Interferon-λ1 Linked to a Stabilized Dimer of Fab Potently Enhances both Antitumor and Antiviral Activities in Targeted Cells

The type III interferons (IFNs), comprising IFN-λ1, IFN-λ2, and IFN-λ3, behave similarly to IFN-α in eliciting antiviral, antitumor, and immune-modulating activities. Due to their more restricted cellular targets, IFN-λs are attractive as potential alternatives to existing therapeutic regimens based on IFN-αs. We have applied the DOCK-AND-LOCK™ method to improve the anti-proliferative potency of IFN-λ1 up to 1,000-fold in targeted cancer cell lines by tethering stabilized Fab dimers, derived from hRS7 (humanized anti-Trop-2), hMN-15 (humanized anti-CEACAM6), hL243 (humanized anti-HLA-DR), and c225 (chimeric anti-EGFR), to IFN-λ1 site-specifically, resulting in novel immunocytokines designated (E1)-λ1, (15)-λ1, (C2)-λ1, and (c225)-λ1, respectively. Targeted delivery of IFN-λ1 via (15)-λ1 or (c225)-λ1 to respective antigen-expressing cells also significantly increased antiviral activity when compared with non-targeting (C2)-λ1, as demonstrated in human lung adenocarcinoma cell line A549 by (15)-λ1 against encephalomyocarditis virus (EC50 = 22.2 pM versus 223 pM), and in human hepatocarcinoma cell line Huh-7 by (c225)-λ1 against hepatitis C virus (EC50 = 0.56 pM versus 91.2 pM). These promising results, which are attributed to better localization and stronger binding of IFN-λ1 to antibody-targeted cells, together with the favorable pharmacokinetic profile of (E1)-λ1 in mice (T1/2 = 8.6 h), support further investigation of selective prototypes as potential antiviral and antitumor therapeutic agents.


Introduction
Interferons (IFNs) are class II cytokines with antitumor and antiviral effects [1], and have been explored extensively as therapeutic agents in a variety of diseases [2][3][4]. However, their clinical usefulness to date, as exemplified by IFN-a, IFN-b, and IFN-c, is limited by a short circulating half-life, systemic toxicity, and suboptimal responses in patients [5]. The discovery of the IFN-l family in 2003 introduced a new opportunity to develop alternative IFN agents [6,7].
A major difference between type III and type I IFNs is the distribution of their respective receptor complexes. IFN-a/b signals through two widely expressed type I interferon receptors, which is at least partially responsible for the systemic toxicity associated with IFN-a/b therapy [10]. In contrast, IFN-ls signal through a heterodimeric receptor complex comprising the IFN-l receptor 1 (IFN-lR1) and the IL-10 receptor 2 (IL-10R2). Whereas IL-10R2 is ubiquitously expressed among hematopoietic and nonhematopoietic cells, IFN-lR1 has a more restricted expression pattern, with the highest levels in epithelial cells, melanocytes, and hepatocytes, and the lowest level in primary central nervous system cells [11]. Although blood immune cells express IFN-lR1, they exhibit impaired response to IFN-ls due to the secretion of a short spliced variant of IFN-lR1 that inhibits the effect of IFN-l1 [12]. The limited responsiveness of neuronal cells and immune cells also contributes to the reduced toxicity of IFN-ls, compared to type I IFN [8,12].
Human IFN-l1 conjugated to a 20-kDa polyethylene glycol (PEG-IFN-l1) is currently under clinical development for the treatment of chronic HCV infection. In a phase Ib study, antiviral activity was observed at each dose (0.5-3.0 mg/kg), with viral load reduced 2.3 to 4.0 logs when PEG-IFN-l1 was administered to genotype 1 HCV patients who relapsed after IFN-a therapy [32,33]. A phase IIb study [34] reported that HCV patients (genotypes 1 and 4) had significantly higher response rates to treatment with PEG-IFN-l1 versus PEGinterferon a-2a, with the additional advantages of PEG-IFN-l1 over PEG-interferon a-2a including lower frequency of adverse events, decreased frequency of flu-like symptoms, anemia, and musculoskeletal symptoms, and rarely observed neutropenia and thrombocytopenia. However, higher rates of hepatotoxicity were seen with the highest-dose PEG-IFN-l1 compared with PEGinterferon a-2a [34].
The DOCK-AND-LOCK TM (DNL TM ) platform technology [35][36][37] is a powerful method to construct novel agents of defined composition and retained bioactivity by site-specific conjugation of two types of modules, one containing the dimerization and docking domain (DDD) of cAMP-dependent protein kinase A (PKA), referred to as the DDD module, and the other bearing the anchoring domain (AD) of an interactive A-kinase anchoring protein (AKAP), referred to as the AD module. Among the distinctive features of DNL are the spontaneous formation of a dimer of the DDD module and the self-assembly of the DDD module with the AD module into a non-covalent complex, which is subsequently rendered covalent with disulfide bonds to enhance stability in vivo [36]. The amino acid sequences of a pair of DDD and AD linkers useful for the DNL conjugation are termed DDD2 and AD2, respectively. We have previously demonstrated in preclinical studies the superiority of antibody-targeted delivery of IFN-a-2b using novel conjugates, made by the DNL method, comprising four molecules of IFN-a-2b covalently linked to a humanized IgG with specificity for CD20 [37] or HLA-DR [38]. Here, we report the development of another class of DNL conjugates, referred to as 2(Fab)-l1 for targeted delivery of IFN-l1 via a stabilized dimer of Fab derived from a humanized or chimeric antibody, and show that the greatly enhanced antitumor as well as antiviral potency observed for the four prototypes can be attributed to their increased localization and stronger binding to target cells.

Expression and Purification of DNL Modules
To produce the AD2-IFN-l1 module used for DNL conjugation, a synthetic cDNA sequence encoding the AD2 peptide, linker peptide (EFPKPSTPPGSSGGAP), and human INF-l1 with a single point mutation (C171S) were cloned into the Msc I and Xho I restriction sites of the pET-26b vector ( Figure 1A). Competent Rosetta-pLysS cells (Novagen, Billerica, MA) were transformed with the expression vector and cultured in shaker flasks at 37uC in Difco 2xYT broth (Becton Dickinson, Franklin Lakes, NJ), supplemented with 100 mg/ml kanamycin sulphate and 34 mg/ ml chloramphenicol. When the cell density reached OD 600 = 1.0, cultures were switched to 30uC and protein expression was induced with 0.5 mM IPTG for 4 h. Cell pellets were frozen, thawed and homogenized in a lysis buffer comprising 2% Triton-X 100, 5 mM MgSO 4 , 10 units/ml benzonase (Novagen), 100 mM 4-(2-aminoethyl) benzenesulfonyl fluoride (Sigma-Aldrich, St. Louis, MO), and 20 mM Tris-HCl, pH 8.0. The insoluble material, containing inclusion bodies, was pelleted by centrifugation, re-homogenized in 1% Triton X-100 in PBS, and re-pelleted. Inclusion bodies were solubilized in 6 M guanidine-HCl, 100 mM Na-phosphate, pH 8.0, and applied to a His-Select affinity column (GE Healthcare, Piscataway, NJ). The denatured protein was eluted in 4 M guanidine-HCl, 100 mM NaH 2 PO 4 , pH 4.5. The eluate was neutralized with 3 M Tris-HCl, pH 8.6, added dithioerythreitol (DTE) to 60 mM, and the resulting solution was held at room temperature overnight. The reduced, denatured protein solution was diluted rapidly into 80-fold volume of 0.5 M arginine, 20 mM oxidized glutathione, 2 mM EDTA; 100 mM Tris, pH 8.0, then dialyzed against 5 L of renaturation buffer Expression vectors for Fab-DDD2, -AD2, or IgG-AD2 modules were engineered from IgG-pdHL2 expression vectors for hRS7, hMN-15, and hL243, as described previously [36] ( Figure 1B). The c225-Fab-DDD2 module was constructed using the sequences from DrugBank (Access number: DB00002) as the variable domains of c225. These Fab or IgG modules were produced in myeloma cell culture of SpESFX-10 cells [39] and isolated from culture broths using Kappa or MabSelect affinity chromatography (GE Healthcare, Piscataway, NJ).
(C2)-l1, respectively, using the DNL method described previously [35,36]. The DNL products were purified by sequential chromatography on Kappa-select and His-select affinity columns. Enhanced Activities of IFN-l1 in Targeted Cells Cell Lines TE-11 was a gift of Dr. Hiroshi Nakagawa (University of Pennsylvania) [40]. Huh-7 was from the Japanese Collection of Research Bioresources (Osaka, Japan). All other cell lines were from the American Type Culture Collection (ATCC, Manassas, VA). ME-180, TE-11, and A375 were grown in RPMI 1640 media (Life Technologies, Grand Island, NY) with 10% fetal bovine serum (FBS, Thermo Scientific HyClone, Logan, UT). HepG2, Huh-7, and SK-MES-1 were grown in EMEM media (ATCC) with 10% FBS. A549 was grown in F12 media (Life Technologies) with 10% FBS.

Flow Cytometry
For cell-binding assays, cells were trypsinized briefly, washed, re-suspended in 1% BSA-PBS, and incubated with humanized mAbs (10 mg/ml) or serially diluted IFN-l1-based agents. Binding was detected with either FITC-labeled goat anti-human IgG or with mouse anti-human IFN-l1 followed by FITC-labeled goat anti-mouse IgG. All incubations were 45 min at 4uC with 1% BSA-PBS washes between incubations. Binding was measured by flow cytometry using FACS Calibur (BD Biosciences, San Jose, CA).
Changes of MHC-I expression were evaluated by flow cytometry on FACS Calibur following treatment of cells with indicated agents for 3 days. Cells were probed with FITC-labeled mouse anti-human HLA-ABC. FITC-labeled non-specific mouse IgG 1k was used as a negative control.
In vitro Proliferation ME-180, SK-MES-1, and TE-11 were seeded in 96-well plates (1,000 cells/well) and held at 37uC overnight prior to incubation with the indicated agents for 4 days. Viable cells were measured using a CellTiter 96 Cell Proliferation Assay agent (Promega, Madison, WI).

Antiviral Assays
The anti-HCV activities of rhIFN-l1, (c225)-l1, and (C2)-l1 were measured in Huh-7-Con1 cells (EGFR + /HLA-DR -), which contain HCV genotype 1b Con1 replicon integrated with a firefly luciferase gene to serve as reporter of viral level [41,42]. Huh-7-Con1 cells were treated with each of the three agents for 3 days, and the viral replication level was determined by measuring luciferase activity. Meanwhile, the cytotoxicity of these agents was also evaluated on parental Huh-7 cells using CellTiter Glo kit (Promega). The anti-HCV assay was performed by HD Biosciences (China) Co., Ltd (Shanghai, China).
In addition, the antiviral activity of (15)-l1 was measured by PBL Interferon Source (Piscataway, NJ) on A549 cells (CEA-CAM6 + ) with EMCV using the cytopathic effect inhibition assay [43]. Included in the assay were hMN-15-Fab-DDD2 as a negative control, rhIFN-l1 standard (PBL Interferon Source) as a positive control, and (C2)-l1 as a non-targeting control for structural counterpart.

SDS-PAGE and Immunoblot Analyses
SDS-PAGE was performed under reducing and non-reducing conditions using 4-20% Tris-Glycine gels (Lonza, Allendale, NJ). To evaluate phosphorylation of STATs, HepG2 cells (5610 5 ) were treated with IFN-l1 agents for 1 hour prior to lysis with PhosphoSafe TM extraction reagent (Novagen). Cell lysates were resolved on SDS-PAGE, transferred to nitrocellulose membranes, and probed with rabbit antibodies against total STAT1, STAT2, or STAT3 or the phosphotyrosine-specific antibodies pY-STAT1, pY-STAT2, or pY-STAT3. Signals were detected with HRP-goat anti-rabbit antibody. An anti-b-actin antibody was used for assessing b-actin as the loading control.

RT-PCR Analysis
HepG2 cells were treated with IFN-l1 agents for 24 h and total RNA was isolated using TRIzolH Reagent (Life Technologies). The mRNA expression of the myxovirus resistance A (MxA) gene was analyzed using SuperScriptH III One-Step RT-PCR System (Life Technologies) with primers of forward 59-AGATCCAG-GACCAGCTGAGCCTGT- 39 and reverse 59-GTGGAACTCGTGTCGGAGTCTGGTA-39 at conditions: cDNA synthesis-55uC/30 min for one cycle and PCR-94uC/ 15 sec, 62uC/30 sec, 68uC/30 sec for 25 cycles. Meanwhile, a 452-bp fragment of the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) cDNA was amplified at similar conditions as an internal control [29].

ELISA
MaxiSorp 96-well plates (Nunc, Roskilde, Denmark) were coated overnight with goat anti-human F(ab')2 antibody and blocked with PBS containing 2% bovine serum albumin (BSA) for 1 h. Serum samples for pharmacokinetic study were serially diluted and added to the coated plates. The captured (E1)-l1 was detected with mouse anti-human IFN-l1, followed by HRPconjugated goat anti-mouse IgG Fc antibodies. Pure (E1)-l1 was included as a standard for quantifying serum samples.

Statistical Analyses
Statistical significance (P,0.05) was determined with F tests for all results using the Prism GraphPad software package (Advanced Graphics Software, Rancho Santa Fe, CA).
The ability of (15)-l1 to up-regulate the cell surface expression of MHC class I antigens (MHC-I) also was investigated in HepG2 cells by flow cytometry (Figure 5B). Whereas treatment of HepG2 cells with hMN-15-Fab-DDD2 up to 1 nM for three days gave no change in the surface levels of MHC-1 (MFI ,50), a more than 3fold increase in MFI (,170) was observed in cells treated with (15)-l1 as low as 1 pM. The expression of MHC-1 reached a maximal level (MFI ,270) with (15)-l1 at 100 pM. Although AD2-IFN-l1 was capable of up-regulating MHC-1, it required a higher concentration (1 nM) to be as effective.
We also examined the capability of (15)-l1 to induce the expression of the myxovirus resistance A (MxA) gene, a known marker for IFN bioactivity, and the results of RT-PCR are shown   Figure 5C. In untreated or hMN-15-Fab-DDD2-treated HepG2 cells, the mRNAs of MxA were undetectable by RT-PCR. The induction of the MxA gene was evident in cells treated with 0.1 pM of (15)-l1 or 10 pM of AD2-IFN-l1, thus further attesting to the advantage of (15)-l1 over AD2-IFN-l1 in particular, and 2(Fab)-l1 over IFN-l1 in general.

Pharmacokinetics in Mice
After a single dose (2.4 nmol) of subcutaneous administration, the mean serum concentration of intact (E1)-l1 reached a high level by 6 h and fell below the limit of ELISA detection by 48 h (Figure 6A). The pharmacokinetic (PK) parameters derived from noncompartmental analysis demonstrate a mean residence time of 12 h with a T 1/2 of 8.6 h, and clearance of 2.2 ml/h. When the concentrations of (E1)-l1 in the serum sample were also measured by its inhibitory activity on ME-180 cells using the MTS assay, the results were largely consistent ( Figure 6B).
As shown in our previous study [37], recombinant IFN-a has a very rapid rate of clearance in mice, showing a mean residence time of only 0.7 h. Thus, 2(Fab)-l1 demonstrates a significantly improved PK that is comparable to PEG-IFN-a [37] and PEG-IFN-l [44].

Discussion
The potential of IFN-l as a therapeutic alternative to IFN-a is being explored with PEG-IFN-l1, which shows an improved safety profile over PEG-IFN-a-2a in clinical studies [4,33]. However, rates of some serious adverse events, including doselimiting hepatotoxicity, are similar for patients treated with PEG-IFN-l1 and PEG-IFN-a-2a, and even more frequent in patients treated with the highest dose of PEG-IFN-l1 [34]. On the other hand, IFN-ls were less effective than type I IFNs against certain cancers [26] or viruses [45]. We postulated that linking IFN-ls to an antibody specific for an abundantly expressed surface antigen could enhance its localization at the target cells, resulting in greater potency and, hopefully, less toxicity to non-target cells. Accordingly, we developed four prototypes of IFN-l-based immunocytokines, each comprising IFN-l1 conjugated sitespecifically to a stabilized dimer of Fab, and demonstrated their superiority compared to the unconjugated parental modules alone or in combination. The improved effects were shown in both replicon expressing firefly luciferase was treated with indicated concentrations of (c225)-l1, (C2)-l1, or rhIFN-l1 agents. After 3 days, luciferase activity was measured and antiviral effects were determined by percent activity reduction relative to untreated cells. Data were analyzed by Graph Pad Prism using a sigmoidal fit (variable slope). Samples were run twice independently in duplicate. (B) Enhanced anti-EMCV potency of (15)-l1 in A549 cells. A549 cells were incubated with serial dilutions of (15)-l1, (C2)-l1, rhIFN-l1, or hMN15-Fab-DDD2 before being challenged with EMCV. A visual cytopathic effect (CPE) determination was performed, and the data were analyzed by GraphPad Prism using a sigmoidal fit (variable slope). Samples were run twice independently in duplicate. Mean 6 SD, 95% confidence interval. doi:10.1371/journal.pone.0063940.g004 Enhanced Activities of IFN-l1 in Targeted Cells PLOS ONE | www.plosone.org In the antitumor studies, we found ME-180 to be the most sensitive cell line to IFN-l1, with over 80% maximum inhibition and a 20-fold lower EC 50 than other cell lines reported in the literature (Table S1) [25][26][27][28][29]. Moreover, the abundant expression of Trop-2 on ME-180 renders these tumor cells highly responsive to (E1)-l1, with growth inhibition detectable at 1 fM and an EC 50 (,0.1 pM) 1000-fold lower than AD2-IFN-l1 (EC 50 ,100 pM). Similar enhancements of inhibitory potency also were observed in TE-11, SK-MES-1, and other cancer cell lines. Because IFN-ls also induce innate and adaptive immune responses, which were not evaluated here, and in view of recent studies showing that the constitutive expression of IFN-l in several murine cancer cell lines, including B16 melanoma, BNL hepatoma, and MCA205 fibrosarcoma, despite the lack of in vitro antiproliferative activity, markedly suppressed tumor growth and metastasis in syngeneic mouse models by recruiting immune cells and related cytokines [30,46,47], we posit that targeted delivery of IFN-l will resemble constitutive expression of IFN-l in cancer cells, resulting in a local immune response and enhanced cytotoxicity in the immunotherapy of cancer [48].
In the antiviral studies, the specific targeting of (c225)-l1 to EGFR-positive Huh-7 cells hosting HCV genotype 1b Con1 replicon exhibited 123-and 163-fold enhancement of antiviral potency over non-targeting rhIFN-l1 and (C2)-l1, respectively. In another assay, the targeting of (15)-l1 to CEACAM6-positive A549 cells challenged with EMCV exhibited 6-and 10-fold improvement of antiviral protection over the non-targeting rhIFN-l1 and (C2)-l1, respectively. The difference in potency between (c225)-l1 and (15)-l1 is likely due to the distinct sensitivity of their targeted cell/virus systems to IFN-l1. In a previous study [17], rhIFN-l1 was 10-fold less potent than rhIFN-a in the Huh-7/ equal loading. (B) Flow cytometric analysis of MHC class I antigen expression. HepG2 cells were cultured for three days in the presence of increasing concentrations of (15)-l1, AD2-IFN-l1, or 15-Fab-DDD2 and then stained with FITC labeled mouse anti-human HLA-ABC or isotopic-matched control Ab. MFI, mean fluorescence intensity. N = 10,000 cells. Data were analyzed by FlowJo software (Tree Star, Ashland, OR). (C) RT-PCR analysis of MxA mRNA expression. HepG2 cells were treated with IFN-l1 agents for 24 h, and the mRNA expression of MxA gene was analyzed in 100 ng of total RNA/ sample. GAPDH cDNA was amplified as an internal control for cDNA amounts. Data are representative of two independent experiments. doi:10.1371/journal.pone.0063940.g005 Enhanced Activities of IFN-l1 in Targeted Cells HCV system, and 210-fold less potent in A549/EMCV [26]. Although the enhancement of (15)-l1-induced antiviral activity is not as high in target cells, compared to (c225)-l1, it is still a significant finding, considering that pegylated IFNs only retain about 30% or less activity of unpegylated IFNs [49], and the specific activity of AD2-IFN-l1 was similar to rhIFN-l1. Thus, 2(Fab)-l1 may allow a lower dose with the same or less frequent dosing schedule than the PEG-IFN-l1 used in current clinical studies.
As a therapeutic agent, recombinant IFN is limited by its very rapid rate of clearance. Based on our previous study, recombinant IFN-a-2b, PEG-IFN-a-2a, and PEG-IFN-a-2b exhibited a halflife of 0.7, 14.9, and 9.3 h, respectively [37]. Thus, a comparable half-life of (E1)-l1 in mice (8.6 h) to the two regulatory approved PEG-IFN-as is encouraging, and merits further in vivo experiments in murine and non-human primate models. conditions. Reducing condition resolved three bands representing polypeptides for Fab-DDD2-heavy chain, kappa light chain, and AD2-IFN-l1, and non-reducing condition resolved a major highrelative mobility band representing the covalent DNL structure.