains to be explored.
Finally, in addition to the 14-3-3 proteins, our UL97 interactome analysis identified 27 potential new UL97 binding partners and/or substrates, as well as 19 previously identified binding partners and/or substrates. The list of new and old targets includes multiple heat shock proteins, as well as DnaJ and TRiC/ CCT family chaperones, providing a wealth of opportunities for further exploring the role of UL97 during HCMV infection.

Viruses and Bac mutagenesis
The bacterial artificial chromosome (BAC)-derived recombinant HCMVs were strain AD169 (pAD/Cre) (65) and strain TB40/E (TB40-BAC4) (66). All BAC mutagenesis was conducted using a two-step Red recombination procedure (67). We generated an HCMV mutant, rBAD-HA97-S13A, that expresses UL97 bearing an HA-tag at the N terminus and an alanine substitution of Ser-13 (pAD/Cre-HA97-S13A). The revertant HCMV, rBAD-HA97, that expresses WT UL97 bearing an HA-tag at the N terminus was prepared by BAC mutagenesis with pAD/Cre-HA97-S13A. Recombinant TB40/ E HCMVs, rTB40/E-HA97, and rTB40/E-HA97-S13A were generated in the same way. DNA fragments for homologous recombinations were obtained as gBlocks gene fragments from Integrated DNA Technologies (IDT). Sequences are available upon request. Recombinant BAC genomes were verified by sequencing the UL97 gene and flanking regions, as well as by genomic DNA digestion patterns with Xba I, which was introduced at the HA-tag site. Virus titers were determined by plaque assays on hTERT-BJ1 cells cultured in media with 5% FBS. HCMV infections were performed under serum-starved conditions if not otherwise specified.

Reagents
Maribavir was obtained from Med Chem Express (catalog no.: # HY-16305). Ganciclovir was obtained from Merck (SML2346). Both reagents were dissolved in dimethyl sulfoxide and added into the infected cells at a final concentration of

D C
Relative HA-UL97 level normalized to GAPDH * B Figure 12. Serine 13 is not required for productive HCMV replication in fibroblasts. A, serum-starved hTERT-BJ1 fibroblasts were infected with the indicated recombinant viruses at an MOI of 1. Cell-free virus was collected at 6 days post-infection (dpi) and titers were determined by plaque assay. Error bars denote SDs from three biological replicates. Statistical analysis utilized a two-tailed unpaired Student's t test. n.s., not significant (p > 0.05). B, analysis as in panel (A) with infections at an MOI of 0.1 and harvest at 10 dpi. C, serum-starved hTERT-BJ1 fibroblasts were mock infected (M) or infected with the indicated recombinant viruses at an MOI of 0.1. Lysates prepared at the indicated dpi were analyzed by Western blotting with the indicated antibodies. A representative image from one of three independent biological replicates is shown. D, the protein levels of HA-tagged UL97 at 9 dpi from the experiment shown in panel (C) were quantified and normalized to GAPDH levels. The value of HA-tagged UL97-S13A (white bars) is presented relative to the value of HA-tagged WT UL97 (gray bars: set at 1). Error bars denote SDs. Statistical analysis utilized a two-tailed unpaired Student's t test. *p < 0.05. E, cell-free virus was collected at 9 dpi of the experiment shown in panel (C) and titers were determined by plaque assay. Error bars denote SDs. Statistical analysis utilized a two-tailed unpaired Student's t test. n.s., not significant (p > 0.05). HCMV, human cytomegalovirus; MOI, multiplicity of infection.  Figure 13. 14-3-3 binding by UL97 minimally effects maribavir sensitivity but not ganciclovir sensitivity. A, serum-starved hTERT-BJ1 fibroblasts were infected with the indicated recombinant AD169-based viruses at an MOI of 0.1. Ganciclovir (GCV, 2 μM) or DMSO were added after 1 h. Cell-free virus was collected at 9 days post-infection (dpi) and titers were determined by plaque assay. The value shows virus titer relative to rBAD-HA97 with GCV. B, analysis as in panel (A) except in the presence or absence of maribavir (MBV, 2 μM). The value shows virus titer relative to rBAD-HA97 with MBV. C, serum-starved hTERT-BJ1 fibroblasts were infected with the indicated recombinant TB40/E-based viruses at an MOI of 0.05. GCV (2 μM) or DMSO were added after 1 h. Cell-free virus was collected at 9 days post-infection (dpi) and titers were determined by plaque assay as in panel (A). The value shows virus titer relative to rTB40/E-HA97 with GCV. D, analysis as in panel (C) except in the presence or absence of MBV (2 μM). The value shows virus titer relative to rTB40/E-HA97 with MBV. Error bars denote SDs from three biological replicates. Statistical analysis utilized a two-tailed unpaired Student's t test. n.s., not significant (p > 0.05). DMSO, dimethyl sulfoxide; MOI, multiplicity of infection.

Halo tag pull downs and LC-MS/MS analysis
Kinase was purified from HEK-293T cells transfected with pFC14a-HA-UL97 WT or pFC14a (Promega) with a 1 h bead incubation as previously described (5). Samples were eluted with TEV protease (Promega) and submitted to the University of Wisconsin Biotechnology Center for LC-MS/MS analysis with a Thermo Fisher Scientific Orbitrap Elite. Two independent biological replicates were analyzed. Results were analyzed using Mascot and Scaffold viewer, and peptides proteins reported had at least one identifying peptide with 1% false discovery rate threshold. In Table 1, the listed proteins were detected in HA-UL97-expressed samples but not in controls (pFC14a) and ordered by score (quantitative value normalized total spectra).

Data availability
The data supporting the findings of this study are available within the article and its supporting information.
Supporting information-This article contains supporting information.