Standardized serum GM-CSF autoantibody testing for the routine clinical diagnosis of autoimmune pulmonary alveolar proteinosis

Abstract

Autoantibodies against granulocyte/macrophage colony-stimulating factor (GMAbs) cause autoimmune pulmonary alveolar proteinosis (PAP) and measurement of the GMAb level in serum is now commonly used to identify this disease, albeit, in a clinical research setting. The present study was undertaken to optimize and standardize serum GMAb concentration testing using a GMAb enzyme-linked immunosorbent assay (GMAb ELISA) to prepare for its introduction into routine clinical use. The GMAb ELISA was evaluated using serum specimens from autoimmune PAP patients, healthy people, and GMAb-spiked serum from healthy people. After optimizing assay components and procedures, its accuracy, precision, reliability, sensitivity, specificity, and ruggedness were evaluated. The coefficient of variation in repeated measurements was acceptable (<15%) for well-to-well, plate-to-plate, day-to-day, and inter-operator variation, and was not affected by repeated freeze-thaw cycles of serum specimens or the reference standards, or by storage of serum samples at -80°C. The lower limit of quantification (LLOQ) of the PAP patient-derived polyclonal GMAb reference standard (PCRS) was 0.78ng/ml. Receiver operating characteristic curve analysis identified a serum GMAb level of 5μg/ml (based on PCRS) as the optimal cut off value for distinguishing autoimmune PAP serum from normal serum. A pharmaceutical-grade, monoclonal GMAb reference standard (MCRS) was developed as the basis of a new unit of measure for GMAb concentration: one International Unit (IU) of GMAb is equivalent to 1μg/ml of MCRS. The median [interquartile range] serum GMAb level was markedly higher in autoimmune PAP patients than in healthy people (21.54 [12.83-36.38] versus 0.08 [0.05-0.14] IU; n=56, 38; respectively; P<0.0001). Results demonstrate that serum GMAb measurement using the GMAb ELISA was accurate, precise, reliable, had an acceptable LLOQ, and could be accurately expressed in standardized units. These findings support the use of this GMAb ELISA for the routine clinical diagnosis of autoimmune PAP and introduce a new unit of measure to enable standardized reporting of serum GMAb data from different laboratories.

Keywords: Autoantibodies; Autoimmune disease; Diagnosis; Enzyme linked immunosorbent assay; Granulocyte/macrophage-colony stimulating factor; MCRS; PAP; PCRS; Pulmonary alveolar proteinosis; monoclonal reference standard; polyclonal reference standard; pulmonary alveolar proteinosis; recombinant, human GM-CSF; rhGM-CSF.

Copyright © 2013. Published by Elsevier B.V.

Figures

Fig. 1.. Effects of the Capture Antigen on GMAb ELISA Performance.

A. Non-Glycosylated Versus Glycosylated GM-CSF Capture Antigen. Serum GMAb concentration measured with the GMAb ELISA using GM-CSF produced in E. Coli (non-glycosylated form) or in yeast (glycosylated form). Each symbol represents the mean of 2 determinations on serum samples from each autoimmune PAP patient or healthy person (n = 20 and 10, respectively). The GMAb concentration measured for each sample with each capture antigen was not different (P=0.796, n = 30; Mann-Whitney Rank Sum Test). B. Bland and Altman Analysis. All data shown in panel A were evaluated by Bland and Altman analysis as described in the methods. C. Source of E. coli-derived GM-CSF Capture Antigen. ELISA plates were coated with GM-CSF (1 μg/ml) produced in E. coli obtained from several suppliers (1–5) and then used in the GMAb ELISA with sham samples (i.e., buffer instead of serum) to determine the effect on background optical absorbance. Each bar represents the mean (±SD) of 4 separate determinations. The background absorbance was different for all comparisons except between suppliers 1, 4, and 5 (P < 0.001, n = 4 each; ANOVA with pairwise multiple comparison procedures by the Holm-Sidak method).

Fig. 2.. Effects of the Secondary Detection Antibody on GMAb ELISA Performance.

A. Use of Intact Goat-Anti-Human IgG or an F(ab’)2Fragment as the Detection Antibody. Serum GMAb concentration measured with the GMAb ELISA in parallel using either an F(ab’)2 fragment (A2290) or intact anti-human IgG (A6029) as the detection antibody. Each symbol represents the mean of 3 determinations on the serum sample from a different autoimmune PAP patient. The median GMAb concentration (not shown) determined with the intact IgG secondary antibody was slightly higher than with the F(ab’)2 secondary antibody (128 [51 – 234] vs 118 [43 – 187] μg/mL; P< 0.001; n = 29 serum samples; Wilcoxon Signed Rank test). B. Bland and Altman Analysis. All data shown in panel A were evaluated by Bland and Altman analysis as described in the methods.

Fig. 3.. Characterization and Performance of a PAP Patient-Derived GMAb Polyclonal Reference Standard (PCRS).

A. Purity and composition of the PCRS. PCRS, prepared as described in the methods, commercially available IgG heavy chain isotype standards (IgGκ 1, 2, 3, or 4), or molecular weight markers (MWM) were subjected to polyacrylamide gel electrophoresis under reducing conditions, Coomassie blue staining, and photography as described in the methods. B. Optical Absorbance of the PCRS as a Function of Concentration. The PCRS was serially diluted and evaluated as the standard in the GMAb ELISA as described in the methods. Optical absorbance increased smoothly in proportion with PCRS concentration. Regression analysis using a quadratic equation yielded a correlation coefficient (R2) of 0.9998. Each point represents the mean (+SD) of 5 separate measurements. C. Effect of Regression Method Used on Percent Error of the PCRS Curve Rit. Results from 5 independent, simultaneously conducted experiments determining the optical absorbance of serial dilutions of the PCRS were subjected individually to linear, quadratic, or logarithmic regression analysis and the mean (±SD) percent deviation at each concentration was determined. The percent error of GMAb concentration was determined as the measured value minus the expected value divided by the expected value and multiplied by 100. The mean (±SD) correlation coefficients for regression analysis of 5 separate experiments (not shown) were 0.9999 ± 0.0001 (quadratic), 0.9969 ± 0.0029 (log10), and 0.9819 ± 0.0043 (linear).

Fig. 4.. Effect of Serum Aliquot Volume on GMAb ELISA Performance.

A. Effect on Serum GMAb Concentration. An initial volume of serum either 10 or 100 μl was used to measure serum GMAb concentration with the GMAb ELISA. Bars represent the mean (±SD) of 20 independent determinations using the same autoimmune patient serum. The gray line represents the mean for the data obtained using the 100 μl aliquots of serum. The difference in median (IQR) GMAb concentration for the assay performed using 10 vs 100 μl initial serum aliquots did not reach statistical significance (64.9 [52.6 – 73.4] vs 59.3 [56.0 – 64.8] μg/mL; P = 0.525, n = 20 samples; Mann-Whitney Rank Sum Test). B. Effect on Percent Deviation of the Serum GMAb Concentration. Data from the experiment shown in panel A were used to calculate the percent deviation of the 20 determinations. The mean (±SD) percent deviation for the assay performed when using 10 μl initial serum aliquots was markedly higher than with 100 μl aliquots (17.1 ± 8.9 vs 7.1 ± 5.2; P < 0.001; n = 20 samples; Student’s t test).

Fig. 5.. Effect of the Choice of Serum Dilution on Determination of GMAb concentration.

A. Effect of the Sample Absorbance to that of the Reference Standard Range. Separate series of standard dilutions (1:100, 1:3000, 1:6000, 1:12000) were prepared from serum from 46 autoimmune PAP patients and 32 healthy people and GMAb concentration was measured for each dilution (184 for autoimmune PAP and 128 for healthy controls) using the GMAb ELISA. Individual measurements were defined as ‘in-range’ if the optical absorbance for the dilution was within the range of the optical absorbance values for the reference standards, ‘low’ (and out of range) if the optical absorbance was less than that of the lowest reference standard, and ‘high’ (and out of range) if the optical absorbance was greater than that of the highest reference standard. Only ‘in-range’ values were used to calculate the mean GMAb concentration. In healthy people, the percent deviation for ‘in range’ values was small (*) and no measurements were greater than the highest reference standard (**). Bars represent the mean (±SD) and the number of determinations is shown above the error bars. B. Effect of Serum Dilution Factor on Percent Deviation. Data are derived from the experiments shown in panel A but shown here as function of the specimen dilution. In healthy people, the percent deviation for ‘in-range’ values was small at the 1:100 dilution (*) and no determinations were in-range at higher dilutions (**). Bars represent the mean (±SD) and the number of determinations is shown above the error bars. C. Effect of Serum Dilution on the Frequency of Obtaining ‘In range’ Measurements. Data are derived from the experiments shown in panels A and B. The percentage of determinations that are ‘in-range’ for each of the standard dilutions (1:100, 1:3000, 1:6000, and 1:12000) is shown (bars). Gray bar represent measurements that were ‘in-range’ and the clear bar represents measurements in which GMAbs were undetectable. The cumulative percent of determinations that are ‘in-range’ at each successively greater dilution is shown (open symbols). In healthy people, GMAb values below the LLOQ were accepted as normal and counted as being in-range.

Figure 6.. Specificity of the GMAb ELISA.

Serum from autoimmune PAP patients (n=3) was subjected to GM-CSF affinity chromatography and the unbound fraction (flow-through) and bound fraction (after elution from the column) from each was collected as previously described (Uchida et al., 2004). The GMAb concentration was measured in serum aliquots taken before chromatography and the unbound and bound fractions. All data was normalized to the serum GMAb concentration for each patient, which was set to 100%. Bars represent the mean (±SD) amount of GMAb present in serum (hatched bar), or recovered in the Unbound (open bar) and Bound (solid bar) fractions.

Figure 7.. Measurement and ROC Analysis of Serum GMAb Concentration in Autoimmune PAP Patients and Healthy Controls.

A. Serum GMAb levels in Autoimmune PAP Patients and Healthy People. Serum GMAb concentration was measured using the GMAb ELISA with the PCRS in 44 autoimmune PAP patients and 38 healthy controls as described in the text. B. Histogram of the Distribution of Serum GMAb Concentrations in Autoimmune PAP Patients and Healthy People. Data represented as a frequency distribution of serum GMAb concentrations in autoimmune PAP patients (filled bars) and healthy people (hatched bars). C.-D. Receiver Operating Characteristic (ROC) Curve Analysis of Serum GMAb ELISA Test Results for 44 Autoimmune PAP Patients and 38 Healthy People. Standard ROC characteristic analysis was performed to determine the sensitivity and specificity for the data shown in Panel A. The area under the curve was 1.0 (C), and, at a cut off value for GMAb of 5.0 μg/ml determined by the software, the sensitivity and specificity of the GMAb ELISA were both 100% (D).

Figure 8.. Characterization and Performance of Pharmaceutical-Grade, GMAb Monoclonal Reference Standard (MCRS).

A. Purity of the MCRS. The MCRS was prepared as described in the methods. MCRS, commercially available IgG heavy chain isotype standards (IgGκ 1, 2, 3, or 4), or molecular weight markers (MWM) were subjected to polyacrylamide gel electrophoresis under reducing conditions, Coomassie blue staining, and photography as described in the methods. B. Optical Absorbance of the MCRS as a Function of Concentration. The MCRS was serially diluted and evaluated as the standard in the GMAb ELISA as described in the methods. Optical absorbance increased smoothly in proportion with MCRS concentration. Regression analysis using a quadratic equation yielded a correlation coefficient (R2) of 0.999. C. Effect of Regression Method used on Percent Error of the MCRS Curve Fit. Results from 6 independent, simultaneously conducted experiments determining the optical absorbance of serial dilutions of the MCRS were subjected to linear, quadratic, or logarithmic regression analysis and the percent deviation at each concentration was determined. The percent error of the [GMAb]PCRS measurement was calculated as [GMAb]PCRS minus [GMAb]MCRS divided by [GMAb]MCRS multiplied by 100; where [GMAb]PCRS are the unconverted values (None) or values after conversion using the linear, quadratic, or cubic regression equation parameters (indicated) and [GMAb]MCRS is the value actually determined using the MCRS (assumed to be the true value). The mean (±SD) correlation coefficients for regression analysis of 6 separate experiments (not shown) were 0.999 ± 0.0004 (quadratic), 0.990 ± 0.008 (log10), and 0.982 ± 0.0168 (linear).

Figure 9.. Conversion of GMAb Concentration Data into International units.

A. Comparison of Serum GMAb concentration Determined with the PCRS and MCRS. Separate sets of standard dilutions (1:100, 1:3000, 1:6000, 1:12000) were prepared from serum from 14 autoimmune PAP patients and 1 healthy person. All standard dilutions (60 total) were used to measure GMAb concentration with the PCRS ([GMAb]PCRS) and MCRS ([GMAb]MCRS) as described in the legends to Figures 3 and 7, respectively, in parallel on the same plate. All in-range values (n = 41; defined in the legend to Figure 5) are shown and were used for analysis. The relationship between GMAb concentration determined using each of the two reference standards was evaluated by linear, quadratic (not shown to improve readability), or cubic regression analysis. The correlation coefficients for regression analysis were 0.980 (linear), 0.998 (quadratic), and 1.00 (cubic). B. Effect of Regression Method and GMAb Concentration Range on the Accuracy of Conversion of [GMAb]PCRS to International Units. Using the data and regression equation parameters determined from the linear, quadric, or cubic regression analysis described above (panel A), the values for GMAb concentration determined with the PCRS ([GMAb]PCRS) were converted mathematically to values that would have been obtained if the MCRS had been used (i.e., equivalent to [GMAb]MCRS). The percent error of the [GMAb]PCRS measurement was calculated as [GMAb]PCRS minus [GMAb]MCRS divided by [GMAb]MCRS multiplied by 100; where [GMAb]PCRS are the unconverted values (None) or values after conversion using the linear, quadratic, or cubic regression equation parameters (indicated) and [GMAb]MCRS is the value actually determined using the MCRS (assumed to be the true value). The dashed line represents a percent error of zero. C. Determining the Equation to Convert [GMAb]PCRS Data to International units. Data from six independent experiments similar to and including the one shown in panel A were combined and cubic regression analysis was done to determine the equation needed to convert data for GMAb concentration determined using the PCRS into units equivalent to that determined using the MCRS. In total, serum samples from 55 autoimmune PAP patients and 12 healthy controls were used for determining the equation. The correlation coefficient (R2) was 0.98343.

Figure 10.. Validation of the Equation to Convert [GMAb]PCRS Data to International Units.

A. Comparison of [GMAb]PCRSData Before and After Conversion to International Units with Data Determined Directly with the MCRS. The converted PCRS-based data described above was plotted against the MCRS-based data and evaluated by linear regression analysis. The correlation coefficient was 0.984. See text for further details. B. Bland and Altman analysis. The data shown in panel A were used to evaluate the error for the comparison between [GMAb]PCRS converted to international units and [GMAb]MCRS by Bland and Altman analysis as described in the methods.

Figure 11.

Normal and Abnormal Ranges for GMAb Concentration Expressed in Units of the PCRS, MCRS, and International Units. Serum GMAb concentration in 56 autoimmune PAP patients and 38 healthy people was measured using the GMAb ELISA with the PCRS and converted into international units using the equation parameters determined by the analysis shown in Figure 9C and was also simultaneously measured with the MCRS on the same plate. Box plots show the median (line) and IQR (box top and bottom), and 5th, 95 CI (circles) for patients with autoimmune 56 PAP patients (aPAP), and 38 healthy people (HP).