Incidence and significance of neutralizing antibodies to interferon beta-1a in multiple sclerosis

Interferon (IFN)-β is established as an effective treatment for patients with relapsing remitting multiple sclerosis (MS). Two forms of recombinant IFN-β, IFN-β-1a (Avonex; Biogen, Cambridge, MA) and IFN-β-1b (Betaseron; Berlex Pharmaceuticals, Richmond, CA), have been licensed in Europe and the United States for this indication.^1,2 IFN-β-1a is produced in mammalian cells using the natural human gene sequence, whereas IFN-β-1b is produced in Escherichia coli bacterial cells using a modified human gene sequence that contains a genetically engineered cysteine-to-serine substitution at position 17. IFN-β-1a is glycosylated; IFN-β-1b is not.

Both IFN-β-1a and IFN-β-1b were shown in carefully controlled clinical trials to decrease the frequency of disease relapses and to reduce disease activity on brain MRI scans.^1,2 In the IFN-β-1a study, designed to measure disability progression, IFN-β-1a also significantly slowed the accumulation of physical disability. Neither drug appeared to improve fixed pre-existing neurologic impairments. For this reason, and because of the slowly progressive nature of the disease, IFN-β therapy may provide maximal benefits through long-term therapy aimed at preventing disease activity and progressive clinical deterioration. This requires continuing biological activity during long-term therapy. However, neutralizing antibodies (NAB) can develop to IFN that may reduce or eliminate drug efficacy.^3-5 Approximately 40% of patients treated with IFN-β-1b developed NAB and reduced clinical efficacy appeared in association with NAB.⁶

Because IFN-β-1a more structurally resembles the natural human protein, the immunogenicity of this drug might be less than that observed with IFN-β-1b. This was suggested by preliminary results from the phase III trial.¹ The purpose of this report is to describe antibody analyses on serum samples from the phase III trial of IFN-β-1a and from an ongoing phase IV safety-extension study.

Methods. Study populations. Phase III study. Details of the clinical trial methodology have been published previously.^1,7 Briefly, 301 patients with relapsing MS were enrolled at four clinical sites in the US. Patients had definite MS for at least 1 year, baseline Expanded Disability Status Score(EDSS) of 1.0 to 3.5 inclusive, at least two documented relapses in the prior 3 years, and no relapses for at least 2 months before study entry, and were between 18 and 55 years of age. The study was a double-blind, placebo-controlled, randomized clinical trial. The primary outcome measure was time to the onset of disability progression using Kaplan-Meier survival curves. Disability progression was defined as worsening from baseline EDSS of at least 1 point persisting for at least 6 months. Study visits were scheduled at baseline and every 6 months, at which time serum was stored for subsequent NAB testing. IFN-β-1a (Biogen) was administered intramuscularly (IM) at a dose of 6.0 million international units (MIU) (30 mcg) weekly for up to 104 weeks.

Open label safety extension study. A multi-center, open-label, 4-year, phase IV study (C94-801) with 382 patients was initiated in May 1995 to obtain long-term safety and antigenicity data in patients treated with IFN-β-1a. IFN-β-1a was administered IM at a dose of 6.0 MIU (30 mcg) weekly. Participants in the safety extension study were divided into four groups: 1) Placebo patients from the phase III study (n = 103); 2) IFN-β-1a recipients from the phase III study or pilot study (n = 115); 3) Patients who did not participate in the phase III and who were previously treated with IFN-β-1b (n = 140); and 4) Patients who did not participate in the phase III study and who were not previously exposed to any IFN-β product (n = 24). Eighty-four patients had no exposure to any IFN-β prior to the start of the safety-extension study. A total of 118 patients were exposed only to IFN-β-1b before the start of the safety-extension study (placebo patients or nonparticipants in the phase III study) and had discontinued IFN-β-1b within 2 weeks of starting the study. Participants were evaluated at baseline and every 3 months for the presence of NAB.

Current two-step assay for NAB. The purpose of the two-step assay system is to identify all patients who develop antibodies to IFN-β and to define the subset of these patients whose antibodies have neutralizing activity. In the two-step assay, all sera were screened for the existence of IFN-β-binding antibodies by enzyme-linked immunosorbent assay (ELISA). All sera positive by ELISA were screened for IFN-β neutralizing activity in a cytopathic effect (CPE) assay. In the CPE assay, NAB+ samples were defined by the ability of serum to block an in vitro biologic activity of IFN-β. This two-step approach is based on the premise that a proportion of sera containing IFN-β-binding antibodies identified by ELISA contain NAB, but that samples without binding antibodies do not contain NAB. This approach was validated by testing 125 ELISA-negative samples from patients who had been treated with IFN-β, the majority for 9 months or longer: 123 (98%) of these were negative for NAB. The two positive samples had NAB titers <5.

For the ELISA, murine monoclonal anti-IFN-βantibody (BO2; Yamasa, Shoyu Co. Ltd., Tokyo, Japan) was adsorbed on microtiter plates(Corning Easy-Wash plates, Fisher Scientific, Medford, MA) at 2 µg/mL in 0.06 M carbonate buffer, pH 9.6. Unoccupied sites were blocked with 0.5% nonfat dry milk (Carnation; Los Angeles, CA), after which IFN-β-1a(Avonex; Biogen) diluted to 150 ng/mL in 0.5% milk/0.2 M phosphate buffered saline (PBS) containing 0.05% Tween-20 (Sigma, St. Louis, MO) was incubated in the wells for 2 hours. Buffer without IFN-β-1a was incubated with control wells. Test samples and positive control sample(primate sera of known titer) were diluted 1:20 in 0.5% nonfat dry milk/0.05% Tween-20/PBS and 50 µL added to IFN-β-coated and sham-coated wells and incubated for 2 hours at ambient temperature. After washing with 0.05% Tween/PBS, 50 µL horseradish peroxidase conjugated antihuman immunoglobulin (309-035-082; Jackson Immunoresearch, West Grove, CA) was added and incubated for 1 hour at ambient temperature. Plates were washed, developed with a hydrogen peroxide solution containing 3,3′,5,5′-tetramethylbenzidine (86,033-6; Aldrich Chemical Company, Milwaukee, WI), and stopped with 1 N H₂SO₄. Optical densities were read on an ELISA plate reader (Model 3550; Bio-Rad, Richmond, CA). The net absorbance for each dilution was determined by subtracting the absorbance for the sham-coated well from the absorbance for the IFN-β-1a-coated well. Samples generating a net absorbance greater than 0.13 were considered positive.

Samples that were positive in the ELISA were subsequently tested in a CPE assay. Human lung carcinoma (A549) cells were diluted to 2.5 × 10⁵ cells/mL and 100 µL added to each well of a 96-well microtiter plate. Plates were incubated at 37 °C, 5% CO₂ for 20 to 24 hours. Test serum samples and positive control serum samples (rabbit serum of known titer) were serially diluted threefold in media and 50µL of each dilution was added to duplicate wells. Fifty µL of 40 IU stock solution of IFN-β-1a were added to each of the sample wells, resulting in a final concentration of 10 IU in the well. Plates were incubated for 15 to 20 hours at 37 °C, 5% CO₂ and then the contents were shaken into a bleach bucket. One hundred microliters of encephalomyocarditis virus were added to all but the cell growth control wells. After incubating the plates for approximately 30 hours at 37 °C, 5% CO₂, contents were discarded and 50 µL of crystal violet stain were added to each well. After 5 to 10 minutes, plates were rinsed and allowed to dry. Plates were read on a microtiter plate reader (Model 3550; Bio-Rad, Richmond, CA) and the 50% CPE determined. Titers were calculated using the Kawade method.⁸

Neopterin and β2 microglobulin detection. Neopterin andβ2 microglobulin concentrations in serum samples were measured at Covance Central Laboratory Services (Indianapolis, IN). Neopterin was measured using a competitive binding immunoassay. β2 microglobulin was measured using microparticle enzyme immunoassay technology.

Neopterin and β2 microglobulin analysis. Neopterin was measured in week 52 serum samples from subjects in the safety-extension study before and 48 hours postinjection of IFN-β-1a. A total of 213 pre and post samples were available for analysis. NAB was measured concurrently at week 52. β2 microglobulin was measured in week 52 serum samples from subjects in the phase III trial before and 24 to 48 hours postinjection of IFN-β-1a. Forty-one samples were available for analysis. NAB was measured concurrently at week 52. Mean change in neopterin and β2 microglobulin concentrations (induction) from the time of injection to post injection was compared among the NAB titer categories (NAB-, NAB titer 1 to 4, 5 to 19, and ≥20) using a one-way analysis of variance. Pairwise comparisons of mean values for NAB- and each of the positive NAB titer categories were isolated in the analysis.

Clinical correlates. Patients who were NAB+ at any time up to and including the week 104 antibody evaluation were considered to be NAB+ for the clinical and MRI correlation analyses. The annual relapse rate was determined as the total number of relapses over the given time interval divided by the total years of patient follow-up over this interval. Comparison between NAB groups was made using the likelihood ratio test. Time to sustained disability progression was compared in the NAB+ and NAB- patients using the Kaplan-Meier approach and the log rank test. Number of gadolinium (Gd)-enhanced lesions were compared between NAB- and NAB+ patients using the Wilcoxon rank sum test. The Gd-enhanced lesion data analysis was previously described.¹

Groups formed for comparing baseline demographics were based on antibody status at year 2. All patients were NAB- at baseline. Differences in baseline demographics and disease characteristics between NAB- and NAB+ patients were compared using the Wilcoxon rank sum test for continuous variables and the Fisher's exact test for categorical variables.

Results. Retesting of samples from the phase III assay. The neutralizing activity data originally reported in the phase III study were obtained by using a CPE assay without an initial ELISA step.¹ The assay was performed at a central laboratory associated with the phase III study. Because the original assay used during the phase III study was no longer available, the current two-step assay was developed (as detailed in the methods section). All samples available from the phase III study were retested in the two-step assay. There was a similar incidence of NAB in the two assays. Using a NAB titer cutoff of 20 or greater, 17% of subjects using the new assay and 19% of subjects using the original assay tested positive at week 104. For the phase III study samples that tested positive in both assays, there was a good correlation in the titres determined by the respective assays (r = 0.80, p= 0.001).

Significance of neutralizing antibodies in vivo. Neutralizing activity may be caused by serum antibodies that recognize the receptor binding region of the IFN-β1a molecule and block its interaction with cell-bound receptors. Through this mechanism, NAB would be expected to block the biological effects of IFN-β, and this effect should be evident both in the in vitro assay and in the in vivo response to IFN-β injections. We therefore determined the relation between NAB and the response to injections as determined by measurement of the IFN-inducible molecules neopterin and β2 microglobulin. Serum neopterin levels were measured before and 48 hours after IM injection of 30 mcg of interferon β-1a at week 52 of the open-label safety extension study (C94-801). The mean preinjection neopterin level was 5.4 nmol/L. The mean neopterin change in subjects without detectable NAB was 4.86 nmol/L, and was not significantly different in subjects with titers of NAB between 1 and 4(figure). At titers of 5 or above, in vivo neopterin induction was progressively attenuated, and the effect was highly significant at titers ≥20. These data indicate that NAB detected with the in vitro assay are associated with attenuation of the in vivo biological responses to IFN-β-1a, and demonstrate that the blunted in vivo response is directly correlated with the titer of NAB.

As a second marker of the in vivo effect of IFN, serum β2 microglobulin levels were analyzed in subjects at week 52 in the phase III study who had a serum sample pre and 24 to 48 hours post IM injection of IFN-β-1a. The mean change in subjects without detectable NAB was 0.37 mg/L (n = 24). There was a significant reduction in the induction of β2 microglobulin in the subjects with NAB 5 to 19 (n = 3; -0.14 mg/L) and in subjects with NAB ≥20 (n = 6; -0.05 mg/L) (p = 0.04 and 0.02, respectively).

The potential significance of non-neutralizing antibodies that bind to IFN-β but do not block receptor binding was evaluated. Non-neutralizing antibodies were identified in the two-step assay as samples that were positive in the binding ELISA but negative in the CPE assay. We found no significant relation between non-neutralizing antibodies and in vivo neopterin induction (table 1). The mean neopterin change in 150 subject without binding or neutralizing antibodies was 4.77 nmol/L, which was comparable to the mean neopterin change of 5.44 nmol/L observed in 17 subjects with binding, non-neutralizing antibodies.

Incidence of NAB in patients treated with IFN-β-1a and IFN-β-1b. To assess the immunogenicity of IFN-β-1a relative to IFN-β-1b when tested in the same assay, we determined the incidence of NAB in subjects enrolled in an ongoing safety-extension study of IFN-β-1a, 30 mcg IM 9 weekly. We determined the incidence over time of study in the 84 IFN-β-naive patients, and at entry into the study in the 118 patients who had previously received IFN-β-1b but not IFN-β-1a and had discontinued IFN-β-1b therapy within 2 weeks of initiating IFN-β-1a injections. Baseline demographic features did not differ between these subjects (table 2). Reasons for discontinuation of IFN-β-1b included side effects and personal preference in 96% (113/118) and disease worsening in 4% (5/118).

Table 3 shows the incidence of NAB in 84 subjects in the safety extension study (C94-801) who had not previously been treated with any IFN-β product. The majority of NAB formation occurred after 9 months of treatment and remained at low levels after that. At 18 months, the incidence of NAB positivity at titers ≥5 and at titers ≥20 was 6%. Of the four subjects who had at least one titer ≥20, two had at least one subsequent titer <5. Table 4 shows the incidence of NAB at entry into the safety extension study in the subjects who had received only IFN-β-1b prior to entry. Because IFN-β-1a was used in the NAB assay, the NAB-positive samples are assumed to contain cross-reacting antibodies that were induced by IFN-β-1b but recognize IFN-β-1a. The incidence of NAB positivity was found to be related to duration of prior treatment with IFN-β-1b, with approximately one-quarter or more of patients testing positive when treatment duration was 1 year or greater. A total of 26% of the patients treated with IFN-β-1b for 18 months or longer were positive at a titer ≥20 and 39% were positive at a titer≥5.

Clinical and MRI correlates of NAB. There was no correlation between NAB status and disability progression or relapse rate in patients on IFN-β-1a treatment for 2 years in the phase III study. Twenty percent of the NAB-patients had at least a 1-point sustained change in EDSS after 2 years compared to 22% of the NAB+ (titer ≥20) patients. The annual relapse rate in the NAB- patients over 2 years was 0.65 and in the NAB+ patients 0.50(NS). No significant differences were found using a titer cutoff of ϵ5. The absence of correlation is most likely due to both the small numbers of patients who developed NAB and the limited period of follow-up. There was a relation between NAB and MRI activity at week 104 (table 5). At week 104, there were more Gd-enhancing brain lesions in the NAB+(titer ≥20) compared with NAB- patients (mean, 1.7 versus 0.6 lesions;p = 0.062).

We compared baseline demographics and disease characteristics between patients who remained NAB negative or became positive for NAB (titer ≥20) during 2 years of IFN-β-1a treatment in the phase III trial. There was no relation between NAB and gender, age at study entry, duration of disease, EDSS, prestudy relapse rate, weeks from last relapse to study entry, time to do the 25-foot walk, T2 lesion load at study entry, Gd-enhancing status at study entry, CSF leukocyte count at study entry, or CSF immunoglobulin (Ig)G index at study entry. The only statistically significant difference was in age at diagnosis, with patients who developed NAB being slightly older (mean, 33.1 years versus 28.9 years; p = 0.034).

Discussion. The current results provide evidence that NAB directed against IFN-β-1a have in vivo biological consequences in patients with MS. The development of NAB to IFN-β-1a was associated with decreased in vivo biological response to interferon (i.e., decreased induction of serum neopterin and β-2 microglobulin) and a trend toward decreased treatment effect on an MRI measure of disease activity. In contrast, there was no evidence that nonneutralizing antibodies had any effect on in vivo responsiveness to IFN-β-1a injections. These findings are understandable in light of the biological mechanisms of action of IFN-β. Binding to cellular IFN receptors is necessary and sufficient to trigger an intracellular signaling cascade that results in expression of IFN-inducible genes. Inhibiting the binding of IFN-β to cellular receptors can be expected to abrogate the cellular response to IFN-β and to subsequent downstream effects on viral resistance, cellular proliferation, and immunologic responsiveness, and ultimately the more distal effects on the autoimmune disease process. The latter was demonstrated in this study by higher MRI activity in patients who developed NAB to IFN-β-1a compared to those who did not.

With IFN-β-1a treatment, the incidence of NAB formation has been significantly lower during the phase IV safety extension study than was seen during the phase III trial. Following the phase III study, the manufacturing process was modified to reduce the amount of IFN-β-1a aggregates in the final product, which likely contributes to the lower levels of NAB formation. Improvements in process specifications with IFN-α aimed at reducing aggregates have led to a similar reduction in immunogenicity.⁹

This is the first report comparing the incidence of NAB in patients treated with IFN-β-1a or IFN-β-1b using the same assay. Although these patients were not randomly assigned to receive IFN-β-1a or IFN-β-1b, the patients had similar demographic and disease characteristics when they were started on IFN therapy, and chose to switch from IFN-β-1b to IFN-β-1a because of side effects or convenience. The incidence of NAB to IFN-β-1b in this study was similar to that reported in the IFN-β-1b phase III study using a different assay.⁶ In this study, NAB (titer ≥5) were identified in 39% of IFN-β-1b-treated patients treated for 18 months or longer, compared with 6% of patients treated with IFN-β-1a for a similar duration. The data suggest that IFN-β-1a is less immunogenic than IFN-β-1b.

Several possibilities exist for the difference in immunogenicity between IFN-β-1a and IFN-β-1b. First, the route and frequency of administration may alter immunogenicity. A recent study of IFN-α in mice demonstrated lower immunogenicity with IM versus subcutaneous injections, and with once a week versus three times a week injections.⁹ Second, the absence of glycosylation with IFN-β-1b may lead to aggregation, which is known to increase immunogenicity.^10,11 Lastly, differences in amino acid sequences can produce differences in immunogenicity, as has been demonstrated with insulin.^12-14 The difference in immunogenicity between IFN-β-1a and IFN-β-1b is clearly not just a matter of dose. The IFN-β-1b study group reported a similar rate of NAB formation with 1.6 MIU every other day as with 8 MIU every other day.⁶

In the phase III trial of IFN-β-1b, patients with NAB showed a relapse rate higher than that of NAB- patients and equal to that of placebo-treated patients, an increased number of enlarging lesions on brain MRI scan compared to NAB- patients, and a lower incidence of side effects(fever) than in NAB- patients.⁶ All three effects of NAB in the IFN-β-1b study were only seen after 18 months of treatment. One possible explanation for the increase in disease activity in NAB+ subjects is that subjects who develop antibodies have inherently more active disease than those who do not. However, in the IFN-β-1b study, prestudy relapse rate and EDSS did not differ between those who did and did not develop NAB.⁶ Likewise, we did not see a difference in EDSS, MRI, or CSF parameters at the start of therapy in patients who became NAB+ and those who remained NAB-. Taken together, these results suggest that patients who develop NAB do not have inherently higher disease activity, and that the higher relapse rates and MRI activity in NAB+ patients is due to reduced efficacy caused by NAB to IFN-β. In this regard, our inability to show significant effects of NAB on relapse rate and disability was probably due to the small number of patients who developed NAB, the variability of the clinical outcome parameters, and the relatively short duration of follow-up after patients developed NAB.

The ability of NAB to inhibit the clinical efficacy of therapeutic proteins is an increasingly well-recognized phenomenon in a variety of clinical settings.^3,4 Although most prominent with distinctly foreign proteins such as mouse monoclonal antibodies (e.g., OKT3), NAB have developed in at least some patients with nearly all genetically engineered human proteins. Because NAB are detected by in vitro tests, the in vivo consequences of such antibodies have not always been evident. However, with the use of protein therapeutics in chronic disease settings, it has become clear that treatment effects may be reversed if sufficient levels of NAB develop. In particular, the relation between relapse during treatment and NAB has clearly been established during IFN-α treatment of chronic viral hepatitis C¹⁵ and chronic myelogenous leukemia.^16,17 In addition, abrogation of antitumor and antiviral effects of IFN-α with the development of NAB has been found in patients with hairy cell leukemia, renal carcinoma, non-Hodgkin's lymphoma, melanoma, B-cell leukemia, and cryoglobulinemia.⁴ Fierlbeck et al. showed that development of NAB during treatment of melanoma with natural IFN-β led to decreased biological activity as measured by two serum markers of interferon activity, β2-microglobulin and 2′,5′-oligoadenylate synthetase.¹⁸ Similarly, Calabresi recently reported recurrence of Gd-enhanced MRI lesions and return of increased soluble vascular cell adhesion molecule levels to baseline coinciding with the development of NAB in some MS patients treated with IFN-β-1b.¹⁹ Prior studies that were unable to demonstrate such relations may have been limited by assay methodology and difficulties in measuring treatment effects. Type II errors also may account for any inabilities to show effects of NAB on clinical parameters.

The clinical implication of the current study is that treatment decisions must account for the possibility of NAB formation. The overall clinical effect of NAB is likely to be a function of both NAB titer and persistence. Given the costs of IFN-β treatment and the inherent difficulty in monitoring treatment responses in relapsing remitting MS, patients should not be continued on therapy in the face of persisting NAB. Ultimately, the effectiveness of chronic therapy with IFN-β in individual MS patients may be limited by high titers of NAB.

Appendix