Search

The cellular response to IFN is an essential part of immune reactions and has been subject to investigations for over 50 years 1. The analyses on IFN function frequently involve the use of neutralizing antibodies to block responses and to document the dependence on IFN signals. In this context, we have previously described an unusual “IFN-like” response initiated by blocking antibodies to type I IFN in primary human endothelial cells (EC) or mononuclear blood cells. In the absence of exogenously added recombinant IFN (rIFN), the exposure of EC to increasing concentrations of IFN-blocking mAb resulted in the dose-dependent induction of IFN response genes at the mRNA and protein level 2. The effect was observed for four different mAb directed against human IFN-α or -β and was dependent on the type I IFN receptor. We concluded that an intrinsic feature of the IFN-blocking antibodies was responsible for the observed “IFN-like” activation of EC; a model was proposed of antibody binding to surface Fc-receptors with sequestration of autocrine IFN and subsequent release to nearby IFN receptors, which would result in the observed “IFN-like” signal. We have now obtained evidence that refutes this hypothesis showing that the “IFN-like” activity associated with IFN-blocking mAb is indeed a discrete component that can be separated from the antibody moiety by sequential cycles of antibody immunoprecipitation (Supporting Information Fig. 1 and Supporting Information Methodology). Furthermore, when the standard two-step procedure for antibody purification as performed by the manufacturer (based on ammonium sulfate precipitation and ion exchange chromatography) was extended by a third step of hydrophobic interaction chromatography, the “IFN-like” activity was lost and the neutralizing capacity of the respective antibodies prevailed (Fig. 1A–C). Figure 1 The “IFN-like” activity in antibody preparations can be eliminated by additional antibody purification (three-step process) and is inhibited by polyclonal anti-IFN-α antiserum. The neutralizing anti-IFN-α mAb MMHA-2 and ... Having established that the “IFN-like” activity was attributable to a discrete contaminant of the applied anti-IFN antibody preparations, the possible contamination with microbial products was first examined. Since the majority of pathogen-associated signals leading to the IFN pathway are mediated by the TLR family 3, 4 we screened for hallmarks of TLR activity. However, we did not observe the induction of the transcription factor NF-κB or the pro-inflammatory activation of EC, strongly arguing against TLR involvement (Supporting Information Fig. 2). We then obtained an indication towards contamination with type I IFN from competition studies showing that the contaminant in mAb preparations was neutralized by rabbit (data not shown) or sheep polyclonal anti-human IFN-α antiserum (Fig. 1D). Polyclonal anti-IFN-β antiserum or control antiserum obtained prior to immunization did not affect the “IFN-like” activity (data not shown). The co-purification (and cross-reactivity) of mouse IFN upon mAb isolation from mouse ascites was a potential source of contamination, which was addressed by cytopathic effect inhibition assays on mouse versus human target cells. There was a significantly higher impact on the human target cells, thus arguing for the presence of human rather than mouse IFN-α (Supporting Information Fig. 3A). However, the question remained as to why the contaminating human IFN-α was not neutralized by the investigated anti-IFN-α-blocking mAb (e.g. MMHA-2). When comparing the neutralizing capacity towards various rIFN-α subtypes, the three-step purified mAb failed to inhibit individual family members (IFN-α subtypes 8, 14, and 16) while the sheep polyclonal antiserum potently repressed all IFN-α subtypes (Supporting Information Fig. 3C). This finding supported the notion that a distinct human IFN-α subtype not neutralized by the respective monoclonal was present in the antibody preparation. In accordance, we found that the purified mAb could not block the “IFN-like” activity present in the contaminated mAb preparation (Supporting Information Fig. 3B). Of note, rIFN-α8 and rIFN-α14 had been produced by PBL prior to the preparation of the contaminated antibody MMHA-2. By applying two anti-human IFN-α ELISA tests (not mouse cross-reactive) with distinct sensitivity towards rIFN-α8 and rIFN-α14 evidence was gained for a predominant antibody contamination by human rIFN-α14 (Supporting Information Table 1). However, a combination of contaminating IFN cannot be excluded. Table 1 Level of detectable contamination with rIFN-α in various antibody preparationsa) Thus, the source of contamination could be traced to the sequential production of rIFN and anti-IFN-blocking antibodies with common equipment. Despite a time window of several months between productions, despite the regular two-step purification procedure, and despite standard equipment cleansing, the contamination of antibody preparations with functional type I IFN was substantial. The importance of our observation was further demonstrated by the frequent occurrence of detectable IFN activity in a considerable number of tested antibodies (Table 1). Apart from various mouse monoclonals against human IFN-α and IFN-β (MMHA-2, MMHA-3, MMHA-9, MMHA-13, MMHB-3, MMHB-12), rat anti-mouse antibodies directed against IFN-α (RMMA-1) or IFN-γ (RMMG-1) also presented with significant amounts of human rIFN. While most of these monoclonals originated from PBL and were supplied by PBL or associated distributors in the contaminated form, further examples for contaminated antibodies were found for an alternative supplier. Two anti-pig IFN mAb (K9, F17) similarly showed contamination with rIFN-α (Table 1). Based on the diverse specificity of contaminated antibodies we propose that unspecific co-purification rather than specific antibody binding accounts for the presence of contaminants. While most of the affected antibodies showed IFN-α contamination, the subtype present may vary. The range of detectable IFN activity varied considerably (by a factor of 1000). The highest levels of anti-viral activity as recorded for the anti-IFN-α mAb clone MMHA-2 equalled a concentration of 800 U/mL of human rIFN-α when applying the antibody at a dilution of 50 μg/mL (common for in vitro experiments). For example, stimulation of target cells with 1000 U/mL of biological or rIFN-α2a in the presence of 50 μg/mL of contaminated blocking mAb MMHA-2 would be expected to result in the complete neutralization of the α2a subtype, while exposing the cells to 800 U/mL of non-neutralized rIFN-α14. The net inhibitory effect on the target cells would be minor leading to the false interpretation of results, especially for an experimental setup where the involvement and concentration of type I IFN is the unknown parameter under investigation. Thus, the information given in this report may be of help in interpreting previously conducted experiments with the listed antibodies. With respect to PBL products, all mAb preparations have been carefully evaluated, and contaminated antibodies were found to date back to the last 2–8 years. More stringent purification and equipment cleaning procedures as well as routine testing for contaminating activity have been put in place at PBL in part due to these experiments. With respect to K9, F17, the company producing these antibodies was informed and has, in the meantime, provided the respective clones to PBL for antibody production. Hence, all products identified in this report to have previously been affected by contamination are now being supplied to the research community in a purified form; however, it is easy to envision that reagent providers who prepare multiple cytokines and mAb could face similar issues as those noted here.

Vertebrates have multiple genes encoding Type I interferons (IFN), for reasons that are not fully understood. The Type I IFN appear to bind to the same heterodimeric receptor and the subtypes have been shown to have different potencies in various experimental systems. To put this concept on a quantitative basis, we have determined the binding affinities and rate constants of 12 human Alpha-IFN subtypes to isolated interferon receptor chains 1 and 2. Alpha-IFNs bind IFNAR1 and IFNAR2 at affinities of 0.5-5 μM and 0.4-5 nM respectively (except for IFN-alpha1 - 220 nM). Additionally we have examined the biological activity of these molecules in several antiviral and antiproliferative models. Particularly for antiproliferative potency, the binding affinity and activity correlate. However, the EC50 values differ significantly (1.5 nM versus 0.1 nM for IFN-alpha2 in WISH versus OVCAR cells). For antiviral potency, there are several instances where the relationship appears to be more complicated than simple binding. These results will serve as a point of reference for further understanding of this multiple ligand/receptor system.

We have begun examining serum levels of Type I and III interferons in normal and autoimmune donors by ELISA and a Type I reporter gene activity assay to determine if there are expression pattern signatures of these diseases. Samples were obtained from normal donors and patients with Systemic Lupus Erythematosus, Rheumatoid Arthritis or Multiple Sclerosis and assayed for IFN-α), IFN-β, IFN-ω, and IFN. Normal controls had little detectable IFN-α, IFN-ω, or Type I IFN activity. Low levels of IFN-β were detected in 15% of the normal controls. Approximately 35% of the normal controls had detectable IFN-λ. In the Lupus samples, 45% had detectable IFN-α. IFN-ω and IFN-β were detected in 20 and 40%, respectively, of the Lupus samples. Overall, greater than 80% of the Lupus samples had detectable Type-I IFN. Rheumatoid arthritis samples may have elevated IFN-β and IFN-ω relative to normal controls, but not IFN-α. Multiples sclerosis samples had a slight tendency towards elevated IFN-β. A low percentage of Lupus and multiple sclerosis samples had detectable Type-I IFN activity by the reporter gene assay. None of the samples displayed significant differences in Type III IFN expression. Multiplex ELISA was used to examine a subset of the samples for cytokine profiles and in Lupus samples IL-13 was decreased. In arthritis samples IL-8 and IL-23 were elevated, while in Multiple Sclerosis samples IL-2, IL-4 and IL-10 were elevated. The implications of these results will be discussed.

The human lymphoma cell line NK-92 has many characteristics of human natural killer cells. We have examined NK-92 activation by a full panel of human alpha Interferons (IFNs) by measuring IFN gamma secretion. We confirmed that NK-92 cells secrete IFN gamma in response to IFN alpha 2 in a dose-dependent manner. The IFN gamma response is enhanced by addition of IL-2, IL-18 or PMA, as has been seen with human donor NK cells. The EC50 for IFN alpha 2 ranges between 700–3000 pg/ml. Inclusion of IL-18 exhibited little effect on IFN potency but significant enhancement of IFN efficacy. EC50 values differ by >1000 fold between the IFN subtypes. Alpha 1 demonstrated the weakest activity while alpha 10 and beta 1a were the most potent. The relative potency of the IFNs correlates well with ability to phosphorylate STAT1 and 2. When NK-92 potency is compared with antiviral activity on A549 cells and antiproliferative activity on OVCAR3 cells, there are subtle but significant differences in the rank orders of potency and potencies relative to IFN alpha 2. These studies will be extended to examination of donor NK cells, additional signal transduction pathways, and the secretion of other cyto/chemokines