Are neutralizing anti–GM-CSF autoantibodies present in all healthy persons?

To the editor:

We read with interest the recent paper of Uchida et al¹  in which they show that neutralizing anti–granulocyte-macrophage colony-stimulating factor (GM-CSF) autoantibodies (AAs) occur in all healthy persons. Previously, high-titer neutralizing anti–GM-CSF AAs have been reported by Uchida et al (2004)²  in sera from most patients with idiopathic pulmonary alveolar proteinosis (PAP) and by us in rare patients with myasthenia gravis,³  but not in sera from healthy controls (HCs). Indeed, even non-neutralizing GM-CSF AA are relatively rare in sera from healthy controls^3-5  or autoimmune disease⁵  and cancer patients.^6,7  Nevertheless, low-titer neutralizing anti–GM-CSF AAs have been found in human intravenous immunoglobulin (IVIg) products.^4,8  These originated from the plasma donated by relatively few donors having high levels of AA; a batch of IVIg that did not contain them was shown to originate from plasma pools devoid of neutralizing activity.⁸  Paradoxically, Uchida et al's data¹  indicates that neutralizing anti–GM-CSF AAs are present in every individual, are bound to circulating GM-CSF and can regulate GM-CSF activities in vivo. Because the procedures used in this study to measure neutralizing anti–GM-CSF AAs are not accompanied by an explanation of the full complement of validation experiments and negative controls used, we have reservations about these conclusions. Here, we provide a short critique of the methods highlighting methodologic differences that could have led to distinctly disparate results.

In Uchida et al's study,¹  all healthy control sera tested were positive for anti–GM-CSF AA using an enzyme-linked immunosorbent assay. Although it is difficult to reconcile this observation with results from previous studies,^3-5,8  variations in assay design, format and serum factors may contribute to nonspecific reactivity and generate a false positive signal in the enzyme-linked immunosorbent assays. We believe that in demonstrating specificity, it is essential to show binding of GM-CSF to the F(ab′)₂ fraction of anti–GM-CSF IgG and absence of binding of antibodies to other proteins (ie, non–GM-CSF). Further confirmation of specificity can be obtained by preincubating the antibody-positive samples with an excess of GM-CSF before inclusion in the assay. A reduction in signal with GM-CSF treated samples as opposed to untreated antibody-positive samples would confirm that the binding is due to the presence of genuine anti–GM-CSF AAs. While the authors have shown that sera containing anti–GM-CSF AAs bind to both yeast- and Escherichia coli–derived recombinant GM-CSF, in our experience using E coli–derived GM-CSF, detection of anti–GM-CSF AAs in sera from cancer patients and healthy persons is very rare.^3,5,6  However, binding to yeast-derived recombinant GM-CSF and other yeast-expressed proteins (non–GM-CSF) may be problematic due to the presence of antibodies to yeast glycans.^9,10 

The authors indicate that GM-CSF AAs are complexed with circulating GM-CSF and are present as immune complexes, but the approaches used for this appear to lack validation. For detecting antibodies that are complexed with the antigen, an acid dissociation step is often included in antibody detection protocols.¹¹  Alternatively, novel electrochemiluminescence-based assays which detect antibodies even in the presence of residual antigen are used.¹²  In our hands, an electrochemiluminescence-based assay tolerant for up to 5 μg/mL GM-CSF detected antibodies against GM-CSF in 8% of sera from healthy controls. Inclusion of an acid dissociation step¹¹  did not increase the number of anti–GM-CSF AA–positive sera. These antibodies did not neutralize GM-CSF–induced proliferation of TF-1 cells and were therefore nonneutralizing.

The authors have provided only scant information regarding the performance of the GM-CSF affinity column used to isolate antibodies. If used repeatedly, is it possible that a proportion of high affinity anti–GM-CSF AAs from PAP sera used as a positive control would be retained on the GM-CSF column and released slowly during subsequent purifications of HC IgG? If PAP anti–GM-CSF IgG contamination did occur, not only would it account for the immunoblot data given in Figure 1a in their paper,¹  where radiolabeled GM-CSF was shown to bind to the bound (IgG) fraction from both HC and PAP IgG, but also for the positive results in bioassays. Evidence of absolute clearance of GM-CSF–bound proteins before column reuse would have been useful in eliminating this possibility.

Finally, the functional assays used for GM-CSF, including those used for measuring neutralizing effects, do not appear to have been adequately controlled for specificity. Assays used for measuring GM-CSF activity are not specific as they can respond to a range of cytokines and can be affected by inhibitory components in sera.¹³  Therefore, if specificity for neutralizing GM-CSF is to be claimed, then it is necessary to show that the affinity-purified antibodies do not neutralize other cytokines, such as interleukin-3, which can be tested using the TF1 cell-proliferation assay.¹³