Estrogen-related receptor α is required for efficient human cytomegalovirus replication

Significance

Viruses use the host cell’s resources, including numerous cellular metabolites, to successfully replicate and produce progeny. Here, we report that estrogen-related receptor α (ERRα), an orphan nuclear receptor and transcriptional regulator, supports the expression of multiple metabolic enzymes that contribute to the metabolic program induced by human cytomegalovirus in cultured primary fibroblasts. Loss or inhibition of ERRα impedes the viral replication cycle at multiple levels, leading to a reduction in progeny virus. Our findings identify ERRα as a host factor that is manipulated by the virus for its replicative needs and establish its potential as a pharmacological target for treatment of cytomegalovirus infection.

Abstract

An shRNA-mediated screen of the 48 human nuclear receptor genes identified multiple candidates likely to influence the production of human cytomegalovirus in cultured human fibroblasts, including the estrogen-related receptor α (ERRα), an orphan nuclear receptor. The 50-kDa receptor and a 76-kDa variant were induced posttranscriptionally following infection. Genetic and pharmacological suppression of the receptor reduced viral RNA, protein, and DNA accumulation, as well as the yield of infectious progeny. In addition, RNAs encoding multiple metabolic enzymes, including enzymes sponsoring glycolysis (enolase 1, triosephosphate isomerase 1, and hexokinase 2), were reduced when the function of ERRα was inhibited in infected cells. Consistent with the effect on RNAs, a substantial number of metabolites, which are normally induced by infection, were either not increased or were increased to a reduced extent in the absence of normal ERRα activity. We conclude that ERRα is needed for the efficient production of cytomegalovirus progeny, and we propose that the nuclear receptor contributes importantly to the induction of a metabolic environment that supports optimal cytomegalovirus replication.

Human cytomegalovirus (HCMV) belongs to the β-herpesvirus subfamily, and although most healthy individuals remain asymptomatic subsequent to infection, the pathogen is a major contributor to birth defects and to life-threatening disease in immunocompromised patients (1–3). As with all viruses, HCMV depends on the host cell to provide macromolecular building blocks for virion production, and throughout the course of its evolution, HCMV has adapted to manipulate numerous fundamental cellular processes, including RNA accumulation (4), translation (5), metabolism (6, 7), secretory pathways (8), and the cell cycle (9).

The nuclear receptor gene family, of which there are 48 members in the human genome, encodes transcription factors (10). Some bind to specific hydrophobic ligands such as steroids, vitamin D, fatty acids, and lipids, providing a means to sense changes in the extracellular or intracellular environment; others are “orphans” with no known ligand, and are often constitutively active (11). As transcription factors that modulate broad gene regulatory networks, nuclear receptors control numerous physiological processes, including aspects of metabolism (cell autonomously and at the whole-organism level), development, and cellular differentiation, cancer, circadian rhythm, and immunity (11).

The first indication that nuclear receptor signaling modulates HCMV replication surfaced after the identification of a retinoic acid response element within the major immediate-early promoter (MIEP) that caused a dose-dependent response to retinoic acid in reporter assays (12). Retinoic acid was further shown to enhance viral DNA replication and progeny production in human embryonal carcinoma cells and foreskin fibroblasts (13, 14). The ligand also exacerbated symptoms of infection, which were reversed by retinoic acid receptor antagonists, in a mouse model using murine CMV (15), arguing that the vitamin A metabolite plays a physiologically relevant role in CMV infection.

Two additional nuclear receptors reported to affect HCMV replication are the glucocorticoid receptor and the peroxisome proliferator-activated receptor γ (PPARγ). Cortisol, the glucocorticoid receptor ligand, enhanced virus production in human embryonic kidney cells (16). Most recently, a PPAR response element was described within the MIEP, and a PPARγ antagonist reduced viral protein expression, lipid droplet formation, and virus production (17).

To replicate efficiently, HCMV induces numerous metabolic and lipidomic changes (18, 19), and the intimate connection between metabolism and nuclear receptors (20, 21) offers a compelling reason to examine the role of these transcription factors during infection more closely. In addition, nuclear receptors are also known to modulate inflammatory signaling (22) and cell cycle progression (23), pathways that are important for successful infection.

To explore potential roles for additional nuclear receptors in the HCMV replication cycle, we performed an shRNA screen in permissive primary fibroblasts to identify host factors that modulate virus production. Multiple hits were identified, and the orphan nuclear receptor, estrogen-related receptor α (ERRα), was further validated and characterized to be essential for optimal replication of both laboratory and clinically derived strains of HCMV.

Results

Nuclear Receptor shRNA Screen Identifies Putative Regulators of HCMV Replication.

Each of the 48 human nuclear receptor genes was targeted by using two independent shRNAs, and their effect on HCMV yield was quantified (Fig. 1A). Nuclear receptor-specific shRNAs and a luciferase-targeting control shRNA were delivered to MRC5 fibroblasts using a lentivirus vector (pLKO) that also expressed a puromycin resistance gene (24). At 24 h posttransduction, fresh medium was applied with puromycin (1 μg/mL) to select for cells containing an shRNA, and drug selection was maintained for a total of 96 h. To monitor transduction efficiency, analysis with a GFP-expressing lentivirus performed in parallel demonstrated that >99% of transduced puromycin-selected cells expressed the marker.

Fig. 1.

Nuclear receptor shRNA screen. (A) Design and timeline of the shRNA screen. MRC-5 cells were transduced with lentiviruses expressing an shRNA plus a puromycin (Puro)-resistance gene; cells were selected with Puro (1 μg/mL) and infected with ADinUL99-GFP (0.3 TCID₅₀ per cell). Infectious virus in culture medium was quantified at 96 hpi. (B) Robust z-score values for individual shRNAs. Dotted lines indicate the cutoff value of ±2.3. (C) Frequency distribution of robust z scores from the screen.

At 6 d after the transduction process was initiated, fibroblasts were infected at a multiplicity of 0.3 50% tissue culture infective dose (TCID₅₀) per cell in the absence of puromycin with a phenotypically wild-type derivative of HCMV, ADinUL99-GFP, expressing GFP fused to the pUL99 late viral protein. At 96 h postinfection (hpi) culture supernatants were assayed for infectious virus, and the yield was determined relative to that obtained from luciferase shRNA-expressing cells. The dataset (Dataset S1) was then normalized by robust z score, which compares each shRNA relative to the distribution of the entire screen and is insensitive to outliers (25), and plotted in ascending order (Fig. 1B). A negative z score indicates that knockdown of a nuclear receptor reduced HCMV yield, and a positive score indicates that knockdown increased yield. When the z-score frequency was plotted (Fig. 1C), a near-normal distribution was observed. The threshold for designating a nuclear receptor as a hit was set as those targets for which a single shRNA achieved an absolute robust z score ≥2.3, which is 10-fold greater than the robust z score of control shRNA (0.23). This ranking identified eight nuclear receptors that might support HCMV replication in cultured fibroblasts (negative z scores) (Table 1).

Table 1.

Nuclear receptors identified as hits in the shRNA screen

Nuclear receptor name	Gene ID	Z score
NR1I2 (PXR, pregnane X receptor)	8856	−8.74
ESR2 (ERβ, estrogen receptor β)	2100	−6.88
ESRRG (ERRγ, estrogen-related receptor γ)	2104	−5.03
HNF4G (HNF4γ, hepatocyte nuclear factor 4 γ)	3174	−3.70
NR2C1 (TR2)	7181	−3.40
ESRRA (ERRα, estrogen-related receptor α)	2101	−3.11
RXRG (RXRγ, retinoid x receptor γ)	6258	−2.69
PPARG (peroxisome proliferator-activated receptor γ)	5468	−2.45

Genes for which one shRNA generated an absolute robust z score ≥ 2.3 were identified as hits in the screen. A complete list of robust z scores is provided in Dataset S1.

Because each score represents results obtained for a single shRNA, it is important to recognize that hits in the screen do not unambiguously identify nuclear receptors that modulate HCMV yield; rather, they are candidates for further validation. Deeper coverage of the screen using additional shRNA sequences would likely result in multiple hits targeting the same gene and help to rule out false positives. In addition, the screen does not rule out a role for nuclear receptors that do not meet the cut off. However, PPARγ, which is already known for its positive role during HCMV infection (17), was classified as a hit in the screen, suggesting our methodology was robust and may identify other host factors that play a role during viral replication.

Among the hits, we focused initially on ERRα, a constitutively active orphan nuclear receptor with no known endogenous ligand or estrogen binding activity (26). Our choice of ERRα was based on its role in cellular metabolism. ERRα is a transcriptional regulator of cellular metabolic networks (26), and it can contribute to a cancer-like metabolic profile (27). HCMV infection also modulates glycolysis, the TCA cycle and fatty acid metabolism (18, 28), and it induces a cancer cell-like metabolic signature (6, 18). Consequently, it seemed possible that ERRα activity might underlie aspects of HCMV metabolic manipulations.

Confirmation of a Role for ERRα During HCMV Infection.

To validate a role for ERRα during HCMV replication, the shRNA sequence used to generate the hit in the screen (shRNA1) plus an additional shRNA construct targeting ERRα (shRNA2) were evaluated for their ability to reduce ERRα levels at various times after infection with the AD169 strain of HCMV (ADwt) (Fig. 2A, Left). Compared with control shRNA, both ERRα-targeting shRNAs reduced the level of ERRα, and shRNA1 almost completely abrogated expression of the nuclear receptor. Whereas shRNA1 did not exhibit cytotoxicity in uninfected fibroblasts, as assayed by trypan blue exclusion assay (>95% viable), shRNA2 displayed limited toxicity (85–90% viable). ERRα knockdown reduced the production of ADwt progeny (Fig. 2A, Right). Fibroblasts transduced with lentivirus expressing shRNA1 or -2 produced ∼300-fold and 60-fold less virus, respectively, compared with control shRNA-expressing cells at 96 hpi. The magnitude of reduction in virus yield correlated with the extent of ERRα suppression.

Fig. 2.

ERRα supports the production of HCMV progeny and is induced posttranscriptionally. (A) shRNAs targeting ERRα reduced the yield of HCMV. Fibroblasts were transduced with control (C) or ERRα-targeting shRNA (1 or 2) and subsequently infected with HCMV AD169wt (3 TCID₅₀ per cell). ERRα was monitored by immunoblot assay at the indicated times (Left), and infectivity (infectious units, IU) in the supernatant was titered at 96 hpi (Right). (B) ERRα RNA is induced to a limited extent by infection. Fibroblasts were infected with AD169wt (3 TCID₅₀ per cell) or mock infected, and RNA samples were harvested for qRT-PCR analysis at the indicated times. AU, arbitrary units. (C) ERRα protein is substantially induced during infection. Fibroblasts were infected or mock infected as in B, and protein samples were analyzed by immunoblot assay. Immediate-early 1 protein (IE1) and pUL99 are HCMV proteins. (D) Specificity of higher molecular weight band (***) of ERRα. Fibroblasts were infected as in B, and the immunogenic peptide used to generate the antibody (molar ratio of antibody to peptide, 1:150) blocked recognition of the 50- and 76-kDa ERRα isoforms. (E) Viral DNA replication is required for induction of higher molecular weight isoform, but not lower molecular weight isoform, of ERRα. Similar to B, except cells were treated with phosphonoacetic acid (PAA, 100 μg/mL) or ethanol solvent control.

Posttranscriptional Regulation of ERRα Isoforms Following HCMV Infection.

To evaluate the effect of HCMV on ERRα mRNA and protein expression, time course experiments were carried out following ADwt infection or mock infection of fibroblasts. qRT-PCR analysis showed that the amount of ERRα mRNA did not change at 24 and 48 hpi and was elevated to a limited extent (∼40%) relative to mock-infected cells at 72 hpi (Fig. 2B). However, the amount of ERRα protein increased beginning at 48 hpi and was markedly elevated at 72 hpi (Fig. 2C), showing that the nuclear receptor is substantially up-regulated by a posttranscriptional mechanism following HCMV infection.

When ERRα was assayed by Western blot, two species were evident: a band migrating at the known 50-kDa size of ERRα plus a 76-kDa doublet moiety that was present only at late times after infection (Fig. 2C). To ensure that the antibody did not spuriously recognize a nonspecific band, samples were immunoblotted with anti-ERRα antibody in the presence or absence of a 150-fold excess of the peptide used as immunogen to produce the antibody. The addition of the peptide abolished reactivity with both the 50-kDa and 76-kDa bands (Fig. 2D), arguing that the larger species is a modified isoform of ERRα.

The accumulation of ERRα protein after 24 hpi temporally coincides with the onset of viral DNA accumulation, which is required for late viral protein synthesis (29). To examine if viral DNA synthesis and subsequent late protein production are required for the posttranscriptional up-regulation of ERRα, cultures were treated with phosphonoacetic acid (PAA), an inhibitor of the viral DNA polymerase (30), or solvent alone as a control. As expected, PAA prevented accumulation of a late viral protein, pUL99, without affecting the level of the immediate-early IE1 protein (Fig. 2E). The level of the 50-kDa ERRα isoform did not change, but accumulation of the 76-kDa isoform was dramatically diminished in response to the drug. Thus, the two ERRα isoforms exhibit differential dependence on viral DNA synthesis and the production of late viral proteins.

Pharmacological Inhibition of ERRα Is Detrimental to HCMV Replication.

Although ERRα is an orphan nuclear receptor with no known endogenous ligand, the drug XCT790 acts as an inverse agonist to counteract the constitutive activity of ERRα (31). To complement the analysis using shRNA knockdown of ERRα, the effect of XCT790 on ADwt production was investigated. In addition to inhibiting the transcriptional activity of ERRα, XCT790 causes its proteasomal degradation (32). Treatment with the compound reduced the protein levels of both ERRα species, and the addition of proteasome inhibitor MG-132 prevented the degradation (Fig. 3A, Left). The compound induced a dose-dependent reduction in virus yield, resulting in a 25-fold decrease at 96 hpi in the presence of 2.5 µM of XCT790 (Fig. 3A, Right).

Fig. 3.

Pharmacological inhibition of ERRα reduces the yield of HCMV. (A) XCT790 reduces the levels of ERRα proteins within infected cells (Left). Fibroblasts were infected with AD169wt (3 TCID₅₀ per cell), and indicated samples were treated with MG-132 (5 μM) at 43 hpi, before harvesting at 49 hpi for immunoblot analysis. XCT790 (2.5 μM) reduced the yield of infectious virus (IU, infectious units) at 96 hpi (Right). Control cultures that did not receive drug were treated with the solvent, DMSO. (B) Genetic and pharmacological inhibition of ERRα reduced the yield of the clinical HCMV isolate, TB40/E-GFP, in epithelial cells. ARPE-19 cells were transduced with control or ERRα shRNA-1, infected with TB40/E-GFP (0.1 TCID₅₀ per cell), and the number of immediate-early 1 protein (IE1)-positive nuclei per plaque was quantified by immunofluorescence at 8.5 dpi. For drug treatment, ARPE-19 cells were similarly infected, treated with indicated concentrations of XCT790 or DMSO, and immunostained for IE1-positive nuclei at 8.5 dpi. Three biological replicates were performed. Error bars represent ±1 SD.

ADwt is a laboratory strain of HCMV that is restricted to replication in fibroblasts. To determine the importance of ERRα during infection of a different cell type with a clinical isolate of HCMV, TB40/E-GFP (28, 33), was used to infect retinal pigmented epithelial cells (ARPE-19) (34). TB40/E spreads in a cell-to-cell manner without release of significant quantities of extracellular infectious particles in these cells (35). shRNA knockdown of ERRα (Fig. 3B, Left) and antagonism with XCT790 (Fig. 3B, Right) both led to a marked decrease in plaque size, when assayed by counting the number of IE1-positive cells per visible plaque. Thus, ERRα supports the production of cell-free virus in fibroblasts by a laboratory strain and direct cell-to-cell spread in epithelial cells by a clinical isolate of HCMV.

Localization of ERRα During Viral Replication.

ERRα is normally localized to the nucleus, and unlike many nuclear receptors, it does not require ligand binding in the cytoplasm for translocation to the nucleus (36). To determine whether infection alters its normal localization pattern, ADwt-infected and uninfected fibroblasts were examined by indirect immunofluorescence assay, using the viral IE2 protein as a nuclear marker for infected cells (Fig. 4A). The change in nuclear morphology of infected cells is indicative of dynein-induced nuclear shape changes that are characteristic of HCMV infection (37). From 8 to 48 hpi, ERRα remained in the nucleus. However, at 72 hpi, ERRα was present in both the nucleus and cytoplasm (Fig. 4A), temporally correlating with the appearance of the 76-kDa isoform of the nuclear receptor (Fig. 2 C–E). Nuclear-cytoplasmic fractionation of infected cells at 72 hpi showed that the 76-kDa ERRα isoform is present exclusively in the cytoplasm, whereas the 50-kDa band was present in both fractions (Fig. 4B). Additional bands reacting with the ERRα antibody, not seen in other immunoblots, may be degradation products of the 76-kDa species generated during cell fractionation.

Fig. 4.

A portion of ERRα is relocalized from the nucleus to cytoplasm during viral replication. (A) Immunofluorescence assay. Fibroblasts were infected with AD169wt (0.1 TCID₅₀ per cell) and processed for immunofluorescence assay at indicated time points. The immediate-early 2 protein (IE2) was monitored as a nuclear viral marker. (B) Nuclear-cytoplasmic fractionation of infected cells. Fibroblasts were infected with AD169wt (3 TCID₅₀ per cell), and at 72 hpi, nuclear and cytoplasmic extracts were separated. PARP1 and tubulin are controls for nuclear and cytoplasmic proteins, respectively. (C) Cytoplasmic ERRα does not colocalize with the viral assembly compartment protein, pUL99. Fibroblasts were infected with ADinUL99-GFP (0.1 TCID₅₀ per cell), fixed at 72 hpi, and processed for immunofluorescence assay. (D) ERRα remains nuclear in the absence of viral DNA replication. Fibroblasts were infected with AD169wt (0.5 TCID₅₀ per cell) and treated with phosphonoacetic acid (PAA) or ethanol. Cells were analyzed as before at 72 hpi. ***, modified isoform.

Late during infection, HCMV induces the formation of a cytoplasmic organelle called the assembly compartment (AC) at which numerous viral and cellular proteins accumulate (38), including the HCMV late protein, pUL99. ERRα did not colocalize with pUL99 (Fig. 4C), arguing that it does not preferentially reside in the assembly compartment. The drug PAA, which blocks HCMV DNA replication and the synthesis of late viral proteins, prevented the accumulation of cytoplasmic ERRα (Fig. 4D), consistent with its ability to block accumulation of the 76-kDa isoform (Fig. 2E) and confirming that the large isoform is induced by a late viral function.

ERRα Is Required for the Efficient Accumulation of Viral Proteins and DNA.

To probe further into the role of ERRα during viral replication, a protein accumulation time course experiment was performed comparing ERRα-deficient and control cells (Fig. 5A, Left). Fibroblasts expressing relevant shRNA constructs were infected with ADwt or mock infected. ERRα normally functions by interacting with PGC-1 family coactivators (39). The level of PGC-1α protein increased following infection, and ERRα knockdown had no effect on its accumulation. Infection did not detectably influence the level of PGC-1β. Several virus-coded proteins were also monitored. The accumulation of the immediate-early protein IE1 and an early protein pUL38 was not influenced by ERRα levels. However, the amount of IE2 at 48 and 72 hpi was considerably reduced upon ERRα knockdown, even though at 24 hpi, there was no difference between the knockdown and control cells. Although IE2 is an immediate-early viral protein, it accumulates to an even greater level as the infection proceeds (40). In addition, there was a delay in the expression of the pUL26 and pUL44 early proteins and the late pUL99 protein.

Fig. 5.

Inhibition of ERRα function suppresses viral protein and DNA accumulation. (A) Analysis of viral proteins and DNA following shRNA-mediated knockdown of ERRα. Fibroblasts were transduced with control or ERRα shRNA-1, infected with AD169wt (3 TCID₅₀ per cell) or mock infected, and protein samples were immunoblotted at the indicated times (Left). IE1, IE2, pUL26, pUL38, pUL44, and pUL99 are HCMV proteins. For viral DNA (vDNA) analysis, total DNA from infected cells was extracted and viral DNA was quantified by using qPCR (Right). (B) Analysis of viral proteins and DNA following inhibition of ERRα with XCT790. Fibroblasts were infected with AD169wt (3 TCID₅₀ per cell) and treated with XCT790 (2.5 μM) or DMSO. Protein samples and viral DNA were analyzed as in A. (C) Analysis of viral proteins and DNA following inhibition of ERRα with AM251. Same as in B, except AM251 (4 μM) or DMSO were applied to infected cells. ***, modified isoform. Three biological replicates were performed. Error bars represent ±1 SD.

The decrease in early and late protein expression suggested that there might be a defect in viral DNA synthesis. This turned out to be the case, as ERRα knockdown caused a reduction in the accumulation of viral DNA throughout the course of the infection; at 72 hpi, knockdown cells contained ∼25-fold less viral DNA than control cells (Fig. 5A, Right).

XCT790 treatment affected viral protein and DNA accumulation in a similar manner as observed in the shRNA knockdown experiment, including pUL69, a viral protein expressed with early-late kinetics (41) (Fig. 5B). Finally, a second drug, AM251, which is an endocannabinoid receptor antagonist that displays additional inhibitory activity toward ERRα and causes its proteasomal degradation (42), also generated a similar protein expression profile (Fig. 5C).

To delve further into how ERRα depletion influenced viral gene expression, a PCR array targeting 32 HCMV genes across all temporal expression classes was used (Fig. 6 and Table S1). Similar to the viral protein expression pattern, at 24 hpi there was little difference in viral mRNA accumulation between ERRα and control knockdown cells. Although we detected late transcripts in samples harvested during the early stage of viral replication, they likely resulted from incoming RNA packaged into virions (43). Differences in transcript levels were evident at later times after infection. Most viral RNAs assayed at these times were present at reduced levels in ERRα-deficient conditions; and UL82, UL83, and UL111A RNAs were reduced to the greatest extent, showing approximately 5- to 10-fold reductions at 72 hpi. These results suggest that the lack of ERRα obstructs the normal progression of the HCMV gene expression program at the level of mRNA accumulation, beginning at some point following 24 hpi.

Fig. 6.

Knockdown of ERRα broadly reduces the steady-state levels of viral RNAs. Fibroblasts were transduced with control or ERRα shRNA-1, infected with AD169wt (3 TCID₅₀ per cell), and processed for PCR array targeting viral genes of all temporal classes. Viral genes are categorized as immediate early (IE), early (E), leaky late (LL), late (L), true late (TL), or unknown kinetics (UN). Two biological replicates were performed and each was analyzed in duplicate. Error bars represent ±1 SD.

ERRα Knockdown Reduces the Levels of Multiple RNAs Encoding Proteins Involved in Central Carbon Metabolism.

ERRα is a constitutively active transcription factor that can regulate numerous genes involved in glycolysis and TCA cycle (26, 27, 36). Hence, ERRα may play a similar role during HCMV infection, contributing to metabolic alterations induced by the virus. Targeted PCR arrays were used to determine if glucose and mitochondrial metabolic gene expression is altered in HCMV-infected cells when ERRα expression is repressed (Fig. 7A). Cells expressing ERRα shRNA or a control shRNA were infected with ADwt at a multiplicity of 3 TCID₅₀ per cell and RNAs were analyzed 48 h later. RNAs corresponding to multiple genes that encode for mitochondrial proteins were down-regulated by at least a factor of two following knockdown of ERRα, including metaxin 2 (MTX2), solute carrier family 25 A4 (SLC25A4), and A25 (SLC25A25), and uncoupling protein 1 (Ucp1). Similarly, multiple glycolytic RNAs were down-regulated, such as those encoding enolase 1 (ENO1), triosephosphate isomerase 1 (TPI1), and hexokinase 2 (HK2). Amylo-1,6-glucosidase,4-alpha-glucanotransferase RNA, whose product is required for glycogen breakdown and mobilization of glucose (44), was also down-regulated. The only RNAs in the PCR arrays that were up-regulated at least twofold upon ERRα knockdown encoded translocase of outer mitochondrial membrane 40 (TOMM40) and Bcl2-binding component 3 (BBC3).

Fig. 7.

ERRα knockdown reduces expression of multiple RNAs encoding metabolic enzymes in infected cells. (A) qRT-PCR analysis of arrays that interrogate RNAs related to glucose metabolism and mitochondrial function. ERRα shRNA or control shRNA expressing fibroblasts were infected with AD169wt (3 TCID₅₀ per cell) and processed 48 h later by using qPCR arrays. Numbers in parentheses indicate fold reduction (green) or induction (red) in ERRα knockdown cells. Dotted lines represent twofold changes. (B) Confirmation of PCR array hits. Fibroblasts were transduced and infected as in A, but RNA samples were harvested at 24, 48, and 72 hpi. Full gene names are provided in Table S2. Values were normalized to cellular PPIA RNA. Four biological replicates were performed. Error bars represent ±1 SD.

PCR array hits were validated by performing qRT-PCR analysis on RNA samples prepared at various times after infection (Fig. 7B). All of the above hits were confirmed, except for TOMM40, which showed a slight reduction in expression instead of up-regulation as indicated by the PCR array. ATP-citrate lyase (ACLY) showed only a minor reduction upon ERRα knockdown, consistent with the failure to detect a change in level of this RNA in the PCR array. ERRα has been reported to regulate the cell-surface epidermal growth factor receptor (EGFR) in certain cell lines (42), but ERRα knockdown had no effect on EGFR during HCMV infection (Fig. 7B). In sum, our data demonstrate that ERRα is required in infected cells for maximal expression of multiple RNAs encoding proteins directly involved in central carbon and mitochondrial metabolism.

Steady-State Levels of Central Carbon Metabolites and Nucleotides Within Infected Cells Are Reduced upon ERRα Knockdown.

The decrease in expression of multiple metabolic genes following knockdown of ERRα suggested that it might contribute to the metabolic perturbations accompanying HCMV infection. To investigate this possibility, a liquid chromatography-mass spectrometry (LC-MS)-based metabolomics experiment was performed. Cells expressing ERRα shRNA or control shRNA were mock infected or infected with ADwt at a multiplicity of 3 TCID₅₀ per cell. In addition, HCMV-infected or mock-infected cells were treated with XCT790 or DMSO. At 48 hpi, samples were harvested and analyzed.

Increased levels of glycolytic and TCA cycle intermediates were evident when infected cells were compared with mock-infected cells (Fig. 8A, columns 1 and 3), as has been observed previously (28, 45). When the levels of metabolites in infected cells expressing ERRα shRNA or treated with XCT790 were compared with infected cells in which the nuclear receptor was not knocked down or pharmacologically inhibited (Fig. 8A, columns 2 and 4), many of the HCMV-mediated increases in metabolite levels were blunted, i.e., the profiles for many metabolites following infection versus ERRα inhibition were negatively correlated (Fig. 8B). Importantly, as the glucose metabolism targeting PCR array data predicted (Fig. 7), we observed a reduction in multiple glycolytic metabolites, including hexose phosphate, 2,3-bisphosphoglycerate, fructose-1,6-bisphosphate, dihydroxyacetone phosphate, and phosphoenolpyruvate. Further, most ribonucleoside mono-, di-, and triphosphates as well as deoxyribonucleoside mono-, di-, and triphosphates assayed were present at reduced levels within infected cells following ERRα inhibition.

Fig. 8.

ERRα is required for HMCV-driven modulation of metabolome. (A) ERRα knockdown or inhibition alters the metabolic profile of HCMV-infected cells. Fibroblasts expressing ERRα shRNA-1 or control shRNA were infected with AD169wt (3 TCID₅₀ per cell) or mock infected. For pharmacological inhibition, cultures were treated with XCT790 or DMSO at 2 hpi. Cells were harvested and metabolites analyzed at 48 hpi. Column 1: ADwt/Mock (shRNAs) = [ADwt + control shRNA (shRNAC)]/[mock + shRNAC]; column 2: ADwt + shRNA1/C = [ADwt + ERRα-specific shRNA (shRNA1)]/[ADwt + shRNAC]; column 3: ADwt/mock (DMSO) = [ADwt + DMSO]/[mock + DMSO]; column 4: ADwt + XC790/D = [ADwt + XC790]/[ADwt + DMSO]. (B) Negative correlation between metabolite levels in fibroblasts treated with ERRα-specific shRNA (Left, R = −0.379, P < 0.001) or an ERRα antagonist, XCT790 (Right, R = −0.493, P < 0.001) versus ADwt-infected fibroblasts. Each point represents a metabolite shown in A.

The congruent results from genetic and pharmacological inhibition indicate that ERRα supports aspects of the metabolic induction that occurs during viral replication.

Discussion

The human genome encodes 48 functional nuclear receptors (10), which sense a multitude of ligands and respond to specific extra- or intracellular stimuli (46, 47). Several nuclear receptors scored in an shRNA screen for effects on the yield of HCMV (Fig. 1B and Dataset S1), some with potential to inhibit and others that might support efficient viral growth. Our screen produced several candidates that require further validation; however, we validated one hit, ERRα, by confirming its effect on HCMV protein expression and yield following treatment with two independent shRNAs as well as two different pharmacological inhibitors (Figs. 2, 3, and 5).

ERRα protein is induced substantially by HCMV infection. The change is posttranscriptional, because ERRα RNA levels increase to only a limited extent (Fig. 2B). Posttranscriptional induction of ERRα has also been shown to occur upon T-cell activation (48). ERRα from mock-infected cells migrates as a well-characterized 50-kDa species, but much of the increase following infection is due to the accumulation of a 76-kDa moiety (Figs. 2, 3, 4, and 5). Three lines of evidence argue that the large species is a bona fide ERRα isoform. First, the peptide that was used to generate the antibody blocked detection of the large species in a Western blot assay (Fig. 2D). Second, shRNA knockdown reduced the level of the high molecular weight isoform (Fig. 5A). Third, both XCT790 and AM251 caused degradation of the 76-kDa isoform (Fig. 5 B and C). As ERRα is known to be SUMO modified (49), the larger species may represent SUMOylated ERRα, although the shift in apparent molecular weight (∼26 kDa) of the isoforms is greater than the 15–17 kDa predicted for this modification (50). Perhaps ERRα undergoes multiple posttranslational modifications or a splice variant is produced in HCMV-infected cells. Accumulation of the 76-kDa variant required viral DNA and late protein synthesis (Fig. 2E); and, whereas the 50-kDa ERRα species localized to both nucleus and cytoplasm during the late phase of HCMV infection, the 76-kDa isoform was restricted to the cytoplasm of infected cells (Fig. 4B). To the best of our knowledge, granular cytoplasmic staining of endogenous ERRα has been shown previously only in a study of poorly differentiated breast cancer tissue (51). Alternatively, the granular ERRα in the cytoplasm might represent vesicular structures and reflect localization to the late endocytic pathway. Because ERRα-dependent transcriptional activity is required for viral replication, the host cell may sequester ERRα in the cytoplasm in an attempt to limit infection. Even though nuclear receptors are most notable for altering gene expression by directly binding to target DNA sequences, additional functions have been described (52, 53). As such, it is also conceivable that cytoplasmic ERRα binds host and/or viral proteins to regulate aspects of infection.

How does ERRα support HCMV replication? The steady-state levels of several immediate-early viral proteins (IE1 and IE2) were not affected at 24 hpi by the loss of ERRα activity, but the accumulation of two early (pUL26 and pUL44) and a late viral protein (pUL99) was delayed and their levels were reduced (Fig. 5). IE2 RNA (Fig. 6) and protein levels (Fig. 5) were reduced at 48 and 72 hpi, and this could in turn reduce the levels of other viral gene products, because IE2 is a potent transcriptional activator that drives the cascade of HCMV gene expression (54). The reduced level of IE2 following knockdown or pharmacological inhibition of ERRα argues that the nuclear receptor is required for its optimal expression during the late phase of the replication cycle.

How does ERRα support late IE2 expression? It is possible that the nuclear receptor activates transcription of viral genes, including the major immediate-early promoter that is responsible for the synthesis of mRNA encoding IE2. The timing of ERRα induction following infection (Fig. 2C) fits well with the reduction in IE2 levels at 48 and 72 hpi in the absence of ERRα function (Fig. 5). ERRα could act directly by binding to the viral major immediate-early control region, although inspection of the sequence does not reveal a clear match to an estrogen-related receptor response element. Alternatively, the nuclear receptor might not bind to the viral genome, but could act indirectly by modulating cellular transcription factors that impact viral gene expression. Parenthetically, it is noteworthy that an in silico search for estrogen-related receptor response elements in the viral genome identified a putative binding site that conforms to a sequence known to confer ERRα responsiveness (TCAAGGTCA) upstream of the UL21A transcription start site. However, qRT-PCR analysis showed that ERRα had no effect on the level of UL21A RNA (Fig. 7B).

ERRα supports aspects of metabolic changes induced by HCMV infection, which may subsequently impact IE2 expression. Although knockdown of ERRα did not completely reverse the metabolic phenotype induced by viral infection, the levels of many metabolites, including nucleoside mono-, di-, and triphosphates, were reduced (Fig. 8, columns 2 and 4), most likely resulting in a suboptimal cellular environment for viral RNA and DNA synthesis, consistent with the observed reductions in accumulation of viral DNA and RNAs (Figs. 5 A and B and 6). Late accumulation of IE2 RNA (and other viral transcripts) requires viral DNA replication (40), and this suggests that IE2 accumulation during the late phase of the HCMV replication cycle is obstructed when ERRα activity is reduced. This may prevent a feed-forward mechanism where viral DNA replication leads to even higher IE2 levels, which would then further stimulate expression of viral genes. Thus, insufficient ERRα activity likely reduces viral RNA and protein expression indirectly, at least in part by inhibiting viral DNA replication.

Besides its impact on the synthesis of nucleic acids within HCMV-infected cells by supporting the accumulation of nucleoside and deoxynucleoside mono-, di-, and triphosphates, ERRα is generally required for cells and tissues with high energy demand (55), and it is likely to impact viral replication and spread through additional effects on metabolism. ERRα promotes a Warburg-like effect by inducing a metabolic shift from oxidative phosphorylation to glycolysis (27), and it has also been shown to maintain the TCA cycle by inducing anaplerotic replenishment of TCA cycle metabolites (27). These changes are broadly similar to alterations observed following HCMV infection of fibroblasts (6, 18, 28, 45), and knockdown of ERRα blunts many of these alterations following infection (Fig. 8), confirming a role for the nuclear receptor in the establishment of the viral metabolic program.

How does ERRα contribute to the metabolic perturbations induced by HCMV infection? PCR array analysis targeting a subset of genes encoding or regulating metabolic enzymes identified several RNAs whose levels were decreased upon ERRα knockdown (Fig. 7). Hexokinase 2 (HK2) and adenine nucleotide translocase 1 (ANT1) were among the hits, and both are known to be regulated by ERRα (36, 56, 57). Both enzymes support core metabolism: HK2 phosphorylates glucose (58) and ANT1 exchanges mitochondrial ATP for cytosolic ADP (59).

Additional candidate hits were identified in the nuclear receptor screen that might influence HCMV replication (Table 1). Of note, knockdown of pregnane X receptor (PXR) reduced virus production. This nuclear receptor is involved in the detoxification of xeno- and endobiotics, including cholesterol-derived compounds (60). HCMV infection increases cholesterol synthesis (61) and intracellular cholesterol levels regulate the infectivity of progeny virus (62), so PXR may be required to prevent accumulation of toxic cholesterol byproducts.

In conclusion, ERRα contributes importantly to the altered metabolic program induced by HCMV infection. When ERRα function is ablated, virus-induced metabolic changes are blunted, the cascade of viral gene expression is perturbed, and the production of infectious progeny is reduced. Our finding that an evolutionarily ancient orphan nuclear receptor is required for both viral gene expression and metabolic perturbation suggests that the coopting of this family of transcription factors may be a common mechanism of viruses to replicate in their respective host niches.

Methods

Cells, Viruses, and Reagents.

HFFs (human foreskin fibroblasts) and MRC-5 cells (human embryonic lung fibroblasts) were cultured in DMEM supplemented with 10% FBS and penicillin/streptomycin. For all experiments except for the initial shRNA screen, cells were maintained in DMEM with 2% (vol/vol) dialyzed FBS and penicillin/streptomycin.

A wild-type isolate of the HCMV AD169 strain (ADwt) was produced from a BAC clone by transfection of BAC plasmid (pAD/Cre) and pp71 expression plasmid into HFFs to generate viral progeny of wild-type growth characteristics (63). ADinUL99-GFP, in which the tegument protein pUL99 is tagged with GFP and exhibits wild-type growth kinetics, has been described (64). TB40/E-GFP (28), derived from an endotheliotropic clinical strain, TB40/E (65), contains a GFP expression cassette inserted between US34 and TRS1. All virus stocks were partially purified by centrifugation through a sorbitol cushion (20% sorbitol, 50 mM Tris⋅HCl,1 mM MgCl2, pH 7.2), concentrated 10-fold, and resuspended in DMEM supplemented with 3% (wt/vol) BSA. Viral stock titers were determined by TCID₅₀ on fibroblasts and stored at −80 °C until use. Infections were performed by treating cells with viral inoculum for 90 min, followed by removal of inoculum and washing with serum-free DMEM.

Cell-free viral titers were determined by assaying for IE1-positive cells on reporter fibroblast plates. Infectious supernatant was applied to uninfected fibroblasts, and 24–36 h later, cells were fixed with 4% (wt/vol) paraformaldehyde and immunostained for IE1 (19, 66). For assaying cell-associated viral spread, infected ARPE-19 cells were directly fixed with 4% (wt/vol) paraformaldehyde at 8.5 dpi and immunostained for IE1. The number of IE1-positive cells per plaque were counted as described previously (67).

Drugs were dissolved in DMSO (XCT790, Tocris; AM251, Cayman Chemicals; MG-132, Cayman Chemicals) or ethanol (PAA, Sigma-Aldrich), stored at −20 degrees, and thawed once for application to cells.

Nuclear Receptor shRNA Screen.

Lentiviral particles (pLKO.1) that target individual nuclear receptor genes were purchased from Sigma-Aldrich (Dataset S1). Lentivirus expressing luciferase-specific shRNA and GFP-expressing lentiviral particles were used as negative controls and to monitor the efficiency of transduction, respectively. A total of 10,000 MRC-5 fibroblasts were plated into each well (96-well plate); 18–20 h later, lentivirus inoculum (six transduction units per cell in 50 μL DMEM containing 10% (vol/vol) dialyzed FBS and 4 μg/mL polybrene) was added to each well, the 96-well plate was centrifuged at 650 × g for 45 min at 25 °C, and placed in a 37 °C incubator for 24 h, after which the inoculum was removed and replaced with fresh DMEM containing 10% (vol/vol) dialyzed FBS. After an additional 24 h, medium containing puromycin (1 μg/mL) was added for 96 h (the medium was changed and fresh drug added after 48 h). When the puromycin selection was completed, the transduced MRC-5 cells were infected with ADinUL99-GFP at a multiplicity of 0.3 TCID₅₀ per cell for 90 min and fresh medium was applied without puromycin. Medium was changed at 48 hpi, and supernatants were assayed for infectious virus at 96 hpi by using the Operetta High-Content Imaging system (Perkin-Elmer) to enumerate IE1-expressing nuclei (66). Values were normalized to the negative control shRNA that was set to 1, log₂ transformed, and robust z scores were calculated (25).

Protein Analysis.

To analyze proteins by immunoblot assay, cells were harvested in buffer with protease inhibitor (Roche, EDTA-free tablets) and phosphatase inhibitors (1 mM sodium fluoride; 1 mM β-glycerophosphate; 1 mM sodium orthovanadate; 10 mM sodium pyrophosphate), and analyzed as described previously (68). Nuclear-cytoplasmic extraction was performed as described previously (69), except 350,000 cells were harvested. Primary antibodies to the following HCMV proteins were used: IE1 (clone 1B12) (70), IE2 (clone 3A9) (71), pUL26 (clone 7H1-5) (68), pUL38 (clone 8D6) (72), pUL44 (clone 10D8; Virusys), pUL83 (clone 8F5) (73), and pUL99 (clone 10B4-29) (74). Cellular proteins were assayed using antibodies to: ERRα (Novus), PGC-1α (Sigma-Aldrich), PGC-1β (Active Motif), PARP-1 (Cell Signaling), β-actin (Abcam), and tubulin (Sigma-Aldrich). Secondary antibodies were HRP-conjugated goat anti-mouse and goat anti-rabbit IgG (Jackson ImmunoResearch).

Protein localization was analyzed by indirect immunofluorescence as described (35). Primary antibodies to ERRα and IE2 are described above; secondary antibodies were goat anti-mouse or goat anti-rabbit conjugated with Alexa Fluor-488 or -555 (Life Technologies), and nuclei were counterstained with DAPI. Images were obtained by using a Nikon A1 confocal microscope.

Nucleic Acid Quantification.

To quantify RNA levels using human mitochondria (PAHS-087ZA) and human glucose metabolism (PAHS-006ZA) PCR arrays (SA Biosciences), total RNA was extracted using a Direct-zol RNA kit (Zymo Research). Genomic DNA elimination and reverse transcription were done according to the manufacturer’s PCR array protocol. Triplicate biological samples were pooled subsequent to reverse transcription for each experimental condition, and PCR arrays were processed according to the manufacturer’s protocol on an ABI PRISM 7900HT Sequence Detection System. Data analyses were performed using the manufacturer’s software (www.sabiosciences.com/pcrarraydataanalysis.php). Results were normalized to GAPDH.

To confirm results of the PCR arrays for selected cellular RNAs, 200 ng of total RNA was reverse transcribed using qScript cDNA SuperMix (Quanta Biosciences) according to the manufacturer’s protocol. Real-time PCR was performed with SYBR Green PCR Mastermix (Applied Biosystems) using indicated primers (Table S2) and values were normalized to cellular PPIA mRNA.

To quantify viral RNAs by using an HCMV qPCR array, primers were designed against the annotated AD169 strain (GenBank: FJ527563) using QuantPrime (75) and checked by BLAST algorithm (blast.ncbi.nlm.nih.gov/Blast.cgi) to assure appropriateness for AD169-BAC (pAD/Cre, GenBank: AC146999) used in this study. Five cellular reference genes were included as controls (PPIA, PGK1, ACTB, HPRT1, and B2M). Primer sets (Table S2) were validated by: (i) standard curve analysis using serial dilutions of purified AD169 BAC DNA to ensure that the PCR efficiency for all primers is between 90% and 110%; (ii) melting curve analysis to ensure single peak of amplified product with expected melting temperature; and (iii) agarose gel electrophoresis to ensure a single band of amplified product with expected size. The same quality control criteria applied to the primers of five cellular reference genes, except using serial dilutions of cDNA sample for the standard curve analysis. For analysis, cells were lysed in TRIzol and RNA was extracted using Direct-zol RNA kit (Zymo Research). A total of 200 ng of total RNA was reverse transcribed with qScript cDNA SuperMix (Quanta Biosciences) in a volume of 20 μL and diluted to a volume of 200 μL. The qPCR reaction included 0.7 μL of 200-fold diluted cDNA sample from the RT reaction, 400 nM of each forward and reverse primer, and SYBR Green master mix in a final volume of 10 μL. The geometric mean of cycle threshold (Ct) values (Table S1) of the five reference genes (PPIA, PGK1, ACTB, HPRT1, and B2M) was used to normalize the viral gene expression at different time points postinfection in control or shRNA knockdown cells.

To quantify accumulation of the viral genome by qPCR, cells were washed once in PBS, and genomic DNA was harvested using the Quick-gDNA kit (Zymo Research) according to the manufacturer’s protocol. Viral genome and cellular genome were quantitated by qPCR using primers recognizing UL123 and β-actin, respectively. Standard curves were generated using 10-fold dilutions of BFX-actin (76).

Metabolomic Analysis.

At 48 h after mock or ADwt infection, medium was removed and cellular metabolism was immediately quenched with 80:20 methanol/water (vol/vol) solution at −80 °C (45). Extracts were scraped into a conical tube held on dry ice and centrifuged at 3,200 × g for 5 min. Extracts were dried under nitrogen gas at room temperature, resuspended in water (HPLC grade), and centrifuged at 16,000 × g for 5 min. Metabolites were identified and measured by an Orbitrap mass spectrometer (Thermo Scientific Exactive) (77), and Metabolomic Analysis and Visualization Engine (MAVEN) software was used for the quantitative analysis of metabolite levels (66, 78, 79). Metabolite identities were determined by exact mass and retention time match to authenticated standards. Isomeric metabolites may or may not separate chromatographically. Reported values reflect the sum of the stated metabolite and any nonseparated isomers. Values acquired from experimental conditions were divided by values from the indicated control conditions to obtain ratios and log₂ transformed; a value of zero indicates no change in metabolite levels.

Statistical Analysis.

Values are expressed as mean ± SD, and significance was determined by Student t test. For metabolic data, correlation coefficients were used to determine the significance of similarities between metabolic profiles.