A Potent In Vivo Antitumor Efficacy of Novel Recombinant Type I Interferon

Type I IFN (IFN1) family of antiviral cytokines comprises 13 different subtypes of IFNα, as well as IFNβ, IFNϵ, IFNκ, IFNω, etc (1–3). Potent antiproliferative, proapoptotic, antiangiogenic, and immunomodulatory effects of IFN1 prompted their use for anticancer treatment (reviewed in refs. 4, 5). However, after more than 40 years of trials, the use of IFN1 against tumors is limited by the suboptimal ratio between clinical efficacy and the severity of its side effects (6), as well as limited response rate, which is often attributed to the downregulation of IFN1 receptor (7). This heterodimeric receptor complex encompassing the IFNAR1 and IFNAR2 chains mediates all effects of IFN1 on cells (8–10). Levels of IFN1 receptor were indeed shown to correlate with IFN1-induced growth arrest (11) and apoptosis in the tumor samples (12, 13).

The levels of IFN1 receptor on cell surface are largely regulated by the ubiquitin-mediated internalization and degradation of IFNAR1 (10, 14–18). Downregulation of IFNAR1 can be accelerated in some cancers (19–22), thereby limiting the antitumorigenic effects of IFN1. Remarkably, although activation of the JAK–STAT pathway is required for both antiviral and antitumor effects of IFN1, lower receptor density still allows efficient antiviral responses while impeding ability of IFN1 to suppress cell proliferation (23). Schreiber and colleagues have proposed that responses to IFN1 could be classified as “robust” (such as antiviral effects) or “tunable” (such as antiproliferative or proinflammatory), the latter being much more sensitive to receptor density (24). Indeed, high cell-surface receptor density and maximal receptor occupancy by relatively high doses of ligands are required to mount an efficient antiproliferative effect (24, 25).

Furthermore, the affinity of IFN1 subtypes for the extracellular domain of IFNAR1 correlates with the ability of these subtypes to elicit specific antiproliferative effect (26–29). Thus, antitumorigenic efficacy of IFN1 may be optimized by increasing cell-surface receptor density and/or by designing novel recombinant IFN1 species that display a greater affinity for IFNAR1. A number of IFN1 variants were generated and shown to be effective against tumor cells. For example, a mutant derivative of IFNα-2, IFNα-YNS exhibited tight binding to IFNAR1 and elicited potent proapoptotic activity and antiproliferative/antiangiogenesis effects in vivo; this mutant surpassed IFNα-2 in antitumorigenic activity in a breast cancer xenograft (28, 30).

Yet, another approach to increase efficacy of IFN1 treatment is to improve its pharmacokinetics and biological activities. Various efforts in this direction include the use of IFNα-2b-albumin fusion protein (31), antibody armed with IFN1 (32), and pegylation of IFN1 (33). Furthermore, given that many of antitumorigenic effects of IFN1 are mediated by the stromal cells, generation of an elegant transgenic mouse model that expresses human IFNAR1 and IFNAR2 subunits, and can be used for transplantation of human tumors, resulted in improved ability to test the antitumorigenic effects of IFN1 (34).

Here, we characterized antitumorigenic properties of a novel recombinant IFN1 derived from human IFNα-2b and other IFN1 subtypes by mutagenesis and termed “super-compound interferon-I” (sIFN-I). Compared with IFNα-2b, sIFN-I exhibited higher anti-HIV activity in SCID mice reconstituted with human peripheral blood leukocytes (35). Current studies revealed that sIFN-I exhibits increased affinity for IFNAR1 and has greatly improved pharmacokinetics and signaling in human and mouse cells. sIFN-I robustly inhibits intratumoral angiogenesis and suppresses growth of transplantable and genetically engineered tumors in immunodeficient and immunocompetent mice. We discuss the direct and indirect mechanisms of potent antitumorigenic effects of sIFN-I and potential perspectives of its use in human cancer treatment.

The novel recombinant super-compound IFN (sIFN-I) and interferon IFNα-2b were provided by Sichuan Huiyang Life Science & Technology Corporation and Shanghai Huaxin Biotechnology, respectively. Human IFNβ (#: 10704-HNAS) and murine IFNβ (#: 50708-M02H), M-CSF (#: 11792-H08H), recombinant type I IFN receptor subunit extracellular domain IFNAR1-EC (#: 13222-H08H) and IFNAR2-EC (#: 10359-H08H) were purchased from Sino Biological Inc. Recombinant B18R protein (vaccinia virus-encoded neutralizing type I interferon receptor) was purchased from eBioscience (#: 14-8185).

Crystals of super interferon (sIFN-I) were grown by the hanging-drop vapor diffusion method (3 mg/mL protein concentration) at 20°C with, in the buffer of 1.2 mol/L Li₂SO₄, 0.1 mol/L 3-(cyclohexylamino)-1-propanesulfonic acid, pH 11.1, 0.02 mol/L MgCL₂. Before data collection, the crystals were equilibrated in a solution containing paraffin oil for a few seconds, and then flash cooled in a liquid nitrogen stream at −173°C. Original data collection to 2.6 Å resolutions was conducted by using the synchrotron radiation from beamline BL5A at a photon factory in Tsukuba, Japan. Primary structural determination was achieved by a combination of molecular replacement method. The position of the sIFN-I was found by molecular replacement using PHASER with the crystal structure of IFNα (Protein Data Bank name: IB5L) used as the search model. The final sIFN-I structure was refined by using molecular modeling techniques and a computerized optimization program, CNS1.1.

On the basis of surface plasmon resonance technology, binding affinities of both IFNα-2b and sIFN-I toward recombinant extracellular (EC) domain of type I interferon receptor subunit IFNAR1-EC or IFNAR2-EC were measured using the Biacore T100 Protein Interaction Array system (General Electric HealthCare Co.). For immobilization of the receptor subunit via binding the carboxylated dextran surface of the chip via amino groups in protein, a CM5 sensor chip was incubated with the IFNAR1-EC subunit and IFNAR2-EC subunit, at 20 and 50 μg/mL, respectively. The two tested IFNs were then injected perpendicularly to ligands at different concentrations within the range of 100 to 3,000 nmol/L for IFNα-2b/IFNAR1 binding, 50 to 1,000 nmol/L for sIFN-I/IFNAR1 binding, and 3.125 to 80 nmol/L for both of them on IFNAR2 binding. During IFNs/IFNAR2 binding, a 5-second regeneration procedure with 2 mol/L NaCl was added between each step of concentration. Data were analyzed by using Biacore T100 software. Dissociation constants K_D were determined from the rate constants according to the Equation K_D = k_d/k_a (d, dissociation; a, association).

Human amnion epithelium WISH cells, all human (A549, HeLa, HT-29, and SMMC-7721) and murine (MC38, LLC, and B16F10) cancer cell lines were cultured in their complete conditional medium; primary murine melanoma cell line YUMM was cultured as reported previously (36). Lentiviral shRNA targeting sequences were used for knocking down expression of IFNAR1 in WISH cells. For the construction of A549-IFNAR1-KO cells, the IFNAR1 gRNA targeting sequences were inserted into the Cas9/gRNA target vector LentiCRISPR (37). Lentivirus was packaged and used to infect parental A549 cells. The IFNAR1-negative cell clones were selected with 0.2 μg/mL puromycin and then confirmed by FACS assay and immunoblot. Detailed information about the cell lines and cell culture, shRNA, and sgRNA sequences are provided in Supplementary Materials and Methods and Supplementary Table S1.

Spleen, lymph nodes (including inguinal, brachial, axillary, bilateral superficial cervical, and mesenteric lymph nodes), liver, and small intestinal epithelial tissue were isolated from C57BL/6 mice. Briefly after organs were mechanically disaggregated, primary splenocytes and liver cells were obtained and resuspended in PBS after depletion of red blood cells. For isolation of small intestinal epithelial tissue cells, the intestinal tube of 3-cm length distant from the connection with stomach was cut out, and the interior side was washed from one end by using syringe and sterile PBS. Cells were scraped off with the edge of a cover glass, counted, and collected for further cell culture or mRNA extraction by TRIzol.

Bone marrow cells were flushed from the femurs and tibias of sacrificed C57BL/6 mice and then depleted for red blood cells using red cell lysing solution. The cells (1 × 10⁷ cells/well) were cultured in 6-well plates in medium supplemented with 20 ng/mL macrophage colony–stimulating factor (M-CSF). Nonadherent cells were carefully removed, and fresh conditional medium was added every 2 days. On day 5, the adherent murine bone marrow–derived macrophage cells were collected for further treatment.

Female nude mice (6–8 weeks old) and female C57BL/6 mice (8 weeks old) were purchased from Shanghai SLAC Company. C57BL/6 Ifnar1^+/+ or Ifnar1^−/− mouse (strain: B6.129S2-Ifnar1^tm1Agt/Mmjax) was purchased from The Jackson Laboratory. More detailed information for nude mice models and syngenic transplantable model is provided in Supplementary Materials and Methods. The experiments and animal procedures conducted at Shanghai Institute of Biochemistry and Cell Biology (Shanghai, China) were approved by the Institution Animal Care and Use Committee (IACUC, protocol recording code: IBCB0029REV1). Experiments and all animal procedures conducted at the University of Pennsylvania (Philadelphia, PA) were approved by the IACUC (protocols # 803995). Female C57BL/6 mice harboring Tyr::CreER^T2; Braf ^CA/+; Pten^f/f alleles (which, upon tamoxifen treatment, were converted into Braf^V600E/+, Pten^Δ/Δ specifically in melanocytes) were kindly provided by Drs. McMahon (University of California, San Francisco, San Francisco, CA) and Bosenberg (Yale University, New Haven, CT). Induction of malignant melanoma by tamoxifen treatment was carried out as described previously (22, 38).

For pharmacokinetics studies, female C57BL/6 mice (8 weeks old; Shanghai SLAC Co.) were injected intraperitoneally with sIFN-I or IFNα-2b. All mice in each IFN-treated group (n = 9, further divided into three subgroups) simultaneously received a dose of 50 μg/kg in PBS. Blood samples from each group were collected after 5, 15, 30 minutes, 1, 1.5, 2, 4, 6, 12, and 24 hours from the retro-orbital sinus (subgroup I: 5 minutes, 1, 4, and 24 hours; subgroup II: 15 minutes, 1.5 and 6 hours; subgroup III: 30 minutes, 2 and 12 hours). Serum was obtained by centrifugation at 10,000 rpm for 10 minutes at 4°C and was stored at −80°C. Untreated mice (n = 3) served as negative control. To determine half-life of the two IFNs in serum, the concentration values, determined from ELISA measurements (VeriKine Human IFN Alpha ELISA Kit, #: 41100, PBL Assay Science Inc.), were plotted against time postinjection and numerically fitted using WinNonlin version 6.2 software (Pharsight) as described elsewhere (39). Noncompartmental models were assumed. Data (including SDs) and curve fits were finally plotted with GraphPad Prism 5.

A549 IFNAR1^−/− cells (3 × 10⁵) were seeded into 6-well plates. After 24 hours, the cells were dissociated with cell dissociation buffer (#: 13151-014, Life Technologies), centrifuged at 1,500 rpm for 5 minutes in FACS tube, and washed with 1× PBS once. Then, cells were stained with the self-made mouse anti-human IFNAR1 antibody (1:1,000 diluted in 1% BSA-PBS) for 30 minutes at room temperature. After washing with PBS, cells were stained with AF488-conjugated goat anti-mouse IgG (1:1,000 diluted in 1% BSA-PBS) for 30 minutes. Cells stained with IgG isotype and secondary Ab only were used as negative control and were then analyzed.

For detection of cell populations in spleen from tumor-bearing mice (Braf^V600E/+, Pten^Δ/Δ), splenocytes were suspended after red cell lysis. Then, cells were incubated with Fc blocker antibody for 15 minutes at room temperature. Subsequently, specific antibodies (listed in Supplementary Table S2) were added and staining was continued for 20 minutes on ice. After a washing step, cells were stained with 0.5 μg/mL DAPI and were then analyzed immediately. Flow cytometry data acquisition was performed by LSRFortessa machine (BD Biosciences), and analysis was performed using FlowJo software.

Immunoblots, immunofluorescent analysis, and other immunologic techniques using antibodies are listed in the Supplementary Information and have been described in our previous publications (15–17). For details on the methods for RNA extraction, cDNA synthesis, quantitative PCR, the information of the synthesized primers, H&E staining, cellular senescence detection of paraffin sections, immunofluorescent analysis of frozen sections, for cell viability assay on human and mouse cells and illustrator image processing, data analyzing, and statistics are described in Supplementary Materials and Methods and our previous publications (22).

Comparisons between experimental groups were performed using the Student t test and GraphPad Prism 5 software (GraphPad Prism software Inc.). All data were shown as Mean ± SEM. Statistically siginificant differences are indicated in figures by single (P < 0.05), double (P < 0.01) or triple (P < 0.001) symbols (such as * or #).

Primary structural analysis showed that sIFN-I has 89% amino acid sequence homology with IFNα-2b (Fig. 1A). The crystal structure of sIFN-I was solved at 2.6 Å resolution; the resulting structure showed that sIFN-I is mainly composed of six helixes (A–F as shown in Fig. 1A) and two distinct loops (AB and BC). This structure was generally comparable with the one previously reported for IFNα-2b (40). Nevertheless, a difference between these proteins was noted in the structure of AB loop (residues 25–33: SPFSCLKDR) and BC loop (residues 44–52: DGNQFQKAQ; Fig. 1A and Supplementary Fig. S1A). Given previously published data regarding putative role of these loops in the interaction with the ligands (30), we next sought to determine relative affinities of sIFN-I for the receptor chains IFNAR1 and IFNAR2.

Surface plasmon resonance assay indeed demonstrated different receptor-binding affinities for sIFN-I and IFNα-2b. Under the condition of steady-analysis model used in this experiment, sIFN-I exhibited greater affinity for the extracellular domain of IFNAR1 [K_D 6.003 × 10⁻⁷ mol/L (0.6 μmol/L)] than IFNα-2b [K_D 2.835 × 10⁻⁶ mol/L (2.8 μmol/L); Fig. 1B and D]. Affinity constant of the extracellular domain of IFNAR2 chain analyzed by the dynamic-analysis model exhibited K_D for sIFN-I of 2.192 × 10⁻⁸ mol/L (21.9 nmol/L) and K_D for IFNα-2b of 1.843 × 10⁻⁹ mol/L (1.84 nmol/L). Compared with IFNα-2b, sIFN-I displayed a higher affinity to IFNAR1 (4.72-fold) but lower affinity for IFNAR2 (11.9-fold; Fig. 1C and D). These properties distinguish sIFN-I from other IFN1 variants, such as IFN-YNS and IFN-YNS-α8tail, which exhibit increased affinities to both IFNAR1 and IFNAR2 (24). In fact, with weaker binding toward IFNAR2 but stronger binding to IFNAR1, sIFN-I is mostly reminiscent of properties reported for IFNα-21 (29) that shared 95 % homology with sIFN-I (Supplementary Fig. S1B).

We next compared signaling elicited by sIFN-I and IFNα-2b in human A549 or HeLa cells. A similar extent of STAT1, STAT2, and STAT3 tyrosine phosphorylation was detected after administering both IFN1 types. However, IFNα-2b–induced signaling was more sensitive to inhibition by the vaccinia virus–derived B18R protein mimicking soluble IFN1 receptor and known to inhibit IFN1 pathway via ligand squelching (41) in both cell lines (Fig. 2A and B). This result suggests that sIFN-I may exhibit an enhanced signaling capacity under signaling limiting conditions.

Recombinant IFN1 proteins were shown to opportunistically bind other receptors besides IFNAR1/2, such as the opioid receptors (42–44). Thus, we sought to determine whether signaling by sIFN-I depends on canonical IFNAR1/2-JAK-STAT pathway. Experiments in human fibrosarcoma 2fTGH cells (sensitive to IFN1) and derivative clones lacking IFNAR2 (U5A) or JAK1 (U4A) revealed that both IFNAR2 and JAK1 are required for sIFN-I–induced phosphorylation of STAT1 and STAT3 (Fig. 3A).

Consistent with these results, sIFN-I did not induce the expression of IFN-stimulated genes (ISG), such as ISG15 or CCL5 in either U4A or U5A cells (Fig. 3B). Furthermore, RNAi-mediated knockdown of IFNAR1 attenuated sIFN-I–induced phosphorylation of STAT1/STAT3 and expression of TRAIL in WISH cells (Fig. 3C and D and Supplementary Fig. S2), suggesting an important role of IFNAR1 in sIFN-I signaling. To corroborate these data, we used CRISPR/Cas9 approach to knock out IFNAR1 in human A549 cells (Supplementary Fig. S3). A robust phosphorylation of STAT1 observed in response to IFNγ (which utilizes type II IFN receptor; ref. 45) in selected IFNAR1^+/− or IFNAR1^−/− clones demonstrated that these cells do not harbor defects in JAK signaling. Importantly, STAT1 phosphorylation in IFNAR1-deficient clones was not induced by sIFN-I (Fig. 3E). These data suggest that sIFN-I signals through the IFNAR1/IFNAR2–JAK pathway in human cells.

Poor sensitivity of mouse IFN1 receptor to human IFN1 species and suboptimal pharmacokinetics of IFN1-based agents pose a challenge for efficient testing of biological effects of human IFN1 (34). Notably, treatment of primary mouse cells with sIFN-I revealed that activity of this ligand in induction of ISGs (Irf7 and Isg15) is superior to that of human IFNα-2b. All these effects were dependent on IFN1 receptor status as evident from the lack of sIFN-I–induced gene expression increase in Ifnar1 knockout mice (Fig. 3F).

We further compared pharmacokinetics of sIFN-I and IFNα-2b in mice after intraperitoneal injection of these agents. To this end, blood samples were taken at fixed time points after IFN administration, and IFN concentrations in serum were assessed by ELISA, followed by numerical analysis using WinNonlin6.2 software (Fig. 4A and B). The pharmacokinetic parameters of sIFN-I and IFNα-2b after administration at the same dose are summarized in Supplementary Table S3. At 15 minutes after injection, the mean serum peak concentration (C_max) for IFNα-2b was 16,730 pg/mL. However, the C_max of sIFN-I with 9,915 pg/mL was delayed to 1 hour after administration. Despite the C_max differences between sIFN-I and IFNα-2b, the area under concentration versus time curve [AUC (0−ι)] for sIFN-I and IFNα-2b exhibited comparable value (27,425 and 24,648 pg per hour/mL, respectively) at the same dosage. Such pharmacokinetics data suggested volume distribution (Vz-F) of sIFN-I (4,384 mL/kg) is more extensive than that of IFNα-2b (2,055 mL/kg) at steady state. In other words, the tissue concentrations of sIFN-I were higher than that of IFNα-2b. Consistent with these data, the induction of expression of IFN-induced genes Irf7 and Isg15 in mouse lymph nodes, spleen, liver, and intestinal epithelial cells was notably greater after treatment of mice with sIFN-I compared with IFNα-2b (Fig. 4C–F) treatment. In all, these data suggest that compared with IFNα-2b, sIFN-I exhibits a greater distribution in mouse tissues and accordingly elicits a greater IFN-stimulated genes induction in these tissues.

We next compared the antitumorigenic properties of sIFN-I and of IFNα-2b. These agents administered at the doses of 50 to 150 μg per mice were reasonably well tolerated by the A549 or HT-29 tumor-bearing immunocompromised mice; these mice did not exhibit body weight loss during the course of treatment (Supplementary Fig. S4). Whereas a modest inhibition of tumor growth was elicited by IFNα-2b, administration of sIFN-I robustly suppressed this growth and led to a stable disease (Fig. 5A and Supplementary Fig. S4). Analysis of tumor tissues revealed that sIFN-I treatment increased cell senescence markers (senescence-associated β-galactosidase) and dramatically decreased the rate of cell proliferation (assessed by Ki67 staining). Accordingly, an increased expression of p53 tumor suppressor protein as well as cyclin-dependent kinase inhibitors p21 and p27 was found in tumor tissues from mice treated with sIFN-I (Supplementary Fig. S5).

When tested for growth inhibition in vitro, both IFNα-2b and sIFN-I exhibited robust effects on human WISH cells at the dose of 1 μg/mL (Fig. 5B). A greater dose (50 μg/mL) was required to detect modest inhibitory effect of either of these IFN1 agents on growth of A549, HT-29 human cancer cells, and MC38 mouse cancer cell line. Under these conditions, sIFN-I was slightly more efficient than IFNα-2b, while growth of some of human (SPC-A4) or mouse (YUMM) cancer cell lines in vitro was not inhibited by IFN1 even at 50 μg/mL (Fig. 5C). Given that IFN1 can act on tumor vascularization and antitumor immunity (45), it is plausible that these indirect mechanisms may contribute to potent antitumorigenic effects of sIFN-I observed in vivo.

Treatment of C57BL/6 mice bearing a syngeneic B16F10 melanoma with sIFN-I but not IFNα-2b resulted in suppression of tumor growth (Supplementary Fig. S6A). sIFN-I also suppressed tumor growth in mice burdened with syngeneic colorectal (MC38) or lung (LLC) adenocarcinomas (Supplementary Fig. S6B and S6C). These results suggest that sIFN-I can elicit its antitumorigenic activities in immunocompetent hosts.

To further understand the antitumor effects of sIFN-I on tumor host, we tested its action in immunocompetent C57BL/6 mice inoculated with syngeneic murine melanoma cell line YUMM (Braf^V600E/+/Pten^Δ/Δ/Cdkn2a^Δ/Δ). Administration of sIFN-I into tumor-bearing Ifnar1^+/+ mice led to a dramatic suppression of growth of transplanted tumor. Importantly, when Ifnar1 knockout animals were chosen as hosts, tumors grew more aggressively and did not respond to treatment with sIFN-I (Fig. 6A). Given this Ifnar1-dependent difference in responses to sIFN-I and the fact that YUMM cells were poorly sensitive to growth inhibition by sIFN-I in vitro (Fig. 5B and C), these results suggest that sIFN-I can suppress tumor growth through affecting tumor stromal compartment.

Consistent with this possibility, compared with untreated animals or treated Ifnar1^−/− mice, tumors from sIFN-I–treated Ifnar1^+/+ mice contained fewer blood vessels and were less positive for endothelial marker CD31 (Fig. 6B and C). Furthermore, these tumors contained a greater number of CD3⁺CD8⁺ cytotoxic lymphocytes (Fig. 6B and C). These results support a notion that sIFN acts on tumor stromal compartment and may impede tumor growth via inhibiting tumor angiogenesis and increasing tumor infiltration by CD3⁺CD8⁺ cytotoxic lymphocytes (indicative of reversing tumor immunosuppression) in an IFNAR1-dependent manner.

Having observed a robust therapeutic effect of sIFN-I in transplanted tumors, we sought to determine whether this agent can also be active in genetically engineered models. To this end, we induced melanoma tumors in Braf^V600E/+; Pten^Δ/Δ mice and started the treatment after establishing tumors with the average size of 51 mm³ in both groups. Administration of sIFN-I notably suppressed growth of these tumors (Fig. 6D). When all control mice receiving vehicle had to be sacrificed for humane reasons (i.e., tumor size reaching the limit required by IACUC), animals receiving sIFN-I exhibited either stable disease or partial/complete tumor regression (Fig. 6D and E).

In this model, sIFN-I did not noticeably affect infiltration of tumors with CD31-positive cells. However, consistent with tumor regression, we observed significant increase of infiltrating cytotoxic lymphocytes in tumors treated with sIFN-I in genetically engineered mouse melanomas (Fig. 6F and Supplementary Fig. S7). Furthermore, sIFN-I notably suppressed metastases of genetically engineered melanoma into the lymph nodes (Fig. 6G). These results strongly suggest that sIFN-I exhibits a potent antitumorigenic effect against primary tumors and metastatic disease.

Endogenous IFN1 plays an important role in protection against tumors due to their antiproliferative, antiangiogenic, and immunostimulating activities (2). The response rate and therapeutic efficacy of IFN1-based pharmaceutical agents is limited, especially in solid tumors (4, 6) because oncogene signaling, tumor microenvironment stress, unfolded protein response, and inflammation can decrease the levels of IFNAR1 available for ligand interaction (19–21, 46, 47). Besides developing means to reverse downregulation of IFNAR1, additional solutions for optimizing IFN1 therapy can be based on the observation that antitumorigenic efficacy of diverse IFN1 subtypes parallels affinity of these types for IFNAR1 (48, 49). Here, we describe sIFN-I, a novel recombinant IFN1 exhibiting increased affinity for IFNAR1 and potent antitumorigenic properties.

Intriguingly, although it tightly binds to IFNAR1, sIFN-I exhibits a lesser affinity for IFNAR2 (normally a chain with greater affinity for endogenous ligands; ref. 48) compared with its “parental” molecule IFNα-2b (Fig. 1), which is different from the other reported IFN variants, such as IFN-YNS and IFN-YNS-α8tail (24). The latter variants displayed enhanced ligand binding affinity to both IFNAR1/2, and also showed enhanced anti-proliferation activity for cancer cells in vitro (28). Whereas in vitro activities of sIFN-I are relatively underwhelming, sIFN-I exerts its potent antitumor effect in vivo (Figs. 5A and 6 and Supplementary Fig. S4A).

sIFN-I elicits notable activation of STAT proteins and ensuing induction of ISGs (Figs. 2 and 3F); importantly, all these effects of sIFN-I depend on integrity of the IFNAR1/IFNAR2–JAK pathway (Fig. 3). Furthermore, tumor-bearing mice lacking Ifnar1 are poorly responsive to antitumorigenic activities of sIFN-I (Fig. 6). These results suggest that despite (or because of) potentially altered ligand–IFNAR1–IFNAR2 complex, sIFN-I robustly activates this receptor and downstream IFN1 signaling pathway.

Remarkably, compared with human IFNα-2b, the effects of sIFN-I appear to transcend the species differences. Data presented here reveal that sIFN-I elicits the IFN1-stimulated gene induction responses in primary mouse cells and mice in vivo (Figs. 4 and 5). Furthermore, in terms of pharmacokinetics in mouse, sIFN-I exhibited longer half-life and lower peak drug concentration in serum compared with IFNα-2b (Fig. 4). Intriguingly, there was a two-step serum increase for sIFN-I; this phenomenon was not observed for IFNα-2b injected into mice. These differences could be attributed to the different binding model for sIFN-I toward plasma protein or lipoprotein in blood, which lead to re-release of sIFN-I from the sIFN-I/plasma protein or sIFN-I/lipoprotein dynamic binding complex (50). Such possibility would be consistent with two peaks in concentration–time curve for serum concentration of 2′5-OAS (a well-known downstream marker of the pharmacodynamic activity of IFN) observed in blood after sIFN-I subcutaneous injection for the healthy volunteers (51). Altered pharmacokinetic characteristics of sIFN-I may contribute to greater ISG induction and improved antitumorigenic activities in vivo (Fig. 5) and, furthermore, may potentially cause lesser side effects. These possibilities in humans will be revealed by clinical trials of sIFN-I in Singapore (CTC1300056) and United States (NCT02464007) that are currently conducted in patients with solid tumors.

Previous published data suggested that sIFN-I can suppress the tumor growth in some isolated clinical cases in human patients (52). Our current data demonstrate greater efficacy of sIFN-I over IFNα-2b against human tumors xenotransplanted into immunocompromised mice (Fig. 5A). Given a robust response of mouse tissues to sIFN-I, this response may at least in part be attributed to the effects of sIFN-I on mouse stromal cells. Indeed, in immunocompetent syngeneic transplantation or genetically engineered mouse melanoma models, sIFN-I notably suppressed angiogenesis and/or increased tumor infiltration with cytotoxic lymphocytes. These antiangiogenic and immunostimulatory effects of sIFN-I are likely to contribute to robust antitumorigenic efficacy of sIFN-I that elicit stable disease or/and tumor regression in very aggressive melanoma tumors (Fig. 6). Detailed studies of the mechanisms underlying immunostimulatory and other effects of sIFN-I are ongoing. These studies will be instrumental in designing clinical trials in humans that will address clinical efficacy of sIFN-I alone or in combination with traditional, molecularly targeted or immune-targeted therapies.

G.-W. Wei holds ownership interest (including patents) in Sichuan Huiyang Life Science & Technology Corp. X.-J. Liu is a consultant/advisory board member for Sichuan Huiyang Life Science & Technology Corp. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.-J. Zhang, X.-Y. Liu, S.Y. Fuchs

Development of methodology: K.-J. Zhang, H.-L. Li, X.-L. Fang, Y.V. Katlinskaya, G.-W. Wei, D.-C. Wang, X.-Y. Liu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.-J. Zhang, X.-F. Yin, Y.-Q. Yang, J. Xiao, K.V. Katlinski, Y.V. Katlinskaya, G.-W. Wei, D.-C. Wang, X.-Y. Liu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-J. Zhang, X.-F. Yin, X.-J. Liu, K.V. Katlinski, D.-C. Wang, X.-Y. Liu, S.Y. Fuchs

Writing, review, and/or revision of the manuscript: K.-J. Zhang, X.-F. Yin, L. Chu, R.-B. Guo, G.-W. Wei, D.-C. Wang, X.-Y. Liu, S.Y. Fuchs

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.-J. Zhang, Y.-N. Xu, L.-Y. Chen, X.-J. Liu, S.-J. Yuan, X.-L. Fang, S. Wu, H.-N. Xu, G.-W. Wei, D.-C. Wang, X.-Y. Liu, S.Y. Fuchs

Study supervision: K.-J. Zhang, X.-Y. Liu, S.Y. Fuchs

We are grateful to Drs. McMahon (UCSF), Bosenberg (Yale University), Jiang (Peking University), Melissa Wong (Oregon Health and Science University), and Stark (Cleveland Clinics) for sharing the reagents and to members of Fuchs and Liu labs for insightful comments.

This work was supported by NIH/NCI grant CA092900(to S.Y. Fuchs), Sichuan Science and Technology project 2013ZZ0004 (to K.-J. Zhang), Shanghai Institutes for Biological Science, Chinese Academy of Sciences, and Sichuan Huiyang Life Science and Technology Corp. research program Y363S21763 (to X.-Y. Liu), National Basic Research Program of China 973 Program, no. 2011CB510104 (to X.-Y. Liu), Zhejiang Sci-Tech University grant 1204807-Y(to X.-Y. Liu), Chinese Ministry of Science and Technology fund 2014CB964704(to X.-Y. Liu), and grant from the Sino-American joint laboratory between Conba Group and Zhejiang Sci-Tech University(to X.-Y. Liu).

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.