Strategy of Human Cytomegalovirus To Escape Interferon Beta-Induced APOBEC3G Editing Activity

APOBEC3 family of proteins plays a pivotal role in intrinsic immunity defense mechanisms against multiple viral infections, including retroviruses, through the deamination activity. However, the currently available data on APOBEC3 editing mechanisms upon HCMV infection remain unclear. In the present study, we show that particularly the APOBEC3G (A3G) member of the deaminase family is strongly induced upon infection with HCMV in fibroblasts and that its upregulation is mediated by IFN-β. Furthermore, we were able to demonstrate that neither A3G knockout nor A3G overexpression appears to modulate HCMV replication, indicating that A3G does not inhibit HCMV replication. This may be explained by HCMV escape strategy from A3G activity through depletion of the preferred nucleotide motifs (hot spots) from its genome. The results may shed light on antiviral potential of APOBEC3 activity during HCMV infection, as well as the viral counteracting mechanisms under A3G-mediated selective pressure.

ABSTRACT

The apolipoprotein B editing enzyme catalytic subunit 3 (APOBEC3) is a family of DNA cytosine deaminases that mutate and inactivate viral genomes by single-strand DNA editing, thus providing an innate immune response against a wide range of DNA and RNA viruses. In particular, APOBEC3A (A3A), a member of the APOBEC3 family, is induced by human cytomegalovirus (HCMV) in decidual tissues where it efficiently restricts HCMV replication, thereby acting as an intrinsic innate immune effector at the maternal-fetal interface. However, the widespread incidence of congenital HCMV infection implies that HCMV has evolved to counteract APOBEC3-induced mutagenesis through mechanisms that still remain to be fully established. Here, we have assessed gene expression and deaminase activity of various APOBEC3 gene family members in HCMV-infected primary human foreskin fibroblasts (HFFs). Specifically, we show that APOBEC3G (A3G) gene products and, to a lesser degree, those of A3F but not of A3A, are upregulated in HCMV-infected HFFs. We also show that HCMV-mediated induction of A3G expression is mediated by interferon beta (IFN-β), which is produced early during HCMV infection. However, knockout or overexpression of A3G does not affect HCMV replication, indicating that A3G is not a restriction factor for HCMV. Finally, through a bioinformatics approach, we show that HCMV has evolved mutational robustness against IFN-β by limiting the presence of A3G hot spots in essential open reading frames (ORFs) of its genome. Overall, our findings uncover a novel immune evasion strategy by HCMV with profound implications for HCMV infections.

IMPORTANCE APOBEC3 family of proteins plays a pivotal role in intrinsic immunity defense mechanisms against multiple viral infections, including retroviruses, through the deamination activity. However, the currently available data on APOBEC3 editing mechanisms upon HCMV infection remain unclear. In the present study, we show that particularly the APOBEC3G (A3G) member of the deaminase family is strongly induced upon infection with HCMV in fibroblasts and that its upregulation is mediated by IFN-β. Furthermore, we were able to demonstrate that neither A3G knockout nor A3G overexpression appears to modulate HCMV replication, indicating that A3G does not inhibit HCMV replication. This may be explained by HCMV escape strategy from A3G activity through depletion of the preferred nucleotide motifs (hot spots) from its genome. The results may shed light on antiviral potential of APOBEC3 activity during HCMV infection, as well as the viral counteracting mechanisms under A3G-mediated selective pressure.

INTRODUCTION

Human cytomegalovirus (HCMV) is a ubiquitous opportunistic betaherpesvirus, which, despite infecting the vast majority of the world's population, can rarely cause symptomatic diseases in healthy, immunocompetent individuals (1). However, reactivation of latent HCMV infection in immunocompromised hosts (e.g., transplant recipients) may result in life-threatening diseases. Likewise, HCMV congenital infection can lead to abortion or dramatic disabilities in the infant, including deafness and mental retardation (2). A hallmark of HCMV pathogenesis is its ability to productively replicate in an exceptionally broad range of target cells such as epithelial, smooth muscle, and endothelial cells as well as fibroblasts (3, 4).

A central component of innate antiviral immunity against HCMV is the rapid activation of multiple interferon (IFN) signaling pathways that upregulate the expression of a rising number of restriction factors committed to counteract virus replication. Such intrinsic immune mechanisms therefore provide a frontline antiviral defense mediated by constitutively expressed proteins, already present and active before the virus enters a cell (5, 6). These intrinsic immune effectors, which were initially discovered as being active against retroviruses, include the apolipoprotein B editing catalytic subunit-like 3 (APOBEC3, or A3) family of cytidine deaminases and tetherin, an IFN-inducible protein whose expression blocks the release of human immunodeficiency virus type 1 (HIV-1) (7). However, it soon became apparent that such effectors were also active against other viruses, such as vesicular stomatitis virus, filoviruses, influenza virus, and hepatitis C virus (8). Moreover, other proteins such as PML, hDaxx, Sp100 (9, 10), viperin, and IFI16 were subsequently identified as restriction factors mediating the intrinsic immune response against HCMV infection (11, 12).

The seven members of the APOBEC3 (A3) family of cytidine deaminases (A, B, C, D, E, F, G, and H) (13–16) catalyze the deamination of cytidine nucleotides to uridine nucleotides in single-strand DNA (ssDNA) substrates. These enzymes are widely acknowledged as fundamental players in the defense against various viral infections (14, 15, 17). Since the identification of APOBEC3G (A3G) as a prototype antiretroviral host restriction factor, A3 subsets have been shown to restrict the replication of retroviruses (18), endogenous retroelements (19), and, more recently, DNA viruses such as hepatitis B virus (HBV) (20, 21) and parvoviruses (22, 23). Moreover, different A3 isoforms deaminate human papillomavirus (HPV) genomes (24) as well as BK polyomavirus (BKPyV) (25). Genomes of some herpesviruses, such as herpes simplex virus 1 (HSV-1) and Epstein-Barr virus (EBV), are edited by APOBEC3 on both strands. Interestingly, the editing is higher on the minus strand, possibly due to the fact that during discontinued replication the lagging strand exposes more viral ssDNA to nuclear APOBEC3s than the leading strand (14–16, 26). Human APOBEC3 proteins are also able to mutate the genome of the murine gammaherpesvirus 68 (MHV68) and, therefore, counteract viral replication. In particular, human A3A, A3B, and A3C proved their ability to restrict MHV68 replication (27).

With regard to HCMV, Weisblum et al. (28) have recently reported an important role of APOBEC3A (A3A) in mediating innate immunity against congenital HCMV infection. In particular, A3A was strongly upregulated following ex vivo HCMV infection of maternal decidua, and overexpression of A3A in epithelial cells hampered HCMV replication by inserting hypermutations into the viral genome through cytidine deamination. A3A induction by HCMV was not observed in HCMV-infected chorionic villi maintained in organ culture, primary human foreskin fibroblasts (HFFs), or epithelial cell cultures, suggesting that HCMV-mediated upregulation of A3A is tissue and cell type specific. Intriguingly, IFN-β but not IFN-γ induced A3A expression in uninfected decidual tissues, suggesting its potential regulation as an IFN-stimulated gene (ISG) during HCMV infection.

However, there still remain a number of issues that need further investigation. For example, in contrast to the aforementioned studies, several reports have demonstrated that members of the A3 family are robustly induced in different cell types in vitro and in different tissues in vivo by either IFNs or viruses (e.g., HIV and HBV). Thus, the question as to whether HCMV is able to induce other A3 family members besides A3A in different cell types remains open. Another important issue stems from the observation that HCMV triggers IFN production during the early steps of infection, but it is still unclear whether A3 induction is mediated by IFN rather than the virus itself. In this respect, IFN production triggered by HCMV induces expression of IFN-stimulated genes, including the A3 family, which are committed to restrict virus replication as observed in other viral models. Thus, it is conceivable that HCMV has developed strategies to escape from APOBEC3 editing activity. Finally, a major issue concerns APOBEC3 antiviral activity. Although APOBEC3 editing activity has been reported for all the viruses analyzed, it is still a matter of debate whether this is also true for other viruses such as influenza viruses, herpesviruses, papillomaviruses, and polyomaviruses. Thus, there is a gap in knowledge concerning the mechanism of HCMV evasion from A3-induced viral genome mutagenesis.

In the present study, we present evidence of the following: (i) that A3G and, to a lesser extent, A3F gene products are induced in HCMV-infected HFFs; (ii) that the induction of A3G appears to be mediated by IFN-β as it is drastically decreased upon treatment with anti-IFN type 1 receptor antibodies; (iii) that neither A3G knockout nor its overexpression appears to modulate HCMV replication, indicating that A3G does not inhibit HCMV replication; and (iv) that A3G exerted a selective pressure that, during evolution has likely shaped the nucleotide composition of the HCMV genome.

RESULTS

HCMV infection stimulates various APOBEC3 expression patterns in different cell subsets.

To assess the role of APOBEC3, we first asked whether HCMV infection could regulate mRNA and protein levels of A3 family members in different cell types. For this purpose, total RNAs from HCMV-infected HFFs, human umbilical vein endothelial cells (HUVECs), macrophage-derived THP-1 cells, or human retinal pigment epithelial (ARPE-19) cells were extracted at 8 and 24 h postinfection (hpi) and subjected to reverse transcription-quantitative PCR (RT-qPCR) analysis. Among all A3 family members analyzed, only A3G and A3F displayed mRNA upregulation in HCMV-infected HFFs compared to levels in mock-infected cells (i.e., ∼25 and ∼12-fold at 8 hpi; ∼10 and ∼6 at 24 hpi, respectively) (Fig. 1A). We also observed similar kinetics of mRNA expression for Mx-1, a well-known IFN-inducible gene (Fig. 1A). Human A3F and human A3G share more than 90% promoter sequence similarity and appear to be transcriptionally coregulated (29, 30). In agreement with these findings, we observed a coregulated induction of A3G and A3F expression by HCMV. Notably, A3F and A3G were also induced upon HCMV infection in differentiated THP-1 cells, although several other members of the APOBEC3 family, namely, A3A and A3H, were highly upregulated in this cell line as well (Fig. 1C). In contrast, mRNA expression levels of all A3 family members including A3G and A3F remained unchanged in HCMV-infected HUVEC and ARPE-19 cells, whereas Mx-1 mRNA was potently induced (Fig. 1B and D), suggesting that induction of A3G and A3F is cell type specific.

FIG 1.

Apolipoprotein B editing enzyme catalytic subunit 3 (APOBEC3) gene expression patterns in human cytomegalovirus (HCMV)-infected cells. Primary human foreskin fibroblasts (HFFs) (A), human umbilical vein endothelial cells (HUVECs) (B), differentiated THP-1 cells (THP-1 macrophages) (C), or human retinal pigment epithelial cells (ARPE-19) (D) were infected with HCMV at an MOI of 1 and subjected to RT-qPCR to measure mRNA expression of various APOBEC3 family members (i.e., A3A, A3B, A3C, A3DE, A3F, A3G, and A3H) and Mx-1. Values were normalized to the level of the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA and plotted as fold induction relative to the level in mock-infected cells. Data are presented as mean values of biological triplicates. Error bars show standard deviations, (*, P < 0.05; **, P < 0.01; one-way ANOVA followed by Bonferroni's posttests, for comparison of infected versus mock-infected cells). Ct, threshold cycle.

HCMV infection induces A3G in HFFs.

Since A3G was the most potently induced A3 family member by HCMV, we decided to focus our attention on this gene in all further analyses. Consistent with the RT-qPCR results (Fig. 1A), A3G protein expression was significantly upregulated in HCMV-infected HFFs (Fig. 2A). Intriguingly, the kinetics of A3G protein induction, which peaked at 72 hpi, were delayed relative to those of A3G mRNA, which peaked at 8 hpi (Fig. 1A). At the moment, however, the mechanisms responsible for the delay in protein expression have not been explored. To get further insight into HCMV-induced A3G DNA deaminase activity, we used an in vitro fluorescence resonance energy transfer (FRET)-based oligonucleotide assay. To this purpose, whole-cell lysates of mock- or HCMV-infected HFFs were incubated with an ssDNA oligonucleotide containing a single CCC trinucleotide, which represents the canonical deamination target of A3G, along with uracil-DNA glycosylase (UDG) and RNase A (31). In the presence of A3G cytosine deaminase activity, the formation of a uracil base results in an abasic site following uracil base excision by UDG. Base hydrolysis of the abasic site then releases a 6-carboxyfluorescein (FAM) signal from the FRET pair. As expected, protein extracted from HCMV-infected cells displayed deaminase activity consistent with the kinetics of A3G protein induction, reaching a peak at 72 hpi, when the deamination activity was ∼5-fold higher than that of mock-infected cells (Fig. 2B). Finally, to verify FRET assay specificity, we included an ssDNA oligonucleotide containing the target motif of A3B (TC) (29) as a negative control. As expected, in this case A3G activity was comparable to that in mock-infected cells, confirming that A3G is selectively activated upon HCMV infection (data not shown).

FIG 2.

HCMV infection upregulates A3G in HFFs. (A) Lysates were prepared at the indicated time points and subjected to Western blot analysis for A3G, IEA, and α-tubulin (left panel). A3G protein was subjected to densitometry and normalized to α-tubulin (*, P < 0.05; ***, P < 0.001; one-way ANOVA followed by Bonferroni's posttests, for comparison of infected versus mock-infected cells) (right panel). (B) FRET assay to measure A3G deaminase activity. The average and standard deviation were calculated from three independent experiments (**, P < 0.01; ***, P < 0.001; one-way ANOVA followed by Bonferroni's posttests, for comparison of infected versus mock-infected cells). RFU, relative fluorescence units. (C) HFFs were infected with HCMV at an MOI of 1 or left uninfected (mock) and subjected to immunofluorescence analysis at the indicated time points. A3G (green)/IEA (red) were visualized using primary antibodies followed by secondary antibody staining in the presence of 10% HCMV-negative human serum. Nuclei were counterstained with DAPI (blue). Images were acquired at ×63 magnification, and representative pictures are shown.

Collectively, these results show that infection of HFFs with HCMV upregulates A3G DNA deaminase activity, which is in good agreement with the increase of A3G mRNA and protein levels.

Although A3G is typically described as a cytosolic protein (32), several groups have shown that A3G is also present in the nucleus of different cell lines (33–35). To determine whether subcellular A3G localization varies during early and late infection with HCMV, we carried out a detailed kinetic analysis using confocal microscopy at time points ranging from 24 to 72 hpi. HFFs were mock infected or infected with HCMV at a multiplicity of infection (MOI) of 1, and intracellular localization of A3G was assessed by confocal microscopy. Consistent with the Western blot results, a substantial accumulation of A3G in the nucleus of HCMV-infected cells was observed at 72 hpi compared to the level in mock-infected cells, where localization of detected A3G seemed evenly distributed among the cytoplasm and nucleus (Fig. 2C). Altogether, these results demonstrate that A3G intranuclear localization is enhanced in HCMV-infected HFFs.

A3G upregulation is IFN-β dependent.

The innate immune response against incoming pathogens plays a key role during primary infection, especially in patients with defects in adaptive immunity. Early during infection, HCMV triggers type I IFN production, leading to the induction of a number of IFN-stimulated genes (ISGs), a process that promotes an antiviral state in infected and neighboring cells (36–39). Stimulation of A3 upon IFN production has been observed in different viral models and cell types (40–44). In particular, A3G is strongly induced by IFN-β in response to influenza A virus infection (43). To assess whether HCMV induces A3G expression through IFN-β induction also in our model, HFFs were incubated for 24 h in the presence of serial dilutions of IFN-β (50 to 500 U/ml), and the mRNA levels of A3G were determined by RT-qPCR (Fig. 3A). As shown in Fig. 3A, IFN-β stimulation led to over 30-fold induction of A3G mRNA. Likewise, IFN-β treatment of HFFs led to an increase in A3G protein expression over time, which peaked at the 24-h time point (Fig. 3B).

FIG 3.

APOBEC3G upregulation is IFN-β dependent. (A) HFFs were stimulated for 24 h with the indicated doses of IFN-β, and the mRNA levels of A3G were determined by means of RT-qPCR. Values were normalized to GAPDH mRNA and plotted as fold induction relative to levels in untreated HFFs. (B) Western blot analysis to assess APOBEC3G protein levels and α-tubulin upon IFN-β treatment (1,000 U/ml) for the indicated time points (hpt, hours posttreatment). One representative experiment of three performed in duplicate is shown. (C) HFFs were mock- and HCMV-infected in the presence of an anti-IFNAR antibody (5 μg/ml) or isotype control. At 8 hpi, cells were processed by RT-qPCR to assess A3G expression. Data presented in panels A and C are mean values of biological triplicates. Error bars show standard deviations (*, P < 0.05; **, P < 0.01; one-way ANOVA followed by Bonferroni's posttests, for comparison of treated versus untreated cells).

To definitively prove a causative link between IFN-β production and A3G upregulation, HFFs, pretreated for 18 h with anti-IFNAR antibody (Ab) or an isotype control Ab, were infected with HCMV for 8 h and analyzed by RT-qPCR (Fig. 3C). As expected, suppression of IFN-β production by anti-IFNAR Ab strongly impaired A3G mRNA induction compared to the level with untreated or isotype control Ab-treated HFFs. Altogether, these results indicate that IFN-β released early during HCMV infection triggers A3G expression similarly to what has been reported for other viruses such as orthomyxoviruses and HPV (43, 44).

HCMV replication is not affected by A3G activity.

Several reports have shown that A3G is able to counteract the replication of HIV-1 (45–51), human T-cell lymphotropic virus type 1 (HTLV-1) (52–56), and HBV (20, 21, 57, 58). In contrast, A3 deaminases do not appear to affect viral replication or production of infectious viral progeny of two other viruses such as influenza A (43) virus or polyomavirus (59). Thus, we sought to determine whether A3G acted as a restriction factor for HCMV replication. For this purpose, CRISPR/Cas9 systems were used to knock out the A3G gene in HFFs (A3G KO) or a scrambled control (scramble Ctrl). Western blot analysis confirmed that the majority of cells were knocked out for A3G (Fig. 4A). HFFs depleted of A3G were then infected with HCMV at an MOI of 0.1 for 24 h, 72 h, and 144 h, and the viral yield was measured by standard plaque assay. As shown in Fig. 4B, the replication of HCMV was not significantly affected following A3G knockout.

FIG 4.

A3G is not a restriction factor for HCMV replication. (A) Knockout gene variants in HFFs for A3G (A3G KO) and the scramble control were generated with CRISPR/Cas9 technology. The efficiency of A3G depletion was measured by Western blotting for A3G and α-tubulin. (B) A3G KO HFFs were infected with HCMV at an MOI of 0.1. The extent of virus replication was measured at the indicated times postinfection by titrating the infectivity of supernatants and cell suspension on HFFs by standard plaque assay. Results are expressed as means ± SD. (C) HFFs were transduced with AdVA3G or AdVLacZ at an MOI of 30 PFU/cell. The efficiency of A3G overexpression was measured by Western blotting for A3G and α-tubulin. (D) HFFs were transduced with AdV vectors as described in panel C. Subsequently, cells were infected with HCMV at an MOI of 0.1. The extent of virus replication was measured at the indicated times postinfection as described in B. Results are expressed as means ± SD.

To further confirm these findings, we transduced HFFs with an adenovirus-derived vector constitutively expressing A3G protein (AdVA3G) or with a control vector (AdVLacZ) at an MOI of 30. As shown in Fig. 4C, AdVA3G efficiently increased the expression of A3G protein compared to expression with both HCMV and AdVLacZ. After 24 h, cells were infected with HCMV at an MOI of 0.1 for an additional 24 h, 72 h, and 144 h and then analyzed by standard plaque assay. The efficiency of A3G protein overexpression was monitored by Western blotting (Fig. 4C). Consistent with the knockout results, A3G overexpression did not exert any antiviral effects on HCMV replication (Fig. 4D), indicating either that A3G is not a restriction factor for HCMV replication or, alternatively, that HCMV has evolved to escape A3G restriction activity.

A3G-mediated selective pressure shaped the composition of the HCMV genome.

Because HCMV infection upregulates A3G expression with no evidence of virus replication restriction, we sought to determine whether, during evolution, A3G-mediated selective pressure might have played a role in shaping the composition of HCMV genomes.

A3G preferentially deaminates the 3′ cytosine within CCC hot spots in single-stranded DNA (60, 61), whereas other members of the A3 family have distinct preferences (TTC for A3F and A3C; TC for A3B and A3H; TCG for A3A) (29, 62–67). We thus assessed the representation of these hot spot motifs in the HCMV genome using the HCMV Towne sequence as a detailed functional map of this strain was constructed by systematic deletion of single open reading frames (ORFs) (68). The representation of CCC-GGG, TTC-GAA, TCG-CGA, and TC-GA motifs was calculated in sliding windows and compared to the expected counts obtained by randomly shuffling the HCMV genome sequence (see Materials and Methods). Results indicated that the CCC-GGG hot spot is strongly underrepresented in several large genomic regions, whereas no such pattern is observed for the other motifs (Fig. 5). In particular, the regions where A3G hot spots are underrepresented broadly correspond to the genomic positions where essential ORFs (i.e., ORFs that impair or strongly reduce HCMV growth in vitro when deleted) cluster (68).

FIG 5.

Sliding window analysis of APOBEC3 hot spot motifs along the HCMV genome. The HCMV Towne sequence was used (GenBank accession number GQ121041). Motifs were analyzed in 1,000-bp windows moving with a step of 100 bp. For each window, the percentile rank of the real motif count in the distribution of counts from reshuffled windows is plotted. The lower the percentile rank, the fewer motifs are detected in the window when base composition is accounted for (by reshuffling). A schematic representation of HCMV open reading frames (ORFs) is shown with color codes indicating essential ORFs (red), nonessential ORFs (green), and ORFs with unknown effect when deleted (gray).

To date, only one origin of replication (oriLyt) has been described for HCMV (69). In contrast, the mechanisms of DNA replication remain largely unknown, although a rolling-circle phase is likely to occur (70). When we analyzed the frequency of CCC motifs in the two strands of the viral genomes, we detected no substantial difference (Fig. 6A), suggesting that the A3G hot spot underrepresentation is not mainly determined by preferential deamination of the lagging-strand template (71–74).

FIG 6.

Occurrence of APOBEC3G motif in HCMV. (A) Sliding window analysis of APOBEC3G hot spot motif along the HCMV genome. The APOBEC3G motif (CCC) was analyzed for both strands in 1,000-bp windows moving with a step of 100 bp. For each window, the percentile rank of the real motif count in the distribution of counts from reshuffled windows is plotted. The HCMV Towne sequence was used (GenBank accession number GQ121041). (B) CCC/CCT motif comparison. A preference index calculated for the CCC-GGG motif is plotted against the preference index for the CCT-AGG motif, both calculated for essential (red) and nonessential (blue) Towne ORFs. Spearman's rank correlation coefficient (rho) is also reported, along with the correlation P value.

Altogether, these observations were consistent with the possibility that HCMV has evolved to limit CCC-GGG motifs in its genome, especially in essential ORFs. To further address this possibility, we used an approach that accounts for the coding capacity of the HCMV genome, as well as for the amino acid composition of single ORFs. In fact, CCC is a codon for proline, and the representation of this hot spot motif in coding genes also depends on the proline content of the encoded proteins. Thus, we counted the frequency of the trinucleotide motifs for A3G, A3A, and A3F/A3C in all HCMV ORFs and obtained expected values by reshuffling codons in each ORF. For each motif in each ORF, we computed a preference index that varies between −1 (underrepresentation) and +1 (overrepresentation), with values close to 0 indicating that the representation of motifs is similar to the expected one (see Materials and Methods). Analysis of preference indexes indicated that CCC-GGG motifs are underrepresented in HCMV ORFs and that the median preference index is well below 0. No such pattern was evident for motifs targeted by other APOBEC3 enzymes, which showed preference indexes close to 0 (Fig. 7A). Also, CCT-AGG motifs, which represent the products of A3G deamination without repair, were not overrepresented in HCMV ORFs, and no negative correlation was observed between the preference indexes for CCC-GGG and those for CCT-AGG motifs (Fig. 6B). Thus, the underrepresentation of A3G motifs is not the result of active A3G-mediated deamination and mutation.

FIG 7.

Occurrence of APOBEC3 motifs in HCMV ORFs. (A) The occurrence of hot spot motifs for A3G, A3F/3C, and A3A was analyzed by calculating a preference index. Preference indexes are shown in standard box-and-whisker plot representation (thick line, median; box, quartiles; whiskers, 1.5× interquartile range). The Kruskal-Wallis tests indicated significant differences among motifs (P < 2.2 × 10⁻¹⁶). P values from post hoc tests (Nemenyi tests) are shown. N.S., not significant. (B) Occurrence of A3G hot spot motifs in HCMV essential and nonessential ORFs. Essential ORFs have significantly fewer CCC-GGG motifs than nonessential ORFs (P value from Wilcoxon rank sum test). (C) Occurrence of A3G hot spot motifs in different HCMV strains and isolates. The preference indexes of Towne ORFs are plotted against the corresponding indexes from other HCMV genomes. Isolates derived from different sources or body compartments were analyzed.

We next sought to determine whether essential and nonessential ORFs displayed a different representation of APOBEC3G motifs. ORFs were categorized based on the mutant growth classification proposed by Dunn and coworkers (68), and preference indexes were compared (see Materials and Methods). We found that CCC-GGG motifs are significantly less likely to occur in essential ORFs than in nonessential ones (Wilcoxon rank sum test, P = 0.014) (Fig. 7B). As selective pressure is expected to be stronger at essential ORFs, the latter are the most depleted of A3G motifs.

Finally, we verified that the underrepresentation of CCC-GGG motifs is a general feature of HCMV genomes and is not limited to the Towne strain. Thus, the preference index for CCC-GGG motifs was calculated for all ORFs of other HCMV strains (including Merlin) and clinical isolates deriving from different sources. No substantial differences were observed between the Towne sequence and the sequences of any of these strains or isolates (Fig. 7C). Overall, these results suggest that A3G exerted selective pressure on HCMV and that the virus evolved to limit A3G hot spots in its genome.

DISCUSSION

In summary, we report that HCMV infection specifically upregulates A3G and, to a lesser extent, A3F expression in primary human fibroblasts (HFFs) and that the virus has evolved an escape strategy to avoid editing activity. Our findings indicate that human A3G is induced upon viral infection as a part of the antiviral response mediated by IFN-β. In this regard, addition of anti-IFN receptor Abs during HCMV infection ablates A3G gene product induction, demonstrating that A3G induction by HCMV is IFN dependent. Moreover, IFN-β treatment of HFFs can upregulate A3G expression within 24 h in the absence of HCMV infection, confirming that A3G is a bona fide ISG family member. Accordingly, two IFN-sensitive response elements, namely, IFN regulatory factor element (IRF-E)/IFN-stimulated response element (ISRE), located upstream of the first A3G exon have been identified (42). Recently, Weisblum et al. (28) found that A3A is strongly upregulated following ex vivo HCMV infection of human decidual tissues but not upon infection of chorionic villi, primary fibroblasts (MRC-5 and HFFs), and epithelial (ARPE-19) cell cultures. In line with our results, IFN-β significantly induced A3A expression in uninfected decidual tissues, suggesting its potential regulation as an ISG early during HCMV infection. Altogether, these findings demonstrate that A3A and A3G are differentially regulated in HCMV-infected cells.

In the same study, Weisblum et al. (28) demonstrated that overexpression of A3A severely impaired HCMV replication in epithelial cells through cytidine deamination of the viral genome. Moreover, exogenous A3A expression in ARPE-19 cells downregulated the expression of viral genes, such as immediate early (IE1) and delayed early (UL89) genes, and reduced HCMV DNA accumulation, suggesting that in this cellular system A3A does restrict virus replication. In contrast to these observations, here we show that neither knockout nor overexpression of A3G can modulate HCMV gene expression and its replication, indicating that A3G does not behave as an HCMV restriction factor in vitro.

Based on this evidence, we hypothesized that during evolution HCMV might have developed strategies to escape A3G editing activity. To test this hypothesis, we assessed whether A3G-mediated selective pressure shaped the composition of HCMV genomes. A3G preferentially deaminates the 3′ cytosine within CCC hot spots in single-stranded DNA, whereas other members of the A3 family have distinct preferences. Notably, the CCC-GGG motif, but not other A3 motifs, was found to be significantly underrepresented in several genomic regions where essential ORFs are located. The decrease in CCC-GGG motifs was not paralleled by an increase in their deamination products, and the A3G hot spot motifs were similarly underrepresented in both genome strands. Thus, these observations suggest that A3G no longer affects the HCMV genome composition because the virus has likely evolved to limit the presence of A3G hot spot motifs especially within essential ORFs. In this respect, it is worth mentioning that, albeit underrepresented, some CCC-GGG motifs do occur in HCMV ORFs, including essential ones. Nevertheless, secondary structures and sequence context are also known to modulate A3G preferences (31), suggesting that extant CCC motifs could represent suboptimal targets.

Our findings are in line with previous studies indicating that target motifs for other A3 enzymes are depleted in the genome of alpha papillomaviruses, most likely as the result of viral evolution to avoid restriction (75). Likewise, A3B exerted selective pressure on BKPyV, which shows an underrepresentation of hot spot motifs for this enzyme (59). Nonetheless, the specific knockdown of A3B had little short-term effect on productive BKPyV infection (59).

Recent results have shown that A3A can restrict HCMV replication in human decidual tissues (28). However, we did not find A3A motifs to be underrepresented in HCMV genomes. One possible explanation for this finding is that decidual tissues do not represent the primary target site of HCMV infection, and vertical transmission, despite being clinically relevant, does not contribute significantly to HCMV spread in human populations. Thus, the selective pressure exerted by A3A on HCMV may be limited. In fact, we did not find this enzyme to be upregulated by viral infection in HFFs and other primary HCMV target cell types.

According to these observations, the following scenario could be envisaged. Early during HCMV infection, DNA sensors including cGAS and IFI16 prime IFN-β production, which in turn stimulates expression of ISGs including A3G. To prevent DNA editing by A3G from yielding CCC-GGG hypermutations, the virus has evolved to limit the presence of A3G target motifs in genes essential for its replication.

Various strategies have been adopted by different viruses to prevent the catastrophic consequences of A3-induced hypermutations. While several DNA viruses have evolved to limit the availability of A3 target sites (59, 75), HIV has adopted a completely different evasion strategy based on the ability of its protein Vif to bind A3G and promote its degradation through the proteasome pathway (76–79).

In conclusion, our studies demonstrate for the first time that (i) early during infection, HCMV upregulates A3G in fibroblasts (HFFs) through IFN-β production, (ii) A3G does not restrict HCMV replication, and (iii) HCMV has evolved mutational robustness against IFN-β by limiting the presence of A3G hot spots in essential ORFs of its genome. Our findings reveal a novel immune evasion strategy by HCMV, which further fuels its fame as master in immune evasion.

MATERIALS AND METHODS

Cells and viruses.

Primary human foreskin fibroblasts (HFFs; ATCC SCRC-1041), human retinal pigment epithelial cells (ARPE-19, ATCC CRL-2302), and human embryo kidney 293 cells (HEK293; Microbix Biosystems Inc.) were cultured in Dulbecco's modified Eagle's medium (Sigma-Aldrich) supplemented with 10% fetal calf serum (FCS; Sigma-Aldrich) as previously described (80). THP-1 cells, cultured as nonadherent monocyte-like cells, were grown in RPMI medium (Sigma-Aldrich), with 10% FCS, 600 μg/ml glutamine, 200 IU/ml of penicillin, and 100 μg/ml streptomycin (Gibco). THP-1 cells were differentiated into macrophage-like cells by addition of 100 nM phorbol myristate acetate (PMA; Sigma-Aldrich). All presented data with THP-1 cells were based on PMA-differentiated cells. Human umbilical vein endothelial cells (HUVECs) were isolated from umbilical veins by chymotrypsin treatment and used for experiments at passage 2 ± 7. HUVECs were cultured in endothelial cell basal medium 2 (EBM-2; Lonza), plus endothelial cell growth medium supplements (EGM-2; Lonza), 2% FCS (Sigma-Aldrich), and 1% penicillin-streptomycin solution (Sigma-Aldrich). HCMV strain Merlin was kindly provided by Gerhard Jahn (University Hospital of Tübingen, Germany), propagated, and titrated on HFFs by standard plaque assay (12, 39).

Recombinant adenoviral vectors.

Adenovirus-derived vectors expressing A3G were generated by means of a replacement strategy using recombineering methods (81). Briefly, the A3G gene was amplified using a specific set of primers (forward, 5′-AACCGTCAGATCGCCTGGAGACGCCATCCACGCTGTTTTGACCTCCATAGAAGACACCGGGACCGATCCAGCCTGGATCCATGAAGCCTCACTTCAGAAA-3′; reverse, 5′-TATAGAGTATACAATAGTGACGTGGGATCCCTACGTAGAATCAAGACCTAGGAGCGGGTTAGGGATTGGCTTACCAGCGCTGTTTTCCTGATTCTGGAGA-3′). In order to accomplish homologous recombination, approximately 200 ng of DNA was electroporated into SW102 bacteria harboring pAdZ5-CV5 vector. Cells were then plated on minimal medium agar plates containing 5% sucrose and chloramphenicol and incubated at 32°C for 1 day. The colonies that appeared were inoculated into LB broth containing ampicillin and chloramphenicol and LB broth containing chloramphenicol only. In the colonies grown in chloramphenicol only, the A3G ORF replaced the ampicillin resistance sequence in multiple cloning sites. Colonies were checked by PCR and sequencing. To obtain the recombinant adenovirus, the AdZ vector was transfected into HEK293 packaging cells. Transfected cells were maintained in a 5% CO₂ incubator at 37°C until an extensive cytopathic effect was obtained. Viruses were then purified from infected cultures by freeze-thaw-vortex cycles and assessed for A3G expression by Western blotting. For cell transduction, HFFs were washed once with phosphate-buffered saline (PBS) and incubated with AdVA3G at an MOI of 30. After 2 h at 37°C, the virus was washed off, and fresh medium was applied. For all the experiments, a recombinant adenovirus expressing the Escherichia coli β-galactosidase gene (AdVLacZ) was used as a control (12).

RNA isolation and semiquantitative RT-qPCR.

Total RNA was extracted using a NucleoSpin RNA kit (Macherey-Nagel), and 1 μg was retrotranscribed using a Revert-Aid H-Minus FirstStrand cDNA synthesis kit (Fermentas), according to the manufacturer's protocol. Comparison of mRNA expression levels between samples (i.e., infected versus untreated) was performed by SYBR green-based RT-qPCR on a Mx3000P apparatus (Stratagene), using the following primers: A3A, GTCTTATGCCTTCCAATGCC (forward [Fw]) and GAGAAGGGACAAGCACATGG (reverse [Rw]); A3B, AATGTGTCTGGATCCATCAGG (Fw) and TGAAGGTCAGCAATTCATGC (Rw); A3C, TCTGCATGACAATGGGTCTC (Fw) and AAACTTGGCTGTGCTTCACC (Rw); A3D, GATCTGGAAGCGCCTGTTAG (Fw) and AGTCGAATCACAGGCAGGAG (Rw); A3F, CCATAGGCTTTGCGTAGGTT (Fw) and AATTATGCATTCCTGCACCG (Rw); A3G, TTCCAAAAGGGAATCACGTC (Fw) and AGGGGCTTTCTATGCAACC (Rw); A3H, AGCTGTGGCCAGAAGCAC (Fw) and CGGAATGTTTCGGCTGTT (Rw); glyceraldehyde-3-phosphate dehydrogenase (GAPDH), AGTGGGTGTCGCTGTTGAAGT (Fw) and AACGTGTCAGTGGTGGACCTG (Rw); Mx1, CCAGCTGCTGCATCCCACCC (Fw) and AGGGGCGCACCTTCTCCTCA (Rw).

Neutralization of type I IFNs.

To neutralize the activity of type I IFNs, specific blocking antibodies against interferon receptor (clone MMHAR-2, diluted 1:100; Millipore) were added to culture medium at a concentration of 5 μg/ml for 18 h prior to infection with HCMV Merlin strain at an MOI of 1 and then left in the supernatant until the end of the respective experiment. Mouse IgG2a (clone MOPC-173, diluted 1:100; BD Biosciences Europe) was used as an isotype control. Human recombinant IFN-β was obtained from PBL (catalog number 11415-1).

Transduction of HFFs with lentiviral CRISPR/Cas9.

The CRISPR/Cas9 system was employed to generate specific gene knockouts in primary human fibroblasts. Recombinant lentiviruses were packaged in HEK293T cells by cotransfection of APOBEC3G subgenomic RNA (sgRNA) with a CRISPR/Cas9 All-in-One Lentivector set (Human) (Applied Biological Materials Inc.) and 2nd Generation Packaging System Mix (Applied Biological Materials, Inc.) for producing viral particles using Lipofectamine 2000 (Invitrogen). Viral supernatants were harvested after 48 h and used to transduce fibroblasts by infection in the presence of 8 mg/ml Polybrene. Transduced cells were selected with puromycin (1 μg/ml) over the course of 14 days postransduction. After selection, successful knockout was confirmed using immunoblotting. CRISPR negative-control lentiviruses were produced with a scrambled sgRNA CRISPR/Cas9 All-in-One Lentivector (Applied Biological Materials, Inc.) in HEK293T cells as described above.

Western blot analysis.

Whole-cell protein extracts were prepared and subjected to Western blot analysis as previously described (82, 83). The following primary mouse monoclonal antibodies were used: anti-A3G (VMA00418, diluted 1:1,000; Bio-Rad), CMV IEA (CH160, diluted 1:1,000; Vyrusis), and α-tubulin (39527, diluted 1:4,000; Active-Motif). Immunocomplexes were detected using sheep anti-mouse antibodies conjugated to horseradish peroxidase (HRP) (GE Healthcare Europe GmbH) and visualized by enhanced chemiluminescence (Super Signal West Pico; Pierce-Thermo Fischer Scientific).

Immunofluorescence microscopy.

Indirect immunofluorescence analysis was performed as previously described (82, 84), using the appropriate dilution of primary antibodies for 1 h at room temperature (RT) in the presence of 10% HCMV-negative human serum followed by 1 h of incubation with secondary antibodies in the dark at RT. The following primary antibodies were used: rabbit polyclonal anti-CMV IEA antibody (diluted 1:500) (Santo Landolfo, University of Turin) or mouse monoclonal antibody anti-A3G (VMA00418, diluted 1:200; Bio-Rad). Conjugated secondary antibodies included goat anti-rabbit Alexa Fluor 568 antibody (A-11011, diluted 1:200; Life Technologies) or goat anti-mouse Alexa Fluor 488 antibody (R37120, diluted 1:200; Life Technologies). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). Finally, coverslips were mounted with Vectashield mounting medium (Vector). Samples were observed using a confocal microscope (Leica TCS SP2). ImageJ software was used for image processing.

FRET-based in vitro A3G deamination assay.

A fluorescence resonance energy transfer (FRET)-based assay was used to detect cytosine deaminase activity of A3G (31). Twenty microliters of the cell lysates was used per assay using 96 assay plates. A separate solution of 20 pmol of oligonucleotide, 10 μg of RNase A, and 0.04 U of uracil DNA glycosylase (UDG) were mixed together in 50 mM Tris (pH 7.4)–10 mM EDTA buffer and adjusted to a total volume of 50 μl and then transferred to the assay well. The assay plate was then incubated at 37°C for 5 h. Next, 30 μl of 2 M Tris-acetate, pH 7.9, was added to each well, and the plate was incubated at 95°C for 2 min and on ice for 2.5 min. The fluorescence was then measured at room temperature using a Victor³ 1420 Multilabel Counter (Perkin-Elmer). Experiments were conducted with three independent replicates.

Statistical analysis.

Statistical tests were performed using GraphPad Prism, version 5.00, for Windows (GraphPad Software), unless specified differently in the text. The data were presented as means ± standard deviations (SD). Means between two or three groups were compared by using a one-way or two-way analysis of variance (ANOVA) with Bonferroni's posttest. Differences were considered statistically significant at P values of <0.05, <0.01, and <0.001, as indicated in the figure legends.

Analysis of A3 hot spot motif representation.

HCMV genome sequences were obtained from the GenBank database. To evaluate the genomic representation of A3 hot spots, we counted the number of each A3 motif in 1,000-bp windows along the HCMV Towne genome, using a sliding window approach with a step of 100 bp, on both genome strands. To assess whether this count is an overrepresentation of A3 motifs, we generated 1,000 shuffled versions of each window and counted the number of each motif within these windows. The number of these occurrences was then used to create distributions of motif counts (in each window), and the percentile rank of the true motif count was calculated. These percentile ranks are plotted in Fig. 5. For instance, a rank of 0 in a window indicates that the real number of motif counts was lower than all those obtained in reshuffled versions of that same window.

To investigate the distribution of A3G motifs in the HCMV genome by also accounting for coding capacity and amino acid composition, we counted the frequency of motifs in each HCMV ORF. We then obtained expected values by reshuffling codons in each ORF; specifically, for each ORF, we generated 1,000 codon-shuffled sequences. We next calculated a preference index for A3 motifs, defined as follows: preference index = (number of motifs observed − number of motifs expected)/(number of motifs observed + number of motifs expected). In practical terms, the preference index varies between −1 and +1, with values equal to 0 indicating that the representation of motifs is equal to the expected; negative and positive values indicate under- and overrepresentation, respectively. ORFs were grouped based on the mutant growth classification proposed by Dunn et al. (68): essential (no growth and severely defective) and nonessential (moderately defective and like wild type).