Anti-tumorigenic effects of Type 1 interferon are subdued by integrated stress responses

Integrated stress signals converging on phosphorylation of eIF2α mediate phosphorylation and downregulation of IFNAR1

PERK-dependent acceleration of IFNAR1 proteolysis in cells undergoing unfolded protein response evoked by viral infection or TG treatment prompted us to investigate the effects of tumor microenvironment-relevant inducers of PERK such as hypoxia (13). As expected, the exposure of WM266-4 human melanoma cells to hypoxia (0.5% oxygen) induced phosphorylation of eIF2α (Figure 1A). Knockdown of PERK attenuated this phosphorylation but not the accumulation of HIF1α protein (Figure S1). Intriguingly, hypoxia triggered a robust IFNAR1 downregulation that could be partially reversed upon the knockdown of PERK (Figure 1A). We next tested sensitivity of cells to Type 1 IFN using a cell viability assay that combines assessment of anti-proliferative and anti-survival effects of these cytokines. Consistent with the importance of IFNAR1 levels for responsiveness of cells to its ligands (15, 16, 20–26), hypoxic conditions also attenuated the inhibition of cell viability by various doses of IFNα (Figure 1B). Importantly, this attenuation could be reversed by knocking down PERK (Figure 1B). Under hypoxic conditions, the IC₅₀ of IFNα for cells that received PERK shRNA was 125 U/mL (versus that of 207 U/mL for control shRNA-treated cells, Figure S2). Collectively these data suggest that activation of PERK by hypoxia triggers downregulation of IFNAR1 and ensuing desensitization of cells to IFNα.

Previous studies suggested that activation of PERK mediates IFNAR1 degradation thorough stimulation of the stress activated p38α protein kinase (27) that mediates phosphorylation of a priming site (Ser532 within human IFNAR1 or Ser523 within the mouse receptor (15)). In turn, this priming phosphorylation increases the efficacy of IFNAR1 degron phosphorylation by casein kinase 1α (28) leading to the recruitment of the β-Trcp E3 ubiquitin ligase, IFNAR1 ubiquitination, internalization and lysosomal degradation (reviewed in (29, 30)). Besides phosphorylating eIF2α, PERK has been also implicated in directly activating other phosphorylation cascades (31, 32). Given that, we sought to investigate how PERK signals towards IFNAR1 phosphorylation. A short term treatment of mouse embryo fibroblasts (MEF) from wild type mice with thapsigargin (TG) led to a robust activation of p38 kinase as well as priming phosphorylation and ubiquitination of IFNAR1 (Figure 1C). Consistent with earlier reports (15, 16, 27), a longer exposure to TG decreased the levels of IFNAR1 in these cells (Figure 1D). Intriguingly, the effects of either short term or prolonged treatment with TG were not evident in MEFs from eIF2α^S51A knock-in mice (Figures 1C–1D). These results suggest that inducers of PERK require phosphorylation of eIF2α for the ubiquitination and downregulation of IFNAR1.

To further test this hypothesis, we used isogenic human HT29-derived cells that were stably transduced with empty retrovirus or made to express the constitutively active fragment of eIF2α phosphatase, GADD34 (33). Consistent with previous reports, basal or TG-induced phosphorylation of eIF2α on Ser51 in these HT29A1 cells was attenuated (Figure 1E). In response to TG these cells neither activated p38 kinase nor phosphorylated IFNAR1 at its priming site (Figure 1E). Furthermore, a prolonged TG treatment did not efficiently downregulate IFNAR1 in cells expressing active GADD34 (Figure 1F). These results collectively suggest that phosphorylation of eIF2α downstream of PERK is required for IFNAR1 phosphorylation, ubiquitination and downregulation in response to PERK inducers.

Phosphorylation of eIF2α on Ser51 can also be stimulated by inducers of the ISR such as amino acid deficit acting via GCN2 or activators of PKR (7, 30). Thus we sought to investigate whether these pathways converge and affect phosphorylation and levels of IFNAR1. To this end, we used a panel of MEFs derived from wild type mice or mice lacking PERK, GCN2, or PKR. This panel was subjected to either TG treatment to induce PERK, or heparin (HE) treatment to induce PKR, or leucine starvation (LS) to activate GCN2. All of these inducers stimulated phosphorylation of Ser51 of eIF2α, activation of p38 kinase and priming phosphorylation of IFNAR1 in wild type MEFs (Figure 2A). Furthermore, a prolonged exposure of wild type MEFs to these inducers noticeably downregulated IFNAR1 (Figure 2B). As expected the phosphorylation of eIF2α and activation of p38α kinase in response to TG was impaired in PERK-null cells. Similarly, knockout of GCN2 impaired eIF2α/p38α phosphorylation in response to leucine starvation, and knockout of PKR had the same effect in cells treated with heparin (Figure 2A). Remarkably, the patterns of either priming phosphorylation of IFNAR1 (Figure 2A) or its downregulation (Figure 2B) followed an analogous trend.

Furthermore, the results of additional experiments using MEFs from eIF2α^S51A knock-in and p38α knockout mice suggested that the integrity of eIF2α Ser51 phosphorylation and the availability of p38α protein kinase were required for IFNAR1 priming phosphorylation and downregulation elicited in response to either heparin (Figure 2C–2D) or leucine starvation (Figure 2E–2F). In all, these results suggest that, in addition to established PERK-dependent regulation, the inducers of the ISR can utilize activation of either GCN2 or PKR to trigger eIF2α/p38α-dependent phosphorylation and ensuing downregulation of IFNAR1. A key role of the stress-activated p38α protein kinase in regulating IFNAR1 levels is consistent with a previous report that activation of this kinase was inversely correlated with levels of endogenous IFNAR1 in clinical malignant melanoma samples (34).

Constitutive PKR activity downregulates IFNAR1 levels and signaling

We further sought to corroborate the role of PKR in regulating IFNAR1 levels and activities. Although linked to induction of apoptosis (35, 36), a high level of constitutively active PKR has been reported in clinical human breast cancer tissues and cell lines including T47D cells (9–11). Whereas priming phosphorylation of endogenous IFNAR1 in these cells could be detected upon treatment with a lysosomal inhibitor (to prevent IFNAR1 degradation), further addition of PKR inhibitor C16 readily decreased this phosphorylation (Figure 3A). Furthermore, merely adding C16 to T47D cells led to an increase in IFNAR1 levels in a dose-dependent manner (Figure 3B). Pre-treatment of T47D cells with this PKR inhibitor also increased the IFNα-induced phosphorylation of STAT1 (Figure 3C).

This pharmacologic analysis was supplemented with genetic experiments using a RNAi approach. Knockdown of PKR in T47D cells increased the levels of endogenous IFNAR1 (Figure 3D) and augmented STAT1 activation assessed by its phosphorylation on Tyr701 (Figure 3E) or Ser727 (Figure S3). Importantly, while viability of T47D cells did not robustly change in response to IFNα, these cells could be significantly sensitized to this treatment by adding PKR inhibitor C16 (Figure 3F). Knockdown of PKR noticeably decreased the ability of T47D cells to invade matrigel and/or to migrate through a membrane in a Boyden chamber invasion/motility assay (Figure 3G). In line with the well-established role of Type 1 IFN in restricting cell motility (37), knockdown of IFNAR1 (Figure S4) noticeably stimulated the breast cancer cells movement through the Matrigel/membrane (Figure 3G). Remarkably, in cells that received shRNA targeting IFNAR1, knockdown of PKR did not significantly decrease cell migration/invasion (Figure 3G) indicating that inhibition of invasive phenotype can at least in part be attributed to IFNAR1 stabilization. These results collectively suggest that high levels of constitutive PKR activity in breast cancer cells may restrict IFNAR1 expression and signaling as well as IFNAR1-mediated anti-tumorigenic effects.

Induction of GCN2-dependent pathways promotes IFNAR1 downregulation to attenuate IFN signaling

Culturing MEFs in medium deficient in the essential amino acid leucine triggered IFNAR1 phosphorylation and downregulation that was dependent on GCN2, eIF2α phosphorylation and p38α kinase (Figures 2A, E, F). Leucine starvation also stimulated priming phosphorylation, ubiquitination and downregulation of endogenous IFNAR1 in WM266-4 melanoma cells (Figure 4A). Similar effects were seen in 1205Lu human melanoma cells (Figure S5). We next investigated the role of activated GCN2 in cellular responses to Type 1 IFN. When WM266-4 cells were cultured in the absence of leucine, the activation of STAT1 in response to IFNα was attenuated (Figure 4B and S6). A similar attenuation was seen in 1205Lu cells treated with IFNα but not with IFN (Figure S7) that signals through a different receptor, IFNGR (19). Furthermore, leucine deficit noticeably blunted the sensitivity of WM266-4 cells to the anti-proliferative effects of IFNα (Figure 4C). Importantly, knockdown of GCN2 in these cells at least partially restored their responsiveness to IFNα as evident from analyses of STAT1 phosphorylation (Figures 4B, S6) and cell viability (Figure 4C). In addition, migration and/or invasion of WM266-4 cells grown without leucine was markedly inhibited by knockdown of GCN2 unless IFNAR1 also was downregulated (Figures 4D and S8). These results collectively suggest that activation of GCN2 triggered by amino acid deficit leads to downregulation of IFNAR1 and attenuation of its signaling and of ensuing anti-tumorigenic effects.

Plasmids, Chemicals and Cells

Short hairpin RNA (shRNA) against GCN2 (2), IFNAR1 (21), or PERK (16) have been described previously. PKR inducer heparin sodium salt (HE, used at 1μg/mL), thapsigargin (TG, used at 1μg/mL) PKR inhibitor Imidazolo-oxindole (C16), and MCDB 153 medium without leucine were purchased from Sigma. ShRNA for knockdown of PKR were purchased from Santa Cruz Biotechnology (sc-36263-SH). Dulbecco’s Modification of Eagle’s medium without leucine was obtained from MP Biomedicals. Human IFNα2 was purchased from Bio-Sidius S.A. Murine IFNβ was expressed in Chinese Hamster Ovary cells, and purified to homogeneity from conditioned culture medium using successive Blue Sepharose, copper chelating Sepharose, lentil lectin Sepharose, and Uno S (Biorad) cationic ion exchange chromatographies. The specific activity of the protein was 2 × 10⁸ units/mg in an in vitro antiviral cytopathic effect assay using murine L929 cells challenged with encephalomyocarditis virus. The purified protein had an endotoxin level of <0.05 EU/mg.

Human breast cancer T47D cells (from Ze’ev Ronai, Burnham Institute, San Diego, CA) and melanoma WM266-4 and 1205Lu cells (from Meenhard Herlyn, Wistar Institute, Philadelphia, PA) were cultured as described previously (47, 48). WM266-4 cells were maintained in MCDB 153 medium supplemented with bovine insulin (5μg/mL). Other human cells as well as mouse embryo fibroblasts (MEFs) from knockout mice lacking p38α (from Angelo Nebreda, Centro Nacional de Investigaciones Oncológicas, Madrid, Spain), PKR (from Antonis Koromilas, McGill University, Montreal, Quebec), PERK or GCN2 (from David Ron, New York University, New York, NY) or knock-in mice harboring the eIF2α^S51A allele (from Randal Kaufman, University of Michigan, Ann Arbor, MI) were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% (v/v) fetal bovine serum (Hyclone), non-essential amino acids and β-mercaptoethanol (55 μM). For knockdown experiments in human cells, T47D or WM266-4 cells were transduced with lentiviral particles encoding shRNA against GFP, PKR or GCN2 and selected in 2 μg/mL of puromycin for 2 weeks. The human colon adenocarcinoma HT-29 cell line and its derivative stably expressing the constitutively active GADD34 C-terminal fragment (HT29-A1) were cultured in the presence of puromycin (1 μg/mL) as described previously (33).Cells were exposed to the hypoxia condition of 0.5% oxygen, 5% carbon dioxide and 94.5% nitrogen for the indicated times through the use of the Invivo2–400 Hypoxia Workstation (Ruskinn Technology Ltd).

Antibodies and Immunotechniques

Analyses of phosphorylated (on Ser532 in human IFNAR1 or Ser523 on mouse IFNAR1) and total levels of human and mouse IFNAR1 were carried out as described elsewhere (22, 49). Commercially available antibodies against pSTAT1, p-eIF2α, p-p38 kinase, STAT1, PKR, (Cell Signaling), eIF2α (Biosources), β-actin (Sigma), HIF1α (BD Biosciences), p38α, and ubiquitin (Santa Cruz Biotechnology) were purchased. Immunoprecipitations, immunoblotting, and in vivo ubiquitination assay using denaturing immunoprecipitation were carried out as described previously (23).

Cell invasion assays

BD Biocoat chambers coated with Matrigel matrix (BD Matrigel, 354234) were overlaid with 0.5 mL of cell suspension (2.5 × 10⁴ cells) to each 24-well invasion chamber in media without serum. Medium with 10% serum (0.75 mL) was then added to the outside chamber and cells were incubated at 37°C with 5% CO₂ for either 24 hours (WM266-4 cells) or for 5 days (T47D cells). Cells were stained with crystal violet and images were obtained using a BX-51 Olympus microscope. The images were quantified using Image J.

Viability assays

Cells transfected with the indicated plasmids were selected with puromycin for 3 days and then seeded at 5 × 10⁴ cells per well in a 96 well plate. The cells were treated with indicated doses of IFNα for 24–48 h and were analyzed using a cell proliferation colorimetric assay kit, CellTiter 96 (Promega). The Student t test was used for analysis of statistical significance.