The Function of the Human Interferon- 1a Glycan Determined in Vivo

Recombinant human interferon(rhIFN) is the leading therapeutic intervention shown to change the cause of relapsingremitting multiple sclerosis, and both a nonglycosylated and a significantly more active glycosylated variant of rhIFNare used in treatment. This study investigates the function of the rhIFN1a glycan moiety and its individual carbohydrate residues, using the myxovirus resistance (Mx) mRNA as a biomarker in Mx-congenic mice. We showed that the Mx mRNA level in blood leukocytes peaked 3 h after s.c. administration of rhIFN1a. In addition, a clear dose-response relationship was confirmed, and the Mx response was shown to be receptormediated. Using specific glycosidases, different glycosylation analogs of rhIFN1a were obtained, and their activities were determined. The glycosylated rhIFN1a showed significantly higher activity than its deglycosylated counterpart, due to a protein stabilization/solubilization effect of the glycan. It is interesting to note that the terminating sialic acids were essential for these effects. Conclusively, the structure/bioactivity relationship of rhIFN1a was determined in vivo, and it provided a novel insight into the role of the rhIFN1a glycan and its carbohydrate residues. The possibilities of improving the pharmacological properties of rhIFN1a using glycoengineering are discussed. Interferons (IFN) are natural cytokines produced by cells in most vertebrates as part of the first response to challenges by foreign agents such as virus, bacteria, parasites, and tumor cells. There are three major classes of IFN (I–III) according to the type of receptor through which they signal. IFNis a type I IFN that is known to exert antiviral, antiseptic, antioncogenic, antiproliferative, and immunomodulatory effects (reviewed in Pestka et al., 2004). Type I IFN binds to a transmembrane receptor called the IFNreceptor (IFNAR), which is present on most cells (Pestka, 1997). Upon binding, type I IFN induces a cascade of intracellular events that lead to the transcription of several genes, including the 2 5 oligoadenylate synthase (Rutherford et al., 1988) and myxovirus resistance (Mx) (Staeheli et al., 1986). The Mx gene is activated specifically by type I and type III IFN (Staeheli et al., 1984; Zhou et al., 2007) to produce a 3.5-kilobase mRNA (Staeheli et al., 1986). Several type I IFNs have been approved or are under clinical evaluation as treatment for different diseases. At present, both the glycosylated (IFN1a) and the nonglycosylated (IFN1b) variant of recombinant human (rh)IFNare approved for treatment of the neurological disorder relapsing-remitting multiple sclerosis (RRMS). Both variants reduce the frequency of clinical exacerbations (Duquette et al., 1993), but IFN1b is administered in much higher protein amounts than IFN1a to achieve the same efficacy; i.e., 22 g of IFN1a versus 250 g of IFN1b administered s.c. three times a week resulted in a similar Mx response in peripheral blood leukocytes (Bertolotto et al., 2001). The short half-life of IFNs in the blood circulation dictates frequent injections, and production of neutralizing antibodies are commonly observed, causing the effect of treatment to diminish (Malucchi et al., 2004). Therefore, the bioactivity of the rhIFN1a treatment in RRMS is clinically monitored for each patient by measuring the Mx response (Pachner et al., 2003). Human IFN1a is a glycoprotein comprising 166 amino This work was supported by grants from Warwara Larsen Foundation, Novo Nordisk/Novozymes, Lundbeck Foundation, and the Danish Multiple Sclerosis Society. L.D.-O. and M.T.-A. contributed equally to this work. Article, publication date, and citation information can be found at http://jpet.aspetjournals.org. doi:10.1124/jpet.108.138263. ABBREVIATIONS: IFN, interferon(s); IFNAR, IFNreceptor; Mx, myxovirus resistance; rh, recombinant human; RRMS, relapsing-remitting multiple sclerosis; ko, knockout; HSA, human serum albumin; hpt, hours posttreatment; HPRT1, hypoxanthine phosphoribosyltransferase 1; SQ, starting quantity; bp, base pair; PAGE, polyacrylamide gel electrophoresis; MeCN, acetonitrile; FA, formic acid; MALDI-TOF, matrix-assisted laser desorption ionization time-of-flight; MS, mass spectrometry; ANOVA, one-way analysis of variance; NT, no treatment. 0022-3565/08/3261-338–347$20.00 THE JOURNAL OF PHARMACOLOGY AND EXPERIMENTAL THERAPEUTICS Vol. 326, No. 1 Copyright © 2008 by The American Society for Pharmacology and Experimental Therapeutics 138263/3358506 JPET 326:338–347, 2008 Printed in U.S.A. 338 at A PE T Jornals on D ecem er 1, 2017 jpet.asjournals.org D ow nladed from acid residues in its matured form (Derynck et al., 1980). It contains a single N-glycosylation site (Asn80), and the glycan structures of the protein from different sources have been determined. For example, low heterogeneity of the glycans of Chinese hamster ovary cell-derived rhIFN1a has been reported, and 95% of the structures were found to be biantennary complex glycans containing terminating galactose sialylation and core fucosylation with the overall carbohydrate residue composition NeuAc2Gal2Man3GlcNAc4Fuc1 (Conradt et al., 1987). In addition, the glycans of Chinese hamster ovary cell-derived IFNshow high similarity to the naturally occurring human (Kagawa et al., 1988) and murine (Civas et al., 1988) counterpart. Protein-conjugated carbohydrates are increasingly being recognized as functional modulators of gene products or by exhibiting independent functions affecting the stability, solubility, clearance, and receptor-binding properties of proteins (Oh-eda et al., 1990; Kodama et al., 1993; Logsdon et al., 2004). Although, the IFN1a glycan has been hypothesized to be involved in receptor binding (Civas et al., 1988), it is generally accepted that the carbohydrate moiety is involved in solubility and stability of the protein. This was supported by a 10 times higher antiviral activity of the glycosylated rhIFN1a compared with the nonglycosylated variant as well as altered physicochemical properties, e.g., increased propensity of aggregation of the deglycosylated rhIFN1a (Watanabe and Kawade, 1983; Conradt et al., 1987; Runkel et al., 1998). Based on the fact that the bioactivity of rhIFN1a is a critical parameter in the treatment of RRMS and to a large extent seems to depend on the presence of a conjugated glycan moiety, we determined the structure/bioactivity relationship to further investigate the carbohydrate role of rhIFN1a. After a careful validation of the Mx biomarker, native rhIFN1a and glycosylated analogs were monitored for activity. We present in vivo data confirming that the glycan moiety of rhIFN1a is crucial for its bioactivity by exerting protein stabilization/solubilization. The major contributing factor for these effects could be pinpointed to the terminating sialic acid residues. Materials and Methods Experimental Animals. This study was performed using adult Mx-congenic mice on a C57BL/6 background (breeding couples were kindly donated by Drs. Otter Haller and Dr. Peter Staeheli, University of Freiburg, Freiburg, Germany) and adult Mx-congenic IFNARknockout (ko) mice (provided by Dr. Peter Staeheli). The C57BL/6 Mx mice were bred at the Biomedical Laboratory at the University of Southern Denmark, and the Mx IFNAR-ko mice were bred at the department of Virology and Microbiology at the University of Freiburg. The mice were cared for in accordance with guidelines of the Danish National Animal Care Committee, thereby meeting international guidelines on the ethical use of animals. Commonly used inbred mouse strains such as C57BL/6 and SJL carry a variant allele of the Mx1 gene, and these strains are phenotypically Mx (Staeheli et al., 1988). Wild mice and mice of the strain A2G have a functional Mx1 locus (Haller et al., 1987). The functional Mx gene was introduced into the C57BL/6 mice by crossing C57BL/6 with A2G mice (genotype Mx /Mx ) and continuing backcrossing the Mx offspring with C57BL/6 until the 11th backcross generation was reached (Horisberger et al., 1983). IFN. Rebif (rhIFN1a, sponsored by Serono Nordic/Merck Serono, Hellerup, Denmark) was used to validate the Mx marker. Several protein chemical approaches were performed to purify rhIFN1a from the rhIFN1a formulation containing high concentrations of mannitol and human serum albumin (HSA), e.g., size exclusion chromatography, and HSA-affinity depletion. However, this was not accomplished due to the high concentration difference (the rhIFN1a/HSA concentration ratio was 1:100) and the adhesive nature of HSA. Hence, commercially available rhIFN1a ( 98% pure; R&D Systems, Minneapolis, MN) was used for the determination of the glycostructure/bioactivity relationship. In addition, rhIFN1a, in a new stabilizing formulation containing a high concentration of the surfactant poloxamer 188 (polyethylene-polypropylene glycol), was used to study the kinetics of the desialylated analog. IFNAdministration. Both rhIFN1a and the commercially available rhIFN1a were diluted in sterile phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 1.4 mM KH2PO4, and 8 mM Na2HPO4, pH 7.4) and injected s.c. (0.5 ml). HSA (Sigma-Aldrich, St. Louis, MO) was added to the glycosidase-trimmed rhIFN1a variants (4 g/ml) before injection. Several controls were performed: 1) a vehicle control, where vehicle solution was administrated exactly as rhIFN1a in a dose corresponding to 100,000 IU; 2) a control for endogenous production of type I and III IFNs, where mice were treated with activity-depleted (denatured) rhIFN1a; and 3) a control using IFNAR-ko mice to show that the response was receptor mediated. Blood Sampling. Screening for IFN-induced transcription of the Mx gene was performed by collecting blood before injection and every 3 h within the first 24 h posttreatment (hpt) using 100,000 IU rhIFN1a (n 2). Three blood samples were collected by eye vein puncture (using EDTA-coded pipettes; Danotherm Electric, Rødovre, Denmark) from each mouse, the last with no postsurvival. The eye vein puncture was performed by a well trained animal-technician, and the blood volume collected from each mouse was below 10% of the total blood volume of the mouse. Based on the results of this initial screen, the Mx level was determined for each hour within the first 6 hpt with 100,000 IU rhIFN1a. Real-Time PCR. RNA purification from whole blood was performed using a Versagene RNA blood kit (Gentra Systems, Inc., Minneapolis, MN). The purification was initiated within 10 min after collections and performed according to the manufacturer’s instructions. The RNA was eluted in diethylpyrocarbonate-treated water, which was used throughout the experiments. As determined by means of optic density using a GeneQuant Pro (GE Healthcare, Chalfont St. Giles, UK), 250 ng of RNA was mixed with 300 ng of random hexamer primers (Roche, Basel, Switzerland), and 0.5 mM/ nucleotide was added (final concentration). After incubation for 5 min at 65°C, the sample was cooled on ice. Subsequently, 50 mM Tris-HCl, 75 mM KCl, 3 mM MgCl2, and 10 nM dithiothreitol were added (final concentrations), and the mixture was incubated for 10 min at 25°C. Next, 200 units of Moloney murine leukemia virus were added, and the sample was incubated for 50 min at 37°C followed by termination for 15 min at 70°C. Microtiter plates (96 wells) were used, and the reactions were detected real-time using SYBR Green dye-based detection on an iCycler (Bio-Rad, Milpitas, CA). The primers were designed using Beacon Designer (BioSoft International, Palo Alto, CA) so that one primer for each transcript spanned over an exon-exon junction. The concentration of Mg was adjusted for each primer design [2 mM for the Mx system and 2.5 mM for the hypoxanthine phosphoribosyltransferase 1 (HPRT1) system]. Specifically, the following primers were used: Mx sense, 5 -TCA GTT TCC TCA AAA GGG GTT GAC-3 ; Mx antisense, 5 -AAT ATT CCG TCT GCA CTC CTG GTA-3 ; HPRT1 sense, 5 -GTT AAG CAG TAC AGC CCC AAA ATG-3 ; and HPRT1 antisense, 5 -AAA TCC AAC AAA GTC TGG CCT GTA-3 . cDNA was mixed with 50% RealQ RT-PCR Master Mix (Ampliqon/Bie and Berntsen, Herlev, Denmark) and 0.3 M of each primer and SYBR Green (diluted according to the manufacturer) (Lonza Copenhagen ApS, Vallensbaek Strand, Denmark), and 10 nM fluorescein (Bio-Rad) were added to a final volume of 25 l in each well. The RealQ RT-PCR Master Mix contained hot start DNA polymerase, deoxynucleotide triphosphates, and 1.5 mM The Structure/Activity Relationship of Human IFN1a 339 at A PE T Jornals on D ecem er 1, 2017 jpet.asjournals.org D ow nladed from MgCl2, requiring additional MgCl2 to be added. A three-step protocol was used for the reactions, initiated by a heating step at 95°C for 15 min followed by 35 cycles of a 95°C step for 10 s, a 60°C step for 20 s, and finally a 72°C step for 30 s. Standard series were included on all plates, and all samples where run as triplets. The starting quantity (SQ), based on a standard curve (4 points), for each sample was calculated by the average of each triplet. Data are presented as the ratio between the SQ of the target transcript (Mx) and the SQ of the reference transcript HPRT1. To verify that the designed primers reacted specifically with the target transcript, melting point detection and gel electrophoresis were used to show that PCR gave rise to a single product. The melting point was established by raising the temperature of the PCR products to 95°C for 1 min and then cooling it down to 55°C followed by 80 cycles of 10 s each in which the temperature was raised 0.5°C per cycle. In addition, the PCR products were run on a 3% NuSieve (Cambrex Bio Science) agarose gel with 0.3 Tris-acetate EDTA buffer together with a 50 base pair (bp) and a 100-bp DNA ladder. Furthermore, real-time PCR of 18S was used to validate the consistency of HPRT1 transcription. A sample with a higher concentration of target mRNA than the experimental samples was chosen for generating the standard curve, and it was diluted to a concentration lower than any of the experimental samples (5 dilution series). Denaturation of rhIFN1a. Two micrograms of rhIFN1a was dissolved in 20 l of alkylation buffer containing 6 M guanidine HCl, 30 mM Tris, and 1 mM EDTA. 1,4-Dithiothreitol was added to a final concentration of 50 mM. This mixture was incubated at 56°C for 45 min. Iodoacetamide was added to a final concentration of 100 mM, and the mixture was placed in the dark for 45 min. Salts and other reagents were removed by washing the samples three times in H2O using a spin column with a 3-kDa molecular mass cutoff (Vivascience, Hannover, Germany). After the final washing step, a minor fraction was kept for protein mass determination, and the remainder was used immediately for bioactivity determination. Carbohydrate Trimming. In a set of parallel experiments, different glycosidases were used to trim the carbohydrate moiety of rhIFN1a, generating different glycosylation analogs. In particular, rhIFN1a was treated with the following: 1) fucosidase, 2) sialidase, and 3) PNGase F. To perform the defucosylation experiment (1), 3 g of rhIFN1a was dissolved in 30 l of 20 mM sodium citrate phosphate buffer, pH 6.0, containing 20 mU of -fucosidase (bovine kidney) (Prozyme, San Leandro, CA). This mixture was incubated for 48 h at 37°C. For the desialylation (2), 3 g of rhIFNwas dissolved in 30 l of 50 mM Na2HPO4, pH 6.0, and 8 mU sialidase A (Arthrobacter ureafaciens) (Prozyme) was added. This mixture was incubated for 2 or 48 h (individual experiments) at 37°C. Finally, for the deglycosylation experiment (3), 4 g of rhIFNwas dissolved in 40 l of 20 mM Na2HPO4, pH 7.2. Subsequently, 4 U PNGase F (Flavobacterium meningosepticum) (Roche, Mannheim, Germany) was added, and the mixture was incubated at 37°C for 18, 48, or 72 h. An additional 4 U PNGase F was added for every 24 h. For all experiments, 10 l of the digested sample was mixed with SDSsample buffer and used for gel electrophoresis, whereas another 10 l was used for protein mass determination. The rest of the samples were immediately used for bioactivity determination. Structural Characterization. Samples were mixed with reducing SDS-sample buffer, heated to 100°C for 5 min, and analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) (4–20% gradient gel). The proteins were visualized using Coomassie Brilliant Blue staining. Selected gel bands were excised, and in-gel was digested overnight at 37°C using trypsin as described previously (Shevchenko et al., 1996). The resulting peptide mixtures were used for protein identification and glycan analysis on the peptide level. To perform peptide mass fingerprinting, 0.5 l of the peptide mixture was deposited directly onto the target, 0.5 l of matrix [ -cyano-4-hydroxycinnamic acid, 10 mg/ml in 70% acetonitrile (MeCN), 5% formic acid (FA)] was added, and the mixture was dried. For the glycan analysis, glycopeptides were enriched before the mass analysis using hydrophilic chromatography. To do this, the peptide mixture was dried and redissolved in 10 l of 80% MeCN, 2% FA, and subsequently loaded onto a custom-made microcolumn packed with hydrophilic material (ZIC-HILIC, 200 Å, 10 m; kindly provided by Sequant, Umea, Sweden) into GELoader tips (Eppendorf GmbH, Hamburg, Germany) as described previously (Thaysen-Andersen et al., 2007). After a thorough washing step using the same buffer, the glycopeptides were eluted directly onto the target using 0.5 l of 2% FA, mixed with 0.5 l of matrix (2,5-dihydroxybenzoic acid), and dried. In contrast, hydrophobic chromatography was used as sample clean-up for intact protein mass measurements. Hence, samples were loaded on microcolumns packed with Poros R1 (Applied Biosystems, Framingham, MA) as described previously (Gobom et al., 1999). The column was equilibrated and washed using 10 l of 5% FA, and the glycoprotein was coeluted with the matrix onto the target using 0.8 l of sinapic acid (20 mg/ml) dissolved in 70% MeCN in 5% FA. Peptide mass fingerprinting and glycan analyses were performed using matrixassisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry (MS) on a Bruker Ultraflex (Bruker Daltonics, Bremen, Germany). Reflector mode was activated, and all samples were analyzed in positive polarity mode. A few species were verified based on their fragmentation pattern using the tandem mass spectrometry mode. Internal two-point calibrations were made when possible using known masses of tryptic peptides or other components. Alternatively, external calibrations using tryptic lactoglobulin were used. Data were viewed with the programs MoverZ (Genomic Solutions, Ann Arbor, MI) or FlexAnalysis version 2.4 (Bruker Daltonics). The mass accuracy was generally below 30 ppm for all MS analyses on peptide levels. Statistics. Results are presented in bar graphs as means S.D., and statistics were performed using Prism, version 4 (GraphPad Software Inc., San Diego, CA), applying a one-way analysis of variance (ANOVA) followed by a Tukey’s multiple comparison test to analyze the result in each graphic presentation. Changes in the Mx level 3 h posttreatment compared with no treatment (NT) were additionally analyzed using a paired t test. Differences were considered statistically significant for P values 0.05.