A Negative Feedback of the HIF-1α Pathway via Interferon-Stimulated Gene 15 and ISGylation Free

Posttranslational modifications (PTM) of proteins contribute significantly to the functional diversity of the proteome (1). For example, the ubiqutin–proteasome pathway (UPP) plays a critical role in hypoxic response and tumorigenesis by regulating protein stability of the master transcription factor HIF-1α (2–5). Both the UPP and IFN-stimulated gene 15 (ISG15) conjugation (ISGylation) pathways, which are regulated in response to IFN, contribute to inflammatory responses and cancer cell-killing activity of drugs (6–9). Notably, hypoxia and inflammation are two microenvironments involved in cancer development (5, 10–12).

The initiation and maintenance of tumor cells rely largely on their ability to interact with and adapt to changes in microenvironments (5, 11). Cells within rapidly growing tissues, whose proliferation rate becomes faster than their angiogenesis rate, such as in embryos and solid tumors, often encounter hypoxic stress. Moreover, hypoxia is a feature of the microenvironment in chronic inflammatory conditions such as arthritis and inflammatory bowel disease, and almost all types of solid tumors encounter hypoxic inflammation (10, 13). These two interdependent microenvironments, thus, impact not only the progression of cancer but also its development (12). However, the potential interaction between cancer-related inflammation and hypoxic environments during tumorigenesis and the underlying cross-talk mechanism remain largely unclear.

Cells respond to hypoxia and inflammation through transcriptional programming, allowing them to adjust to the optimal cellular context (4, 5). Hypoxia induces expression of a defined set of genes through hypoxia-inducible factors (HIFs; especially HIF-1α), thereby being associated with erythropoiesis, glycolysis, the epithelial–mesenchymal transition (EMT), angiogenesis, tumor metastasis, therapy resistance, and a poor prognosis (14). Several regulatory mechanisms affecting the activity of HIF-1α have been identified: (i) HIF-1α stability is regulated by changes in oxygen tension. HIF-1α is degraded under normoxia by UPP, which involves prolyl hydroxylase (PHD1/2/3)-mediated hydroxylation and VHL-mediated ubiquitinylation (3, 15); (ii) Factor inhibiting HIF-1 (FIH) hydroxylates HIF-1α on Asp 803, leading to blockages of the recruitment of the p300/CBP coactivator and subsequent transcription inactivation (16, 17); (iii) HIF-1α functions as a transcription factor in a heterodimeric form composed of two subunits, HIF-1α and HIF-1β (18); (iv) HIF-1α protein expression depends on both cap- and IRES-directed translation (19, 20); (v) IFN-τ, one of the Type I IFN produced only in ruminants and/or progesterone regulate HIF mRNA expression in the ovine endometrium and sheep conceptuses (21, 22). Cytokines and growth factors could stimulate HIF-1α protein synthesis and that cytokines induce the NF-κB–dependent expression of HIF-1α mRNA under hypoxia (23).

Similar to the role of HIF-1α in the hypoxia pathway, IFN functions in immune and inflammatory responses via inducing the expression of ISGs (4, 24). Among the ISGs, there is a group that functions through conjugation of ISG15 with proteins in a process termed ISGylation. Studies addressing ISG15 expression in primary tumor tissues and cell lines have indicated a link between ISG15 and tumorigenesis (7, 25–28). Defects in the IFN system (e.g., due to gene mutations) are traits associated with several tumor types and may affect cancer development (29, 30). Consequently, IFN, either alone or in combination with other anticancer drugs, have been used as a therapeutic strategy against malignancies (7, 31).

ISG15, which is known as ubiquitin cross-reactive protein precursor, is composed of two ubiquitin-like domains (∼30% homology to ubiquitin; ref. 4), and ISG15 retains the LRLRGG sequence required for its conjugation with proteins, thus conferring ISGylation on its substrates in a manner akin to ubiquitinylation (4, 27, 28, 32). The expression of both the ISGylation and ubiquitinylation systems is induced by IFN, and these two PTMs interact functionally with each other (33, 34). Moreover, ISGylation competes and/or interferes with specific and global ubiquitinylation levels (35). However, the regulatory role of the ISG15 and the ISGylation system remains to be defined.

The importance of hypoxia and inflammation during tumorigenic processes prompted us to investigate the potential physical, functional, and genetic interactions among the key components of the above two microenvironments. Here, we examined the interactions between the IFN-induced inflammatory response and the hypoxia pathway. IFN treatment modulated the levels of free HIF-1α expression. Unexpectedly, HIF-1α not only interacts physically with ISG15 but is also an ISGylation substrate. Further studies determined that ectopic ISG15 expression disrupts the formation of the HIF-1α and -1β heterodimers. Functional studies revealed that ISGylation of HIF-1α compromises the HIF-1α–induced gene expression as well as tumorigenic growth in both soft-agar plates and xenograft mouse models.

All chemicals and the pFLAG-CMV2 plasmid were purchased from Sigma unless otherwise indicated. The pHis-ISG15 plasmid expressing His-tagged ISG15 was a generous gift from Dr. S.M. Shih (Academia Sinica, Taiwan), and the plasmid expressing the ISG15_G→A mutant was constructed from pHis-ISG15 via site-directed mutagenesis. ISG15 (amplified by PCR) was cloned into pFlag-CMV2 (Sigma) to generate pFlag-ISG15. The reporter plasmid pXP2-Twist-HRE, control reporter pRL-tk, and the pSUPER-based plasmid expressing sh-HIF-1α were generous gifts from Dr. K.J. Wu (National Yang Ming University, Taiwan). The Ube1L-, UbcH8-, and HA-Herc5–expressing plasmids came from Dr. R.M. Krug (University of Texas, TX). The plasmids including pHA-HIF1α-DM, phISG15, and pBabe-HA-VHL were obtained from Addgene. Two plasmids, pHA-HIF-1α and pHA-HIF-1α ΔODD, were provided by Dr. Huang LE (University of Utah, UT). The plasmid pEBB-HIF-1β was from Dr. E. Burstein (University of Texas Southwestern, TX). Fragments of HIF-1α (D1, D2, bHLH, bHLH-PASA, PASA-PASB, PASB, ODD, D2O, D2F1, F1, and F2) were generated via PCR and cloned into the pcDNA3.0-HA vector to generate plasmids expressing the HIF-1α–truncated mutants. GST-ISG15, GST-ISG15-N, and GST-ISG15-C were constructed via PCR and cloned into pGEX4T-1 (GE Healthcare) for expression. The HA-ubiquitin–expressing plasmid was obtained from Dr. C. Sasakawa (University of Tokyo, Tokyo, Japan). Polyclonal ISG15 antibodies were raised against recombinant ISG15 proteins, and were generated by LTK Biolaboratories. Antibodies against HIF-1α, HIF-1β (BD Bioscience), Flag M2, His, Fibronectin, Vimentin, tubulin, actin (Sigma), HA (Covance), Glut1 (Genetex), and Slug (Anaspec) were obtained commercially. The human 769-P, Caki-1, and 293T renal carcinoma cell lines were obtained from the American Type Culture Collection (ATCC), and were cultivated in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% FBS, penicillin (100 units/mL), streptomycin (100 μg/mL), and glutamine (2 mmol/L) at 37°C in an incubator with 21% O₂ and 5% CO₂, to provide normoxic conditions. Hypoxic treatment was performed by harvesting cells in a hypoxia incubator (ASTEC) with 5% CO₂, 1% O₂, and 94% N₂. Simulated hypoxia was achieved by exposure to 260 μmol/L DFX or 200 to 500 μmol/L CoCl₂.

Plasmids expressing short-hairpin RNA (shRNA) sequences targeting sh-ISG15 and sh-Luc were obtained from the National RNAi Core Facility, Taiwan (http://rnai.genmed.sinica.edu.tw). The targeted sequences were as follows: sh1-ISG15, 5′-GTGGTGGACAAATGCGACGAA-3′; sh2-ISG15, 5′-CATGTCGGTGTCAGAGCTGAA-3′ and sh-luciferase (sh-Luc), 5′-CCTAAGGTTAAGTCGCCCTCG-3′.

Cells were seeded into 6 cm dishes at a density of 0.3 to 1 × 10⁶ cells per plate. After various treatments, the cells were washed with 1× PBS buffer twice and lysed; then, immunoblotting and coimmunoprecipitation experiments were conducted as described. Briefly, the cells (∼10⁶ cells) were lysed in 200 μL of TEGN buffer (10 mmol/L Tris, pH 7.5, 1 mmol/L EDTA, 420 mmol/L NaCl, 10% glycerol, and 0.5% NP-40) with a freshly added Na₃VO₄ (1 mmol/L) and 1 × protease inhibitor cocktail (1 mmol/L; Roche). For immunoprecipitation assay, the lysates were mixed with 600 μL of TEG buffer (10 mmol/L Tris, pH 7.5, 1 mmol/L EDTA, and 20% glycerol), 0.6 to 1 μg of antibodies and 20 μL of 50% protein G beads (Pierce), then incubated at 4°C for 3 hours with rotation. The immunoprecipitates were washed 4 times in a coimmunoprecipitation buffer (TEG:TEGN = 3:1), boiled in 1× SDS-sample buffer for 7 minutes, and analyzed via SDS–PAGE, followed by immunoblotting analysis (36).

Briefly, 20 μL of GST-Sepharose beads was incubated with GST or GST-ISG15 fusion proteins in a final volume of 300 μL of sonication buffer [50 mmol/L Na₂H₂PO₄ pH 8.0, 300 mmol/L NaCl, 20% glycerol, 1 mmol/L DTT, and 1 mmol/L phenylmethylsulfonylfluoride (PMSF)] at 4°C for 1 hour with rotation. The beads were washed 5 times with 0.8 mL of sonication buffer containing 1% Triton X-100 and once with 0.8 mL of coimmunoprecipitation buffer supplemented with a 1× protease inhibitor cocktail, 1 mmol/L DTT, and 1 mmol/L PMSF. The beads were then reacted with a 300 μL mixture containing the IVTT lysates, BSA (2 mg/mL), 1× protease inhibitor cocktail, 1 mmol/L DTT, and 1 mmol/L PMSF at 4°C for 1 hour with rotation, washed 3 times with coimmunoprecipitation buffer, boiled for 7 minutes in 1× SDS sample buffer, and then subjected to immunoblotting analysis (37).

The 293T cells were transfected with His-ISG15-, Ube1L-, UbcH8-, and HA-HERC5–expressing plasmids (the ISGylation system) using polyject (SignaGen) or Lipofectamine 2000 (Invitrogen). After various treatments, the cells were lysed in lysis buffer B (1% NP40, 20 mmol/L imidazole, 10 mmol N-ethylmaleimide (NEM), 1 mmol/L Na₃VO₄, and 1× protease inhibitor cocktail in 1× PBS buffer). After the addition of 40 μL of nickel-nitrilotriacetic acid (Ni-NTA) agarose (Qiagen), the mixtures were incubated at 4°C for 3 hours. The pull-down products were washed 4 times with lysis buffer B supplemented with 5 mmol/L NEM, boiled in 1× SDS sample buffer, and analyzed via immunobloting analysis (38).

Cells (769-P, 4 × 10⁴; Caki-1, 3 × 10⁴) were seeded in 6 cm dishes for 4 days in triplicate for each measurement. At 24-hour intervals, cell growth was determined by Trypan blue exclusion assay.

Cells (4 × 10⁴, 769-P; 5 × 10³, Caki-1) were mixed with 2 mL of 0.3% agarose (Genetex)-containing DMEM medium, and overlaid on a precoated layer of 0.6% agarose DMEM (2 mL) in 6-well plates. After incubation for 3 to 4 weeks, the colonies were stained using 0.02% crystal violet, and the number of colonies was counted.

The protocol for animal experiments was designed in accordance with the guidelines of the Institutional Animal Care and Use Committee and approved by the same (No. 20060072). For the tumorigenic xenograft models, 4- to 5-week-old nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice were supplied by the Animal Center at the National Taiwan University College of Medicine and maintained in the specific pathogen-free (SPF) facility (5–6 mice per cage). A total of 5 × 10⁶ cells resuspended in serum-free DMEM were subcutaneously injected into the dorsal regions of the NOD/SCID mice. The injected mice were examined every 3 to 4 days for the appearance of tumors. Tumor volumes were estimated (V = 0.4ab²) from length (a) and width (b) measurements that were obtained using calipers.

Quantitative measurements of the obtained results were carried out using the ImageQuant program according to the manufacturer's instructions (Molecular Dynamics), and presented as the mean ± standard error of mean (S.E.M.) of three independent experiments (n = 3). Statistical analyses were conducted using the simple Student t test. The data were considered to be significant if the P value was less than 0.05 (P < 0.05).

IFN stimulates ISG gene expression and one of the most strongly induced genes, ISG15, reduces the overall levels of ubiquitin-conjugated proteins (ubiquitinylates; ref. 35). We observed that IFN-α-2a (1,000 U/mL) stimulated ISG15 expression and decreased the overall ubiquitinylate level in HeLa cells (Supplementary Fig. S1A), and ectopic ISG15 expression also reduced the ubiquitinylate levels in HeLa and 769-P cells (Supplementary Fig. S1B and S1C). Because HIF-1α stability is affected by the UPP, we determined the free HIF-1α levels after IFN exposure. As shown, HIF-1α levels were affected by IFN treatment under normoxia or hypoxia-mimetic desferoxamine (DFX) treatment (Supplementary Fig. S1D and S1E). Notably, the use of IFN-resistant 293T cells resulted in lower ISG15 expression and only a slight change in HIF-1α level being detected after IFN exposure.

There are four hypoxia-responsive elements (HRE; acgtg/cacgt) at positions upstream of ISG15 promoter (Supplementary Fig. S2A): −705 to −701; −549 to −545; −30 to −26; and −19 to −15. Interestingly, we found that DFX treatment induced ISG15 mRNA expression but reduced ISG15 protein level in 769-P cells (Fig. 1A). A similar reduction of ISG15 gene expression in response to DFX was observed in Caki-1 cells, but not in 293T cells (Supplementary Fig. S2B). In addition, a time-dependent reduction of free ISG15 level was observed in 769-P cells (Supplementary Fig. S2C). When knockdown of HIF-1α expression was present, DFX only minimally affected free ISG15 level in 769-P sh-HIF-1α cells (Supplementary Fig. S2D). Consistently, ectopic expression of HIF-1α ΔODD, a nondegradable mutant of HIF-1α, reduced the endogenous levels of free ISG15 and slightly increased the overall ISGylate levels (Supplementary Fig. S2E and S2F). Both ectopic HIF-1α ΔODD expression and DFX treatment reduced the level of ectopically expressed ISG15 in 293T cells, and the HIF-1α ΔODD–mediated reduction of ISG15 expression could only be partially rescued using proteasome inhibitors (data not shown). These results suggest that both ISG15 mRNA expression and ISGylation activity are enhanced during hypoxia through HIF-1α, indicating a potential cross-talk between HIF-1α–mediated hypoxic response and the IFN-induced inflammatory pathway.

Because ISG15 reduces the overall ubiquitinylate levels and the camptothecin-induced downregulation of topoisomerase I (TOP1; ref. 35) and the UPP plays a main role in HIF-1α stability and TOP1 downregulation of topoisomerases, we examined the impact of ISG15 on the DFX- and CoCl₂-induced accumulation of HIF-1α proteins. As shown, ISG15 expression reduced the levels of free HIF-1α in 293T cells (Fig. 1B). In addition, similar results were observed in both Caki-1 and 769-P cells (Fig. 1C). Moreover, gene dosage-dependence of the ISG15-mediated reduction of HIF-1α level was observed (Fig. 1D).

Next, we determined the molecular mechanism(s) underlying the ISG15-mediated reduction of free HIF-1α level. Because this reduction was detected in VHL-deficient 769-P cells, the VHL-dependent UPP is likely not involved. Moreover, ISG15 could still reduce the levels of free HIF-1α in 769-P cells in which the proteasome was inhibited by MG-132 (Fig. 1E). In Fig. 1F, we showed that ISG15 expression only minimally affected HIF-1α mRNA expression, suggesting a transcription-independent reduction. Together, our results suggest that ISG15 reduces the protein level of free HIF-1α. This notion is further supported by the fact that 769-P sh-ISG15 cells exhibit a higher HIF-1α level than the control sh-Luc cells (Fig. 1G). Interestingly, the reduction of HIF-1α levels by ISG15 was detected in the immunoblotting assay (to detect free HIF-1α), but not seen in a dot-blotting assay (to detect overall HIF-1α; Fig. 1H). Similar results were obtained in 293T cells treated with IFN (Supplementary Fig. S2G). These findings, thus, suggested that the reduction of free HIF-1α levels caused by ISG15 expression or IFN treatment might be due to an increase in PTMs (possibly ISGylation) of HIF-1α, which then leads to mobility change and reduction of free HIF-1α level.

To examine interaction(s) between HIF-1α and ISG15 proteins, co-immunoprecipitation experiments were conducted. We showed that HIF-1α interacts physically with ISG15 under hypoxic treatments (Fig. 2A). Similarly, when 293T cells were transfected with plasmids expressing Flag-tagged ISG15 and HA-tagged HIF-1α, anti-Flag antibodies were able to effectively precipitate both ISG15 and HIF-1α from extracts of cells treated with DFX (Fig. 2B) or CoCl₂ (Supplementary Fig. S3A). Ectopically expressed HIF-1α colocalized with ISG15 in cells exposed to IFN and DFX (Supplementary Fig. S3B).

We generated HIF-1α C′- (D1) and N'-truncated (D2) mutants, and showed that the ISG15-interacting domain is located within the N'-terminus of HIF-1α (Fig. 2C). A similar conclusion was also obtained via GST pull-down assay using recombinant proteins obtained from bacterially expressed GST-tagged ISG15 fragments and hemagglutinin (HA)-fused HIF-1α mutant proteins produced in rabbit reticulocyte lysates (Fig. 2D). We mapped the ISG-interacting domain within the PASB domain of HIF-1α (Supplementary Fig. S3C). Differing from full-length ISG15, the N′- or C′-terminus of ISG15 alone cannot interact with full-length HIF-1α (Fig. 2E), HIF-1α ΔODD, and D1 fragments (Supplementary Fig. S3D).

Because HIF-1α interacts with ISG15, we examined whether HIF-1α is an ISGylation substrate. First, functional expression of components of the ISGylation system, including Ube1L (E1), UbcH8 (E2), Herc5 (E3), and ISG15, caused reduction of HIF-1α levels in cells (DFX-treated; Supplementary Fig. S4A; ectopically expressed, Supplementary Fig. S4B). Notably, the ISGylation of HIF-1α was nearly undetectable in 293T cells, possibly due to low endogenous HIF-1α expression and inefficient ISGylation. With MG-132 and DFX treatments to increase HIF-1α protein level, we could then detect potential band(s) of ISG15-conjugated HIF-1α upon IFN treatment and ectopic expression of the ISGylation system (including the ISG15 and E1/E2/E3 enzymes, Fig. 3A compare lanes 1 and 2). In addition, we observed potential ISGylated HIF-1α bands with ectopic expression of HIF-1α (Fig. 3B). Coupled with immunoblotting assays, pull-down experiments revealed that HIF-1α bands with abnormal mobility are likely ISG15-conjugated. Consistently, knockdown of ISG15 expression and the expression of an ISG15_G→A conjugation mutant reduced the levels of ISGylates and ISGylated HIF-1α bands (Fig. 3C, compare lanes 2 and 3; Fig. 3D, compare lanes 3 and 4). Furthermore, we have found that additional expression of E3 ligase Herc5 promoted greater HIF-1α ISGylation compared with another E3 ligase EFP (Supplementary Fig. S4C).

HIF-1α hydroxylation (at P402/564) allows recognition and ubiquitination by the VHL E3 ligase (15). Next, a DM plasmid containing HIF-1α mutated at two hydroxylation sites (P402/564A) was transfected into cells. HIF-1α DM resulted in a higher overall level of HIF-1α, and HIF-1α was similarly ISGylated (Fig. 3E). In addition, HIF-1α ISGylation occurred in VHL-deficient 769-P cells, and complementation with functional VHL did not affect HIF-1α ISGylation (Fig. 3F). In sum, we suggest that the ISGylation of HIF-1α plays a role in reducing the free HIF-1α levels upon cellular exposure to IFN, and the ISGylation of HIF-1α is distinct from the ubiquitinylation of HIF-1α.

Plasmids expressing different HIF-1α deletion mutants were cotransfected with a set of plasmids expressing the ISGylation system, followed by pull-down-immunoblotting assays to determine the potential ISGylated domain(s) of HIF-1α. We showed that the ISGylation system caused mobility shifts of HIF-1α ΔODD, D1, and D2 fragments (Fig. 3G). Different HIF-1α mutants containing specific domains were constructed and used to map the ISGylation sites in HIF-1α (Supplementary Fig. S4D). Corresponding domains of these different HIF-1α mutants and ISGylation results are illustrated and summarized in Fig. 3H. Thus, our results suggest that HIF-1α is ISGylated at multiple lysine residues.

Next, we sought to determine the impact of ISG15 and ISGylation on the activity of HIF-1α as a transcription factor. We conducted luciferase reporter experiments using a Twist promoter construct (pTwist-Luc) containing HREs (39). Ectopic expression of ISG15 decreased HIF-1α–driven luciferase expression (Fig. 4A). In addition to hypoxia-targeting genes (e.g., Glut1), HIF-1α is known to regulate expression of genes involved in the EMT (e.g., vimentin; ref. 4). In Fig. 4B, expression of the ISGylation system in 293T cells caused a reduction in the VIM mRNA expression. A similar reduction in the HIF-1α–mediated vimentin protein expression by the ISGylation system was observed under normoxia or DFX treatment (Fig. 4C). These data suggest that ISG15 and the ISGylation system compromise HIF-1α transcriptional activity.

Functional expression of HIF-1α caused elevated expression of EMT markers, such as vimentin, fibronection, Glut1, and Slug, depending on the cell lines involved (Fig. 4D and E, compare lanes 1 and 2). Expression of the ISGylation system reduced the levels of EMT proteins in 293T (Fig. 4D, compare lane 2 to lanes 3 and 4) and 769-P cells (Fig. 4E, compare lanes 2 and 3). We further knocked down ISG15 expression in 769-P cells, and the expression levels of EMT markers including vimentin, Glut1 and fibronectin were found to be significantly higher in 769-P sh-ISG15 cells than sh-Luc cells (Fig. 4F). Similarly, IFN treatment also reduced the hypoxia-induced expression of vimentin and Glut1 (Fig. 4G). Together, our results showed that both hypoxia and IFN stimulate the ISGylation system, and that interactions of ISG15 with HIF-1α reduce the transcriptional activity of HIF-1α.

We studied the potential mechanism(s) underlying the ISG15- and ISGylation-mediated reduction of HIF-1α activity. Co-immunoprecipitation experiments showed that ISG15 disrupted HIF-1α/HIF-1β dimerization (Fig. 5A). Unexpectedly, our results revealed that ISG15 expression, like VHL expression, promoted HIF-1α ubiquitinylation (Fig. 5B). Under inhibition of protein synthesis, the expression or knockdown of ISG15 shortened (from ∼50 to ∼15 minutes) or lengthened (from ∼50 to ∼90 minutes) the half-lives of HIF-1α (t_1/2), respectively (Fig. 5C and data not shown). In Fig. 5D, dot-blotting analyses revealed that the overall levels of HIF-1α were reduced with expression of the ISGylation system. Notably, the ISG15 conjugation ability plays a role in that it reduces free HIF-1α level, because the conjugation-defective ISG15_G→A mutant has no effect on free HIF-1α levels (Fig. 5E). Proteasome inhibitors (MG-132 and lactacystin) both restored HIF-1α level (Fig. 5F) and the expression of ISGylation efficiently reduced the expression of HIF-1α P402/564A double mutant (DM) (Fig. 5G). The accumulation of HIF-1α due to DFX treatment appears to be slowed down in cells with ISG15 expressed (data not shown). In sum, our data suggest that ISG15 and/or ISGylation regulate HIF-1α–mediated gene expression via the following mechanisms: (i) prevention of the dimerization of HIF-1α with HIF-1β; (ii) promotion of the ubiquitinylation of HIF-1α; and (iii) accelerated degradation of HIF-1α.

HIF-1α regulates the expression of more than 100 genes encoding proteins that function in various biologic processes such as tumorigenesis (4). We next determined the effect of ISG15 expression on tumor growth. ISG15 expression reduced clonogenic growth (Fig. 6A) and cellular proliferation (Fig. 6B) of cells. Importantly, ISG15 expression in 769-P and Caki-1 not only reduced anchorage-independent growth but also tumor growth in a mouse xenograft model (Fig. 6C and D). We confirmed the elevated levels of ISG15 and/or ISGylates in tumor samples extracted from mice (Fig. 6E). Consistently, HIF-1α ΔODD expression promoted anchorage-independent growth ability in Caki-1 cells (∼2.4-fold), and this HIF-1α–promoted tumorigenic growth ability was abolished by ISG15 expression (Fig. 6F). In conclusion, our results provide the first experimental evidence that ISG15 and ISGylation reduce transcriptional activity of HIF-1α, and subsequently affect HIF-1α-mediated biologic functions, such as EMT and tumorigenic growth.

Under hypoxia, HIF-1α regulates the transcription of a set of genes. In addition, IFN promotes gene expression and two PTM pathways in inflammation, namely ISGylation and ubiqutinylation. Notably, both PTM pathways play regulatory roles in IFN response, and UPP contributes to hypoxic response through regulation of HIF-1α stability (31, 40, 41). Abnormalities in hypoxic and inflammatory IFN responses are frequently observed during tumorigenesis (4, 26–28, 35). Here, we showed that the IFN response regulates the hypoxic response via novel interactions of ISG15 and ISGylation with HIF-1α. We had identified the physical interaction of ISG15 with HIF-1α and that HIF-1α is a novel ISGylation substrate. In the presence of increased ISG15 levels, HIF-1α transcriptional activity and tumor growth are greatly compromised.

Activation of HIF-1α proteins in response to hypoxic stress induces transcriptional programming associated with various biologic processes, such as angiogenesis and glycolysis. A complex of regulatory mechanisms involved in the hypoxic response via modulation of the stability and transcriptional activity of HIF-1α has been reported (42–48). Notably, regulators of HIF-1α also contribute to hypoxia-mediated biologic processes through fine-tuning of HIF-1α expression and activity. Among these regulators that include PHD2/3, FOXO3a, CITED2, and RBX2, their gene expression levels are also regulated by HIF-1α, thus establishing a negative feedback loop of HIF-1α activity. Moreover, HIF-1α is ubiquitinylated and/or SUMOylated during hypoxia, leading to protein degradation (41). The present study addressing the interaction between HIF-1α and ISG15 has uncovered novel mechanisms by which IFN might impact the hypoxia response through the: (i) physical interaction between HIF-1α and ISG15, (ii) ISGylation of HIF-1α, and (iii) modulation of HIF-1α protein stability.

The ISGylation of RIG-I reduced both basal and virus-induced IFN promoter activity (49). We found that ISG15 can be conjugated to HIF-1α, thus leading to reductions in both the cellular level of the free-form HIF-1α and the HIF-1α function. We have further determined that ISG15 conjugates with multiple domains on HIF-1α. However, we were unable to map the lysine residues within HIF-1α that are involved in ISGylation. With regard to the molecular mechanism(s) responsible for the reduced transcriptional activity of HIF-1α with elevated cellular ISG15 level, we found that ISG15 reduces the heterodimer formation of HIF-1α and -1β. Moreover, the ISGylation on the DNA-binding and transactivation domains of HIF-1α might contribute to the reduction of transcriptional activity. Unexpectedly, we also observed that the ISGylation system indirectly promotes HIF-1α ubiquitinylation and, possibly, degradation. These interactions of ISG15 with HIF-1α suggest the existence of cross-talk between inflammatory and hypoxic environments.

Under hypoxia, activation of HIF-1α induces regulated gene expression including genes encoding negative regulators of HIF-1α (42–48). We, thus, examined the promoter regions of genes encoding the ISGylation system, and found HREs in the promoter regions of most of ISGylation/deISGylase genes including ISG15 (4 HREs), Ube1L (1), UbcH8 (1), Herc5 (2), EFP (1), Ubp43 (2), and Ubp21 (1), suggesting a potential transcription regulation of the ISGylation activity by HIF-1α. Consistently, we found that hypoxic conditions and HIF-1α affect ISG15 mRNA expression and the overall ISGylate levels. Together, our findings suggest that the hypoxia response and the ISGylation system regulate each other. Our results support a feedback regulation of HIF-1α–mediated biologic functions, in part, through the HIF-1α–induced upregulation of the ISG15 and ISGylation system, and the negative impact of ISG15 and ISGylation on the HIF-1α pathway.

Abnormal expression and activity of IFN and hypoxic responses have been detected in malignancies (26–29, 35), and a subset of IFN responses have been reported to act as key cytotoxic determinants for DNA-damaging agents (6, 7). Correlations between elevated expression of ISGs, slower tumor growth, and radioresistance have been demonstrated in an animal model (25). However, it remains unresolved why the increased expression of ISGs can slow down cancer cell proliferation but favor resistance to chemotherapy. Our demonstration on the negative regulatory role of ISG15 to HIF-1α activity might provide such an explanation. A negative regulator of HIFs, PHD2, has recently been demonstrated to not only inhibit tumorigenic growth but also to promote resistance to chemotherapy (50). Our results, therefore, suggest that a deficiency of ISG15 and/or ISGylation deficiency may not only affect cancer development but also augment chemotherapeutic responses.

No potential conflicts of interest were disclosed.

Conception and design: Y.-H. Yeh, M.-Y. Hsieh, Y.-C. Yeh, T.-K. Li

Development of methodology: Y.-H. Yeh, Y.-C. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-H. Yeh, Y.-C. Yang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-H. Yeh, M.-Y. Hsieh

Writing, review, and/or revision of the manuscript: Y.-H. Yeh, T.-K. Li

Study supervision: T.-K. Li

This work was financially supported by a Frontier Grant from the National Taiwan University as well as by grants from the National Science Council (NSC-96-2320-B-002-081-MY3) and the National Health Research Institute (NHRI-EX101-9939NI).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.