Synthetic lethal mutations in the cyclin A interface of human cytomegalovirus

Generally, the antagonism between host restriction factors and viral countermeasures decides on cellular permissiveness or resistance to virus infection. Human cytomegalovirus (HCMV) has evolved an additional level of self-imposed restriction by the viral tegument protein pp150. Depending on a cyclin A-binding motif, pp150 prevents the onset of viral gene expression in the S/G2 cell cycle phase of otherwise fully permissive cells. Here we address the physiological relevance of this restriction during productive HCMV infection by employing a cyclin A-binding deficient pp150 mutant virus. One consequence of unrestricted viral gene expression in S/G2 was the induction of a G2/M arrest. G2-arrested but not mitotic cells supported viral replication. Cyclin A destabilization by the viral gene product pUL21a was required to maintain the virus-permissive G2-arrest. An HCMV double-point mutant where both pp150 and pUL21a are disabled in cyclin A interaction forced mitotic entry of the majority of infected cells, with a severe negative impact on cell viability and virus growth. Thus, pp150 and pUL21a functionally cooperate, together building a cell cycle synchronization strategy of cyclin A targeting and avoidance that is essential for productive HCMV infection.

Introduction Control of the cell division cycle by cyclins, cyclin-dependent kinases (CDKs) and CDK inhibitors (CKIs) is fundamental for proliferation, development and homeostasis of multicellular organisms [1,2]. To reprogram the cell cycle for their own benefit, viral pathogens have evolved, or acquired from their host, genes and sequence motifs facilitating direct interaction with the cyclin-CDK protein network [3].
Herpesviruses are particularly well suited for multifaceted interactions with the cell cycle machinery, owing to the large coding capacity of their genomes. The repertory of herpesviral cell cycle regulators comprises on the one hand factors leading to constitutive activation of the cell cycle. This is exemplified by the β and γ-herpesviral orthologs of cyclins [4] and CDKs [5], which release CDK substrate phosphorylation from the control of cellular cyclins and CKIs [6,7]. On the other hand, herpesviruses target cellular cyclin-CDK activity to arrest the cell cycle at stages conducive to virus replication [8]. A recent example is the UL21a gene product (pUL21a) of human cytomegalovirus (HCMV), which is required to block DNA synthesis and mitotic entry of infected cells [9,10]. Like CKIs of the Cip/Kip family (p21, p27, p57), pUL21a contains a high affinity RXL-type cyclin binding motif but is only a poor CDK substrate [10]. In contrast to CKIs, however, pUL21a does not act as a stoichiometric inhibitor of cyclin-CDK complexes but specifically recruits cyclin A (also referred to as cyclin A2) for proteasomal degradation [9,10].
Viral interactions with the cell cycle are not necessarily unidirectional. HCMV encodes a second RXL-type cyclin A-binding protein, pp150 (also referred to as pUL32), that is neither an activator nor an inhibitor of the cell cycle but is itself subject of cyclin A-CDK-dependent regulation [11]. PP150 enters the host cell as part of the HCMV virion and blocks de novo viral gene expression in a cyclin A and CDK-dependent manner [12,13]. In fibroblasts and other permissive cell types, this mechanism restricts the onset of viral replication to the G0/G1 phase of the cell cycle where cyclin A expression is low or absent. S/G2 cells, though, do not abrogate but only delay infection as pp150-mediated repression is relieved once cells loose cyclin A protein after cell division [14,15]. Animal CMVs, including chimpanzee CMV, the closest relative of HCMV, lack RXL sequence motifs in their pp150 homologues and accordingly initiate viral gene expression independent of the cell cycle position at the time of infection [11,16]. Thus, it is yet unclear what function, if any, the pp150-dependent restriction serves in the context of productive HCMV infection.
Here, we show that the pp150-RXL motif, alone, is dispensable for efficient viral growth. However, genetic disruption of both pp150 and pUL21a-RXL motifs dramatically enhances the mitotic phenotype and growth defect of a pUL21a-RXL single mutant virus. Thus, the cyclin A antagonist pUL21a and the cyclin A sensor pp150 are part of a virus-host interface, that functions as a fail-safe system securing undisturbed HCMV replication under non-mitotic conditions.

Results
Besides its role as a cell cycle-dependent restriction factor, pp150 has a well-documented function in the late phase of HCMV infection where it is required for capsid trafficking and stability [17][18][19], virion maturation and egress [20][21][22]. Before we began to use the cell cycleindependent HCMV-pp150-RXL mutant for investigating the consequences of unrestricted viral gene expression in S/G2, we made sure that the essential late functions of pp150 are not hampered by the cyclin binding deficiency. To this end we infected G0/G1-arrested fibroblasts, which, independent of the pp150-RXL mutation status, supported the ganciclovir-sensitive de novo synthesis of viral DNA ( Fig 1A) and all stages of HCMV protein expression (Fig 1B). By measuring the accumulation of infectious progeny in the supernatant, we confirmed that pp150-RXL mutants were capable to grow to almost the same high levels as the corresponding WT and revertant viruses ( Fig 1C). Furthermore, we made sure that the pp150-RXL mutation does not negatively affect virion infectivity (S1 Fig). Thus, in principle, pp150-RXL mutant HCMV is fully competent in virus replication and release.
We then infected proliferating cells, that were partially synchronized in S phase by release from contact inhibition. As expected, only pp150-RXL mutant HCMV was able to initiate viral immediate early (IE) gene expression in S phase cells (S2 Fig). Whereas HCMV-WT infected S phase cells, like non-infected control cells, were able to complete the cell cycle and divided between 6 and 24 h post infection, the pp150-RXL mutant virus blocked cell division leading to an accumulation of cells with a G2/M DNA content (Fig 2A-2C). At later times, we observed a shift of the G2/M-arrested population to a > 4n DNA content, resembling the gain of DNA content seen in the G1-arrested fractions of HCMV-WT and pp150-RXL mutant   (Figs 1A and 2A). The late increase in DNA content was of viral origin as it was for the most part prevented by treatment with ganciclovir, an inhibitor of the HCMV DNA polymerase (Fig 2D). This indicated that the pp150-RXL mutant virus is able to replicate its genome in G2 cells with similar speed and efficiency as in G1 cells.
This view was further supported by analysis of viral IE, early and late gene expression. We employed a 5-ethynyl-2'-deoxyuridine (EdU) pulse labeling strategy [23] to separately track G1 and S phase-infected cells. The negative impact of S phase infection on HCMV-WT gene expression was still evident 3 to 4 days later by reduced levels of viral early (gB) and late (pp28) gene products in EdU-positive cells. In contrast, the pp150-RXL mutation allowed the cascade of lytic gene expression to proceed with similar strength and kinetics in EdU-positive and negative cells (Fig 2E). The presence of pp28 which, as a "true" late gene product, depends on viral DNA synthesis [24], supported our conclusion from the ganciclovir experiment that efficient viral replication can occur at late stages of the cell cycle, if only the cyclin A-dependent block of IE gene expression is overcome.
We then had a closer look at the cell cycle position of pp150-RXL mutant infected cells. To check whether the observed block in cell division takes place in G2 or M phase, we analyzed histone H3 serine-10 phosphorylation. Although this phosphosite has a dual role in chromosome condensation [25] and transcriptional elongation [26], and was recently found increased in HCMV-Ad169 infected interphase cells [27], the high abundance of histone H3 de novo phosphorylation during M phase [28] makes it a reliable and well-accepted marker of mitosis in flow cytometry and immunocytochemistry. We identified a small but, compared to WT, statistically significantly increased population of pp150-RXL mutant infected M phase cells that in contrast to G2 cells showed no signs of viral DNA replication (Fig 2F and 2G). Thus, in the absence of pp150-mediated restriction, HCMV is fully competent to replicate from the S/G2 cell cycle compartment but has an increased risk to enter into an abortive mitotic state.
The finding of non-permissive mitotic cells was reminiscent of the phenotype of pUL21a-RXL mutant HCMV, which has lost the capacity to block the G1/S transition by cyclin A down-regulation and in consequence, forces up to 30% of infected cells into a non-productive and genetically unstable metaphase arrest [10]. This prompted us to ask whether both mechanisms, the pp150-cyclin A-dependent restriction of viral gene expression to G0/G1 and the pUL21a-cyclin A-dependent cell cycle block may cooperate to protect HCMV from fatal entry into mitosis. To address this question we constructed a virus carrying double-point mutations of both, the pUL21a and pp150 RXL motifs and compared the effects on cell cycle progression and virus replication side by side with the corresponding single mutants and HCMV-WT. During the first 12 to 24 h after virus entry, the pp150 status clearly dominated the phenotypic outcome of HCMV infection. The pUL21a-RXL single mutant, like WT virus, was unable to start IE gene expression in S/G2 (S3 Fig), and hence also to block cell division ( Fig 3A). In contrast, the pp150/pUL21a-RXL double mutant behaved like the pp150-RXL single mutant virus in these respects. From 24 h on, the consequences of uncontrolled cyclin A expression (S4 Fig) became apparent in RXL double mutant and, with a delay of further 24 h, also in pUL21a-RXL single mutant infected cells: i) cells moved from G1 towards G2/M ( Fig 3A); ii) mitotic untreated. DNA histograms were obtained by flow cytometry. (E) EdU incorporation and the expression of selected viral immediate early (IE1/2), early (gB) and late (pp28) gene products were determined by flow cytometry at different time points post infection. EdUpositive cells are displayed in red, EdU-negative cells in gray. (F) IE-positive cells were analyzed by flow cytometry for phosphorylation of histone H3 at serine 10 (pH3(ser10)), a marker of mitotic chromatin condensation. The relative proportion of mitotic cells (mitotic index) is given in percent of total cells. (G) The averages and standard deviations of mitotic indices were calculated from six independent experiments. Statistically significant differences, based on a two-tailed, paired Student's t test, are marked with asterisks; ** (p < .01); * (p < .05).
doi:10.1371/journal.ppat.1006193.g002  (Fig 3B, S5 Fig); iii) expression of mitotic kinases cyclin B, aurora A and B was strongly up-regulated above the already increased levels in WT and pp150-RXL mutant infected cells (S4 Fig). The latter observation was consistent with our previous finding [10] that loss of pUL21a-cyclin A interaction enhances the long known stimulatory effect of HCMV on cyclin B expression [29].
Not only was the timing of mitotic entry accelerated by the simultaneous deletion of both cyclin A interaction sites, its extent was greatly elevated as well (Fig 3B, S5 Fig). The mitotic index peaked at 48 h in case of the RXL double mutant, reaching 60% of IE-positive cells. At this time, only 10-12% of the pp150 or pUL21a-RXL single mutant infected cells had entered mitosis (Fig 3B), demonstrating a synergistic pro-mitotic effect of the RXL double mutation. The high and early incidence of mitotic cells had a huge negative impact on viral replication, which was reflected by the near absence of cells accumulating with a greater than 4n DNA content ( Fig 3A, S3 Fig) and by an about 500-fold growth defect (Fig 3C) of the RXL double mutant. In contrast, the pp150-RXL single mutation conferred even a growth advantage on HCMV during the first 4 to 6 days after S-phase infection, compared to WT virus. This suggests that in the presence of a functional cyclin A degradation mechanism, the fast, cyclin Aresistant onset of viral gene expression outweighs the negative consequences of a moderately increased mitotic index.
Notably, the pUL21a-RXL single mutant showed a greater lag in accumulation of infectious progeny (Fig 3A) than previously reported for G0/G1 phase infection experiments [10]. That was reflected at the protein level by reduced expression of the essential viral trans-activator IE2 and of early and late gene products (S4 Fig). The changes are most likely due to the continuous presence of cyclin A in pUL21a-RXLmut infected S phase cells (S4 Fig) which in the presence of pp150-WT is known to exert a negative effect on viral gene expression [9,12,30]. Analysis of early (gB) and late (pp28) protein profiles in the fraction of IE-positive S/G2 cells confirmed that, in contrast to pp150-RXLmut infection, the G2 arrested state was only semi-permissive for the pUL21a-RXL single mutant in terms of viral replication (S6 Fig). Remarkably, up to 30% of all pUL21a-RXLmut and WT infected cells remained refractory to IE gene expression for 2 days, even after re-entry into G1 phase (S3 Fig). Whereas WT closed the gap to pp150-RXL mutant infections between 2 and 4 days post infection (S3 Fig), the number of IE-deficient pUL21-RXL single mutant infected cells only slowly decreased, consistent with the delayed growth of this mutant (Fig 3A).
Although mitotic entry of infected cells was characterized by a shutdown of viral gene expression (S6 Fig), the RXL double mutant was still capable to maintain viral gene expression and replication in G2 (S4 and S6 Figs). To clarify what causes the 500-fold growth defect, we analyzed the influence of mitotic entry on the stability and viability of RXL-double mutant infected cells. In fact, those cells showed, similar to the pUL21a-RXL single mutant [10], chromosomal damage that progressively leads to a complete pulverization of the chromosomal material ( S7 Fig). Furthermore, the decrease of mitotic cells seen after 2 dpi (Fig 3B) was paralleled by a large die-off of double mutant infected cells (Fig 3D, S8 Fig) catastrophe and cell death. Regarding this instability of mitotic cells, the total number of double mutant infected cells entering mitosis during the time course of the experiment probably was greatly underestimated by flow cytometry (Fig 3B, S5 Fig), which gives only a snap-shot of the relative cell cycle distribution of viable cells.
We conclude that in the absence of the pp150/pUL21a-cyclin A interface, cyclin A and cyclin A-resistant viral gene expression together lead to the fast kinetics and high penetrance of mitotic entry seen for RXL-double mutant infected cells, with severe consequences for cell survival and virus growth (Fig 3E).

Discussion
HCMV encodes two cyclin A interacting proteins, the cyclin A-CDK substrate pp150 and the cyclin A destabilizing module pUL21a. Here we show that both proteins act synergistically, together constituting a control circuit required for the synchronization of viral replication with the cell division cycle. The tegument protein pp150 senses the cellular cyclin A status at the beginning of infection and restricts the onset of IE gene expression to the G0/G1 phase, where cyclin A2-CDK is inactive. The early gene product pUL21a maintains the status of low cyclin A2-CDK activity by targeting cyclin A for proteasomal degradation. Thus, pp150-mediated restriction of viral gene expression in S/G2 phase has a genuine function in the productive replication cycle of HCMV instead of being merely an inevitable by-product of a silencing mechanism that contributes to establishing quiescent infection in undifferentiated cells [11].
The pp150-dependent restriction to cyclin A-negative cells appears important enough to justify a significant delay of virus gene expression and replication after infection of proliferating cells (Figs 2E and 3C). This is particularly remarkable in view of the high overall growth rates of a cyclin A sensor-less pp150-RXL mutant virus in both G1 and G2 cell cycle compartments (Figs 1C and 3C). In that respect, pp150 mutant HCMV behaves like animal CMVs, which lack a pp150-cyclin A interface and therefore can efficiently replicate in S/G2 cells [11,16]. Taken together, this may point to a model where cyclin A sensing by pp150 alone is not absolutely required for productive HCMV infection but has been developed to strengthen and support cell cycle synchronization by pUL21a. Because pUL21a is expressed with early kinetics and therefore not available during the first hours of infection [31], a pre-synchronization step by a protein like pp150, delivered by the incoming virion, makes sense and keeps lytic gene expression in safe distance from G2/M transition, giving pUL21a time to install a stable cell cycle arrest in interphase. If this presynchronization is missing the pUL21a function is not sufficient, or is not present early enough, to tightly inhibit mitotic entry in every infected cell (Figs 2F, 2G and 3B).
The described synchronization strategy of HCMV is in striking analogy to how another human pathogen, human papilloma virus (HPV) coordinates its replication with the host cell cycle. Just as HCMV, many HPV strains encode two RXL motif containing cyclin A interactors, E1 and E1˄E4 [32,33]. The early protein E1 is a DNA helicase that, due to its cyclin A-CDK-dependent nuclear localization, is only able to initiate viral DNA synthesis in its CDK-phosphorylated form [33,34]. This makes sense from the viewpoint of HPV given that this small DNA virus in contrast to HCMV heavily depends on the cellular DNA replication machinery, which is only available and active at times of high cyclin A-CDK activity. The late protein E1˄E4 re-localizes cyclin A-CDK to the cytoplasm, thereby preventing the onset of mitosis [35,36]. Thus, both HCMV and HPV have evolved two layers of cyclin A interaction. The first layer consists of cyclin A-sensitive CDK substrates that confine the start of viral replication to the most suitable cell cycle phase-G1 in case of HCMV, S phase in case of HPV. The second layer consists of potent, negative regulators of cyclin A-CDK that provide stable conditions for virus growth by arresting the cell cycle in interphase (HCMV: G1 arrest, HPV: G2 arrest). It is fascinating, that two only distantly related viruses like HCMV and HPV have developed such similar strategies to best synchronize their life cycle with that of virally favorable phases of the cell cycle for reaching highest efficacy and efficiency of viral replication.
Although HCMV, HPV and many other viruses have evolved sophisticated mechanisms to prevent mitotic entry in productively infected cells [37], it is important to note that in other contexts, such as virus entry or persistence, viruses use mitosis to their own advantage. Herpesviruses as well as papillomaviruses encode mitotic chromosome tethering factors to maintain latent viral genomes episomally in dividing cells [38][39][40][41]. Papillomaviruses and select retroviruses require the nuclear envelope breakdown in mitosis for nuclear import of viral genomes [42].
Why then mitosis presents a problem when it occurs in the middle of the productive replication cycle of HCMV? Certainly, the global shutdown of gene transcription [43] and mRNA translation [44] in mitosis would seem to be counterproductive for a virus that hijacks the host cell machinery to achieve maximum replication. Also, structural changes like the nuclear envelope breakdown are potentially harmful for the functional integrity of viral replication and assembly compartments [10]. However, if mitosis would represent only a short and transient interruption in the replication cycle of HCMV, the virus could possibly cope with such perturbations. But, HCMV, as many viruses, encode potent inhibitors of the anaphase promoting complex (APC/C) to stabilize substrates of this E3 ubiquitin ligase that play a role in viral replication [45][46][47][48]. Due to the essential function of APC/C in mitosis, productively infected cells cannot simply traverse through this cell cycle phase-they become arrested at the metaphaseanaphase transition [10]. Another aspect is that HCMV expresses DNA damaging enzymes [49,50] and subverts cellular DNA repair [51,52]. It is unclear, as yet, to what extent virusinduced DNA breaks contribute to the chromosomal fragmentation visible in mitotic cells (S7 Fig). However, unrepaired DNA damage and prolonged metaphase arrest are known to cause cell death by mitotic catastrophe [53,54] and, though HCMV is a master in preventing premature cell death during interphase [55], it is evidently not equally prepared to protect productively infected cells in mitosis.

Cells and viruses
Human embryonic lung fibroblasts (Fi301) were maintained as described previously. To synchronize them in S phase, fibroblasts were first synchronized in G1 phase by contact inhibition and then seeded at lower cell density to allow reentry into the cell cycle. Thirteen to seventeen hours after re-plating when most cells had reached early S phase, they were infected. The following recombinant viruses were used: the parental WT virus HCMV-TB40-BAC4 [56], HCMV-TB40-pUL21a-RXLmut [10], the HCMV-TB40-pp150-RXL mutant RV1659 and revertant RV1677 [11]. To obtain a HCMV-TB40-pp150/pUL21a-RXL double mutant, the pUL21a-RRL ARA mutation was introduced into RV1659 by traceless BAC mutagenesis [57]. Viruses were propagated on Fi301 cells and titered by flow cytometry of IE1/IE2-positive cells as described [10]. A multiplicity of infection (MOI) of 5 to 10 IE-protein-forming units (IU) per cell was used for all experiments. To determine particle-to-IU ratios in virus stocks, virion DNA was prepared by ultracentrifugation and proteinase K/SDS treatment essentially as described [58], and quantified by real-time PCR using the UL123-specific primer pair 5'-GCCT TCCCTAAGACCACCAAT-3' / 5'-ATTTTCTGGGCATAAGACATAATC-3'. For detection of S phase cells, cells were pulse-labelled (60 min) with 10 μM 5-ethynyl-2´-desosxyuridine (EdU) before infection. Where indicated, ganciclovir was used at a final concentration of 50 μM. Giemsa staining was performed essentially as described [10].