Response: Granulocyte/macrophage colony-stimulating factor autoantibodies and myeloid cell immune functions in healthy persons

Bazin and colleagues present an interesting alternative hypothesis (ie, acid activation of cryptic granulocyte/macrophage colony-stimulating factor [GM-CSF] immunoglobulin autoreactivity) to explain our observation that GM-CSF autoantibodies are present at low levels in healthy persons. Notwithstanding, several lines of evidence support our conclusion that GM-CSF is bound to low levels of naturally occurring autoantibodies in healthy persons. First, (Figure 1D, our study¹ ), GM-CSF autoantibodies were readily detected in sera of all of 72 healthy persons without prior acidification using an enzyme-linked immunosorbent assay (ELISA) with a sensitivity and specificity approaching 100% that is used to measure GM-CSF autoantibodies in patients with autoimmune pulmonary alveolar proteinosis (PAP). Second, we isolated GM-CSF autoantibodies directly from serum using GM-CSF affinity chromatography without prior acidification in some assays, for example, IgG subtyping analysis (Figure 1C, our study¹ ). Third, when IgG was collected from the sera of healthy persons using protein G, washed exhaustively to remove unbound proteins, and evaluated by Western blotting to detect GM-CSF (ie, a “pull-down” type approach), GM-CSF bound to IgG was detected in the sera from all healthy persons evaluated (Figure 2A, our study¹ ). Fourth, using a novel ELISA developed to detect GM-CSF whether bound to autoantibodies or free in solution (Figure 2, our study¹ ), the total serum GM-CSF concentration in healthy persons was 3084 (± 484) pg/mL (mean ± SEM, n = 11). In contrast, levels were less than 1 pg/mL using a commercial ELISA detecting only unbound GM-CSF. Thus, the majority of GM-CSF in serum is in bound form, which is undetectable by a commonly used assay. Fifth, Bazin et al²  did not determine whether the cryptic GM-CSF autoreactive antibodies were neutralizing or not, and the GM-CSF autoreactivity in 6M urea-treated IVIG detected by Bouvet et al³  had a binding avidity (1.23μM) far lower than that of GM-CSF autoantibodies from PAP patients (20pM)⁴  and healthy persons (similar to PAP patients¹ ). Incidentally, the report by Watanabe et al⁵  evaluated granulocyte colony-stimulating factor (G-CSF) autoantibodies, not GM-CSF autoantibodies. Thus, while our current data do not rule out the possibility of acidification-mediated activation of cryptic GM-CSF binding activity, Bazin's hypothesis does not explain the multiple lines of evidence supporting the conclusion that GM-CSF is present within immune complexes in the serum of healthy persons.

Meager and colleagues raised primarily methodologic questions regarding our report. For details regarding the validation data and negative controls for our GM-CSF autoantibody assay, readers are referred to Uchida et al,⁴  in which we demonstrated this assay specifically detects human GM-CSF, but does not detect murine GM-CSF, carboxymethylated human GM-CSF (alters tertiary structure), macrophage colony-stimulating factor, G-CSF, interleukin-3 (IL-3), tumor necrosis factor α, IL-4, IL-10, or interferon-γ. The assay's accuracy, precision, and lower limit of quantification are included in supplemental Table 1 of our study.¹  Our experience using this assay in PAP patients, various other diseases and in healthy persons has been reported.^6-8  Together, these data demonstrate the assay to be accurate, precise, highly sensitive and specific for detection of GM-CSF autoantibodies in human serum. In our report,¹  the authenticity of GM-CSF autoantibodies in healthy persons was demonstrated by using far-Western blotting, liquid chromatography and tandem mass spectroscopy, IgG class subtyping, and by the ability of highly purified GM-CSF autoantibodies to inhibit the growth of TF-1 cells (Figure 3A, our study¹ ). Meager et al previously found GM-CSF autoantibody detection problematic when using yeast-derived GM-CSF as the capture antigen and attributed this to the presence of yeast glycans and yeast expressed proteins other than GM-CSF. We used GM-CSF produced in Escherichia coli as the capture antigen in our study.¹  Thus, our results cannot be attributed to nonspecific binding to yeast glycans. Notwithstanding, we compared results using E coli–derived (unglycosylated) and yeast-derived (glycosylated) GM-CSF as the capture antigen and found no significant differences (see supplemental Figure 1C, our study¹ ). Meager et al suggest our results might be explained use of GM-CSF affinity columns previously used to isolate GM-CSF autoantibody from autoimmune PAP patients (who have high levels of GM-CSF autoantibodies⁶ ). However, we used new GM-CSF affinity columns for isolation of GM-CSF autoantibodies from healthy persons. Thus, our results cannot be explained by leaching of GM-CSF autoantibodies from previously used affinity columns.

Meager et al suggest our experiments showing that GM-CSF exists in the form of immune complexes lack validation and suggest an alternative method. We used multiple experimental approaches to demonstrate that GM-CSF is bound to autoantibodies in the sera of healthy persons. First, we developed and validated a novel ELISA capable of detecting GM-CSF whether bound to autoantibodies or free in solution (Figure 2, our study¹ ). Second, we isolated IgG from the sera of healthy persons using protein G, washed it exhaustively to remove unbound proteins and evaluated GM-CSF in the column eluate by Western blotting (Figure 2A, our study¹ ). Although use of an additional electrochemiluminescence approach may provide interesting results confirming our findings, it likely would not change the conclusions of our study. Meager et al suggest our GM-CSF function assays were not adequately controlled for specificity. We showed that GM-CSF autoantibodies purified from IVIG blocked the GM-CSF–stimulated increase in neutrophil CD11b levels but had no effect on stimulation by IL-8 (Figure 3F, our study¹ ). These autoantibodies also specifically inhibited the GM-CSF–stimulated growth of TF-1 cells in culture (Figure 3A, our study¹ ) and blocked GM-CSF–dependent STAT5 phosphorylation (Figure 3B-C, our study¹ ). GM-CSF autoantibodies isolated from IVIG or from PAP patient serum reduced the CD11b stimulation index to a similar degree (Figure 3E, our study¹ ).

Finally, based on their prior report⁹  that neutralizing GM-CSF autoantibodies were readily detectable in 20 of 21 batches of commercial IVIG and that one batch without these antibodies originated from plasma pools in which GM-CSF–neutralizing activity was not detected, they conclude that the neutralizing GM-CSF autoantibodies in IVIG originated from the plasma donated by relatively few donors having high levels of GM-CSF autoantibodies. A high serum GM-CSF autoantibody level is highly sensitive and specific for autoimmune PAP,^{1,4,6,7,10,11}  the prevalence of which is 7 per 1 million in the general population.¹¹  Thirty-one percent of these patients are asymptomatic¹¹  and, thus, might not be excluded by the usual screening procedures used in selecting suitable blood donors. If each batch of IVIG represents pooled donated blood from 1000 apparently healthy persons, then the probability of a batch including even one donor with asymptomatic PAP is 0.0022. Thus, simple probability calculations demonstrate the chance of including one such person per batch in 20 of 21 batches is extremely low (actually, Pr = 1.48 × 10⁻⁵⁴). Since the mean plus or minus SE serum GM-CSF autoantibody concentration in 158 PAP patients was 113 (± 7) μg/mL,^6,8  inclusion of more than one asymptomatic PAP patient would be required to generate levels equivalent to that observed in healthy controls in our study. Thus, we believe their conclusion is unlikely, and we accept the alternative hypothesis that low levels of GM-CSF autoantibodies are present in IVIG and in healthy persons.