Distinct Roles for Human Cytomegalovirus Immediate Early Proteins IE1 and IE2 in the Transcriptional Regulation of MICA and PVR/CD155 Expression

Human CMV (HCMV) is an endemic β-herpesvirus that does not cause clinically obvious disease in healthy individuals, where it establishes a life-long latency. In immunocompromised hosts, such as AIDS patients and organ transplant recipients, infection often becomes clinically apparent and can cause life-threatening diseases. HCMV is also the leading viral cause of congenital infections and birth defects (1, 2). HCMV disseminates throughout the body, with a broad range of different cell types supporting productive viral infection (3). Additionally, it induces a plethora of immunomodulatory pathways to subvert the host innate and adaptive immune responses (2). To date, few antiviral drugs are available, but long-term treatment is frequently associated with toxic side effects and the emergence of drug-resistant mutants (4, 5).

Clearly, in the absence of an effective and preemptive HCMV vaccine, additional therapeutic agents are urgently needed, and strategies to potentiate an anti-HCMV immune response could also be a valuable alternative approach.

With this rationale, we investigated whether molecules capable of activating cytotoxic lymphocytes may be positively regulated following HCMV infection, thus enhancing the recognition and elimination of infected cells. In particular, we focused on the ligands of NKG2D and DNAM-1/CD226, two activating receptors expressed by all cytotoxic lymphocytes. NKG2D delivers a potent activating signal and plays a prominent role in the recognition and elimination of infected cells (6, 7). In humans, NKG2D ligands (NKG2DL) are the MHC class I–related chain (MIC) molecules MICA and MICB and the UL16-binding proteins (ULBP)1–6, whose expression is restricted in normal cells but can be rapidly induced upon cellular stress, including a viral infection (6, 7). DNAM-1 receptor is essential to NK cell–dependent antitumor immunity (8), and its role in the response to viral infections is also starting to emerge (9–12). It is an adhesion molecule, and the binding to its ligands, poliovirus receptor (PVR) (CD155) and Nectin-2 (CD112), promotes leukocyte migration as well as effector responses of both NK and T cells (8, 13). HCMV evolved specific strategies to block the functions of NKG2D and DNAM-1. Indeed, there is an array of viral molecules (UL16, UL141, UL142, US18, and US20, US9, and microRNA-UL112) targeting both NKG2DL and DNAM-1 ligands (DNAM-1L) and impairing recognition and elimination of HCMV-infected cells by NK cells and other NKG2D⁺ and DNAM-1⁺ cells (14–17). In contrast, it is still debated whether and how HCMV upregulates NKG2DL, whereas for DNAM-1L it has not been investigated.

Immediate early (IE) proteins are the first to be expressed during HCMV lytic infection and play crucial roles in regulating viral gene expression and in dysregulating host cell physiology to dictate an intracellular environment conducive to viral replicative cycle, as well as in counteracting host immune responses (2). The 72-kDa IE1, 86-kDa IE2, and 55-kDa IE55 proteins share identical N-terminal 85 aa resulting from differentially spliced transcripts, and their expression does not require de novo protein synthesis (18, 19). IE1 and IE2 are absolutely critical for the temporal cascade of viral gene expression, as they transactivate early and late genes, and either positively or negatively autoregulate their own expression (18, 19). Whereas IE1 is a relatively weak transactivator, IE2 is the most important HCMV regulatory protein and is a strong transcriptional activator of viral and cellular gene expression. It binds to DNA directly, represses its own promoter (the major IE promoter) (20), and cooperates with cellular transcription factors via protein–protein interactions. These IE2 activities are crucial for transcriptional activation of viral and host genes, as well as for regulation of several cellular functions (19). The IE55 protein is a splice variant of IE2 gene product, with a deletion between residues 365 and 519 in the C terminus, a region required for many IE2 functions, including transcriptional activation and DNA binding (19, 21–25).

Among the cellular pathways activated by IE proteins is the DNA damage response (DDR) (26, 27), involved in cell cycle checkpoint control, DNA replication and repair, and apoptosis (28). DDR is activated by many viruses, including HCMV, and although its functional relevance in HCMV infection has not been clarified, this virus induces DDR, including activation of ataxia telangiectasia mutated (ATM), ataxia telangiectasia and Rad3–related protein (ATR), and the downstream protein H2AX (26, 27, 29–34). Interestingly, expression of some NKG2DL and DNAM-1L can be dependent on the activation of DDR and on ATM/ATR kinases (35–43).

In this study, we investigated the role and the mechanisms of IE protein–mediated regulation of NKG2DL and DNAM-1L, as well as the potential of DDR in stimulating activating ligand expression. This study provides new mechanistic insight into the regulation of antiviral immunity by HCMV IE proteins.

The following mAbs were used in flow cytometry: anti-MICA (M673) and anti-ULBP4 (M475) (Amgen); anti-MICA (AMO-1) (BamOmaB); anti-MICB (MAB236511), anti-ULBP1 (MAB170818), anti-ULBP2 (MAB165903), and anti-ULBP3 (MAB166510) (R&D Systems); anti–Nectin-2 (CD112) and mouse control IgG1-FITC (BD Biosciences); anti-PVR (SKII.4), provided by Dr. M. Colonna (Washington University, St. Louis, MO); Alexa Fluor 488–conjugated anti-IE Ags (MAB810X) and FITC-conjugated anti–phospho-histone H2AX (γH2AX) (Ser¹³⁹; clone JBW301; Merck Millipore); mouse control IgG (BioLegend); allophycocyanin-conjugated goat anti-mouse (GAM) (Jackson ImmunoResearch Laboratories); and GAM-FITC (Cappel). In cytotoxicity assays, the following blocking mAbs were used: anti-NKG2D (MAB149810; R&D Systems), anti–DNAM-1 (clone DX11; Bio-Rad), and mouse IgG1 isotype control (BioLegend). The following Abs were used in immunoblotting: anti-p85 subunit of PI3K and anti-IE Ags (MAB810R; Merck Millipore); anti-ATM (D2E2) (Cell Signaling Technology); and anti-ATR (sc-1887) and anti–DNA-dependent protein kinase (DNA-PK)_CS (sc-5282; Santa Cruz Biotechnology). Other reagents used were: caffeine, methylcellulose, phosphonoformic acid (PFA; Foscarnet), gelatin, and crystal violet (Sigma Aldrich); Lipofectamine 2000 (Invitrogen); and Dharmafect from Dharmacon (GE Healthcare). The phosphorothioate oligodeoxynucleotide fomivirsen (also known as ISIS 2922) complementary to IE2 mRNA (44, 45) was synthesized by Metabion International.

Primary human foreskin fibroblasts (HFFs), the retinal epithelial cell line ARPE-19, and the human embryo kidney 293T cells were purchased from the American Type Culture Collection. HFF and 293T cells were grown in DMEM containing 10% FCS, 2 mM glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin, and 100 μg/ml streptomycin sulfate, and ARPE-19 cells were grown in a 1:1 mixture of DMEM and Ham’s F-12 medium (Invitrogen) containing 10% FCS, 15 mM HEPES, 2 mM glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin, and 100 μg/ml streptomycin sulfate. HFFs were used at passages 14–28. Human microvascular endothelial cells (HMVECs) (dermal origin, CC-2543) were obtained from Clonetics and cultured in endothelial growth medium corresponding to endothelial basal medium (Clonetics), containing 10% FCS, human recombinant vascular endothelial growth factor, basic fibroblast growth factor, human epidermal growth factor, insulin growth factor 1, hydrocortisone, ascorbic acid, and heparin. Cells were seeded onto culture dishes coated with 0.2% gelatin. Experiments were carried out with cells at passages 4–15. Fibroblasts derived from an ataxia telangectasia–mutated patient and not expressing ATM protein (AT^−/−) were provided by Drs. M. Fanciulli and T. Bruno (Regina Elena National Cancer Institute, Rome, Italy) (46). They were grown in DMEM containing 15% FCS and used at passages 5–8. All cells were maintained at 37°C in a 5% CO₂ atmosphere.

The HCMV AD169 strain (ATCC VR538) was prepared by infecting semiconfluent monolayers of HFF cells at a virus-to-cell ratio of 0.01 and cultured until a marked cytopatic effect was seen. Stocks were then prepared after three rounds of cell freezing and thawing, subjected to centrifugal clarification, and frozen at −80°C. Virus titers were measured by standard plaque assays on HFF cells. Stock solutions used in all experiments contained ∼2 × 10⁷ PFU/ml. Standard plaque assays were used also in different experiments to determine viral titers in the supernatants harvested from infected cells. HCMV TR was derived from an ocular specimen (47), and after a few passages on fibroblasts, it was cloned into a bacterial artificial chromosome (48, 49). Reconstitution of infectious TR was performed as previously described (50) by cotransfecting HFFs with the corresponding TR bacterial artificial chromosome and a plasmid expressing HCMV pp71. Reconstituted infectious virus retained the ability to infect endothelial and epithelial cells, as well as monocytes and macrophages (49, 50). HCMV VR1814 is a derivative of a clinical isolate recovered from a cervical swab of a pregnant woman (51). This strain was propagated in HUVECs and titrated as previously described (52).

Cells were infected at ∼80–90% confluence at a multiplicity of infection (MOI) of 1 PFU/cell, unless otherwise specified, in their respective culture medium, without FCS, and after 2 h (AD169 or TR strains) or 5 h (VR-1814 strain) at 37°C, virus inoculum was discarded and replaced with fresh growth medium (day 0). Mock-infected control cultures were exposed for the same amount of time to an equal volume of medium. At various days after infection, cells were harvested and analyzed. In some experiments, PFA was added after virus inoculation at a final concentration of 200 μg/ml, whereas fomivirsen was added 1 h before viral inoculum and maintained in the culture medium during the infection and then throughout the assay (44, 45). The DDR inhibitor caffeine (53, 54) was added 2 d postinfection (dpi) at a final concentration of 10 mM.

Recombinant adenoviral vectors (AdV) encoding HCMV IE2 (AdV-IE2) and Escherichia coli β-galactosidase (AdV-LacZ) have been previously described (55, 56). AdV-IE72 (AdV-IE1) was provided by Dr. Timothy F. Kowalik (University of Massachusetts Medical School, Worcester, MA) (27). Recombinant AdV stocks were generated, purified, and titrated as previously described (27, 55, 56). For adenoviral transduction, HFFs were infected at ∼80–90% confluence at an MOI of 4 PFU/cell in DMEM without FCS for 2 h at 37°C. When the viral proteins were not expressed in combination, the total MOI was equalized to 4 with AdV-LacZ. After 2 h, the virus inoculum was discarded and replaced with fresh growth medium (day 0) and analyzed at the indicated day after infection. Mock-infected cells served as control cultures. Following infection, cultures were maintained in growth medium and analyzed at the indicated day after infection.

Mock-infected or infected cells were harvested at the indicated day after infection and stained with mAbs specific for MICA, MICB, ULBP1–4, PVR, and Nectin-2, followed by GAM-APC or by GAM-FITC (for experiments with PFA) and analyzed by flow cytometry on a FACSCalibur (BD Biosciences). The mean fluorescence intensity (MFI) value of the isotype control Ab was always subtracted from the MFI relative to each molecule. For intracellular staining of IE Ags or γH2AX, cells were fixed in 1% formaldehyde, permeabilized with 70% ethanol, and then incubated with Alexa Fluor 488–conjugated anti-IE mAb (MAB810X) or with FITC-conjugated anti-γH2AX (JBW301), respectively.

Cell-mediated cytotoxicity was assessed in 4-h ⁵¹Cr-release assays. Polyclonal NK cells, generated as previously described (57), were used as effectors and incubated at different ratios with 5 × 10³ target cells in U-bottom, 96-well microtiter plates at 37°C in a 5% CO₂ atmosphere. To block NKG2D and DNAM-1 receptors, effector cells were preincubated with 1 μg/10⁶ cells of specific or isotype control mAbs for 15 min at room temperature. Cells were then washed and used in the assays. Percentage of lysis was determined by counting an aliquot of supernatant and using the formula: 100 × [(sample release − spontaneous release)/(total release − spontaneous release)]. Mean inhibition of lysis (%) ± SE by anti-NKG2D, anti–DNAM-1, or isotype control mAb treatment was calculated in comparison with untreated NK cells (no Ab) using the formula: [1− (% specific lysis by mAb treatment/% specific lysis of no Ab) × 100].

Cells were lysed for 20 min at 4°C in a lysis buffer containing 0.2% Triton X-100, 0.3% Nonidet P-40, 1 mM EDTA, 50 mM Tris HCl (pH 7.6), 150 mM NaCl, and protease inhibitors to obtain whole-cell protein extracts. Lysates (30–40 μg) were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Merck Millipore). Membranes were blocked with 5% milk and probed with the indicated Abs. Immunoreactivity was revealed using an ECL kit (Amersham).

The ON-TARGETplus SMARTpool small interfering RNA (siRNA) specific for ATM and ATR (siATM and siATR), and the ON-TARGETplus nontargeting pool (siCtrl) were purchased from Dharmacon (Thermo Fisher Scientific). siRNA specific for DNA-PK_CS (sc-35200) (siDNA-PK) was from Santa Cruz Biotechnology. HFFs (70–80% confluence) were transfected with 100–200 nM siRNA using DharmaFECT siRNA transfection reagent (Thermo Fisher Scientific), according to the manufacturer’s recommendations. One to two days after transfection, cells were infected with AD169, as indicated in the figure legends. Cells and supernatants were harvested and analyzed at 2 or 3 dpi, as indicated. Densitometric analysis was performed with ImageJ software.

Total RNA was extracted using TRI Reagent solution (Life Technologies), according to the manufacturer’s instructions, and 1 μg of total RNA was used for cDNA first-strand synthesis in a reaction volume of 25 μl. Real-time PCR was performed using the ABI Prism 7900 sequence detection system (Applied Biosystems); cDNAs were amplified in triplicate with primers for MICA (Hs00792195_m1), ULBP3 (Hs00225909_m1), PVR (Hs00197846_m1), and GAPDH (Hs03929097_g1) using specific TaqMan gene expression assays (Applied Biosystems). Relative expression of each gene versus GAPDH was calculated according to the 2^−ΔΔCt method.

The pGL3-MICA promoter vector was previously described (58) and provided by Dr. J. Bui (University of California at San Diego, La Jolla, CA). The MICA −270 bp promoter plasmid was obtained as previously described (59). Mutant MICA −270 bp–CG construct was generated using a QuickChange site-directed mutagenesis kit (Stratagene). Primer sequences used were: −92 bp 5′-CGGTCGGGGGACCG-3′ −78 bp; primers for mutagenesis were: forward, 5′-CCAGTTTCATTGGATGAGATGTCGGGGGACATGGCCAGGTGACTAAG-3′, reveverse, 5′-CTTAGTCACCTGGCCATGTCCCCCGACATCTCATCCAATGAAACTGG-3′. Inserted mutations were verified by sequencing. pGL2-PVR (−571 bp fragment) promoter luciferase reporter vector and progressive deletions were provided by Dr. G. Bernhardt (Hannover Medical School, Hannover, Germany) (60). pSG5-IE1, pSG5-IE2, and pSG5-IE55 were previously described (24). The IE2 cDNA cloned in the pRSV vector and the zinc finger mutant of IE2, with cysteines 428 and 434 mutated into serine residues (pRSV-IE2-Zn mut), were a gift of Prof. Jay Nelson (61).

In all transfection experiments, 3 μg of luciferase reporter, 0.25 μg of pRL-CMV-Renilla, and 2 μg of IE protein vectors or pSG5 empty vector were cotransfected into 80–90% confluent cells growing on a 10-cm² area using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocols. In some experiments, the pSG5-IE2 vector was replaced by pSG5-IE55, pRSV-IE2, or pRSV-IE2-Zn-mut, as indicated. Forty-eight hours after transfection, cells were harvested and prepared for the luciferase assays, using the Dual-Luciferase reporter assay kit and the GloMax-Multi detection system (Promega), following the manufacturer’s instructions. Relative luciferase activity was calculated by dividing the luciferase activity of pGL3-MICA or pGL2-PVR reporter, cotransfected with IE proteins, by the respective pGL3- or pGL2-Basic to remove the unspecific effect of IE proteins on the reporter vector. The unspecific modulation of the reporter empty vector activity was probably due to a general activation of the transcriptional machinery by IE proteins and was more evident for IE1. This correction allowed us to better appreciate the specific effect of the viral proteins on ligand promoters.

In chromatin immunoprecipitation (ChIP) assays, 293T cells were cotransfected with 5 μg of MICA -270 promoter plasmid (wild-type or mutated) and pSG5-IE1 (10 μg) and pSG5-IE2 (10 μg), or pSG5 empty vector (20 μg), using Lipofectamine 2000. In ChIP assays on the endogenous MICA promoter, 293T cells were transfected with pSG5-IE1, pSG5-IE2, or pSG5 empty vector. After 48 h, cells were processed for ChIP assays following the manufacturer’s protocol for the Magna ChIP A/G chromatin immunoprecipitation kit (Merck Millipore). Chromatin was immunoprecipitated with a polyclonal rabbit anti-IE Ab, recognizing a segment of IE2 (aa 1–143) or control polyclonal rabbit serum. PCR primers used were: MICA forward, 5′-AGGTCTCCAGCCCACTGGAATTTTCTC-3′, reverse, 5′-CGCCACCCTCTCAGCGGCTCAAGC-3′. Results are expressed as relative enrichment as compared with the input. Negative control (polyclonal rabbit serum) values were subtracted from the corresponding samples. Quantifications were obtained by serial dilutions of the input DNA samples. The analysis was performed using the SDS version 2.4 software (Applied Biosystems). PCRs were validated by the presence of a single peak in the melt curve analysis, and amplification of a single specific product was further confirmed by electrophoresis on agarose gel.

For staining of cell surface MICA, HFFs were grown to semiconfluence on glass coverslips in 24-well plates and infected with AD169 and TR at an MOI of 1 PFU/cell for 2 h at 37°C. After 4 dpi, cells were washed with PBS, fixed in 1% paraformaldehyde for 15 min at room temperature, and blocked in 1% FCS diluted in PBS (20 min, room temperature), but not permeabilized. Indirect immunofluorescence analysis was performed by incubating fixed cells with the anti-MICA mAb AMO-1 (1:40) for 2 h at 37°C, followed by secondary Ab incubation with CF594-conjugated rabbit anti-mouse IgG (Sigma-Aldrich) for 1 h at room temperature. Samples were then visualized with an Olympus IX70 inverted laser scanning confocal microscope, and images were captured using FluoView 300 software (Olympus Biosytems).

Statistical analyses of the data were performed using a paired Student t test or a one-way ANOVA, where indicated. A p value <0.05 was considered significant.

Increased or de novo expression of T/NK cell–activating ligands on infected cells represents a crucial host immune defense mechanism to sense and react against different pathogens (7, 12). Therefore, we examined the expression of NKG2DL and DNAM-1L in several cell models with multiple HCMV strains. First, human primary foreskin fibroblasts (HFFs) were infected with the HCMV laboratory strain AD169. We observed higher levels of MICA and ULBP3 on infected HFFs, with maximal expression at ∼3 dpi, whereas no changes in MICB, ULBP1, or ULBP4 expression were detected. In contrast, ULBP2 was down-modulated by HCMV (Fig. 1). The DNAM-1L PVR, but not Nectin-2, was also upregulated by HCMV, with a maximal increase at 3 dpi (Fig. 1).

To verify that the augmented expression of MICA, ULBP3, and PVR was not restricted to a particular viral strain, their modulation was also examined in HFFs infected with the low-passage strains VR-1814 and TR. Consistently, an induction of MICA was observed upon infection with VR-1814, independently of the MOI (Fig. 2A, 2B). A low, but statistically significant, level of MICA induction was also observed with TR (Fig. 2C) and was confirmed by confocal microscopy (Supplemental Fig. 1). ULBP3 and PVR ligands were also upregulated on HFFs infected with VR-1814 (Fig. 2A, 2B) or TR (Fig. 2C). Taken together, these results demonstrated that by 3–4 dpi, MICA, ULBP3, and PVR were upregulated on infected primary fibroblasts in an HCMV strain-independent manner.

Next, we extended our investigation to other cell type/viral strain combinations by infecting primary endothelial (HMVEC) and epithelial cells (ARPE-19) with TR and VR-1814 strains (Fig. 2C). MICA expression was either down-modulated on TR-infected HMVECs or not affected in the other combinations, whereas ULBP3 and PVR were always upregulated independently of the cell type and/or the viral strain used.

Thus, these results demonstrate that, despite few exceptions, HCMV positively regulates the expression of MICA-, ULBP3-, and PVR-activating ligands, with a pattern that generally overcome cellular- or viral strain–related differences.

Next, we tested whether the observed upregulation of NKG2D and DNAM-1 ligands upon HCMV infection had consequences on NK cell–mediated cytotoxicity. Chromium-release assays were performed using polyclonal NK cell cultures from different donors as effectors, and uninfected or AD169- and TR-infected HFFs as targets. HFFs infected with the low-passage strain TR became more resistant to NK cell–mediated lysis, whereas infection with AD169 resulted in a variable pattern, with an either increased, unchanged, or decreased sensitivity (Fig. 3A, data not shown). These results are in line with previous observations on both laboratory and low-passage HCMV strains, which demonstrated that cells infected with low-passage strains were more resistant to NK cell–mediated cytotoxicity, compared with AD169-infected cells (14, 15, 57, 62). Nevertheless, despite the increased resistance of TR-infected cells to NK cell lysis, blocking NKG2D or DNAM-1 receptors resulted in a significant inhibition that was comparable to that observed with AD169-infected cells or uninfected cells in all experiments performed (Fig. 3B).

Overall, these data indicate that NKG2D and DNAM-1 receptors contribute to the elimination of HCMV-infected cells. Moreover, despite the increased resistance to NK lysis of HFFs infected with the TR strain, NKG2D and DNAM-1 ligands still contribute to the recognition of these target cells, in accordance with the increased expression of MICA, ULBP3, and PVR ligands upon TR infection.

As previous studies reported that HCMV manipulates the DDR (26, 27, 29–34), a pathway able to stimulate NKG2DL and DNAM-1L expression as well (35–43), we examined the involvement of DDR signaling in the HCMV-mediated upregulation of activating ligands using genetic and pharmacological approaches.

Upon HCMV infection, the levels of γH2AX, the phosphorylated form of the histone variant H2AX, a well-known substrate of DDR kinases (28), increased of ∼2-fold, demonstrating activation of the DDR pathway in our experimental settings (Supplemental Fig. 2A, 2B). Next, we determined the contribution of the three main DDR kinases (ATM, ATR, and DNA-PK) on ligand expression, IE expression, and viral replication. First, the role of ATM was investigated in fibroblasts derived from a patient affected by ataxia telangiectasia (AT^−/−), where ATM is not detectable. HCMV infection still increased the expression of MICA, ULBP3, and PVR, although with delayed kinetics compared with normal HFFs (Supplemental Fig. 2C). Moreover, both progeny virus production and IE expression were only partially affected in AT^−/− cells, but not in a statistically significant manner (data not shown). Then, we used specific siRNA to transiently deplete ATM (siATM) (Supplemental Fig. 2D–G), and consistently to AT^−/− fibroblasts, there was no effect on MICA, ULBP3, and PVR expression induced by HCMV (Supplemental Fig. 2D) and on the percentage of IE⁺ cells and viral replication (Supplemental Fig. 2E, 2F). Similar results were obtained with siRNA specific for ATR (Supplemental Fig. 3A–D) or DNA-PK (Supplemental Fig. 3E–H), as well as with a triple gene silencing with the three siRNAs specific for ATM, ATR, and DNA-PK (siDDR) (Fig. 4). Finally, activating ligands were still upregulated in AD169-infected HFFs treated with caffeine, a well-known and broad spectrum inhibitor of DDR (data not shown).

Altogether, these results suggest that DDR activation does not play a role in the HCMV-induced upregulation of MICA, ULBP3, and PVR.

To identify the molecular mechanisms underlying ligand upregulation in HCMV-infected cells, we hypothesized that some events in the early stages of infection could be responsible. To verify this hypothesis, HFFs were infected with HCMV and treated with PFA, a selective inhibitor of viral DNA polymerase (63). As shown in Fig. 5, MICA, ULBP3, and PVR levels were increased on the surface of infected cells even in the presence of PFA, indicating that viral DNA replication and expression of delayed early and late genes are dispensable for ligand upregulation.

Next, to investigate whether the increase in ligand cell surface levels was a consequence of a virus-induced transcriptional activation, we measured ligand mRNA content by real-time PCR at different hours after infection. MICA, ULBP3 and PVR mRNA progressively increased during the course of infection, with a maximal expression at 24–48 h postinfection (hpi) (Fig. 6).

These data suggest that upregulation of MICA, ULBP3, and PVR cell surface levels by HCMV is the outcome of a transcriptional activation of the corresponding genes.

Because early steps of infection were crucial for ligand up-regulation, we investigated whether the major viral IE proteins, IE1 and IE2, were involved in the modulation of MICA, ULBP3, and PVR expression by transducing HFFs with recombinant AdV encoding for IE1, IE2, or their combination and analyzing ligand mRNA and cell surface levels at 24, 48, and 72 hpi. There was a significant upregulation of MICA mRNA at all time points only in IE2 transduced cells, whereas IE1 did not affect MICA mRNA levels, neither when used alone nor in combination with IE2 (Fig. 7A). Similar results were obtained for MICA cell surface expression, which showed an IE2-dependent increase, particularly evident at 72 hpi (Fig. 7B, 7C, data not shown). In contrast, PVR mRNA content and membrane expression was mostly upregulated by the coexpression of IE1 and IE2, whereas IE proteins alone had a weaker effect (Fig. 7). ULBP3 mRNA and cell surface expression were instead not affected by IE proteins (Supplemental Fig. 4). Thus, whereas the HCMV-induced upregulation of ULBP3 may be the consequence of other virus-related effects than solely the overexpression of IE1/IE2, MICA and PVR increase could be reproduced by expression of IE proteins, although with a different requirement.

Then, to further sustain the role of IE proteins in MICA and PVR upregulation, we inhibited their expression by using fomivirsen (also known as ISIS 2922), an antisense oligodeoxynucleotide complementary to IE2 mRNA and able to prevent both IE1 and IE2 protein expression when used at certain concentrations (44, 45). This approach allowed us to specifically address the role of IE proteins in regulating ligand expression within the context of HCMV infection. To this end, HFFs were treated with different doses of fomivirsen, from 1 h before and throughout the entire infection (Fig. 8). At the highest dose of fomivirsen (500 nM), expression of both IE1 and IE2 was inhibited (Fig. 8C, 8D), as previously observed by Azad et al. (44), and, as expected, MICA and PVR upregulation could not be detected (Fig. 8A, 8B). By progressively decreasing the concentration of fomivirsen (to 5 and 1 nM), we could rescue IE1 protein expression (which was the first IE protein to reappear) and IE2 (Fig. 8C, 8D). At these low concentrations of fomivirsen, recovery in HCMV-induced ligand upregulation was observed (Fig. 8A, 8B).

These results clearly demonstrated that the specific inhibition of IE protein expression in the context of HCMV infection prevented MICA and PVR increase, therefore supporting the importance of these viral proteins in the HCMV-mediated ligand regulation.

Next, we further examined the possibility that IE proteins could activate MICA and PVR gene promoters. Thus, we cotransfected HFFs with pGL3-MICA (58) or pGL2-PVR (60) luciferase reporter plasmids, harboring, respectively, −1 kb and −571 bp MICA and PVR promoter regions, together with IE1 or IE2 expression vectors. We observed that only IE2 transactivated the MICA promoter, up to ∼3-fold compared with the control. Transfection of IE1, alone or together with IE2, did not significantly affect MICA promoter activity, compared with IE2 alone (Fig. 9A). These results are in line with previous observations obtained on the regulation of MICA mRNA and cell surface expression in cells transduced with AdV IE2.

We then analyzed IE2 structural requirements and its interaction with MICA promoter sequences. First, we observed that expression of IE55, which lacks the transcriptional activation and DNA binding properties of IE2, was a poor transactivator of MICA, either in combination with IE1 (Fig. 9B) or alone (Fig. 9D). Then, a zinc finger mutant of IE2, which cannot bind to DNA but retains the ability to transactivate early gene promoters by protein–protein interactions (61, 64), did not significantly increase MICA promoter activity, neither with IE1 nor alone (Fig. 9C, 9E). These results indicate that the IE2 functional domains located primarily toward the C-terminal end of the protein are required to transactivate MICA gene promoter.

Then, we used a shorter MICA construct (MICA −270 bp) to map the regions targeted by IE proteins. This fragment was indeed activated by IE1 and IE2 at similar levels compared with the longer MICA −1 kb region, indicating that the IE-responsive region was contained within the 270-bp fragment (Fig. 10A).

IE2-binding sites identified on viral and cellular promoters contain invariant CG residues at both ends of a 10-nucleotide sequence (CG-N₁₀-CG) (20, 25, 65, 66), and we found a similar sequence within MICA promoter between residues −92 and −78 (Fig. 10B). To evaluate the contribution of this putative IE2-binding site to the overall IE2-dependent transactivation of MICA, we changed by site-directed mutagenesis this unique CG-N₁₀-CG motif into an AT-N₁₀-AT sequence within the context of the MICA −270 bp construct. The introduced mutations significantly reduced IE2-dependent transactivation of MICA, thus supporting an involvement of the putative IE2-binding site in the regulation of this promoter (Fig. 10A, 10B).

We then addressed the capability of IE1/IE2 proteins to directly bind to MICA promoter by ChIP assays, using the wild-type or the CG mutant form of MICA, in highly transfectable 293T cells. Using an anti-IE Ab and specific primers to amplify the region containing the putative IE2-binding site, we observed that IE1/IE2 were recruited to the MICA promoter. The interaction was not detectable with the empty vector pSG5 or using normal rabbit serum as a negative control (Fig. 10C). Disruption of the putative IE2-binding site of MICA reduced IE binding by ∼60%, further demonstrating that this sequence is involved in the IE2-dependent transactivation of MICA (Fig. 10D). The binding was confirmed on the endogenous MICA promoter as well, and it was detectable only when IE2 was expressed (Fig. 10E).

Taken together, these results demonstrate the capability of IE2 to directly bind sequences within MICA gene promoter, and that this binding is required for MICA transcriptional activation.

In relationship to PVR, we performed similar transient cotransfection assays with a PVR −571 bp construct (60) and vectors expressing IE proteins. Although IE1 activated the PVR promoter up to 10-fold over the control, the combination of IE1 and IE2 induced a prominent transcriptional activation that exceeded significantly the effect of IE1 alone. IE2 was instead ineffective in stimulating PVR (Fig. 11A). In contrast to what was observed for MICA, expression of IE55 and of the IE2 zinc finger mutant did not affect PVR promoter activity (Fig. 11B, 11C). Finally, to identify the IE-responsive regions, we cotransfected IE1 and IE2, alone or in combination, with progressive deletions of PVR promoter (Fig. 11D, 11E) (60) and observed a significant drop in luciferase activity with the truncated sequences between −281 and −213 bp, indicating that this fragment mediated most of the transactivating activity resulting from the combination of IE1 and IE2, and only in minor part from IE1 alone (Fig. 11E).

Collectively, our results indicate that the increase in cell surface expression of MICA and PVR upon HCMV infection is mediated by IE proteins through the transcriptional activation of their gene promoters.

The molecular mechanisms driving the expression of NKG2DL and DNAM-1L remain largely unknown, particularly in virus-infected cells. In this study, we investigated the impact of HCMV infection on their expression and showed that MICA, ULBP3, and PVR are upregulated on infected cells in different cell type/viral strain combinations. For MICA, data suggest that its increased or de novo expression may be restricted to certain cell types, as it was observed on infected fibroblasts independently from the strain used, but not in endothelial or epithelial cells. Information on a cell type–specific regulation of MICA expression are currently not available, and further investigations would be of unquestionable interest for a better characterization of this molecule. However, the evidence that in primary fibroblasts MICA was induced by both laboratory and low-passage HCMV strains suggests that the downmodulating activity exerted by the viral proteins UL142, US9, US18, and US20 on this ligand (14–17) was not sufficient to prevent its overall cell surface expression. Similarly, although UL142 was described to prevent expression of ULBP3 as well (67), in our settings this ligand was always increased, consistent with previous findings (57). These discrepancies may be related to different experimental conditions and/or to the considerable polymorphism in the UL142 sequence among different strains (68, 69). Thus, some variants of viral proteins may be less efficient at down-modulating NKG2DLs than others. At the same time, polymorphisms in both the coding and noncoding regions of MICA and ULBP3 (70–73) may also impact their expression upon HCMV infection. Thus, a prediction deriving from the presence of NKG2DL on the cell surface of HCMV-infected targets would be that blocking the receptor in cytotoxicity assays results in a decreased NK cell lysis. Indeed, this was the outcome of blocking experiments (Fig. 3), which demonstrated that the NKG2D receptor plays a role in the elimination of infected cells, as previously shown (57).

In relation to PVR, at present there are few reports on its regulation by HCMV (74–76). In particular, its expression was down-modulated in fibroblasts infected with the low-passage strain Merlin (74, 75). In contrast, our results show, to our knowledge for the first time, that PVR can be upregulated by HCMV infection in different cell types and with different viral strains, thus offering the immune system the opportunity to detect and react against infected cells through the activating receptor DNAM-1. Indeed, blocking of DNAM-1 in killing assays resulted in a significant inhibition of target cell lysis, similar to what we observed for NKG2D (Fig. 3). Thus, from a functional point of view, the numerous HCMV immunoevasion strategies evolved against NKG2D and DNAM-1 ligands seem to be not completely successful, because these activating receptors still play a role in eliminating infected cells, including those infected with low-passage strains, which are per se less susceptible to NK killing (this study and Refs. 14, 15, 57, 62). In line with our data, DNAM-1 plays a relevant role in NK cell recognition of HCMV-infected myeloid dendritic cells early in infection, whereas the effect of viral-mediated downregulation of DNAM-1L prevails at later stages, thus underlying the importance of the kinetics of immune evasion mechanisms (76). Moreover, a recent study demonstrated that DNAM-1Ls are rapidly induced during murine CMV infection in vivo, and the engagement of DNAM-1 is essential for the optimal NK cell–mediated host defense against the virus (11). Of note, because DNAM-1 is also expressed by many other leukocyte subsets and is an important activator of their effector functions, it may affect a wide range of immunological responses (8, 12, 13, 77).

To gain insights into the molecular mechanisms regulating the expression of activating ligands in infected cells, we investigated the role of DDR, a host cell pathway that positively affects the expression of activating ligands (35–43) and that it is activated by HCMV (26, 27, 29–34). Nevertheless, in HCMV-infected HFFs, MICA, ULBP3, and PVR were still increased even when ATM, ATR, and/or DNA-PK were knocked down, thus indicating that these DDR kinases are not involved in the HCMV-mediated ligand stimulation, similar to what has been reported for murine NKG2DL during murine CMV infection (78).

HCMV IE proteins have been suggested to be implicated in the regulation of MIC proteins (16, 79), but the molecular mechanisms are unknown. Moreover, no data have been reported on the regulation of PVR by HCMV. Our results show that ectopic expression of IE1 and IE2 induced a significant increase of MICA and PVR, both at the mRNA and cell surface level. In particular, IE2 emerged as the main transactivator of MICA promoter, with the effect strictly dependent on its DNA binding activity, because it was lost in the presence of the IE55 isoform or the zinc finger mutant form of IE2. Accordingly, through ChIP and mutagenesis approaches, we identified an IE2 consensus sequence within the MICA gene promoter that turned out to be critical for MICA promoter transactivation by IE2.

This observation contributes to challenge the prevailing view that activation of cellular genes by IE2 depends on interactions with basal transcription factors, whereas nucleotide-specific binding of IE2 is the predominant mode of regulation of HCMV promoters (19–21, 25, 65). Moreover, this finding also suggests that the IE2-binding sites on cellular versus HCMV promoters are different, with the 10 internal nucleotides of the CG-N₁₀-CG motifs being GC rich, rather than AT rich, as previously suggested for the cyclin E promoter (66), and it supports the idea that IE2 is relatively sequence tolerant (25, 65, 66).

In regard to PVR, our results demonstrate a different mechanism of the HCMV-induced upregulation. In fact, PVR mRNA and protein upregulation required the coexpression of both IE1 and IE2. Furthermore, by using progressive deletions of PVR promoter, we mapped a region between −281 and −213 bp mostly responsive to IE1/IE2 combination. This fragment contains a potential IE2-responsive CG-N₁₀-CG element (from −271 to −257: 5′-CG-CAGGCGCAGG-CG-3′), but it is unlikely that IE1/IE2 proteins bind to the PVR promoter because the IE55 isoform and the zinc finger mutant of IE2 retained the capability to activate PVR promoter, and IE1 seems not to bind DNA directly (18). Accordingly, in ChIP assays we were unable to observe any detectable binding to the PVR promoter either of the single IE proteins or of their combination (data not shown). Thus, it is more likely that the −281/−213 bp region contains the binding sites of cellular transcription factors recruited and/or activated by IE proteins. In fact, this 68-bp region contains putative binding sites for several transcription factors, such as E2F, Sp1, AP-2α, Nrf-1, and NF-κB (C. Fionda and C. Cerboni, unpublished observations), but further studies should be undertaken to identify which are the cellular proteins involved in the IE-mediated activation of the PVR promoter.

As a final consideration on the importance of IE proteins in the regulation of MICA and PVR gene expression, we should also emphasize that it was observed not only by IE overexpression, but also in the context of HCMV infection. Indeed, by using fomivirsen (44, 45), we observed that the inhibition of IE protein expression prevented the HCMV-induced MICA and PVR upregulation (Fig. 8). Conversely, regaining IE protein expression by lower doses of fomivirsen resulted in a recovery of ligand upregulation as well. These data thus clearly demonstrate that inhibition of IE protein expression in HCMV-infected cells prevents MICA and PVR increase.

In regard to ULBP3 regulation, although we could detect a significant increase in its mRNA and cell surface level upon HMCV infection, overexpressing IE1/IE2 by adenoviral vectors did not have a major effect on the expression of this ligand (Supplemental Fig. 4), suggesting that IE1/IE2 were not sufficient for ULBP3 upregulation.

From our study, two questions arise. First why should a virus increase the expression of molecules involved in the elimination of infected cells? A possible answer could derive from the absolute requirement of IE proteins for a productive viral replication (18, 19), with the induction of NKG2DL and DNAM-1L being an unavoidable side effect of the strong transactivating activity of IE2. In this scenario, upregulation of activating ligands in HCMV-infected cells may represent an acceptable toll to pay to survive. Moreover, the IE2-consensus sequence we identified is conserved among different allelic variants of MICA promoter (C. Fionda and C. Cerboni, unpublished observations, and Ref. 72, 73), suggesting that during the virus/host coevolution, a positive selection of promoter sequences in MICA alleles carrying the IE2 DNA binding site occurred, with the host likely making IE2 useful for its own cellular gene expression as well. The second question is how can we reconcile the observed HMCV-triggered increase of activating ligands with the immunoevasion strategies evolved by the virus to target the same molecules? There could be a window of opportunity, a temporal frame in the early phases of HMCV infection during which the unavoidable upregulation of NKG2DL and DNAM-1L by IE proteins precedes the late expression of virus-encoded immunoevasion proteins. Thus, with elevated, functionally relevant levels of activating signals, the immune surveillance against the viral infection could be sufficiently robust, allowing recognition of infected cells by cytotoxic lymphocytes even at early times of infection. Moreover, HMCV diversity and tropism could have an important role as well. In fact, a hallmark of HCMV infections is its dissemination to a wide range of host tissues and cell types (3) with significant differences in the level of virus diversity between different compartments (80, 81). Although neither the mechanism explaining HCMV compartmentalization and intrahost genetic diversity nor their effects on clinical disease are clear at present, one possibility is that the generation of mutants may influence NK cell and/or T cell recognition, depending on the compartment (81).

In conclusion, our findings contribute to improve the understanding of the mechanisms underlying the regulation of the expression of NKG2D and DNAM-1 ligands, and consequently affecting immune responses mediated by their activating receptors expressed on all cytotoxic lymphocytes. This knowledge may be exploited to take full advantage of this potent immune pathway for therapeutic purposes.

We thank Jay A. Nelson, Tim Kowalik, Marco Colonna, Jack D. Bui, Günter Bernhardt, Giuseppe Gerna, Andrea Gallina, and Maurizio Fanciulli for reagents, members of the Santoni laboratory for discussions, and John Hiscott for a critical reading of the manuscript.

The authors have no financial conflicts of interest.