Characterization of the Self-Association of Human Interferon-α2b, Albinterferon-α2b, and Pegasys

Characterization of the Self-Association of Human Interferon-α2b, Albinterferon-α2b, and Pegasys YIMING LI,1 WALTER F. STAFFORD,2 MARK HESSELBERG,1 DAVID HAYES,2 ZHUCHUN WU,1 MICHAEL BYRNE1 1

Human Genome Sciences, Inc., Rockville, Maryland 20850

2

Boston Biomedical Research Institute, Watertown, Massachusetts 02472

Received 17 May 2011; revised 8 August 2011; accepted 16 August 2011 Published online 4 October 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jps.22751 ABSTRACT: The self-association of human interferon-"2b (hIFN-"2b), albinterferon-"2b (a recombinant protein with human serum albumin and hIFN-"2b peptides fused together in a single polypeptide chain), and Pegasys (PEGylated hIFN-"2a) was characterized by analytical ultracentrifugation analyses. By examining the apparent sedimentation coefficient distribution profiles of each protein at different concentrations, it was concluded that the above three proteins are self-associating in albinterferon-"2b formulation buffer. By model fitting of sedimentation data using SEDANAL software, the stoichiometry and equilibrium constants of the self-association of these proteins were characterized. The self-association of hIFN-"2b results in the formation of stable dimers, fast-reversible tetramers, octamers, and hexadecamers. In contrast, although both albinterferon-"2b and Pegasys are self-associated, their self-association stoichiometries are significantly different from that of hIFN-"2b. The selfassociation of albinterferon-"2b results in the formation of reversible dimers and trimers, whereas the self-association of Pegasys gives only reversible dimers. The self-association behaviors of hIFN-"2b and albinterferon-"2b involves attractive electrostatic forces, which can be suppressed to a negligible level in low pH (pH 4.0–4.5) and high salt concentration (400 mM NaCl) buffer, allowing quantification of their size variant contents by sedimentation velocity analysis. © 2011 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 101:68–80, 2012 Keywords: analytical ultracentrifugation; biopharmaceuticals characterization; macromolecular prodrugs; physicochemical properties; physical characterization; proteins

INTRODUCTION Interferons (IFNs) are a family of pleiotropic cytokines with antiviral, antiproliferation, antitumor, and immunomodulatory properties.1 Recombinant human interferon (hIFN)-"2a and hIFN-"2b were the first two IFNs licensed by the US Food and Drug Administration for treatment of hairy cell leukemia. Subsequently, these IFNs were approved for clinical use for a variety of viral and cancer indications.2 Owing to their rapid clearance from the body, frequent dosing (daily or three times weekly) of these two IFNs over an extended period (6–12 months or more) is necessary for some indications such as hepatitis Correspondence to: Yiming Li (Telephone: +301-398-5254; [email protected]), Walter F. Stafford (Telephone: +617-6587808; [email protected]) Yiming Li and David Hayes’ present address is Analytical Biochemistry, MedImmune, LLC., Gaithersburg, Maryland 20878. Journal of Pharmaceutical Sciences, Vol. 101, 68–80 (2012) © 2011 Wiley Periodicals, Inc. and the American Pharmacists Association

68

B and C.3 To prolong the serum half-life, PEGylated forms of these two IFNs were developed. PEGylated hIFN-"2a (Pegasys), manufactured by Hoffmann La Roche, Inc. (Basel, Switzerland), is a covalent conjugate of hIFN-"2a with a single 40 kDa branched bis-monomethoxy-polyethylene glycol (PEG) chain.4–6 PEGylated hIFN-"2b (Pegintron), manufactured by Schering Corporation (now Merck & Co.) (Kenilworth, New Jersey), is a covalent conjugate of hIFN-"2b with a linear 12 kDa PEG chain.7 Both products have a half-life much longer than their unmodified counterparts.8 A different approach to improve the pharmacokinetic properties of IFNs was adopted by Human Genome Sciences, Inc. This approach is based on the expression of the recombinant “gene” formed by fusing DNA sequences encoding human serum albumin (HSA) and hIFN-"2b. The fusion protein manufactured in this manner is referred to as albinterferon-"2b. The clinical trial data showed that albinterferon-"2b has a half-life significantly longer than PEGylated IFNs.9–11

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

SELF-ASSOCICATION OF HUMAN INTERFEREON-"2b , ALBINTERFERON-"2b, AND PEGASYS

Self-association of a protein pharmaceutical may impact its stability, biological function, and bioavailability.12–16 Moreover, there is a concern that self-association may favor the formation of stable, long lived aggregates, which might cause immunogenicity.17 Therefore, characterizing the selfassociation of a drug protein is important for drug development. Although the structures and biological functions of various IFN-" species have been extensively studied, the self-association of various forms of IFN-" in aqueous environment has not been characterized. In this paper, the work on characterizing the self-association of albinterferon-"2b in comparison with hIFN-"2b and Pegasys is reported.

EXPERIMENTAL Materials Albinterferon-"2b, hIFN-"2b, HSA, albinterferon-"2b formulation buffer, and Pegasys formulation buffer were manufactured in-house in Human Genome Sciences, Inc. Pegasys was from Hoffmann La Roche, Inc. Albinterferon-"2b formulation buffer composition: 10 mM sodium phosphate, 200 mM mannitol, 60 mM trehalose, 0.01% polysorbate 80, pH 7.2. Pegasys formulation buffer composition: 8 g/L of NaCl, 0.05 g/L of polysorbate 80, 10 g/L of benzyl alcohol, 2.62 g/L of sodium acetate trihdrate, 0.0462 g/L acetic acid, pH 6.0. Preparation of Pegasys Pegasys sample (8 × 0.6 mL) in its own formulation buffer (containing polysorbate 80) was subjected to cation-exchange chromatography as given below. The sample (8 × 0.6 mL) was buffer exchanged into 20 mM sodium acetate (pH 4.5) and concentrated to approximately 950 :L by ultrafiltration with an Amicon Ultra-4 30 kDa centrifugal filter device (Millipore, Billerica, Massachusetts). Chromatography of the buffer-exchanged Pegasys was performed using a Gold HPLC System (Beckman Coulter, Fullerton, California) with a 4.0 × 250 mm Propac WCX-10 column (Dionex, Sunnyvale, California). The chromatography was performed at a flow rate of 1 mL/min. For each of the four injections, the column was equilibrated with mobile phase A (25 mM sodium acetate, pH 4.5). After sample application, the column was washed 10 min with mobile phase A. Pegasys was eluted from the column with a 30 mL gradient of 0%–90% mobile phase B (mobile phase A containing 400 mM NaCl). The eluted Pegasys was pooled, concentrated, and buffer exchanged to the buffer for ultracentrifugation. Sample Preparation for Ultracentrifugation All four proteins, HSA, IFN-"2b, albinterferon-"2b, and Pegasys (purified by cation-exchange chromatography), were dialyzed against the buffers used for DOI 10.1002/jps

69

ultracentrifugation. Concentrations of the dialyzed proteins were determined by measuring sample absorbance at 280 nm using an 8453 diode-array spectrophotometer (Agilent, Santa Clara, California). After protein concentration determination, the sample proteins were diluted to desired concentrations with dialysate. Analytical Ultracentrifugation Sedimentation velocity (SV) and sedimentation equilibrium (SE) were performed in a ProteomeLabTM XLA protein characterization system analytical ultracentrifuge (Beckman Coulter, Brea, CA). Except for the SV run of Pegasys in a low concentration range (0.019, 0.056, and 0.167 mg/mL), the SV runs of all the proteins were monitored with absorbance scans at 280 nm. The SV run of Pegasys in low concentrations was monitored with absorbance scans at 220 nm for higher signal/noise ratio. To keep the loading absorbance values within the suitable range for SV run, a cell assembled with a 3-mm centerpiece was used for the samples with high protein concentrations. These samples include HSA at 3 mg/mL, albinterferon-"2b at 3 and 2.7 mg/mL, hIFN-"2b at 1.2 mg/mL, and Pegasys at 0.167 mg/mL (for run monitored by 220 nm scans) and at 1.5 mg/mL (for run monitored by 280 nm scans). All other samples were loaded into cells with a 12-mm centerpiece. For cells with a 12-mm centerpiece, 440 :L of reference buffer was loaded into the reference sectors and 430–435 :L of sample was loaded into the sample sectors. For cells with a 3-mm centerpiece, 110 :L of reference buffer was loaded into the reference sector and 108 :L of sample was loaded into the sample sector. Test scans at a rotor speed of 3000 rpm were performed to ensure proper loading without leaking before an SV run was performed. All SV runs were performed at 20◦ C. The rotor speed and protein concentrations for SV analysis were varied for different proteins and will be given in the Results and Discussion section. Sedimentation equilibrium runs of albinterferon"2b and hIFN-"2b were performed at 20◦ C, at three loading concentrations (0.3, 0.9, and 2.7 mg/mL for albinterferon-"2b; 0.2, 0.6, and 1.2 mg/mL for IFN"2b) and three rotor speeds (8000, 12,000, and 16,000 rpm). Determination of the Physicochemical Parameters for Data Analysis and Model Fitting The densities of the buffers were measured using a DMA 5000 density meter (Anton Paar, Graz, Austria). The viscosities of the buffers at 20◦ C were measured using an Ostwald viscosmeter (VWR, West Chester, Pennsylvania). The ε280 (absorbance extinction coefficient at 280 nm) value of albinterferon-"2b was JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

70

LI ET AL.

determined to be 0.62 AU-mL/mg] using the Edelhoch method.18 The ε280 and specific volume (¯v) values of HSA and hIFN-"2b were calculated based on their amino acid sequences using the Sednterp software (available from the RASMB web site at http:// rasmb.org and from Dr. John Philo’s web site at http:// www.jphilo.mailway.com) .19 The ε280 and v¯ values of Pegasys were taken from the literature.4–5 Data Analysis and Model Fitting The ls-g∗ (s) profiles of the test proteins were used to evaluate whether self-association of each protein exists. An ls-g∗ (s) profile displays the distribution of the apparent sedimentation coefficient of the test protein. It is calculated directly from a least squares modeling of the sedimentation boundary by superposition of sedimentation profiles of ideal nondiffusing particles using the Sedfit software developed by Schuck and Rossmanith.20–21 To characterize the self-association of the test proteins, model fittings of the SV and SE data were performed using the SEDANAL software (available from the RASMB website http://rasmb.org) developed by Stafford and Sherwood.22

RESULTS AND DISCUSSION In a protein self-association system, protein molecules are in a dynamic equilibrium between monomers and higher oligomers. When subjected to a centrifugal force, complexes sediment faster than the monomers and larger complexes sediment faster than the smaller complexes. Consequently, the sedimentation speed of the measurable protein macroscopic boundary depends on the self-association parameters (stoichiometry and self-association equilibrium constants). Therefore, with the development of computational software for data analysis, SV analytical ultracentrifugation has become a powerful tool for characterizing protein self-association. In our studies, the first step was to determine whether the sample proteins undergo selfassociation. This is carried out by comparing the lsg∗ (s) profiles and the weight-average sedimentation coefficients (sw ) of the protein system at different concentrations. A shift in the ls-g∗ (s) profile toward a higher sedimentation coefficient (s) range and an increase in the sw value with an increase in the protein concentration are strong indications of protein selfassociation. The second step was to build the selfassociation models (stoichiometry parameters) based on the sw values using the SEDANAL software. The third step was to determine the stoichiometry and self-association equilibrium constants of the system by model fitting of the SV and SE data to the proposed models using the SEDANAL software. This software has two advantages over other software programs for characterizing molecular interaction properties. It alJOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

lows users to build various models with different stoichiometry and equilibrium constants. Moreover, it can take both hydrodynamic and thermodynamic nonideality into account in the model fitting. hIFN-α2b, Albinterferon-α2b, and Pegasys Undergo Self-Association in Albinterferon-α2b Formulation Buffer To test whether the studied proteins are selfassociated in albinterferon-"2b formulation buffer, SV runs of these proteins at different concentrations were performed. The ls-g∗ (s) profiles and the sw values of each protein at different concentrations were calculated using the Sedfit software. Figure 1a shows the ls-g∗ (s) profiles of hIFN-"2b at 0.3, 0.6, and 1.2 mg/mL in albinterferon-"2b formulation buffer. The ls-g∗ (s) profile shifted significantly toward higher s range with increase in hIFN-"2b concentration. The sw value of hIFN-"2b increased from 4.12 to 4.89 and 5.62 S, respectively, as hIFN-"2b concentration increased from 0.3 to 0.6 and 1.2 mg/mL. These results demonstrate that hIFN-"2b is undergoing self-association in albinterferon-"2b formulation buffer. On the basis of the sw values at the three concentrations, it is reasonable to estimate that selfassociation complexes larger than hexamer may exist in significant amounts in this system. Initial estimates of the size of the oligomers were based on the calculation of a minimum molar mass from the observed sedimentation coefficients using standard hydrodynamic relationships. Significant shift of ls-g∗ (s) profile toward higher s values and a considerable increase in the sw values with an increase in the concentration were also observed in albinterferon-"2b (Fig. 1b). The sw values of albinterferon-"2b at 0.45, 1.35, and 2.7 mg/mL were 5.72, 6.31, and 6.75 S, respectively. These results demonstrate that this protein is self-associating in its formulation buffer. On the basis of the sw values at the three concentrations, it is reasonable to estimate that dimer and/or trimer may be the dominant selfassociation complexes. Therefore, the fusion of hIFN"2b with HSA seems to reduce the self-association of hIFN-"2b dramatically. In contrast to albinterferon-"2b and hIFN-"2b, Pegasys exhibited a very different pattern of ls-g∗ (s) profile change in response to the concentration increase. With a concentration increase from 0.167 to 1.5 mg/ mL (Fig. 1c), the ls-g∗ (s) profile of Pegasys shifted significantly toward lower instead of higher s range, and the sw value of Pegasys decreased from 1.47 to 1.18 S. This result reflects the nonideality of the Pegasys solution. The nonideality of the Pegasys system may be explained by the well-known, large excluded volume effects23 associated with the 40 kDa PEG polymer that is covalently attached to hIFN-"2a. During the SV process, the hydrodynamic interaction DOI 10.1002/jps

SELF-ASSOCICATION OF HUMAN INTERFEREON-"2b , ALBINTERFERON-"2b, AND PEGASYS

71

Figure 1. The ls-g∗ (s) profiles of (a) hIFN-"2b, (b) albinterferon-"2b, and (c and d) Pegasys in albinterferon-"2b formulation buffer. The rotor speed was 50,000 rpm (for hIFN-"2b and Pegasys) or 44,000 rpm (for albinterferon-"2b). The SV run of Pegasys at low concentrations (0.019, 0.056, and 0.167 mg/mL) was monitored with 220 nm scans, whereas all other runs were monitored with 280 nm scans. In each run, the sample with the highest concentration was loaded into a cell assembled with a 3-mm centerpiece and the other two samples were loaded into cells with a 12-mm centerpiece, respectively.

of the PEG moiety with the buffer causes nonideality of the SV analysis of Pegasys. This hydrodynamic drag is primarily due to the large hydrodynamic radius of Pegasys, as the PEG moiety behaves as an essentially random coil extension of hIFN-"2a.5 During sedimentation, the large hydrodynamic radius of Pegasys causes displacement of the solvent, resulting in backward flow of the solvent, significant enough to reduce the speed of the sedimentation of Pegasys. The higher the Pegasys concentration, the larger the backward flow generated. In the concentration range appropriate for sedimentation analysis of Pegasys with absorbance scans at 280 nm, the reduction in sedimentation coefficient by nonideality masks the increase in the sedimentation coefficient caused by self-association (Fig. 1c). Consequently, self-association of Pegasys cannot be detected by comDOI 10.1002/jps

paring the ls-g∗ (s) profiles at different concentrations within this concentration range (0.167–1.5 mg/ mL). To overcome this problem and detect the selfassociation of Pegasys, SV analysis at lower concentrations (0.019, 0.056, and 0.167 mg/mL) was carried out. In order to allow the data from low concentration SV analysis runs to have signal/noise ratios high enough for reliable model fitting, the moving boundaries in the cells were monitored by scans at 220 nm (at this wavelength, the absorbance extinction coefficient of Pegasys is 11.53 AU-mL/(mg-cm). Owing to the strong absorbance of polysorbate 80 at 220 nm, albinterferon-"2b formulation buffer without polysorbate 80 was used as the running buffer for an SV run with low Pegasys concentrations (0.019, 0.056, and 0.167 mg/mL). Figure 1d shows the ls-g∗ (s) profiles of Pegasys from this run. The nonideality of the Pegasys JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

72

LI ET AL.

solution was significantly reduced and the ls-g∗ (s) profile shifted to a higher s value region with increasing protein concentration. This result demonstrates that Pegasys is self-associating in albinterferon-"2b formulation buffer without polysorbate 80. To evaluate the effect of polysorbate 80 (0.01%) on the selfassociation of Pegasys, the ls-g∗ (s) profiles and the sw values of Pegasys at 0.167 mg/mL in albinterferon"2b formulation buffer were compared with those in albinterferon-"2b formulation buffer without polysorbate 80. In these two buffers, Pegasys exhibits similar ls-g∗ (s) profiles (data not shown here) and has nearly the same sw values (1.46 and 1.45 S in buffer with and without polysorbate 80, respectively). This indicates that polysorbate 80 at 0.01% has negligible impact on the self-association of Pegasys. Effect of Ionic Strength and pH on the Self-Association of Albinterferon-α2b and hINF-α2b In order to study the effects of ionic strength and pH on albinterferon-"2b self-association, SV runs of this protein were carried out in its formulation buffer and buffers with the components similar to the formulation buffer, but with different pHs and NaCl concentrations. The sw values from these SV runs are shown in Table 1. The net charge values of the protein in different buffers in the table were calculated using the Sednterp software. The results indicate that attractive electrostatic forces play an important role in albinterferon-"2b self-association because the self-association can be greatly suppressed (indicated by decrease in the sw value) by an increase in the NaCl concentration. However, net charge increase in albinterferon-"2b molecules due to buffer pH change decreases the sw value of the protein. This might be due to an increase in the expulsive electrostatic force among the albinterferon-"2b molecules. In the buffer with lower pH and high salt concentration (25 mM citric acid–NaOH, 400 mM NaCl, pH 4.1), the self-association of albinterferon-"2b and hIFN"2b was suppressed to a negligible level, as indicated by the fact that the sw values of both proteins did not increase with increase in the protein concentration (data not shown). Thus, the apparent sedimentation coefficient values of the monomers of these two proteins in low pH and high salt buffer can be determined by SV analysis. These values are useful for model fitting of the self-association of these proteins. In addition, SV analysis of these proteins in this buffer can be used as an orthogonal method to evaluate size-exclusion high-performance liquid chromatography (SE-HPLC) for quantitating the size variants of these IFNs. The limitations of SE-HPLC for quantitating protein size variants24–26 and the advantages of SV in this context have been reported previously.27–28 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

Characterization of the Self-Association of hIFN-α2b in Albinterferon-α2b Formulation Buffer Model construction and selection for fitting the hIFN"2b sedimentation data were based on the ls-g∗ (s) profiles and the sw values. Because the sw of hIFN-"2b was 5.62 S at 1.2 mg/mL in albinterferon-"2b formulation, it is likely that the self-association complexes with a size equal to or larger than hexamer were present at a significant level. Therefore, various models with one or more self-association complexes with a size equal to or larger than hexamer were constructed for the model fitting. The self-association of hIFN-"2b was characterized by model fitting of the SE and SV data with the SEDANAL software. SE analysis was used in conjunction with SV analysis to provide corroborating and complementary types of information. Although both these techniques use the analytical ultracentrifuge, they provide fundamentally different types of information and, therefore, constitute independent approaches to the characterization of protein interaction. SV data contain more information for calculating sedimentation coefficients and association equilibrium constants (Keq ). But model fitting of the SV data is more difficult than the model fitting of the SE data. The SV data have to be fitted for not only the Keq values, but also the sedimentation coefficients. The fitted values of the sedimentation coefficients of the monomer and oligomer are correlated with the values of the fitted Keq values. Moreover, one cannot determine the sedimentation coefficients of the monomer and oligomer without making assumptions about their shapes. In contrast, only the Keq values for each reaction need to be fitted in model fitting of the SE data when molar masses of monomer and oligomers are known. Thus, model fitting of the SE data is generally preferred over the model fitting of the SV data for studying the stoichiometry of protein self-association. However, because of the difficulties in deconvoluting the sum of more than three exponential terms, SE data from a complex association system often do not contain enough information to find unique, best solutions of molar masses and Keq values. To overcome this problem, the stoichiometry of hIFN-"2b self-association system was determined by model fitting of the SE data. The Keq values for different reactions were calculated by model fitting of the SV data based on the known stoichiometry from SE model fitting. The model fitting of the SE data was carried out to tackle the self-association stoichiometry of hIFN"2b. SE was performed at three concentrations (1.2, 0.6, and 0.2 mg/mL) and three rotor speeds (8000, 12,000, and 16,000 rpm). Among many constructed models, the SE data fit to only two models with acceptable root mean square deviation (RMSD). These DOI 10.1002/jps

SELF-ASSOCICATION OF HUMAN INTERFEREON-"2b , ALBINTERFERON-"2b, AND PEGASYS

Table 1.

73

Changes of Albinterferon-"2b sw Values with Ionic Strength and pH Changes

Buffer

Protein Concentration (mg/mL)

Net Charge

sw (S)

0.9 0.9 0.9 0.9 1.0 1.0 1.0

22.89 –4.85 –15.53 –18.80 –15.53 –15.53 –15.53

5.43 6.43 6.38 6.11 6.27 5.23 4.55

FB (pH adjusted to 5.2) FB (pH adjusted to 6.2) FB (pH 7.2) FB (pH adjusted to 8.2) FB (pH 7.2) FB containing 150 mM NaCl FB containing 500 mM NaCl FB, formulation buffer.

two models are (1) dimer–tetramer–octamer–hexadecamer rapidly reversible self-association model (referred to as 2-4-8-16 model in the following text); and (2) dimer–hexamer–dodecamer fast-reversible selfassociation model (referred to as 2-6-12 model in the following text). The model fitting of the SE data showed that the fitting of the 2-4-8-16 model gives a significantly lower RMSD than the fitting of the 2-612 model (Table 2). Therefore, the 2-4-8-16 model is the better model. Figure 2 shows a representative fitting plot of the SE data to 2-4-8-16 model. This model was further tested by fitting the data from a SV run in albinterferon-"2b formulation buffer. This SV run was performed at 50,000 rpm at three loading concentrations (0.3, 0.6, and 1.2 mg/mL). To facilitate model fitting of hIFN-"2b SV data from the run performed in albinterferon-"2b formulation buffer, a SV run was performed in a low pH and high salt buffer (25 mM citric acid–NaOH, 400 mM NaCl, pH 4,1) at 50,000 rpm. The SV data from this run was used to determine the apparent sedimentation coefficient value of hIFN-"2b monomer. As indicated above, in this buffer, hIFN-"2b is not self-associated and, therefore, the apparent s value of IFN-"2b monomer can be determined as 1.69 S by SV analysis. This value was used to estimate the s values of self-association complexes according to the following formula:

where S1 is the sedimentation coefficient of hIFN"2b monomer, Sn is the value for the hIFN-"2b selfassociation complex containing n molecules of hIFN"2b monomer. The calculated s values of hIFN-"2b self-association complexes were entered as the initial s values of these complexes for model fitting. Fitting of the 2-4-8-12 model was performed with floated s values and self-association equilibrium constant values.

SV

DOI 10.1002/jps

monomer is not detectable) 2A2  A4 Keq2 = 2.46 × 106 M 2A4  A8 Keq3 = 2.06 × 104 M

−1

−1

2A8  A16 Keq4 = 9.83 × 105 M

−1

where A, A2 , A4 , A8 , and A16 are the monomer, selfassociation dimer, tetramer, octamer, and hexadecamer, respectively. Keq1 , Keq2 , Keq3 , and Keq4 are the equilibrium constants for the first, second, third, and forth self-association reactions, respectively. Because monomers are not detectable, the value of Keq1 cannot be determined by model fitting. Its value must be extremely high as the dimers are very stable.

To characterize the self-association of Pegasys, model fitting of the data from all SV runs at various concentrations (0.019–1.5 mg/mL) was performed using the SEDANAL software. Because the nonideality of the Pegasys solution affects the sedimentation speed of Pegasys, it must be taken into account in model fitting. This type of hydrodynamic nonideality is usually represented by the following equations29 :

Results of Model Fitting of hIFN-"2b Ultra-centrifugation Data to 2-6-12 and 2-4-8-16 Models

Sedimentation Mode SE

2A  A2 (Keq1 is so high that the

Characterization of the Self-Association of Pegasys in Albinterferon-α2b Formulation Buffer

Sn = S1 × n2/3

Table 2.

As shown in Figure 3 and Table 2, the SV data fit well to the 2-4-8-16 model with a RMSD value of 0.00427. According to the model fitting of the SV data, the selfassociation behavior of hIFN-"2b in albinterferon-"2b formulation buffer is characterized by the following equations:

Model

Keq1 (M)

Keq2 (M)

Keq3 (M)

Keq4 (M)

RMSD (×10–3 ) AU

2-6-12 2-4-8-16 2-4- 8-16

Very high Very high Very high

1.48 × 109 1.02 × 106 2.46 × 106

1.27 × 106 3.46 × 104 2.06 × 104

– 5.01 × 105 9.83 × 105

7.64 4.64 4.56

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

74

LI ET AL.

Figure 2. The fitting plots from global fit of hIFN-"2b SE data to “dimer–tetramer– octamer–hexadecamer reversible self-association” model. Nine data sets (1–9) were used for this global fitting, giving fitting plots a–i. Data sets 1–3 were from the sample of 0.2 mg/mL, analyzed at 8000, 12,000, and 16,000 rpm, respectively. Data sets 4–6 were from the sample of 0.6 mg/mL, analyzed at 8000, 12,000, and 16,000 rpm, respectively. Data sets 7–9 were from the sample of 1.2 mg/mL, analyzed at 8000, 12,000, and 16,000 rpm, respectively. In each of these fitting plots, the green curve is the calculated hIFN-"2b concentration (represented by absorbance at 280 nm, corresponded to left y-axis) distribution along the radius in the direction of the centrifugation force based on the model fitting. The red circles are the experimental data points of hIFN-"2b concentration distribution along the radius (cm), along the direction of the centrifugation force (also corresponded to left y-axis). Blue dots at the bottom are the deviation of the experimental concentration distribution data points from the theoretical values. All fitting plots have a random distribution of residues. From each data set, 30 scans were used. The RMSD value for this global fit was 0.00464 AU.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

DOI 10.1002/jps

SELF-ASSOCICATION OF HUMAN INTERFEREON-"2b , ALBINTERFERON-"2b, AND PEGASYS

75

s = so /(1 + K sc) D = Do (1 + 2BMc)/(1 + K s c) where s is the sedimentation coefficient at concentration c, so is the sedimentation coefficient at zero concentration, Ks is the hydrodynamic nonideality parameter. BM is the second virial coefficient. To take the effect of the nonideality into account in model fitting, the hydrodynamic nonideality parameter Ks was included as a floating parameter. The use of the data from low concentration loadings (0.019, 0.056, and 0.167 mg/mL) increases the weight of the data with lower nonideality and facilitates model fitting. Because polysorbate 80 in the buffer has negligible impact on the self-association of Pegasys, the data collected from the SV runs in the buffer with or without polysorbate 80 can be combined together for global model fitting. Among the tested models, the data fit well only to the monomer-dimer reversible self-association model (Fig. 4). Acceptable model fitting is indicated by an acceptable RMSD value (8.64 × 10−3 AU) along with reasonable fitted s values of the monomer and dimer (0.87 and 1.56 S). The somewhat higher noise level of this fit as compared with the others results from the combination of the absorbance data from 280 nm with the noisier data at 220 nm. According to the results of model fitting, the selfassociation of Pegasys in Albinterferon-"2b formulation buffer is characterized by the following equations: −1

2A  A2 (Keq = 3.16 × 105 M ; Ks = 0.39 mL/mg)

Figure 3. The fitting plots from a global fit of hIFN"2b SV data to “dimer–tetramer–octamer–hexadecamer reversible self-association” model. Representative plots from a global fit of total 60 absorbance scans from three separate cells (loading concentrations of 0.3, 0.6, and 1.2 mg/mL) in SV experiments. Shown for each cell are the first and last time-difference scans from each 20-scan set contributing to the global fit from that cell. The red circles are the concentration differences, C(obs), at constant radius between two absorbance scans taken at different times. The green lines are the concentration differences, C(calc), calculated from the parameters being fit. The C(obs) and C(calc) correspond to the left y-axis in units of absorbance at 280 nm. The plotted deviations between the observed and calculated concentration differences are on the right y-axis using the same units and scale (absorbance at 280 nm), but offset for the visibility of the residuals. The x-axis is the radius of the cell in centimeters and is the same for both y-axes. The SV run was performed with a rotor speed of 50,000 rpm at 20◦ C. Panels a, b, and c show the fitting of the data from cells with loading concentrations of 0.3, 0.6, and 1.2 mg/mL, respectively. DOI 10.1002/jps

where A and A2 are the monomer and dimer, respectively. Keq is the equilibrium constant of the selfassociation reaction. The value of Ks obtained by fitting was Ks = 0.39 mL/mg. A calculated value for BM of 0.113 mL/mg was used in the fitting using a Stokes radius of 8.73 nm. The self-association of Pegasys in its own formulation was also studied by ultracentrifugation. The results (not shown here) indicate that Pegasys is selfassociating in its formulation buffer, with the same self-association stoichiometry given above.

Characterization of the Self-Association of Albinterferon-α2b in its Formulation Buffer Because the sw of albinterferon-"2b at 2.7 mg/mL was 6.91 S in its formulation buffer, it is likely that the self-association complexes with a size as large as or larger than the dimer are present at significant levels. Therefore, various models with one or more of the self-association complexes such as dimers, trimers, and tetramers were evaluated. Among these JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

76

LI ET AL.

Figure 4. The fitting plot from a global fit of Pegasys SV data to “monomer–dimer reversible self-association” model. Representative plots from a global fit of a total of 120 absorbance scans from six separate cells (loading concentrations of 0.5, 1.5, 0.167, 0.056, 0.167, and 0.019 mg/mL) in SV experiments. Shown for each cell are the first and last time-difference scans from each 20-scan set contributing to the global fit from that cell. The solid points are the concentration differences, C(obs), at constant radius between two absorbance scans taken at different times. The solid lines are the concentration differences, C(calc), calculated from the parameters being fit. The C(obs) and C(calc) correspond to the left y-axis in units of absorbance at either (a–c) 280 nm or (d–f) 220 nm. The plotted deviations between the observed and calculated concentration differences are on the right y-axis using the same units and scale (absorbance at either 280 or 220 nm), but offset to be at the bottom of each plot for visibility. The x-axis is the radius of the cell in centimeters and is the same for both y-axes. The SV run was performed with a rotor speed of 50,000 rpm at 20◦ C. Panels a, b, c, d, e, and f show the fitting of the data from cells with loading concentrations of 1.5, 0.5, 0.167, 0.167, 0.056, and 0.019 mg/mL, respectively. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

DOI 10.1002/jps

SELF-ASSOCICATION OF HUMAN INTERFEREON-"2b , ALBINTERFERON-"2b, AND PEGASYS

77

Figure 5. The fitting plot from global fit of albinterferon-"2b SE data to “monomer–dimer–trimer reversible self-association” model. The SE run was performed with three loading concentrations (0.3, 0.9, and 2.7 mg/mL) at three rotor speeds (8000, 12,000, and 16,000 rpm). Panels a–c show the fit of the data from the cell with a loading concentration of 0.3 mg/mL at rotor speeds of 8000, 12,000, and 16,000 rpm, respectively. Panels d–f show the fit of the data from the cell with a loading concentration of 0.9 mg/mL at rotor speeds of 8000, 12,000, and 16,000 rpm, respectively. Panels g–i show the fit of the data from the cell with a loading concentration of 2.7 mg/mL at rotor speeds of 8000, 12,000, and 16,000 rpm, respectively. The y-axis is in units of absorbance at 280 nm and the x-axis is the radius in units of centimeters.

DOI 10.1002/jps

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

78

LI ET AL.

models, only the monomer–dimer–trimer reversible self-association model fit well with albinterferon-"2b SE data. Because this model is relatively simple, known mass value for monomer was entered and floated during model fitting. Figure 5 shows the global fit of the SE data to this model. Good model fitting was achieved as indicated by (1) small RMSD (0.0055 AU), (2) random distribution of the residuals, and (3) small deviation (0.08%) of the fitted albinterferon-"2b monomer molecular weight (MW, 85757 Da) from the theoretical MW (85685 Da). According to the model fitting, albinterferon-"2b self-association is characterized by the following equations: −1

A + A  A2 (Keq1 = 6.96 × 104 M ) −1

A + A2  A3 (Keq2 = 6.94 × 104 M )

Figure 6. The fitting plot from a global fit of albinterferon"2b SV data to “monomer–dimer–trimer reversible selfassociation” model. Representative plots from a global fit of total 60 absorbance scans from three separate cells (loading concentrations of 2.7, 1.35, and 0.45 mg/mL) in SV experiments. Shown for each cell are the first and last timedifference scans from each 20-scan set contributing to the global fit from that cell. The solid points are the concentration differences, C(obs), at a constant radius between two absorbance scans taken at different times. The solid lines are the concentration differences, C(calc), calculated from the parameters being fit. The C(obs) and C(calc) correspond to the left y-axis in units of absorbance at 280 nm. The plotted deviations between the observed and calculated concentration differences are on the right y-axis using the same units and scale (absorbance at 280 nm), but offset to be at the bottom of each plot for visibility. The x-axis is the radius of the cell in centimeters and is the same for both y-axes. The SV run was performed with a rotor speed of 44,000 rpm at 20◦ C. Panels a, b, and c show the fitting of the data from cells with loading concentrations of 0.45, 1.35, and 2.7 mg/mL, respectively. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

where A, A2 , and A3 are the monomer, dimer, and trimer, respectively. Keq1 and Keq2 are the equilibrium constants for the first and second reactions, respectively. The authenticity of this model for the characterization of the self-association of albinterferon-"2b was supported by the model fitting of the SV data. Figure 6 shows the global fit of the SV data to this model. The model fitting was carried out with Keq1 and Keq2 fixed to the fitted values from the model fitting of the SE data (Keq1 = 6.96 × 104 M−1 , Keq2 = 6.94 × 104 M−1 ). The SV data fit well with the “monomer–dimer–trimer reversible self-association model,” as indicated by (1) small standard deviation (0.0034 AU), and (2) random distribution of the residuals. In order to evaluate the role of the HSA domain in the self-association of albinterferon-"2b, a SV run of HSA at 0.33, 1, and 3 mg/mL in albinterferon-"2b formulation buffer was performed. The ls-g∗ (s) profile of HSA did not shift to the higher s region, and the sw values did not increase with an increase in the HSA concentration (data not shown). These results indicate that HSA is not self-associating. Therefore, the HSA domain is not responsible for the self-association of albinterferon-"2b. Instead, fusion of hIFN-"2b with HSA dramatically reduces the self-association stoichiometry of hIFN-"2b because the self-association of albinterferon-"2b results in the formation of only reversible dimers and trimers. In the biopharmaceutical industry, there are concerns about self-association of protein pharmaceuticals because of its potential to lead to a decrease in the drug protein solubility and an increase in aggregate formation. Aggregate formation might result in development of immunogenicity. Fusion of HSA with hIFN-"2b dramatically reduces the self-association of hIFN-"2b, which may be attributed to the high DOI 10.1002/jps

SELF-ASSOCICATION OF HUMAN INTERFEREON-"2b , ALBINTERFERON-"2b, AND PEGASYS

solubility and low viscosity of albinterferon-"2b. Although albinterferon-"2b is self-associated, it is unlikely that the self-association of this fusion protein is problematic as discussed below. There is little, if any, evidence that links the fastreversible self-association of protein molecules with protein instability. Instead, it has been observed that self-association stabilizes many proteins, including some pharmaceuticals. For example, self-association is required for stabilizing transthyretin to prevent human neurodegeneration.16 The stability of insulin can be increased by promoting self-association and, therefore, regular insulin preparations contain zinc ions in sufficient amount to promote self-association to hexamers.14 However, protein self-association might result in formation of complexes that are capable of inducing immunogenicity against the drug protein. According to current view, protein aggregates are defined very broadly as high MW proteins composed of multimers of natively conformed or denatured monomers. Such species may be soluble or insoluble, and reversible or irreversible, within the given environment.30 Although low valence aggregates such as dimer and trimer appear inefficient in inducing immune response, large aggregates with a highly repetitive arrayed structure (e.g., aggregates containing more than 10 monomers) are more likely to be able to induce immunogenicity.30–31 Therefore, large self-association complexes, although not stable, are of concern for their potential in inducing immunogenicity. Although electrostatic interaction seems to be responsible for the self-association of hIFN-"2b and albinterferon-"2b, as indicated by the results of the experiments discussed above, there are notable differences between the self-association of hIFN-"2b and the self-association of albinterferon-"2b. Selfassociation of hIFN-"2b results in the formation of stable or slowly reversible dimers and fast-reversible tetramers, octamers, and hexadecamers. In contrast, the self-association of albinterferon-"2b results in the formation of only fast-reversible dimers and trimers. The self-association of Pegasys results in the formation of only dimers. Therefore, fusion of hIFN-"2b with HSA or PEGylation of hIFN-"2b prevents the formation of self-association complex with highly repetitive arrayed structures. Such a change in the self-association behavior might be beneficial in reducing the immunogenicity of drug proteins. Moreover, according to calorimetric and spectroscopic studies, albinterferon-"2b has higher thermal structure stability as compared with hIFN-"2b (unpublished data). Further studies are needed to clarify whether the increase of thermal structure stability in albinterferon-"2b are related to the change of selfassociation properties due to the fusion of hIFN-"2b with HSA. DOI 10.1002/jps

79

REFERENCES 1. Maher SG, Romero-Weaver AL, Scarzello AJ, Gamero AM. 2007. Interferon: Cellular executioner or white knight? Curr Med Chem 14(12):1279–1289. 2. Bekisz J, Schmeisser H, Hernandez J, Goldman ND, Zoon KC. 2004. Human interferons alpha, beta and omega. Growth Factors 22(4):243–251. 3. Osborn BL, Olsen HS, Nardelli B, Murray JH, Zhou JX, Garcia A, Moody G, Zaritskaya LS, Sung C. 2002. Pharmacokinetic and pharmacodynamic studies of a human serum albumin-interferon-alpha fusion protein in cynomolgus monkeys. J Pharmacol Exp Ther 303(2):540–548. 4. Bailon P, Palleroni A, Schaffer CA, Spence CL, Fung WJ, Porter JE, Ehrlich GK, Pan W, Xu ZX, Modi MW, Farid A, Berthold W, Graves M. 2001. Rational design of a potent, long-lasting form of interferon: A 40 kDa branched polyethylene glycol-conjugated interferon alpha-2a for the treatment of hepatitis C. Bioconjug Chem 12(2):195–202. 5. Dhalluin C, Ross A, Leuthold LA, Foser S, Gsell B, Muller F, Senn H. 2005. Structural and biophysical characterization of the 40 kDa PEG-interferon-alpha2a and its individual positional isomers. Bioconjug Chem 16(3):504–517. 6. Foser S, Schacher A, Weyer KA, Brugger D, Dietel E, Marti S, Schreitmuller T. 2003. Isolation, structural characterization, and antiviral activity of positional isomers of monopegylated interferon alpha-2a (PEGASYS). Protein Expr Purif 30(1):78–87. 7. Wang YS, Youngster S, Grace M, Bausch J, Bordens R, Wyss DF. 2002. Structural and biological characterization of pegylated recombinant interferon alpha-2b and its therapeutic implications. Adv Drug Deliv Rev 54(4):547–570. 8. Karnam US, Reddy KR. 2003. Pegylated interferons. Clin Liver Dis 7(1):139–148. 9. Bain VG, Kaita KD, Yoshida EM, Swain MG, Heathcote EJ, Neumann AU, Fiscella M, Yu R, Osborn BL, Cronin PW, Freimuth WW, McHutchison JG, Subramanian GM. 2006. A phase 2 study to evaluate the antiviral activity, safety, and pharmacokinetics of recombinant human albumin-interferon alfa fusion protein in genotype 1 chronic hepatitis C patients. J Hepatol 44(4):671–678. 10. Balan V, Nelson DR, Sulkowski MS, Everson GT, Lambiase LR, Wiesner RH, Dickson RC, Post AB, Redfield RR, Davis GL, Neumann AU, Osborn BL, Freimuth WW, Subramanian GM. 2006. A phase I/II study evaluating escalating doses of recombinant human albumin–interferon-alpha fusion protein in chronic hepatitis C patients who have failed previous interferon-alpha-based therapy. Antivir Ther 11(1):35– 45. 11. Subramanian GM, Fiscella M, Lamouse-Smith A, Zeuzem S, McHutchison JG. 2007. Albinterferon alpha-2b: A genetic fusion protein for the treatment of chronic hepatitis C. Nat Biotechnol 25(12):1411–1419. 12. Chi EY, Krishnan S, Kendrick BS, Chang BS, Carpenter JF, Randolph TW. 2003. Roles of conformational stability and colloidal stability in the aggregation of recombinant human granulocyte colony-stimulating factor. Protein Sci 12(5):903–913. 13. Clodfelter DK, Pekar AH, Rebhun DM, Destrampe KA, Havel HA, Myers SR, Brader ML. 1998. Effects of non-covalent selfassociation on the subcutaneous absorption of a therapeutic peptide. Pharm Res 15(2):254–262. 14. DeFelippis MR, Chance RE, Frank BH. 2001. Insulin self-association and the relationship to pharmacokinetics and pharmacodynamics. Crit Rev Ther Drug Carrier Syst 18(2):201–264. 15. Marianayagam NJ, Sunde M, Matthews JM. 2004. The power of two: Protein dimerization in biology. Trends Biochem Sci 29(11):618–625. JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

80

LI ET AL.

16. Wiseman RL, Green NS, Kelly JW. 2005. Kinetic stabilization of an oligomeric protein under physiological conditions demonstrated by a lack of subunit exchange: Implications for transthyretin amyloidosis. Biochemistry 44(25):9265– 9274. 17. Cromwell MEM, Felten C, Flores H, Liu J, Shire SJ. 2006. Selfassociation of therapeutic proteins: Implications for product development. In Misbehaving proteins: Protein (Mis)folding, aggregation, and stability; Murphy RM, Tsai AM, Eds. New York: Springer. 18. Edelhoch H. 1967. Spectroscopic determination of tryptophan and tyrosine in proteins. Biochemistry 6(7):1948–1954. 19. Laue TM, Shah BD, Ridgeway TM, Pelletier SL. 1992. Computer-aided interpretation of analytical sedimentation data for proteins. In Analytical ultracentrifugation in biochemistry and polymer science; Harding SE, Rowe AJ, Horton JC, Eds. Cambridge, UK: Royal Society of Chemistry. 20. Schuck P. 2003. On the analysis of protein self-association by sedimentation velocity analytical ultracentrifugation. Anal Biochem 320(1):104–124. 21. Schuck P, Rossmanith P. 2000. Determination of the sedimentation coefficient distribution by least-squares boundary modeling. Biopolymers 54(5):328–341. 22. Stafford WF, Sherwood PJ. 2004. Analysis of heterologous interacting systems by sedimentation velocity: Curve fitting algorithms for estimation of sedimentation coefficients, equilibrium and kinetic constants. Biophys Chem 108(1–3):231– 243.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 101, NO. 1, JANUARY 2012

23. Gokarn YR. 2003. Hydrodynamic behavior and thermal stability of a pegylated protein: Studies with hen egg lysozyme. Ph.D. Thesis. Durham: University of New Hampshire. 24. Barth HG, Garlin FJ. 1984. A review of polymer shear degradation in size-exclusion chromatography. Journal Liq Chromatogr 7:1717–1738. 25. Harpur AG, Layton MJ, Das P, Bottomley MJ, Panayotou G, Driscoll PC, Waterfield MD. 1999. Intermolecular interactions of the p85alpha regulatory subunit of phosphatidylinositol 3kinase. J Biol Chem 274(18):12323–12332. 26. Kuhlman B, Yang HY, Boice JA, Fairman R, Raleigh DP. 1997. An exceptionally stable helix from the ribosomal protein L9: Implications for protein folding and stability. J Mol Biol 270(5):640–647. 27. Gabrielson JP, Brader ML, Pekar AH, Mathis KB, Winter G, Carpenter JF, Randolph TW. 2007. Quantitation of aggregate levels in a recombinant humanized monoclonal antibody formulation by size-exclusion chromatography, asymmetrical flow field flow fractionation, and sedimentation velocity. J Pharm Sci 96:268–279. 28. Philo J. 2006. Is any measurement method optimal for all aggregate sizes and types? AAPS J 8:E564–E571. 29. Schachman HK. 1959. Ultracentrifugation in biochemistry. New York: Academic Press. 30. Rosenberg AS. 2006. Effects of protein aggregates: An immunologic perspective. AAPS J 8(3):E501–E507. 31. De Groot AS, Scott DW. 2007. Immunogenicity of protein therapeutics. Trends Immunol 28(11):482–490.

DOI 10.1002/jps