Ankyrin Repeat Proteins of Orf Virus Influence the Cellular Hypoxia Response Pathway

ABSTRACT

Hypoxia-inducible factor (HIF) is a transcriptional activator with a central role in regulating cellular responses to hypoxia. It is also emerging as a major target for viral manipulation of the cellular environment. Under normoxic conditions, HIF is tightly suppressed by the activity of oxygen-dependent prolyl and asparaginyl hydroxylases. The asparaginyl hydroxylase active against HIF, factor inhibiting HIF (FIH), has also been shown to hydroxylate some ankyrin repeat (ANK) proteins. Using bioinformatic analysis, we identified the five ANK proteins of the parapoxvirus orf virus (ORFV) as potential substrates of FIH. Consistent with this prediction, coimmunoprecipitation of FIH was detected with each of the ORFV ANK proteins, and for one representative ORFV ANK protein, the interaction was shown to be dependent on the ANK domain. Immunofluorescence studies revealed colocalization of FIH and the viral ANK proteins. In addition, mass spectrometry confirmed that three of the five ORFV ANK proteins are efficiently hydroxylated by FIH in vitro. While FIH levels were unaffected by ORFV infection, transient expression of each of the ORFV ANK proteins resulted in derepression of HIF-1α activity in reporter gene assays. Furthermore, ORFV-infected cells showed upregulated HIF target gene expression. Our data suggest that sequestration of FIH by ORFV ANK proteins leads to derepression of HIF activity. These findings reveal a previously unknown mechanism of viral activation of HIF that may extend to other members of the poxvirus family.

IMPORTANCE The protein-protein binding motif formed from multiple repeats of the ankyrin motif is common among chordopoxviruses. However, information on the roles of these poxviral ankyrin repeat (ANK) proteins remains limited. Our data indicate that the parapoxvirus orf virus (ORFV) is able to upregulate hypoxia-inducible factor (HIF) target gene expression. This response is mediated by the viral ANK proteins, which sequester the HIF regulator FIH (factor inhibiting HIF). This is the first demonstration of any viral protein interacting directly with FIH. Our data reveal a new mechanism by which viruses reprogram HIF, a master regulator of cellular metabolism, and also show a new role for the ANK family of poxvirus proteins.

INTRODUCTION

Cells must be able to respond rapidly to low oxygen levels in order to survive, and triggering of the heterodimeric transcription factor hypoxia-inducible factor (HIF) is key to the cellular adaptation to hypoxia. Under normoxic conditions, the HIF alpha subunits (HIF-1α and HIF-2α) are tightly suppressed by posttranslational modifications, which are relaxed with decreasing oxygen availability. These posttranslational modifications are carried out by a family of nonheme Fe(II)- and 2-oxoglutarate (2-OG)-dependent dioxygenases that hydroxylate conserved proline or asparagine residues within the HIF-1α proteins (1). The hydroxylation of two prolines within HIF-1α by HIF prolyl hydroxylases (PHDs) leads to the degradation of HIF-1α by the ubiquitin-proteasome system (2). Asparaginyl hydroxylation in the HIF-1α C-terminal transactivation domain (CAD) by the hydroxylase factor inhibiting HIF (FIH) prevents the recruitment of the coactivators p300/CBP, thereby blocking complete assembly of the complex and reducing HIF's transcriptional activity (3, 4). When oxygen levels are low, HIF-1α avoids hydroxylation, is stabilized, translocates to the nucleus, and forms the functional HIF transcription factor by binding its constitutive partner protein ARNT (also known as HIF-1β) and recruiting p300/CBP. This complex binds hypoxia response elements (HRE) and activates a battery of target genes that have various roles, including roles in energy metabolism, angiogenesis, the apoptotic cascade, the NF-κB signaling pathway, and cell cycle regulation (reviewed in reference 5).

Activation of the HIF pathway is also an important target during viral infection. HIF activation by various mechanisms has been identified following infection with hepatitis B and C viruses, Epstein-Barr virus, Kaposi's sarcoma-associated herpesvirus (6, 7), and, more recently, vaccinia virus (VACV) (8). In many cases, the mechanisms and downstream effects of HIF activation are still unclear, although there are clear links between HIF activation and progression of viral infection. In persistent viral infections, including those caused by hepatitis B and C viruses and human T-lymphotrophic virus, downstream targets of the HIF pathway, such as angiogenesis and antiapoptotic programs, play key roles in viral pathogenesis (9). VACV was recently shown to induce hypoxic signaling via inhibition of HIF prolyl hydroxylases, and thereby stabilization of HIF-1α, in normoxia (8).

Although activation of HIF is emerging as a target of a range of viruses, to date, there have been no reports of FIH being subverted during viral infection. While FIH was initially characterized as a protein targeting HIF-1α, it was subsequently shown to hydroxylate numerous ankyrin repeat (ANK)-containing proteins (10). It has been proposed that, in most cases, hydroxylation of ANK proteins by FIH indirectly regulates HIF activity, with the ANK proteins sequestering FIH, resulting in reduced HIF-1α hydroxylation and elevated HIF activity (10).

Several viruses, including members of the Poxviridae, Polydnaviridae, Iridoviridae, Phycodnaviridae, and Mimiviridae families, encode ANK proteins, and many of these proteins are predicted to function as host range-adaptive factors (11–15). However, the mechanisms of action and targets of these ANK proteins remain unknown in most cases. Most chordopoxviruses encode multiple ANK proteins (12, 16), but information on the functions of this largest protein family in chordopoxviruses is only slowly becoming available (reviewed in reference 12). Approximately 80% of poxviral ANK proteins also have an F-box-like motif near the C terminus (17). Cellular F-box proteins act as target recognition subunits in the ubiquitin-proteasome pathway. It has been demonstrated that the five ANK proteins encoded by the parapoxvirus orf virus (ORFV) are able to bind to the cellular SCF1 complex via their F-box-like motif (18), and similar data have been presented for other poxviral ANK proteins (reviewed in reference 12). ORFV causes a highly contagious, eruptive skin disease in sheep and goat populations, and it is readily transmitted to humans (19).

In this study, bioinformatic analysis identified the five ANK proteins of ORFV as potential substrates of FIH. We demonstrate that these viral proteins are able to interact with FIH and that at least three of them are hydroxylated by FIH, within the ANK domains. Importantly, this interaction leads to derepression of the HIF-1α CAD via sequestration of FIH, which correlates with an upregulation of HIF target genes in virus-infected cells.

RESULTS

ORFV ANK proteins interact with endogenous FIH in an ANK domain-dependent manner.

A bioinformatic search using the FIH consensus sequence derived from HIF and ANK substrates (10) identified all five ORFV ANK proteins (ov008, ov123, ov126, ov128, and ov129) as potential substrates of FIH, with each containing one to three predicted sites of hydroxylation (Fig. 1).

FIG 1.

ORFV ANK proteins each contain potential sites of FIH hydroxylation. The amino acid sequences flanking potential FIH hydroxylation sites within ORFV ANK proteins ov008, ov123, ov126, ov128, and ov129 were aligned with a consensus sequence for known FIH hydroxylation sites and with FIH hydroxylation sites of specific human FIH substrates. Key residues that form the consensus sequence for hydroxylation by FIH (10) are shown in bold. π, small, uncharged amino acid; ϕ, hydrophobic amino acid.

To determine first if these ANK proteins can interact with FIH, the coding region of each ORFV ANK protein was subcloned into an expression vector such that the protein would be expressed with an N-terminal Flag tag, and the constructs were transiently transfected into HeLa cells. Immunoprecipitation with anti-Flag antibody and subsequent examination by Western blotting showed efficient coprecipitation between endogenous FIH and ov008, ov123, ov126, and ov128 but only limited coprecipitation with ov129 (Fig. 2A), while control reactions confirmed the specificity of the immunoprecipitations (Fig. 2B). Other ANK substrates have also been shown to form a tight interaction with FIH that is amenable to coprecipitation (20).

FIG 2.

ORFV ANK proteins interact with endogenous FIH in an ANK domain-dependent manner. (A) Immunoprecipitation assays of HeLa cells transiently transfected with the indicated Flag-tagged ORFV ANK protein or the vector-only control (N) and immunoprecipitated (IP) with anti-Flag agarose. Precipitates and 20% of total cell lysates (cell lysate) were analyzed by SDS-PAGE and immunoblotting with anti-Flag and anti-FIH antibodies. (B) Immunoprecipitation assays of HeLa cells transiently transfected with Flag-tagged ov008 and immunoprecipitated with anti-Flag agarose (Flag) or protein G affinity gel (Prt G), as a control. Precipitates were analyzed by SDS-PAGE and immunoblotting with anti-Flag antibody. (C) Schematic representation of Flag-tagged ov008 expression constructs. Predicted molecular masses are shown. (D) Immunoprecipitation assay of HeLa cells transiently transfected with the indicated Flag-tagged ov008 constructs or the vector-only control (N) and immunoprecipitated with anti-Flag antibody. Precipitates and 20% of total cell lysates (cell lysate) were analyzed by immunoblotting with the indicated antibodies. (E) ov008 is able to interact with FIH during ORFV infection. HeLa cells were infected with rORFV-N-Flag-ov008 at an MOI of 5 in the presence or absence of 1 mM DMOG, as indicated. At 8 hpi, Flag-ov008 was immunoprecipitated with anti-Flag antibody. Precipitates and 20% of total cell lysates (cell lysate) were analyzed by immunoblotting with the indicated antibodies. (F) Endogenous FIH colocalizes with ov008. HeLa cells were transiently transfected with Flag-tagged ov008 and examined by confocal microscopy using anti-Flag antibody, anti-FIH antibody, and DAPI (4′,6-diamidino-2-phenylindole). The merged image shows the combined signals from three channels. The results shown are representative of three independent experiments.

To elucidate the region(s) of the ANK proteins responsible for the interaction with FIH, we constructed truncated versions of ov008 as a representative protein. The 516-amino-acid ov008 protein consists of a 358-amino-acid N-terminal ANK domain containing 10 repeats and a 34-amino-acid C-terminal F-box-like domain, separated by a linker region of approximately 124 amino acids. All five of the ORFV ANK proteins and the majority of poxviral ANK proteins have similar domain arrangements (17). Flag-tagged deletion constructs of ov008 lacking either the ANK domain (ov008ΔA) or the C-terminal F-box-like domain (ov008ΔF) (Fig. 2C) were used in coimmunoprecipitation experiments (Fig. 2D). Endogenous FIH was readily observed to coprecipitate with full-length ov008 (57 kDa) or ov008ΔF (52 kDa) but not with ov008ΔA (24 kDa) or a vector-only control (N). This result indicates that the ov008 ANK domain, but not the F-box-like domain, is required for the interaction with FIH.

To examine the interaction in ORFV-infected cells, a recombinant ORFV expressing Flag-tagged ov008 was used. The recombinant ORFV retained the ov008 genomic location and transcriptional regulatory sequences of the parental ORFV strain NZ2 (21), and expression of the N-terminally Flag-tagged ov008 protein was detected as early as 4 h postinfection (hpi) (not shown). Cell lysates prepared at 8 hpi were immunoprecipitated with anti-Flag antibody and examined for coprecipitation of FIH. Dimethyloxalylglycine (DMOG), a 2-oxoglutarate (2-OG) analogue that competitively inhibits many 2-OG-dependent oxygenases, including FIH, was used to simulate hypoxic conditions (4). HeLa cells infected with recombinant ORFV were treated with 1 mM DMOG for 7 h prior to harvest. The Western blot results showed that coprecipitation of FIH and ov008 was detectable in untreated cell lysates but was substantially enhanced in the DMOG-treated sample (Fig. 2E). The treatment appeared to somewhat reduce the expression level of ov008 during infection but clearly increased the interaction between FIH and ov008. Together, these results suggest that ov008 interacts with FIH in ORFV-infected cells. This DMOG-enhanced binding is reminiscent of that of other FIH substrates, such as IκBα (10), and supports ov008 as a likely substrate of FIH.

To further examine the interaction between FIH and the ORFV ANK proteins, Flag-tagged ov008 was expressed in HeLa cells by transient transfection and its subcellular distribution analyzed by fluorescence microscopy. Endogenous FIH generally localizes in the cytoplasm (22), as observed in untransfected cells (Fig. 2F). Flag-tagged ov008 localized in the nucleus (Fig. 2F), and in these cells, FIH also localized predominantly in the nucleus, suggesting that overexpression of ov008 alters the localization of FIH, sequestering it in the nucleus.

Three of the five ORFV ANK proteins are substrates of FIH.

The interaction assays were consistent with the prediction that ORFV ANK proteins might be substrates of FIH. We next examined the ability of FIH to hydroxylate the ANK proteins in vitro. ANK domains of each ORFV ANK protein fused to a maltose binding protein (MBP) tag were expressed in Escherichia coli, purified via amylose affinity chromatography, and used in an in vitro CO₂ capture assay, which indirectly assesses substrate hydroxylation by FIH. This assay measures ¹⁴CO₂ produced by decarboxylation of a radiolabeled cosubstrate, 2-oxoglutarate, by FIH during hydroxylation. The ovine FIH (oFIH) enzyme was used in hydroxylation assays because ORFV primarily infects sheep. Thus, the oFIH ortholog was amplified from Ovis aries cDNA and cloned into a bacterial expression vector. Following sequence analysis, a ClustalW2 multiple-sequence alignment was performed on both human FIH (hFIH) and oFIH, which revealed 97% identity between the two proteins. Thioredoxin (Trx)-6His-tagged oFIH expressed, purified, and analyzed by SDS-PAGE and Coomassie blue staining showed intense bands at approximately 58 kDa, corresponding to recombinant oFIH (Fig. 3A). In vitro CO₂ capture assays revealed that oFIH had an activity similar to that of hFIH with the mouse Notch ankyrin repeat domain (ARD) as the substrate and no activity with the negative control, i.e., the mouse Notch1 RBP-Jκ-associated module (RAM) domain (Fig. 3B). For analysis of the ORFV ARDs, purified Trx-6His-tagged oFIH and positive and negative controls (mouse Notch1 ARD and Notch1 RAM domain) were used for analysis of the purified MBP-tagged ORFV ANK proteins (Fig. 3C). Three of the five ANK proteins (ov008, ov126, and ov129) supported enzymatic turnover by FIH at levels similar to those with the Notch1 ARD, whereas very low to negligible activity was detected for ov123 and ov128 (Fig. 3D). To confirm hydroxylation, the recombinant ov008, ov126, and ov129 proteins were hydroxylated in vitro with purified FIH, and trypsin-digested samples were subjected to nanoflow ultra-high-pressure liquid chromatography–mass spectrometry (nUHPLC-MS) analysis with an LTQ-Orbitrap-Velos Pro machine. nUHPLC-MS/MS unambiguously demonstrated the presence of tryptic peptides corresponding to the hydroxylation by FIH of Asn40 in ov008 (Fig. 3E), Asn285 in ov126, and Asn44 in ov129 (data not shown).

FIG 3.

ORFV ANK proteins are substrates of FIH in vitro. (A) Recombinant Trx-6His-tagged Notch ankyrin repeat domain (ARD), Notch RAM domain, ovine FIH, and human FIH as analyzed by SDS-PAGE with Coomassie blue staining. (B) The Notch ARD and the Notch RAM domain were tested as substrates of FIH in CO₂ capture assays, using a 20 mM concentration of each ANK protein and a saturating amount (1 mM) of recombinant 6His-oFIH for each assay. Data are means ± standard deviations (SD) for experiments performed in triplicate and are representative of 3 independent experiments. (C) Recombinant MBP, MBP-tagged ORFV ARDs, 6His-tagged Notch ARD, Notch RAM domain, and ovine FIH as analyzed by SDS-PAGE with Coomassie blue staining. (D) The ARDs of ORFV ANK proteins were tested as substrates of ovine FIH in CO₂ capture assays, using a 20 mM concentration of each ANK protein and a saturating amount (1 mM) of recombinant 6His-oFIH for each assay. Data are means and SD for experiments performed in triplicate and are representative of 3 independent experiments. **, P < 0.05; ***, P < 0.01. (E) Confirmation of FIH hydroxylation of ORFV ANK proteins via nUHPLC-MS/MS. Linear ion trap MS/MS spectra for hydroxylated precursor ions at m/z 654.3328²⁺, corresponding to residues Ala³¹ to Arg⁴², with modification at Asn⁴⁰, were observed for the nUHPLC fractions of in-gel tryptic digests derived from ov008. Ions corresponding to “y” and “b” fragments are labeled on the spectra and denoted in the corresponding sequence insets. Fragment ions that carried the hydroxylated asparagine are highlighted with asterisks.

The 1st ANK of ov008 is required for binding to FIH.

Using SWISS-MODEL protein structure homology modeling with human ANK-1 (PDB entry 1N11) as the template, a model of the ANK domain of ov008 was generated (Fig. 4A). Previous reports using primary sequence analysis and site-directed mutagenesis indicated that multiple residues of ANK domains are important for FIH recognition and interaction, specifically residues in close proximity to the hydroxylated Asn (10). It has also been shown that nonhydroxylated ankyrin repeats can interact with FIH (23). Using this information, we identified 3 repeats of ov008 that have sequences similar to the FIH consensus sequence and may contribute to FIH binding, namely, the 1st (containing hydroxylated Asn40), 4th, and 8th repeats (Fig. 4B). A series of plasmid-based expression constructs of ov008 were made, in which each of these repeats was removed (ov008Δ1A, ov008Δ4A, and ov008Δ8A) or, in the case of the key asparagine residue in repeat 1 or 4, mutated to an alanine (ov008N40A, ov008N151A, and ov008N40A/N151A).

FIG 4.

The 1st ANK of ov008 is required for interaction with FIH. (A) The ankyrin repeat domain of ov008 was modeled using SWISS-MODEL and the human ankyrin-1 structure (PDB entry 1N11). Each repeat is numbered. (B) Amino acid sequences of the 1st, 4th, and 8th repeats of ov008, which were deleted to generate ov008Δ1A, ov008Δ4A, and ov008Δ8A, respectively. The underlined asparagine (N) residues in the 1st and 4th repeats were replaced with alanine to generate ov008N40A and ov008N151A, respectively, as well as ov008N40AN151A when both asparagines were replaced. (C) Lysates of HeLa cells transiently transfected with the indicated Flag-tagged ORFV ANK proteins, i.e., ov008 (008), ov008Δ1A (Δ1A), ov008Δ4A (Δ4A), ov008Δ8A (Δ8A), or ov008ΔA (ΔA), or with the vector-only control (N) were immunoprecipitated with anti-Flag agarose. Precipitates and 20% of total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-Flag and anti-FIH antibodies. (D) Lysates of HeLa cells transiently transfected with Flag-tagged ORFV ANK construct ov008 (008), ov008N40A (m1), ov008N151A (m4), or ov008N40AN151A (m1m4) or with the vector-only control (N) were immunoprecipitated with anti-Flag agarose. Precipitates and 20% of total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-Flag, anti-FIH, and anti-actin antibodies. (E) Immunoprecipitation assays were conducted as described for panel D, except that cells were treated with 1 mM DMOG for 6 h before harvest. Precipitates and 20% of total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-Flag, anti-FIH, and anti-Skp1 antibodies. The results shown are representative of three independent experiments.

Coimmunoprecipitation assays carried out in HeLa cells with each deletion mutant revealed that the 1st and 4th repeats were required for the interaction with FIH, whereas the 8th repeat was not (Fig. 4C). All the constructs showed similar abilities to interact with Skp1, a known binding partner of the C-terminal F-box-like domain of the ORFV ANK proteins (18), suggesting that each of the modified forms of ov008 adopted the natural conformation and that the failure of ov008Δ1A or ov008Δ4A to pull down FIH was not the result of a generalized disruption of the protein's three-dimensional fold.

Next, the three asparagine point mutants of ov008 were transiently expressed in HeLa cells for immunoprecipitation assays. Both ov008N40A and ov008N40A/N151A showed reduced coprecipitation of FIH, whereas the N151A mutation alone did not affect FIH binding (Fig. 4D). Addition of DMOG during the last 6 h before the cells were harvested had no discernible effect on the interaction between FIH and ov008N40A (Fig. 4E). All three variants were able to pull down Skp1. Together, these observations indicate that both the 1st and 4th ANKs of ov008 are required for interaction with FIH but the 8th repeat does not play a significant role. The Ala substitution studies further support the significance of the 1st repeat and its verified hydroxylation site in the interaction between ov008 and FIH.

Transient transfection of ov008 or ORFV infection does not affect FIH levels.

Our data clearly identify an interaction between ORFV ANK proteins and cellular FIH. In light of the established ability of the ORFV ANK proteins to associate with SCF1 ubiquitin ligase complexes (18), we examined FIH levels in ORFV-infected cells to address the possibility that the viral proteins might direct the degradation of FIH. Lamb testis (LT) or HeLa cells were infected with ORFV NZ2 at a multiplicity of infection (MOI) of 5, and cell lysates were collected at times up to 48 hpi. Western blot analysis of infected cell lysates showed that compared to the actin loading control, FIH levels in both cell types were not affected by ORFV infection at any time up to 12 hpi (Fig. 5A and B). However, in LT cells, the levels of FIH compared to the actin loading control decreased gradually from around 16 hpi. A previous transcriptional analysis showed that ov008 is expressed early in infection (24). The transition to ORFV late gene expression occurs around 10 hpi and is marked by a broad decrease in the synthesis of host cell proteins (25), and it seems likely that the reduced FIH levels seen very late in infection reflect a generalized effect on many cellular proteins rather than a specific effect on FIH, which is supported by a concomitant decrease in actin.

FIG 5.

Neither ORFV early in infection nor its ANK proteins alter levels of FIH. Primary lamb testis (LT) cells (A) or HeLa cells (B) were infected with ORFV at an MOI of 5, and total cell lysates harvested at the indicated hpi were analyzed by SDS-PAGE and immunoblotting with anti-FIH and anti-actin antibodies. (C) HEK 293T cells were transiently transfected with the indicated ORFV ANK proteins, and total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-FIH and anti-paxillin antibodies. Results shown are representative of three independent experiments.

We also examined FIH levels in HEK 293T cells transfected with the full-length ORFV ANK expression constructs, and in no case did transient overexpression of a viral ANK protein result in altered levels of endogenous FIH (Fig. 2A and 5C). Together, these results showed no evidence that ORFV or its ANK proteins direct the degradation of FIH.

ORFV ANK proteins derepress the HIF-1α CAD and induce HIF target genes.

We next employed a HIF-1α C-terminal transactivation domain (CAD) reporter assay to determine if the interaction of the ORFV ANK proteins with FIH might alter HIF-1α activity. Given the very high conservation between human and ovine HIF-1α (96% identity overall, with a higher identity for the CAD), HEK 293T cells were used for these assays. Cells were transiently transfected with each ORFV ANK protein and cotransfected with a plasmid containing the luciferase reporter gene under the control of the GAL4 response element and an expression plasmid encoding the CAD of HIF-1α fused to the GAL4 DNA binding domain (Fig. 6A). FIH-dependent hydroxylation represses the HIF-1α CAD (4), so sequestration of FIH by the ORFV ANK proteins would be predicted to derepress the HIF-1α CAD, as observed with other ANK substrates (26). Consistent with this prediction, both ov123 and ov126 derepressed the HIF-1α CAD to the same extent as that for DMOG treatment. Derepression, although at lower levels, was also seen with ov129, ov008, and ov128.

FIG 6.

HIF activation by ORFV or its ANK proteins. (A) Reporter assays in which HEK 293T cells were transiently transfected with empty pGALDBD or pGALDBDhHIF-1α CAD together with p-G5E1B-Luc, pRLTK-renilla, and the indicated Flag-tagged ORFV ANK protein or empty vector control (empty). Transfected cells were treated for 16 h with normoxia or 1 mM DMOG (positive control), and firefly luciferase activity was quantified relative to Renilla luciferase activity. Data are the means and SD for three independent experiments performed in triplicate. Statistical analysis compared sample activities to the activity seen with p-Apex-empty-HIF-CAD (empty). (B) Analysis of HIF target genes by quantitative PCR (qPCR) was conducted on LT cells infected with ORFV at an MOI of 3 for the indicated times. Total RNA was harvested from infected cells, and equivalent levels of RNA were reverse transcribed and subjected to qPCR. Expression levels were normalized to ACTB and expressed as fold changes relative to the levels in uninfected (phosphate-buffered saline [PBS]-treated) cells. Data are means and SD for experiments performed in triplicate and are representative of 3 independent experiments. (C) CRISPR-Cas9 technology was used to generate two knockout lines in HeLa cells. Total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-FIH and anti-paxillin antibodies. Both lines were found to be FIH deficient and were used in subsequent assays. FIH-overexpressing cell lines were generated via transient transfection with the pEF-hFIH-puro vector. Total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-FIH and anti-paxillin antibodies to visualize increased FIH levels. (D) qPCR analysis of FIH-overexpressing or FIH knockout HeLa cells virally infected as described for panel A. Data are the means for three independent experiments performed in triplicate. Expression levels were normalized to RPLPO and expressed as fold changes relative to the levels in uninfected (PBS-treated) cells. (E) Immunofluorescence of HeLa cells transiently transfected with Flag-tagged ov008. After 24 h, cells were examined by confocal immunofluorescence microscopy using anti-Flag antibody, anti-HIF-1α antibody, and DAPI. Note that cells on the same slide that showed no Flag signal served as an internal, nontransfected control. (F) HeLa cells were infected with rORFV-N-Flagov008. At 8 hpi, cells were examined by confocal immunofluorescence microscopy as described for panel E. (G) Immunofluorescence assays conducted as described for panel F, except that cells were incubated in 3% oxygen after infection. (H) HeLa cells were infected with ORFV, and total cell lysates were analyzed by SDS-PAGE and immunoblotting with anti-HIF-1α and anti-tubulin antibodies to visualize any HIF protein stabilization. Where indicated (+), cells were treated with 1 mM DMOG for 6 h before harvest. **, P < 0.05; ***, P < 0.01.

These findings led us to hypothesize that in ORFV-infected cells the viral ANK proteins inhibit FIH-dependent hydroxylation of residual HIF-1α, leading to expression of HIF target genes. To investigate this, ORFV-infected LT cells were analyzed for HIF target gene expression. Three independent experiments suggested an increase in vascular endothelial growth factor (VEGF) expression at 24 hpi, but this varied considerably between experiments (between 2.5- and 13-fold), and the combined data were not statistically significant (Fig. 6B). Similarly, PHD3 (also known as EGLN3) appeared to be induced in all three experiments, with increases of 2- to 5-fold, but again, the combined data were not statistically significant. In contrast, there was no indication of induction of GLUT1 in any of the experiments. Similar responses were observed in ORFV-infected HeLa cells (Fig. 6D).

To further demonstrate that this HIF target gene induction was dependent on FIH, HeLa cell lines stably overexpressing FIH or FIH knockout lines generated using clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 (Fig. 6C) were analyzed after ORFV infection. Compared to wild-type (WT) cells, the FIH knockout lines showed similar or, in most cases, decreased induction of the HIF target genes VEGF and PHD3 following ORFV infection (Fig. 6D). These lower levels of viral induction are consistent with HIF being derepressed, independently of infection, due to the lack of FIH in the knockout cells (as supported by the observation of elevated VEGF expression in the uninfected knockout cells [data not shown]). The absence of FIH would be expected to negate the ability of the ORFV ANK proteins to sequester FIH. Importantly, the FIH-overexpressing line showed reduced induction of the two HIF target genes (Fig. 6D), as expected given that saturating levels of FIH should overcome sequestration by ORFV ANK proteins.

To demonstrate that the upregulation of HIF target genes was not due to HIF protein stabilization, as observed with some other viruses, including VACV, HIF protein levels were also assessed after ORFV ANK protein expression. Due to its fast turnover in normoxia, the HIF-1α protein is difficult to detect in most cell types under normoxic conditions. Consistent with this, HIF-1α was not detected at significant levels in HeLa cells grown under normoxic conditions and was not detected in ov008-transfected cells (Fig. 6E). To further examine the possible effects of ov008 on HIF-1α levels, HeLa cells were treated with DMOG to stabilize HIF-1α levels and then transfected with ov008. Although DMOG treatment allowed the ready detection of HIF-1α in cell nuclei, ov008 transfection had no detectable effect on HIF-1α stabilization or its distribution.

Although the transient expression of ov008 did not produce evidence of HIF-1α stabilization, it remained important to investigate if ORFV infection triggered HIF stabilization. HeLa cells infected with rORFV-N-Flag-ov008 at an MOI of 0.5 were examined by HIF-1α immunofluorescence microscopy. Again, no HIF-1α stabilization was seen in infected cells (Fig. 6F). The association of ORFV infection and HIF activation might be revealed under conditions mimicking the mildly hypoxic environment of the skin, where the virus naturally replicates. HeLa cells were therefore infected as described above but were incubated in 3% oxygen and without DMOG treatment. In this case, the HIF-1α stabilization occurred as expected, but there was no sign that HIF-1α levels were altered by ORFV infection (Fig. 6G). Furthermore, immunoblotting revealed no evidence of HIF-1α stabilization in HeLa cells infected with ORFV, although the HIF-1α protein was readily detected when HeLa cells were treated with DMOG (Fig. 6H). Together, these data are consistent with ORFV infection leading to a sequestration of FIH by the ORFV ANK proteins, resulting in elevated HIF-1α CAD activity and consequent induction of HIF target genes independently of changes in HIF-1α protein levels.

DISCUSSION

This study was based on the observation that each of the five ORFV ANK proteins contained motifs that might make them substrates of the asparaginyl hydroxylase FIH. Coimmunoprecipitation showed that each of the ORFV ANK proteins was able to bind to endogenous FIH, although ov129 showed substantially less coprecipitation of FIH. This is the first demonstration of viral proteins interacting directly with FIH. Further evidence of interaction was provided by the translocation of FIH from the cytoplasm to the nucleus in the presence of ov008, where it colocalized with ov008. The interaction between ov008, as a representative of the ORFV ANK proteins, and FIH was shown to be dependent on the ANK domain and was not affected by deletion of the F-box-like domain at the C terminus of ov008. Replacement of the FIH-targeted Asn residue in the 1st repeat but not the 4th repeat of ov008 reduced the interaction substantially. Together, these data indicate a dominant role of Asn40 and ANK 1 in the interaction of ov008 with FIH and in ov008's hydroxylation by FIH. Protein structural modeling suggests that some distortion of the structure of the central region of ov008 occurs as a result of deletion of the 4th repeat. It is possible that the 4th repeat has a role in maintaining the overall structure of the ov008 ANK domain but perhaps is only indirectly involved in the interaction with FIH. Similar observations have been made in studies of the VACV ANK protein K1L, which suggests that some repeats in the ANK domain do not provide a substrate interaction function but serve as basic scaffoldings and can be replaced with similar but unrelated synthetic ANKs (27). Despite the previously demonstrated interaction between ORFV ANK proteins and the cellular E3 complex SCF1, we did not detect any effect on the level of FIH, suggesting that the viral proteins do not direct the degradation of FIH.

In ORFV-infected cells, the interaction between ov008 and FIH was readily detected in the presence of the hypoxia mimic DMOG, consistent with the viral protein being a substrate of FIH's enzymatic activity. CO₂ capture assays further demonstrated that the ANK domains of ov008 as well as ov126 and ov129 are substrates of FIH and are subject to hydroxylation. Moreover, LC-MS/MS data directly confirmed that Asn40 in the 1st ANK of ov008 is hydroxylated in the presence of FIH.

In light of the role that FIH plays in the regulation of HIF activity, we addressed the possible involvement of ORFV ANK proteins in the activation of HIF and detected HIF-1α CAD derepression induced by expression of each of the 5 ORFV ANK proteins. In addition, ORFV infection was shown to stimulate expression of a range of HIF-responsive genes. Although prolyl hydroxylation is the predominant mechanism by which HIF is regulated, the role of FIH in fine-tuning the HIF response is nonetheless important. Depletion of endogenous FIH abolishes hydroxylation of the HIF-1α CAD and leads to the induction of several HIF target genes, including VEGF and PHD3, with inconsistent results for GLUT1 (28–30). Similar trends were seen here, with FIH-overexpressing and FIH knockout HeLa lines showing altered induction of the FIH-dependent HIF target genes VEGF and PHD3 but not GLUT1. ORFV lesions show extensive proliferation of vascular endothelial cells and dilation of blood vessels, which have been attributed to the ORFV VEGF-E gene (31) but may also be magnified by the observed induction of host cell VEGF expression. It is also interesting that GLUT1 expression appeared to be independent of FIH and, in contrast to that of VEGF and PHD3, decreased after ORFV infection. This may reflect the metabolic changes commonly induced in cells by viral infection (32, 33) and may be related to the general shutdown of host transcription and translation by poxviruses (25).

Together, these observations are consistent with a model in which the ORFV ANK proteins bind FIH, preventing its hydroxylation of HIF-1α and thereby allowing formation of a transcriptionally active HIF complex. HIF-1α is stabilized by a range of pathogens, including VACV (8), but intriguingly, we did not see evidence of increased levels of HIF-1α in ORFV-infected cells or in cells overexpressing the ORFV ANK proteins. Although HIF-1α levels increase in response to hypoxia, studies have revealed that residual HIF-1α is present under normoxic conditions (34). The activity of this residual HIF-1α is suppressed by FIH such that inhibition of FIH causes increased HIF transcriptional activity (35).

Our data suggest a scenario in which ORFV ANK proteins block FIH-mediated inhibition of the residual HIF-1α that exists in cells even in the presence of active PHDs. In this manner, ORFV might achieve HIF activation without a need to inhibit PHD activity. This is consistent with the proposed role for FIH-dependent hydroxylation of cellular ANK proteins in regulating HIF activity through sequestration of FIH (36). The ORFV ANK proteins appear to act as competitive inhibitors of FIH, with each protein showing different affinities and levels of hydroxylation, although there is no obvious correlation between the two. For example, while ov126 bound FIH with a relatively high affinity and was efficiently hydroxylated in vitro, ov129 bound FIH with a low affinity but displayed efficient hydroxylation. This is reminiscent of other FIH substrates whose binding does not correlate with hydroxylation, including the ANK protein Notch4, which is able to bind FIH but is not hydroxylated (37). It is also important that the hydroxylation data reflect hydroxylation of the isolated ARDs by purified FIH and may not reflect the efficiency of hydroxylation of full-length ORFV ANK proteins in a cellular context. The HIF CAD, for example, demonstrates significant differences in FIH-dependent hydroxylation efficiency in vitro and within cells (38).

In addition, in vivo, ORFV replicates in the mildly hypoxic environment of the skin. There the low oxygen concentration can allow low levels of constitutive HIF pathway activation (39). FIH activity can occur at lower oxygen levels than those for the PHDs (40), such that even when HIF-1α is no longer degraded following ubiquitination via the PHD-VHL pathway, FIH still functions to inhibit HIF coactivator complex formation. Thus, in the natural environment of ORFV, PHD hydroxylation is likely to be partially interrupted, and the interaction of the viral ANK proteins with FIH may further enhance HIF pathway activation. However, evidence that ORFV may increase expression of a HIF-responsive gene even in cells lacking FIH (Fig. 6C) raises the possibility that ORFV may encode additional, as yet unidentified factors able to enhance HIF activation. Further work will be required to explore that possibility.

Oxygen-dependent hydroxylation by FIH of targets other than HIF-1α has been reported (10, 41, 42), and posttranslational modification by hydroxylation is a characterized biological mechanism to improve protein stability (43). One report revealed that the melting temperature of a synthetic 3-repeat ANK domain was raised after the hydroxylation of an Asn residue (44). Another study showed that mutations that interrupted FIH hydroxylation sites on IκBα's ANK domain reduced its ability to associate with NF-κB but prolonged the half-life of unpaired IκBα (45). These studies demonstrate that the hydroxylation of ANK domains by FIH can assist with conformational stability, but the functional effects remain unclear. These observations raise the possibility that the interaction of ORFV ANK proteins with FIH may have additional consequences distinct from the postulated role in activation of HIF-1α.

The ability of the ORFV ANK proteins to bind FIH has implications for the subcellular localization of FIH, and consequently for its function. For example, the coexpression of ov008 sequesters FIH away from the cytoplasm, where it is normally located, and into the nucleus. Given that some FIH substrates do not enter the nucleus (for example, TRPV3 [41]), this is likely to alter the efficiency of FIH-dependent hydroxylation of specific substrates. It is also worth noting that altered subcellular localization of FIH has been correlated with prognosis for certain cancers, although the mechanism remains unclear (46, 47).

The possible effects on viral replication of the ORFV ANK protein interaction with FIH have yet to be explored. There have been no reports of relevant gene deletion studies other than one of a transposition-deletion variant of ORFV strain NZ2 in which three genes, one of which encodes ov008, were found to be deleted (48). Inoculation of sheep with this variant produced lesions that were reduced in size and resolved more quickly than wild-type lesions, but it was not possible to determine if ov008 contributed to these differences. Future studies may need to consider the deletion of multiple combinations of the 5 ORFV ANK protein-encoding genes (49).

There is growing recognition of the extent of viral manipulation of cell metabolism, and a number of viruses have been shown to induce aerobic glycolysis. The most extensively studied poxvirus, VACV, however, appears not to induce glycolysis but rather glutaminolysis, and this is in part linked to a protein (C16) known to stabilize HIF-1α (42, 50). There is no evidence that ORFV has a C16 homolog, and it remains unclear why VACV targets a PHD to achieve HIF activation while ORFV targets FIH for a similar outcome. However, VACV encodes multiple ANK proteins, and these may play a role similar to that described here for the ORFV ANK proteins. HIF-1α is stabilized by a range of different pathogens, and the downstream effects may be different from those typically seen in hypoxia. However, it seems likely that enhanced HIF activity would increase the energy available to support viral replication. These results show the ability of FIH to bind and hydroxylate ORFV ANK proteins. The sequestration of FIH by ORFV ANK proteins causes derepression of HIF activity in reporter gene assays, and ORFV-infected cells show upregulated HIF target gene expression. These findings describe a previously unknown mechanism of HIF activation by viruses that may extend to other members of the poxvirus family.

MATERIALS AND METHODS

Cells, viruses, and plasmids.

Primary lamb testis (LT) cells, HeLa cells, and HEK 293T cells were grown in minimum essential medium (MEM) or Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum and a penicillin-streptomycin-kanamycin mixture at 37°C and 5% CO₂. ORFV strain NZ2 was propagated in and titrated on LT cells (25). Flag-tagged viral ANK protein genes cloned into the expression plasmid pApex-3 were described previously (18). Derivatives of ov008 in which specific ANKs were deleted were constructed using PCR splicing. Alanine substitutions in ov008 were made with custom-synthesized (GenScript) DNA sequences subcloned into ov008-pApex-3 so as to replace the sequence between the NheI and BamHI sites. G5E1B-luciferase (51), GalDBD-hHIF-1 737–826, Gal-O empty (22), and control Renilla luciferase (Promega) reporter plasmids have been described previously. The ovine FIH open reading frame was recovered by PCR from Ovis aries cDNA and cloned into the NcoI/XhoI-digested pET-32a (Novagen) vector to generate a thioredoxin–6-histidine amino-terminal fusion (pET-32aoFIH). pEF-5myc-6His-hFIH was constructed by subcloning human FIH into pEF-IRES-PURO5 along with oligonucleotides coding for a myc-6His tag. For bacterial expression of MBP-tagged ORFV ANK proteins, ORFV ANK domain-encoding regions were subcloned via PCR from pApex-3 vectors into pMBP. The following domains were subcloned using the listed restriction sites: ORFV ANK 008-1-325aa, KpnI/SalI; ORFV ANK 123-1-366aa, XhoI/NcoI; ORFV ANK 126-1-350aa, BamHI/SalI; ORFV ANK 128-1-367aa, BamHI/HindIII; and ORFV ANK 129-1-363aa, BamHI/HindIII. All constructs were verified by sequencing.

Transfection.

HeLa cells were seeded at 60% confluence 24 h before transfection by use of Fugene HD (Promega) following the manufacturer's guidelines. Routinely, 2 μg of plasmid DNA and 3 μl of Fugene HD were applied to 3 × 10⁵ cells, and cells were harvested 24 h later.

Western blotting.

Cells were lysed in SDS-PAGE loading buffer, and equal amounts of protein were resolved by 10% Tris-glycine SDS-PAGE and transferred to nitrocellulose membranes (GE Healthcare). Membranes were probed with horseradish peroxidase-conjugated mouse anti-Flag (Sigma), rabbit anti-FIH (Novus), rabbit anti-tubulin (Abcam), rabbit anti-actin (Santa Cruz), mouse anti-paxillin (Thermo Fisher), mouse anti-HIF1 (BD Biosciences), or rabbit anti-Skp1 (Santa Cruz) antibody followed by horseradish peroxidase-conjugated goat anti-rabbit IgG antibody (Dako). Proteins were visualized by chemiluminescence, and images were acquired on an Odyssey Fc imaging system (Li-Cor).

Coimmunoprecipitation.

Cells were lysed in immunoprecipitation buffer (50 mM Tris, pH 7.6, 150 mM NaCl, 1% NP-40, and protease inhibitor cocktail) on ice for 30 min. Insoluble material was removed by centrifugation at 14,000 × g and 4°C for 10 min. Cleared lysates were incubated with agarose-conjugated mouse anti-Flag antibody (20 μl per sample; Sigma) or EZview Red protein G affinity gel (20 μl per sample; Sigma) for 16 h at 4°C. Samples were washed and then incubated in SDS-PAGE loading buffer for 10 min at 95°C. Proteins were resolved by SDS-PAGE and detected by Western blotting.

Confocal microscopy.

Cells on 13-mm-diameter coverslips were fixed (100% methanol) and permeabilized (100% acetone) before being incubated with mouse anti-Flag (Sigma), goat anti-FIH (Santa Cruz), or rabbit anti-HIF-1α (Novus) antibody followed by Alexa 488-conjugated goat anti-mouse, Alexa 594-conjugated goat anti-rabbit, or Alexa 594-conjugated rabbit anti-goat IgG antibody (Invitrogen). Finally, cells were stained with Hoechst dye (Invitrogen), and coverslips were mounted onto slides with Slow Fade Gold antifade reagent (Life Technology). Fluorescence images were acquired on a Zeiss LSM 710 confocal laser scanning microscope and analyzed with Zen software (Zeiss) or ImageJ, version 1.48v (NIH).

Reporter assay.

HEK 293T cells were seeded in a 24-well tray and the following day transfected at 30% confluence by use of Lipofectamine 2000 (Invitrogen) with 100 ng of 5× GalRE-luciferase reporter (52), 5 ng of control Renilla luciferase reporter (Promega), 100 ng GalDBD-hHIF-1 727–826 CAD (4), and 100 ng of a pApex3 full-length ORFV ANK construct, with empty vector controls where required. Cells were harvested 16 h later and analyzed using a dual-luciferase system (Promega) according to the manufacturer's instructions.

Cell line generation.

FIH knockout cell lines were generated using CRISPR-mediated genome editing of the FIH (HIF-AN) coding region. Guide RNAs targeting two different sites were designed at http://tools.genome-engineering.org to generate two independent lines. Oligonucleotides for site 1 (CACCgGACGCGGAATGGGCCTAGTC and AAACGACTAGGCCCATTCCGCGTCc) and site 2 (CACCgGCAGTTATAGCTTCCCGACT and AAACAGTCGGGAAGCTATAACTGCC) were annealed and cloned into BbsI-digested pSpCas9(BB)-2A-GFP (PX458) to generate sg-FIH-ex1-g1-pX458 and sg-FIH-ex1-g2-pX458. Each construct was transfected into HeLa cells by use of Fugene HD (Promega), and 48 h later, cells were diluted and single cells plated. Each clone was expanded for genotyping by PCR and sequencing and for protein analysis by Western blotting (Fig. 6C). One clone from each targeted site was used in subsequent experiments. HeLa cell lines overexpressing FIH were generated using pEF-myc-6His-hFIH and were selected with 2 mg/ml of puromycin. FIH levels were analyzed via Western blotting (Fig. 6C).

Quantitative real-time PCR.

RNA was isolated using an RNeasy minikit (Qiagen) and reverse transcribed using Superscript III (Invitrogen). Real-time PCR was performed in triplicate, using the following primers: oVEGF F, TTGCCTTGCTGCTCTACCTT; oVEGF R, AGATGTCCACCAGGGTCTCA; oPHD3 F, TTACCTCCTGTCCCTCATCG; oPHD3 R, GTTCCATTTCCTGGGTAGCA; oGLUT1 F, ATCCTCATTGCCGTGGTG; oGLUT1 R, GCCTCCTCGAAGATGCTTGT; oACTB F, CTGCCGGACGGGCAGG; oACTB R, GATTCCATGCCCAGGAAGG; hVEGF F, CCTTGCTCTACCTCCAC; hVEGF R, GCAGTAGCTGCGCTGATAGA; hPHD3 F, ACTTCGTGTGGGTTCCTACG; hPHD3 R, AGCTACATGGTGGGATCCTG; hGLUT1 F, GATTGGCTCCTTCTCTGTGG; hGLUT1 R, TCAAAGGACTTGCCCAGTTT; hRPLP0 F, CTCACTGAGTCAGGGACAT; and hRPLP0 R, GTGATACCTAAAGCCTGGAA (the primers spanned an intron where possible). Reaction mixtures were prepared with Fast SYBR green master mix (Applied Biosystems) and run on a StepOnePlus thermocycler (Applied Biosystems). Expression levels were normalized to that of ACTB for ovine samples and that of RPLPO for human samples and are presented as fold inductions over the levels in the uninfected cell line, with P values calculated using unpaired Student's t test. Melting curves for PCR products were used to confirm the presence of single amplicons.

Protein expression and purification.

All thioredoxin (Trx)-His6- and MBP-tagged proteins were expressed in Escherichia coli BL21(DE3), purified by Ni²⁺ and amylose affinity chromatography (20), and stored at 4°C in 150 mM NaCl, 20 mM Tris, pH 8.0. Protein concentrations were determined using calculated extinction coefficients and the absorbance measured at 280 nm, and purity (approximately 90%) was assessed by SDS-PAGE with Coomassie blue staining.

In vitro hydroxylation assay.

ANK domain hydroxylation was assayed by a CO₂ capture assay as described previously (20). Each 40-μl reaction mixture contained a saturating amount of Trx-6His-oFIH (1 μM), 20 μM ORFV ANK substrate, 500 μM FeSO₄, 40 μM [¹⁴C]2-oxoglutarate, 4 mM ascorbate, 500 μM dithiothreitol, 0.4 mg bovine serum albumin, and 50 mM Tris-HCl (pH 7.0) and was incubated at 37°C for 30 min. Filters were dried, Ultima-Gold XR scintillation fluid was added, and counts were performed on a MicroBeta 2450 machine (PerkinElmer).

Mass spectrometry.

In-gel trypsin digestion of excised SDS-PAGE gel bands and mass spectrometry analysis of the extracted peptides were performed as described previously (53). Nanoflow ultra-high-pressure liquid chromatography (nUHPLC)–tandem MS (MS/MS) analysis was performed using a Waters (Milford, MA) NanoAcquity system (maximum pressure, 10,000 lb/in²) interfaced with a linear ion trap (LTQ)-Orbitrap-Velos Pro hybrid mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). The tryptic digests were acidified, loaded on the trap, and separated using nUHPLC as described previously (53, 54). Peptides were separated by a set of three linear gradients: up to 27% solvent B over 45 min, up to 60% solvent B over 4 min, and up to 95% solvent B over 5 min. After the final gradient, the column was held at 95% solvent B for 5 min. The LTQ-Orbitrap-Velos Pro mass spectrometer acquisition and database search parameters were essentially as described previously (37). Asparagine hydroxylation was included as a variable modification, and the Sequest HT database search engine in Proteome Discoverer (version 1.4.1.14) software (Thermo Fisher Scientific) was used to search against a custom-made ORFV ANK protein domain database. Tandem MS spectra for putative hydroxylated peptide sequences (precursor ions) were validated manually.

Protein modeling.

The ANK domains of ORFV ANK proteins were modeled using the SWISS-MODEL protein structure homology modeling server (http://swissmodel.expasy.org/), with human ankyrin (SMTL ID 1n11.1.A) as the template. Predicted results were selected with a rainbow color scheme to verify the N and C termini, and snapshots were taken directly from the three-dimensional molecule image.