Human Cytomegalovirus Can Procure Deoxyribonucleotides for Viral DNA Replication in the Absence of Retinoblastoma Protein Phosphorylation

ABSTRACT

Viral DNA replication requires deoxyribonucleotide triphosphates (dNTPs). These molecules, which are found at low levels in noncycling cells, are generated either by salvage pathways or through de novo synthesis. Nucleotide synthesis utilizes the activity of a series of nucleotide-biosynthetic enzymes (NBEs) whose expression is repressed in noncycling cells by complexes between the E2F transcription factors and the retinoblastoma (Rb) tumor suppressor. Rb-E2F complexes are dissociated and NBE expression is activated during cell cycle transit by cyclin-dependent kinase (Cdk)-mediated Rb phosphorylation. The DNA virus human cytomegalovirus (HCMV) encodes a viral Cdk (v-Cdk) (the UL97 protein) that phosphorylates Rb, induces the expression of cellular NBEs, and is required for efficient viral DNA synthesis. A long-held hypothesis proposed that viral proteins with Rb-inactivating activities functionally similar to those of UL97 facilitated viral DNA replication in part by inducing the de novo production of dNTPs. However, we found that dNTPs were limiting even in cells infected with wild-type HCMV in which UL97 is expressed and Rb is phosphorylated. Furthermore, we revealed that both de novo and salvage pathway enzymes contribute to viral DNA replication during HCMV infection and that Rb phosphorylation by cellular Cdks does not correct the viral DNA replication defect observed in cells infected with a UL97-deficient virus. We conclude that HCMV can obtain dNTPs in the absence of Rb phosphorylation and that UL97 can contribute to the efficiency of DNA replication in an Rb phosphorylation-independent manner.

IMPORTANCE Transforming viral oncoproteins, such as adenovirus E1A and papillomavirus E7, inactivate Rb. The standard hypothesis for how Rb inactivation facilitates infection with these viruses is that it is through an increase in the enzymes required for DNA synthesis, which include nucleotide-biosynthetic enzymes. However, HCMV UL97, which functionally mimics these viral oncoproteins through phosphorylation of Rb, fails to induce the production of nonlimiting amounts of dNTPs. This finding challenges the paradigm of the role of Rb inactivation during DNA virus infection and uncovers the existence of an alternative mechanism by which UL97 contributes to HCMV DNA synthesis. The ineffectiveness of the UL97 inhibitor maribavir in clinical trials might be better explained with a fuller understanding of the role of UL97 during infection. Furthermore, as the nucleoside analog ganciclovir is the current drug of choice for treating HCMV, knowing the provenance of the dNTPs incorporated into viral DNA may help inform antiviral therapeutic regimens.

INTRODUCTION

Two members of the family of conserved protein kinases encoded by herpesviruses (UL13 of herpes simplex virus 1 [HSV-1] and BGLF4 of Epstein-Barr virus [EBV]) were found to phosphorylate two substrates on a residue also targeted by cyclin-dependent kinase 1 (Cdk1) (1), indicating that these proteins mimic at least some activities of cellular Cdks. Subsequently, UL97 (2) and then the other beta- and gammaherpesvirus conserved protein kinases (3) were shown to display bona fide Cdk activity, establishing them as viral Cdks (v-Cdks). UL97 sits at the center of pharmacological anti-human cytomegalovirus (HCMV) therapy. It phosphorylates and thus activates the antiviral drug ganciclovir, a nucleoside analog that is currently the first-line treatment for HCMV infections (4, 5). It is also the target of the experimental inhibitor maribavir (MBV), which has yet to prove effective in phase III clinical trials (6–8). The central role of UL97 in HCMV drug therapy, the significant medical burden that HCMV infection represents, and the failure of common strategies to produce an effective vaccine against HCMV make understanding the role of UL97 during HCMV infection paramount.

Viruses deficient for UL97 synthesize less viral DNA (vDNA), export fewer capsids from the nucleus into the cytoplasm, and grow to much lower titers than wild-type (WT) viruses (9–11). Many substrates for UL97 have been identified or proposed (12, 13), but the role that phosphorylation of these proteins plays during HCMV infection is not understood, despite the fact that the kinase activity of UL97 is the critical component of current and investigatory therapies. One of the most prominent UL97 substrates is the retinoblastoma (Rb) tumor suppressor (2, 14), also a target of the cellular Cdks. Hypophosphorylated (active) Rb restrains the transactivation potential of the cellular E2F transcription factors, whose target genes comprise many of the enzymes required to synthesize DNA, including those specifically required for the synthesis of the deoxyribonucleotide triphosphates (dNTPs) that serve as the substrates of DNA replication (15–19). Hyperphosphorylated (inactive) Rb disassociates from E2F, allowing the expression of E2F-responsive genes. Many DNA viruses, including those classified as tumor viruses, inactivate Rb, and it is a long-held contention that Rb inactivation is required for the efficient replication of these DNA viruses, in part because Rb controls the expression of the enzymes that mediate both deoxyribonucleotide biosynthesis and polymerization (20–24).

Purine and pyrimidine ribonucleosides (rNs) (glycosylamines comprised of nitrogenous bases and ribose sugars) and their phosphorylated (ribonucleotides) and/or reduced (deoxyribonucleosides [dNs]) derivatives can be generated through overlapping de novo and salvage pathways (Fig. 1) (25). The de novo pathway utilizes metabolites like glucose and glutamine to synthesize the nucleotide bases and sugars (26), whereas the salvage pathway scavenges imported extracellular macromolecules and the by-products of nucleic acid catabolism (and other degradative processes) to resynthesize all classes of nucleosides from partial products (25, 27). Salvage pathways predominate during quiescence, providing the nucleotides required for the limited transcription and DNA repair occurring in these noncycling cells (28), whereas the de novo pathway can satisfy the high demand for dNTPs required for cellular DNA synthesis during the S phase of the cell cycle (19).

FIG 1.

Nucleotide biosynthesis. The black text and arrows denote the de novo pathway of nucleotide synthesis, and the red text and arrows denote the salvage pathway of nucleotide synthesis. The black and red boxes show that the indicated metabolites can be derived from both the de novo and salvage pathways. The involvement of HCMV in these processes is denoted by blue text and arrows. dNDPs, deoxyribonucleotide diphosphates.

Several circumstances led us to hypothesize that UL97 promoted viral DNA synthesis through the inactivation of Rb and promotion of nucleotide biosynthesis. While HCMV encodes DNA synthesis proteins, it does not encode any known nucleotide-biosynthetic factors, unlike other herpesviruses that encode both DNA replication and nucleotide synthesis enzymes (29). However, during infection, UL97 can phosphorylate Rb and activate the expression of E2F-responsive nucleotide-biosynthetic enzymes [NBEs]), such as the R2 subunit of ribonucleotide reductase (A. J. Hume and R. F. Kalejta, unpublished observations), UMP synthase (UMPS), and thymidylate synthase (TS) (30–34). Another E2F-responsive NBE, dihydrofolate reductase (DHFR) (19), is known to be induced during HCMV infection (31, 32), and HCMV infection has been shown to increase metabolic flux through the pyrimidine nucleotide synthesis pathway (30, 32, 34, 35). In the absence of UL97, vDNA synthesis is reduced (9), but DNA replication can be rescued through expression of the human papillomavirus (HPV) E7 protein, an oncoprotein that can also induce E2F-mediated gene expression through the inactivation of Rb (30). Thus, a model in which UL97's contribution to HCMV vDNA synthesis is mediated through the inactivation of Rb and induction of E2F-responsive NBE gene expression fits the paradigm of the DNA tumor virus oncoproteins with which it shares functionality.

However, we found that HCMV utilizes both the de novo and salvage pathways to generate the dNTPs it needs for DNA synthesis. Furthermore, our results show that dNTP levels are limiting for HCMV DNA replication, regardless of UL97 activity or the phosphorylation status of Rb, and that dNTP availability is not the only source of the DNA replication defect observed in the absence of UL97 activity. Therefore, the contribution of UL97 activity to HCMV DNA replication must be at least partially independent of nucleotide synthesis, highlighting our incomplete understanding of the complex role of UL97 during HCMV infection.

MATERIALS AND METHODS

Viruses and cells.

The wild-type HCMV strain was AD169. The AD169 UL97-null virus has been described previously (11). The ΔUL97* mutant arose during continued passage of the UL97-null virus. The HSV-1 strain was KOS and was a gift from Curtis Brandt. Normal human dermal fibroblasts (NHDFs) were maintained in Dulbecco's modified Eagle medium (DMEM), 10% fetal bovine serum (FBS), 1× pen strep glutamine (PSG) (Gibco). Prior to all experiments, cells were plated subconfluently and serum starved in DMEM, 0.1% FBS, PSG. Infections were performed with the indicated viruses at a multiplicity of infection (MOI) of 0.1 as determined by plaque assay.

Antibodies and inhibitors.

Antibodies against Rb (Cell Signaling; 9309), phospho-Rb 807/811 (Cell Signaling; 8516), phospho-Rb 249/252 (Affinity Bioreagents; OPA1-03896), tubulin (Sigma; T6199), R2 (Abcam; ab172476), p53R2 (Abcam; ab8105), and R1 (Abcam; ab137114) were from the commercial sources indicated. Antibodies against IE1 (36) and UL97 (37) have been described previously. The inhibitors maribavir (10 μM in dimethyl sulfoxide [DMSO]; Acme Biosciences; A4028), leflunomide (50 μg/ml in DMSO; Enzo Life Sciences; ALX-430-095), 6-methylmercaptopurine riboside (10 μM in phosphate-buffered saline [PBS]; Sigma; M4002), hydroxyurea (HU) (10 μM in PBS; Sigma; H8627), methotrexate (MTX) (10 nM in PBS; Sigma; M9929), and phosphonoacetic acid (PAA) (250 mg/ml in water; Sigma; 284270) were from the commercial sources indicated. Hydroxyurea (38) and methotrexate (39) are not toxic at the concentrations used.

Nucleosides.

All nucleosides and nucleotides were purchased from Sigma. 2′-Deoxyguanosine was dissolved in 0.1 M NaOH. All the other nucleosides and nucleotides were dissolved in warm medium at a concentration of 1 mM for each nucleoside/nucleotide. The dN-containing medium was made by diluting the 2′-deoxyguanosine solution at a ratio of 1:50 by addition to medium containing the other 3 dNs to yield a final concentration of 1 mM for each nucleoside. After the addition of all the nucleosides/nucleotides, the medium was filter sterilized.

Western blots and immunofluorescence.

Cells were lysed in a buffer containing 2% SDS and 4% beta-mercaptoethanol, and equal volumes of the lysate were separated on SDS-PAGE gels, followed by transfer to nitrocellulose membranes (GE Healthcare; 10439396). The membranes were stained in Ponceau S to visualize total protein levels and then blocked in 5% bovine serum albumin (BSA) in Tris-buffered saline with 0.5% Tween 20 (TBST). The membranes were then incubated with the indicated primary antibody, washed with TBST, incubated with secondary antibody, washed again with TBST, and visualized by fluorescence on a Li-Cor Odyssey Fc system. Cells grown on glass coverslips were serum starved and infected with the indicated viruses, followed by processing for indirect immunofluorescence as previously described (3).

siRNA transfection.

NHDFs were serum starved for 48 h, trypsinized, and counted. The cells were collected by centrifugation and resuspended with 1 μg of small interfering RNA (siRNA) per 10⁶ cells and 100 μl of electroporation reagent (Mirus; MIR50114). The cells were then nucleofected using the U-20 program on an Amaxa Nucleofector II machine, resuspended in 0.1% serum medium, and plated. Twenty-four hours posttransfection, the medium was changed, and 24 h following that, the cells were infected. DNA and protein were harvested at 96 h postinfection. siRNAs were obtained from GE Healthcare/Dharmacon (OnTarget Plus Control siRNA [D-001810-10], R2 Smartpool [L-0103790-00] p53R2 [GGG AAA GAG UGG UGG CCU UUU]).

DNA replication assays.

The entry time point sample was collected 5 h postinfection (hpi), at which point medium containing inhibitor and/or nucleoside treatments was added. The medium was changed every 24 h, and the endpoint time point sample was collected at 96 h postinfection, unless otherwise noted. Viral and cellular DNAs were extracted with an IBI DNA isolation kit (IB47202). DNA levels were quantitated by real-time PCR (ABI 7900HT), using primers and a TaqMan probe specific for IE1 (F, CGA CGT TCC TGC AGA CTA; R, TCC TCG GTC ACT TGT TCA; probe, TGG GAG ACC CGC TGT TTC CA) to measure the number of viral genomes, and primers and a TaqMan probe specific for β-actin (F, CGG AAC CGC TCA TTG CC; R, ACC CAC ACT GTG CCC ATC TA; probe, CCT CCC CCA TGC CAT CCT GC) to measure the number of cellular genomes. After quantitation, viral genome numbers were normalized to cellular DNA levels by dividing IE1 by actin. This copy number ratio was used to measure DNA replication as follows: endpoint ratio/entry ratio = DNA replication. Each experiment was performed for at least three biological replicates. Statistical analyses were performed with Student's t test or the Mann-Whitney U test as appropriate.

Reverse transcription-PCR.

RNA was isolated from infected cells at 96 hpi with an IBI RNA isolation kit (IB47323) and DNase treated according to the manufacturer's instructions (Promega RQ1 DNase; M6101). Equal amounts of RNA were then subjected to reverse transcription-PCR using primers specific for the viral UL54 gene (F, GGC AGC GCA GCT ACT TTT; R, ACC TCG TAC ACG GGA AAA), the viral UL27 gene (F, GCG ACT GAT CTC TGG CAA GTA; R, ATG AAC CCC GTG GAT CAG), or the cellular GAPDH (glyceraldehyde-3-phosphate dehydrogenase) gene (F, GAG CCA AAA GGG TCA TC; R, GTG GTC ATG AGT CCT TC) by using a one-step reverse transcription-PCR kit (Promega Access RT PCR; A1281). Reactions were then run on a 1.2% agarose gel, and DNA was visualized through ethidium bromide staining and recorded on a Li-Cor Odyssey Fc.

HCMV growth assay.

Serum-starved NHDFs were infected at an MOI of 0.01 for 9 days. Cell-free and cell-associated virus was harvested and serially diluted, and the titer was determined on NHDFs by indirect immunofluorescent detection of IE1-positive cells. Counting the IE1-positive nuclei quantitates viral stocks with an accuracy comparable to that of a standard plaque assay (40). Statistical analyses were performed with Student's t test.

RESULTS

Deoxyribonucleotides are limiting for HCMV DNA replication.

To explore the roles of UL97, de novo nucleotide biosynthesis, and nucleotide salvage in HCMV vDNA replication, we tested the abilities of exogenously added rNs, dNs, or dNTPs to stimulate HCMV DNA replication at 4 days postinfection (dpi) (96 hpi), when wild-type vDNA accumulation was 17-fold higher than that of the UL97-null virus (Fig. 2A). vDNA replication was curtailed by drugs that inhibit nucleotide-biosynthetic enzymes shared by the de novo and salvage pathways (HU, which inhibits ribonucleotide reductase, and MTX, which inhibits dihydrofolate reductase) (Fig. 1) and was rescued by equimolar amounts of all four dNs added to the culture medium (Fig. 2B). Exogenous dNs failed to rescue vDNA replication suppressed by PAA, an inhibitor of the viral DNA polymerase not known to affect nucleotide biosynthesis (Fig. 2B). Even in the absence of an inhibitor, dNs stimulated HCMV vDNA replication (Fig. 3A). However, vDNA replication in cells infected with herpes simplex virus 1 (strain KOS), which encodes viral NBEs, including ribonucleotide reductase (R1 and R2 subunits) and thymidine kinase (29), was not stimulated by exogenous dNs (Fig. 3B). We conclude that dNTP levels are limiting in HCMV- but not HSV-1-infected cells.

FIG 2.

Pharmacological inhibition of nucleotide synthesis inhibits HCMV DNA replication. (A) Serum-starved NHDFs were infected with the indicated viruses at an MOI of 0.1. DNA was harvested at the indicated times postinfection and quantitated by real-time PCR. The data displayed show the increase of viral DNA levels compared to day 1. (B) Serum-starved NHDFs were infected with WT HCMV at an MOI of 0.1 in the presence of the indicated treatments. The data are displayed as percentages of DNA replication compared to WT infection with no added nucleotides. The error bars represent standard errors, and the statistical significance of differences is noted: *, P < 0.05; NS, not significant.

FIG 3.

Exogenous dNs enhance HCMV but not HSV-1 DNA replication. (A) Serum-starved NHDFs were infected in the presence of the indicated nucleosides and nucleotides and measured for DNA replication at 96 h postinfection. The data are displayed as percent DNA replication compared to WT infection with no added nucleotides. (B) Serum-starved NHDFs were infected with HSV-1 in the presence or absence of additional deoxyribonucleosides and measured for DNA replication at 72 h postinfection. The error bars represent standard errors, and the statistical significance of differences is noted: *, P < 0.05; NS, not significant.

While exogenous dNs enhanced vDNA replication in HCMV-infected cells, rNs and dNTPs did not (Fig. 3A). dNTPs cannot enter cells but can bind to the P2Y family of extracellular receptors, eliciting signaling responses that alter cellular biology (41), though they fail to enhance vDNA replication (Fig. 3A). Enhancement by dNs, but not rNs, indicates not only that dNTPs are limiting even during a wild-type infection, but also that the conversion of ribonucleotides to deoxyribonucleotides by ribonucleotide reductase is a rate-limiting step in the process of vDNA synthesis in HCMV-infected cells. The ribonucleotide reductase enzyme is a heterodimeric tetramer consisting of a dimer of the solitary and constitutively expressed R1 subunit and a dimer of either of two subunits termed R2 and p53R2 (42). Expression of the R2 subunit is stimulated during S phase through E2F-mediated transcriptional activation and yields a stable protein that affords a high catalytic turnover rate, providing the primary source of ribonucleotide reductase activity during S phase (19, 43, 44). R2 accumulation during HCMV infection is blocked by maribavir inhibition of UL97 (Fig. 4A), as predicted, because R2 is an E2F-responsive gene. p53R2, in contrast, is constitutively expressed and active in quiescent cells (28), though its expression is stimulated by p53 in response to DNA damage (45). p53R2 is expressed in HCMV-infected cells even in the presence of maribavir (Fig. 4A). Hydroxyurea inhibits ribonucleotide reductase activity whether R1 is complexed with R2 or with p53R2 (46, 47), so either subunit could provide dNTPs for HCMV vDNA replication. Knockdown of either R2 or p53R2 (Fig. 4A) suppressed replication (Fig. 4B). Simultaneous knockdown of both R2 and p53R2 showed similar levels of inhibition, which was rescued in all cases by exogenous dNs (Fig. 4B). Knockdown of both R2 and p53R2 inhibited HCMV vDNA replication considerably less than the ∼100-fold inhibition seen with hydroxyurea (Fig. 2B), likely because drug treatment is more efficient than double knockdown for fully inhibiting enzyme activity. Thus, we conclude that HCMV utilizes both R2 and p53R2 to produce dNs, and thus dNTPs, for use during vDNA replication, but combined, their activity is still rate limiting.

FIG 4.

HCMV utilizes both the R2 and p53R2 subunits of ribonucleotide reductase. NHDFs were serum starved and transfected with the indicated siRNAs or treated with maribavir; 96 h postinfection, DNA and protein samples were harvested. (A) Protein samples were analyzed by Western blotting with the indicated antibodies. The small decrease in R1 levels observed in the presence of MBV was not seen in other experiments. (B) DNA replication levels were quantitated by real-time PCR and are presented as percentages of the DNA replication observed with WT virus in scrambled siRNA-transfected cells. Rb p807, phosphospecific antibody for Rb phosphorylated on serines 807/811; Tub, tubulin; IE1, HCMV IE1; SCR, scrambled siRNA. The error bars represent standard errors, and the statistical significance of differences is noted: *, P < 0.05; NS, not significant.

The vDNA replication deficit in the absence of UL97 protein or activity is not fully compensated for by exogenous dNs.

As lower levels of E2F-responsive NBEs are expressed in cells infected with UL97-null or kinase-deficient viruses compared to the wild type (30), we expected that exogenous dNs would increase vDNA replication in the absence of UL97 protein or activity. Indeed, we found that vDNA replication in cells infected with a UL97-null virus (ΔUL97) was enhanced by exogenous dNs in the presence of inhibitors of nucleotide biosynthesis (Fig. 5A). Likewise, vDNA replication in cells infected with WT virus in the presence of the UL97 inhibitor MBV was enhanced by exogenous dNs in the absence of inhibitors of nucleotide biosynthesis (Fig. 5B). Exogenous dNs did not rescue vDNA replication suppressed by the viral DNA polymerase inhibitor PAA (Fig. 5A). dN addition also qualitatively increased the rate at which UL44-positive vDNA replication compartments formed (48) and the size they achieved (Fig. 5C), which is also indicative of increased DNA replication in the presence of exogenous dNs.

FIG 5.

Exogenous dNs enhance HCMV ΔUL97 DNA replication. (A) Serum-starved NHDFs were infected with HCMV ΔUL97 and measured for DNA replication in the presence of the indicated treatments. The data are displayed as percent DNA replication compared to ΔUL97 infection with no added nucleotides. The error bars represent standard errors, and the statistical significance of differences is noted: *, P < 0.05; NS, not significant. (B) Serum-starved NHDFs were infected with WT HCMV in the presence of MBV or HCMV ΔUL97 in the presence and absence of exogenous dNs and measured for DNA replication. The data presented show the increase in DNA replication observed in the presence of dNs compared to their absence. (C) NHDFs were grown on coverslips, serum starved, and infected with the indicated viruses at an MOI of 0.1 in the presence or absence of exogenous dNs, as indicated. At the indicated times postinfection, coverslips were harvested and analyzed by indirect immunofluorescence with an antibody to HCMV UL44.

The boost that exogenous dNs afforded vDNA replication in the absence of UL97 function was quantitatively identical to that in the presence of UL97 function (Fig. 6A), even though the levels of DNA synthesized were different (Fig. 6B). Furthermore, when both de novo and salvage pathways were inhibited with hydroxyurea or methotrexate, thus making vDNA replication fully unsupported by cellular nucleotide biosynthesis and fully dependent on exogenous dNs, the wild-type virus synthesized ∼15-fold more DNA than the UL97-null virus (Fig. 6A). From this, we conclude that the vDNA replication defect in the absence of UL97 remains even when the vDNA replication capability is uncoupled from nucleotide biosynthesis. Thus, the defect in vDNA replication seen in the absence of UL97 is at least partially independent of any effects the v-Cdk might have on nucleotide biosynthesis.

FIG 6.

The HCMV ΔUL97 DNA replication defect remains when vDNA replication capacity is uncoupled from nucleotide biosynthesis. (A) Serum-starved NHDFs were infected with the indicated viruses at an MOI of 0.1 in the presence or absence of exogenous dNs or inhibitors, as indicated. The data are presented as the increase in viral DNA from input DNA compared to the DMSO control, with exact values indicated. (B) Serum-starved NHDFs were infected as for panel A, but the data are presented as the absolute levels of DNA replication [endpoint vDNA (IE/actin ratio)/input vDNA (IE/actin ratio)] observed under the indicated conditions. Exact values are indicated. The data depicted were also used in Fig. 5A and 2B.

Both de novo and salvage pathways generate deoxyribonucleotides for use during HCMV vDNA replication.

Because the salvage pathway of nucleotide synthesis is active in quiescent cells and because UL97 activates the expression of enzymes in the de novo pathway but is not necessary for the generation of the dNs required for vDNA replication, we asked whether the de novo pathway was utilized during HCMV infection. Small-molecule inhibitors of critical enzymes of the de novo purine (6-methylmercaptoriboside [6MMPR], which inhibits phosphoribosyl pyrophosphate amidotransferase) (49) and pyrimidine (leflunomide, which inhibits dihydroorotate dehydrogenase) (50) pathways inhibited HCMV vDNA replication, and this was rescued by exogenous dNs (Fig. 7A and B). Others have shown that leflunomide inhibits infectious HCMV production, which can be partially reversed by adding a nucleoside precursor (51). Inhibition of de novo purine or pyrimidine synthesis did not appear to affect the accumulation of the mRNA encoding the HCMV DNA polymerase UL54 (Fig. 7C), indicating that the effects of the drugs are more likely manifested during DNA synthesis than during RNA synthesis, though both processes require nucleotides. The decrease in vDNA replication seen with inhibitors of the de novo pathway (∼10-fold) (Fig. 7A and B) is substantially less than that seen with inhibitors that target enzymes common to both the de novo and salvage pathways (∼100-fold) (Fig. 2C). Thus, both de novo nucleotide biosynthesis and nucleotide salvage appear to contribute to production of the dNTPs required for HCMV vDNA synthesis.

FIG 7.

HCMV utilizes de novo NBEs for DNA replication. (A and B) Serum-starved NHDFs were infected with WT or ΔUL97 HCMV in the presence of the indicated treatments and measured for DNA replication by real-time PCR. The data show DNA replication as a percentage of vehicle control (DMSO). The error bars represent standard errors, and the statistical significance of differences is noted: *, P < 0.05; NS, not significant. (C) Serum-starved NHDFs were infected with HCMV in the presence of the indicated treatments, and RNA was harvested at 96 h postinfection. Equal amounts of RNA were analyzed by reverse transcription-PCR with primers for the indicated targets in the presence (+RT) or absence (−RT) of reverse transcriptase. LEF, leflunomide.

The vDNA replication deficit in the absence of UL97 protein or activity still occurs in the presence of Rb phosphorylation.

Both UL97 and de novo NBEs contribute to the efficiency of HCMV vDNA replication, and UL97 induces the accumulation of NBEs through Rb inactivation (2, 19, 30). However, our data indicate that UL97 must enhance HCMV vDNA synthesis in a manner independent of (and likely in addition to) the stimulation of NBE synthesis. To determine if this additional mechanism of vDNA synthesis stimulation required UL97-mediated Rb phosphorylation, we took advantage of a spontaneous mutant virus that allows Rb phosphorylation in the absence of the UL97 protein. During the course of growing the ΔUL97 virus, we inadvertently generated a ΔUL97 virus stock that, upon infection of serum-starved fibroblasts, resulted in Rb phosphorylation yet still failed to synthesize UL97 (Fig. 8A and B). This virus, which we call ΔUL97*, is similar to the parental ΔUL97 virus in terms of the magnitude of defects in vDNA replication (Fig. 8C) and infectious-progeny output (Fig. 8D), but it does induce Rb phosphorylation, unlike parental ΔUL97. Cells infected with ΔUL97* showed inactivating Rb phosphorylation on serines 807/811 and increased expression of the E2F-responsive NBEs TS and the R2 subunit of ribonucleotide reductase (Fig. 8A and B), but the virus still showed a vDNA replication defect (Fig. 8C). Because the vDNA replication defect in the absence of UL97 remains even when Rb is phosphorylated on inactivating residues and when the expression of E2F-responsive genes is induced (Fig. 8A and B), we conclude that UL97 contributes to vDNA replication, at least in part, in a manner independent of Rb phosphorylation.

FIG 8.

HCMV ΔUL97 has a DNA replication defect when Rb is phosphorylated by cellular Cdks. (A and B) Serum-starved NHDFs were infected with the indicated viruses at an MOI of 1, stimulated with serum (Serum), or mock infected (Mock). At the indicated times postinfection, the protein lysates were harvested and analyzed with the indicated antibodies by Western blotting. Rb p249, Rb phosphorylated on serine 249 and threonine 252; Rb p807, Rb phosphorylated on serines 807/811. (C) Serum-starved NHDFs were infected with the indicated viruses and measured for DNA replication by real-time PCR. The data are presented as percentages of WT HCMV DNA replication. (D) Serum-starved NHDFs were infected with the indicated viruses at an MOI of 0.1 for 9 days, after which progeny virion production was quantitated with an IE1 production assay. Titers are presented relative to the wild type. The error bars represent standard errors, and the statistical significance of differences is noted: *, P < 0.05; NS, not significant.

Continued growth of UL97-null viruses selects for second-site mutations in the viral UL27 gene (52). UL27 induces the degradation of the cellular acetyltransferase Tip60, which inhibits the expression of p21, a Cdk inhibitor (53). Thus, in UL27-positive cells, p21 levels are high, cellular Cdks are inhibited, and Rb is phosphorylated during infection only if UL97 is present and active. In UL27-negative cells, p21 levels are low (54), and as a result, cellular Cdks are active and able to phosphorylate Rb. Rb phosphorylation by cellular Cdks or UL97 can be easily distinguished by monitoring the residues serine 249 and threonine 252, which are phosphorylated by cellular Cdks but not UL97 (2). In ΔUL97*-infected cells, Rb was phosphorylated on serine 249 and threonine 252 (Fig. 8A and B), indicating phosphorylation by cellular Cdks. These residues were not phosphorylated in cells infected by low-passage-number UL97-null virus (Fig. 8B) or wild-type HCMV in the presence of maribavir (Fig. 8A). Therefore, we suspected our ΔUL97* virus acquired a UL27 mutation, as had been previously observed. However, sequencing analysis of the UL27 gene of ΔUL97* found no differences from the parental ΔUL97 virus (Fig. 9C), RNA analysis indicated that ΔUL97* expresses UL27 mRNAs (Fig. 9A), and protein analysis indicated that p21 levels were not decreased (Fig. 9B). Because ΔUL97* does not encode UL97 (Fig. 9C) but carries a wild-type UL27 gene (Fig. 9C) and transcribes UL27 (Fig. 9A), and the translated UL27 protein appears to be functional (Fig. 9B), it is likely that another, unidentified gene of ΔUL97* is responsible for permitting cellular Cdk phosphorylation of Rb during these infections.

FIG 9.

HCMV ΔUL97* expresses functional UL27. (A) Serum-starved NHDFs were either stimulated with serum or infected with the indicated viruses for 96 h. Equal amounts of RNA were analyzed by reverse transcription-PCR with primers for the indicated target genes in the presence (+RT) or absence (−RT) of reverse transcriptase. (B) Serum-starved NHDFs were either mock infected or infected with the indicated virus at an MOI of 1 for the indicated times. Lysates were harvested and analyzed with the indicated antibodies by Western blotting. (C) The UL97 and UL27 genes of the ΔUL97 and ΔUL97* viruses were sequenced, and the results are presented diagrammatically. The percentages represent the degrees of sequence identity between WT virus and the mutant viruses (when sequence is present), and numbers indicate the genetic positions of the indicated genes or deletions in the indicated genes in the reference WT sequence (X17403). The open bars represent the engineered deletion in the original ΔUL97 virus that remains absent in ΔUL97*.

DISCUSSION

Quiescent cells maintain low levels of nucleotides that can be generated by salvage pathways to support transcription and DNA repair (28, 55). Even the modest needs of retroviruses for dNTPs during reverse transcription overburden this supply (56, 57). Based on genome size and accumulation, an HCMV infection in a quiescent cell requires over 1,000-fold more DNA synthesis than a retroviral infection. The origin of the nucleotides between the two possible pathways, de novo synthesis and salvage, was addressed here.

Perhaps not surprisingly, we found that both de novo and salvage pathways are utilized to produce dNTPs in HCMV-infected cells. However, what was surprising was the dispensability of Rb phosphorylation in this process. HCMV encodes a kinase, UL97, that mimics the activity of the cellular Cdks that phosphorylate and thus inactivate Rb to induce the expression of E2F-responsive genes, among which are multiple NBEs (2, 19). Specific NBEs upregulated upon HCMV infection were not induced upon infection in the absence of UL97 activity, implying a role for UL97 in nucleotide availability (Fig. 8A and B) (30). However, the vDNA replication defect observed in the absence of UL97 was not fully corrected with exogenous dNs or by cellular Cdk-mediated Rb phosphorylation (Fig. 6A and B and 8B). While it is possible that the cellular Cdk activity toward Rb apparent in our ΔUL97*-infected cells is itself inhibitory to HCMV vDNA replication, such a defect was not observed in UL27-null cells that also display cellular Cdk-mediated Rb phosphorylation (53).

Inducing the expression of NBEs through Rb phosphorylation likely plays a part in the ability of UL97 to stimulate vDNA synthesis, and it would be interesting to compare the kinetic flux of nucleotide pools (34) in the presence and absence of UL97. However, any effects UL97 has on nucleotide pools cannot be the whole story. Thus, the complete role of UL97 in viral DNA replication remains an enigma. UL97 phosphorylates the viral DNA polymerase processivity factor UL44 (37), but that phosphorylation event does not correlate with the efficiency of HCMV vDNA synthesis (58). Likewise, phosphorylation of no other known or potential UL97 substrate has been directly linked to positive effects on HCMV vDNA replication. Interestingly, both expression of the papillomavirus E7 oncoprotein and the addition of serum to quiescent, subconfluent cultures can rescue the vDNA replication defect observed in the absence of UL97 (30). While both of these scenarios lead to Rb inactivation, myriad other cellular changes will also be enacted, one or more of which may be necessary to correct the defect in vDNA replication displayed by the UL97-null virus. Because an E7 protein unable to degrade Rb failed to rescue the defect (30), either this mutant papillomavirus protein must also be deficient in some other activity, or Rb inactivation is necessary but not sufficient to correct the vDNA replication defect in the absence of UL97.

Irrespective of how UL97 contributes to vDNA replication, it is clear that HCMV procures nucleotides through mechanisms in addition to, and likely redundant with, UL97. The viral pp71 protein targets hypophosphorylated Rb for degradation (59), and the viral IE2 protein appears to directly activate E2F-responsive gene expression (60). The multiple modulators of the Rb-E2F pathway may make the absence of any single one insignificant. Also, in addition to the Rb/E2F pathway, nucleotide biosynthesis is also controlled by the c-myc protein (61). As HCMV activates c-myc expression at very early times postinfection (62), it is likely that c-myc also contributes to providing nucleotides in support of HCMV vDNA replication. The critical need for nucleotides in cellular and viral processes makes it unsurprising that HCMV (and normal cellular processes) utilizes multiple pathways to ensure that sufficient levels are available for DNA replication.

The finding that dNTPs are limiting for wild-type HCMV infection was initially surprising. HCMV activates multiple pathways leading to the synthesis of NBEs and, based on the amount of DNA generated over the course of infection, synthesizes DNA at a rate that must be substantially lower than that of an uninfected cell during S phase (34, 63, 64). It is possible that nucleotide levels are kept at lower than maximal levels and thus become limiting in HCMV-infected cells because the precursors for their synthesis are also key building blocks of the lipids that HCMV appears to covet (65, 66). Both glucose and glutamine are key early substrates for both nucleotide and fatty acid synthesis (26) (Fig. 1). Metabolomic analysis indicates that while HCMV infection increases the carbon flux through nucleotide synthesis pathways by 3-fold, it increases flux through fatty acid synthesis pathways by 20-fold (65). Conversely, metabolic flux in HSV-1-infected cells is steered more toward nucleotide biosynthesis than fatty acid synthesis (67), likely explaining why dNTPs are not limiting for HSV-1 but are for HCMV (Fig. 3). Thus, it appears likely that HCMV regulates the flow of precursors into nucleotide and fatty acid biosynthesis pathways to coordinate the rates and final amounts of DNA and lipid synthesis. As fatty acid synthesis inhibitors are being explored as antiviral drugs, the interactions of these compounds with nucleotide biosynthesis inhibitors should be evaluated.