Stability assessment of anti-bacterial antibodies in immunoglobulin G-depleted serum with validated immunoassays

Patients suffering from primary immunodeficiency (PID) share defects in their immune system that could affect both the innate and/or the adaptive immune system. PID diseases are highly heterogenic [1–3]. Knowledge of their molecular structure is progressively increasing with now at least 300 different forms of PIDs identified. Targeted next generation sequencing has been demonstrated to provide a rapid molecular diagnosis of severe PID [4] and constantly reveals more gene defects associated with PID. A recent study [5] systematically evaluated data from January 1981 to June 2020 and identified 104,614 registered PID patients and at least 10,000 additional patients, coming mainly from Asian and African countries. The molecular gene defect(s) that cause PID were identified in about 13% of registered patients. Over half of patients with PID suffer from antibody deficiencies. Perez et al. [6] segregated the different PID phenotypes for which immunoglobulin G (IgG) lifelong replacement therapy [7,8] is indicated: Agammaglobulinemia due to absence of B cells, hypogammaglobulinemia with poor antibody function, normal immunoglobulins with poor antibody function, hypogammaglobulinemia with normal antibody function, isolated IgG subclass deficiency with recurrent infections, and finally recurrent infections due to a complex immune mechanism. Bruton reported the first IgG replacement therapy, which was based on the intramuscular administration of IgG in 1952 [9]. After plasma-derived IgG preparations that could be safely injected intravenously became available, intravenous IgG (IVIG) administration emerged as the leading form of the IgG replacement therapy in the 1980s and 1990s. However, the effective development of highly concentrated IgG preparations [10,11] and successful introduction of facilitated IgG administration using human recombinant hyaluronidase [12,13] established the subcutaneous IgG (SCIG) administration as an effective treatment option. The so-called ‘rapid push’ administration approach [14] that is favored by some PID patients because of its simplicity completes the various treatment options available. All the options aim to increase circulating IgG levels to reach at least temporarily that found in healthy individuals. Different concentration-time profiles will, however, be obtained by the different administration methods. For example, infusion of IVIG results rapidly in an IgG concentration peak, followed by a decrease over time until the so-called IgG through level is reached before the next infusion, while a clearly lower and later increase in IgG concentration will occur after subcutaneous IgG administration. Regardless of the particular treatment option the patient and physician select, IgG replacement therapy aims to reduce the occurrence of severe infections. Furthermore, patient-tailored, optimized IgG replacement therapy also has the potency to alleviate complications often associated with PID, including bronchiectasis, autoimmune or digestive tract disorders. Typically, antibody deficiency increases the risk of recurrent infections affecting ears, nose, throat and the respiratory tract. Encapsulated bacteria such as Streptococcus or Haemophilus have often been identified to be involved in these infections [15,16]. Streptococcus pneumoniae [17] is a gram-positive bacterium responsible for pneumonia, meningitis and otitis media, while Haemophilus influenzae is a gram-negative bacillus, associated with a variety of airway mucosal infections and bacterial meningitidis. These human pathogenic bacteria share the characteristic that complex well-known polysaccharide structures composing the capsules of the bacteria have been identified as major virulence and antigenic factors. The presence of IgG antibodies targeting these antigens has been repeatedly demonstrated in humans who have and have not been vaccinated against these pathogens, in IVIG products [18–20] and in PID patients receiving IVIG replacement therapy [21]. The latter study showed a linear relation between serotype-specific anti-pneumococcal antibodies in pediatric PID patients and the administered IVIG product, as well as a good correlation with adult plasma levels. More specifically, Orange et al. [22] carried out a meta-analysis to identify a possible relation between IgG through levels and the incidence of pneumonia in PID patients. Increasing the through level, that is, the plasma IgG level measured in the patient immediately before the next IgG replacement dose, progressively reduced the risk of pneumonia, and thus confirmed the protective action of IVIG. Checking the immune response after vaccination with polyvalent pneumococcal vaccine is now often part of the toolbox for a comprehensive PID diagnosis [23–25]. Finally, because of the increasing therapeutical use of biologicals modulating immune functions, secondary hypogammaglobulinemia (SID) [26] frequently observed for example after B-cell depletion, is gaining in importance. Also, SID patients suffer from recurrent infections mainly caused by the pathogens mentioned above.

It is interesting that despite of the importance of the protective antibodies which are mainly directed against complex carbohydrate epitopes making up the outer capsules of these pathogens data is so sparse on their stability in human serum. Therefore, and to support clinical testing, we set up a 15-month stability study. This study addressed the stability of anti-haemophilus influenzae type b (Hib) IgG and anti-pneumococcocal IgG in human IgG-depleted serum, kept frozen at -20°C. Two modified commercially available enzyme-linked immunosorbent assays (ELISAs) were validated and used for the measurement.

Materials & methods

Materials

The human intravenous IgG preparation GAMMAGARD Liquid/KIOVIG (#LE12K355; GGL; Takeda Manufacturing Austria AG, Vienna, Austria) was used as the source for the anti-Hib and anti-PCP IgG antibodies [27]. GGL is formulated as a ready-for-use sterile, liquid preparation of highly purified human IgG antibodies at 100 mg/ml in 0.25 M glycine, pH 4.6–5.1. IgG-depleted human serum was obtained from Sigma (S5143, #032M4771V; Vienna, Austria). The quality of IgG depletion was confirmed by nephelometric IgG measurement (<0.5 mg/ml; data on file) and total IgG enzyme-linked immunosorbent assay (ELISA; 1.9 μg/ml, data on file). The ELISA kits for the measurement of anti-Hib IgG (VaccZyme Anti-Hib IgG MK016) and of anti-PCP IgG (VaccZyme anti-PCP IgG MK012) were purchased from the Binding Site (Birmingham, UK). Both kits contain antigen-coated ready-to-use microwells, five prediluted assay calibrators, two assay control samples, the specific sample diluent, peroxidase-labeled antibody to human IgG and the reagents required for the colorimetric measurement of peroxidase (i.e., 3,3′,5,5′-tetramethylbenzidine (TMB) substrate and phosphoric acid as stopping solution).

We furthermore used NaCl, KCl, KH₂PO4, and Na₂HPO4 (all were of analytical grade from VWR, Vienna, Austria) together with Tween 20 (EIA-grade, Bio-Rad, Vienna, Austria). In addition, the plate shaker PHMP-4 (Grant Instr. Ltd), the ELISA reader EL808 (Bio-Tek), and the 96-well plate washer ELx405 (Bio-Tek, all supplied by Szabo, Vienna, Austria) were applied.

Methods

Anti-Hib IgG ELISA

The VaccZyme™ anti-Hib IgG ELISA kit is designed to measure specific IgG antibodies against Hib capsular polysaccharide. Microwells are pre-coated with the Hib capsular polysaccharide antigen polyribosylribitol phosphate [28,29] conjugated to human serum albumin. The calibrators, controls and diluted test samples are loaded to the wells and antibodies recognizing the Hib antigen bind during the first incubation step. After washing the wells to remove unbound proteins, peroxidase-labeled rabbit anti-human IgG conjugate is added. The conjugate binds to the captured human antibody. After a washing step, the bound conjugate is visualized with 3,3′,5,5′-tetramethylbenzidine (TMB) substrate, and the enzymatic reaction is stopped by addition of phosphoric acid. Within a defined range, the color intensity of the reaction product depends on the anti-Hib IgG concentration and is determined by measurement of the optical density (OD) at 450 nm. The standards supplied with the kit are calibrated against the standard from the Center for Biologics Evaluation and Research, US Food and Drug Administration, Lot 1983.

The ELISA was carried out according to the manufacturer's instructions, introducing the following modifications: we used a different washing buffer (PBST; 0.8% NaCl, 0.02% KCl, 0.02% KH₂PO₄, 0.126% Na₂HPO₄ × 2H₂O, pH 7.2, 0.05% Tween 20), established a six-point instead of the five-point assay calibration curve, prepared the serial dilution series of the calibrator, the samples, and the controls on the plate and finally, we implemented the reference wavelength of 620 nm for the measurement. In particular, we used the assay standard 4 supplied with the kit that had assigned an anti-Hib IgG concentration of 3 μg/ml for preparing the assay calibration curve. The six twofold serial dilutions, prepared in duplicates, ranged from 1/100 to 1/3,200 and covered the anti-Hib IgG concentration range from 0.9 to 30 ng/ml instead of 110 to 9,000 ng/ml as in the original assay. Samples were diluted to an approximate anti-Hib IgG concentration of 30 ng/ml, but at least 1/100. These dilutions were used to start the serial 1 + 1 dilution series. All incubations were done under constant shaking of 500 rpm at room temperature (RT) using the plate shaker PHMP-4. The data evaluation was based on a log-log calibration curve that correlated the calibrators' anti-Hib IgG concentrations with their mean blank-corrected ODs. Only ODs within the OD range covered by the calibration curve were transformed to concentrations. The anti-Hib concentrations thus obtained for the individual dilutions were averaged to yield the final result. A detailed assay protocol is provided in the Supplementary Materials.

Anti-PCP IgG ELISA

The VaccZyme™ anti-PCP IgG ELISA kit is designed to measure IgG responses to 23 pneumococcal polysaccharides isolated from S. pneumoniae. Microwells are pre-coated with 23 PCP antigens (1, 2, 3, 4, 5, 6B, 7F, 8, 9N, 9V, 10A, 11A, 12F, 14, 15B, 17F, 18C, 19A, 19F, 20, 22F, 23F, 33F). Calibrators and controls are pre-adsorbed against capsular polysaccharide (CPS) and samples are diluted in a diluent containing CPS. The calibrators, controls and diluted samples are loaded to the wells and antibodies recognizing the PCP antigen bind during the first incubation. After washing the wells to remove unbound proteins, peroxidase-labelled rabbit anti-human IgG conjugate is added and binds to the captured human antibody. After a washing step, TMB and phosphoric acid was used to measure the bound conjugate.

The ELISA was carried out according to the manufacturer's instructions, introducing the modifications already described for the anti-Hib IgG ELISA: We used PBST as the washing buffer, prepared the dilution series of standard, controls and samples on the plate, used the reference wavelength of 620 nm for the measurement, and established an about 300-times more sensitive six-point instead of the five-point assay calibration curve. In particular, we used the assay standard 3 supplied with the kit that had assigned an anti-PCP IgG concentration of 30 μg/ml for preparing the assay calibration curve. The six twofold serial dilutions ranged from 1/100 to 1/3,200 and covered the anti-PCP IgG concentration range from 9.4 to 300 ng/ml instead of 3.3 to 270 μg/ml as in the original assay. Samples were diluted to an approximate anti-PCP IgG concentration of 300 ng/ml, but at least 1/100. These dilutions were used to start the serial 1 + 1 dilution series. All incubations were done under constant shaking of 500 rpm at RT using the plate shaker PHMP-4. The data evaluation was based on a log-log calibration curve that correlated the calibrators' anti-PCP IgG concentrations with their mean blank-corrected ODs. Only ODs within the OD range covered by the calibration curve were transformed to concentrations. The anti-PCP concentrations thus obtained for the individual dilutions were averaged to yield the final result. A detailed assay protocol is provided in the Supplementary Materials.

Preparation of the validation samples

We prepared three validation samples by adding GGL to the commercially available IgG-depleted human serum. Thus, we obtained samples with anti-Hib and anti-PCP IgG concentrations ranging from 0.1 to 2.8 μg/ml and from 5 to 101 μg/ml, respectively (Supplementary Table 1). Aliquots of 60 μl were prepared and kept at -20°C before analysis. In addition, we prepared aliquots of the GGL preparation used for the spiking, and the IgG-depleted serum, which we also kept at -20°C until analysis.

Design of the assay validation

The three validation samples, the GGL preparation and the IgG-depleted serum were measured in six independent runs. The results obtained were checked against the guideline on bioanalytical method validation issued by the European Medicines Agency (EMA) [30]. In particular, we addressed accuracy, precision, linearity, selectivity/specificity, and the samples' stability on the bench. For the latter, we kept the samples with the highest anti-bacterial IgG levels at RT for up to 72 h and subjecting these sample to up to four repeated freezing-thawing cycles.

Preparation of the stability samples & design of the stability study

To mimic clinical study samples from PID patients as closely as possible, we spiked a commercially available IgG-depleted serum with GGL at a ratio of 1/10 to obtain stability samples with anti-Hib and anti-PCP IgG concentrations of about 3 and 90 μg/ml, respectively. Aliquots of these samples were measured after 3, 6, 9, 12 and 15 months at -20°C with six individual dilution series prepared for the samples taken at each time point. These serial twofold dilution series started at 1/100 and 1/1,000 for the measurement of anti-Hib and anti-PCP IgG, respectively. In addition, the GGL preparation was placed at -80°C and measured at the same time points with four individual dilution series prepared. These serial twofold dilution series started at the dilutions of 1/200 and 1/2,000 for the anti-Hib and anti-PCP IgG measurement, respectively. For the data evaluation, we ran ANOVA followed by Dunnett's multiple comparison test and regression analysis using GraphPad Prism 9.2.0. Significant differences were defined by α <0.05. The 100 ± 20% range was used to indicate stability [30].

Results

The result section provides assay validation data followed by the stability data, separately reporting the data for the anti-Hib IgG and the anti-PCP IgG measurement.

Assay validation data for the anti-Hib IgG ELISA

The mean log-log six-point calibration curve obtained in 18 independent runs carried out with the modified assay ranged from 0.94 to 30 ng anti-Hib IgG/ml (Figure 1A). This modification increases the sensitivity of the assay by a factor of 100. Furthermore, the modified calibration curve also met the EMA guidance for ligand-binding assay calibration curves requiring six non-zero assay standards [30]. The calibration curves showed adequate reproducibility as demonstrated by relative standard deviations (RSDs) calculated for the mean slopes and y-intercepts, which did not exceed 4.5%. The good quality of the log-log fitting applied was confirmed by the mean correlation coefficient of 0.9985, with none of the individual values being lower than 0.9966. Also, the more stringent mean relative total error (RTE) of 13.7% further demonstrated the accurate curve fitting. RTE, a combined measure of accuracy and precision, was obtained by backfitting the blank-corrected ODs measured for the individual calibration curve standards and calculating the agreement with their respective nominal concentrations. Then, the absolute difference between the mean agreement, calculated for the six assay calibrators, and the optimal 100% agreement was determined. This difference, providing the overall accuracy of the calibration curve fitting, was then summed up with the twofold standard deviation determined for the mean agreement. The sum obtained represents the RTE. Individual RTEs did not exceed 22%. Consequently, the results of the back-fitting exercise met generally accepted requirements because all relative agreements were within the 100 ± 20% range. For the low and the high control sample, supplied with the ELISA kits, we determined in 18 runs mean concentrations of 94.6 ± 6.2% and 90.2 ± 7.1% of their labeled anti-Hib IgG concentrations that were provided by the manufacturer (Figure 1B). The values were normally distributed as shown by the D'Agostino & Pearson test with p-values of 0.6556 and 0.8569 for the low and the high control, respectively (Figure 1C). Table 1 shows the anti-Hib IgG concentrations measured in six independent runs for the three validation samples with high, medium, and low anti-Hib IgG levels, the GGL preparation and the IgG-depleted serum. The RSDs determined did not exceed 11% and thus met the EMA precision criterion (Figure 2A). Also, low total errors were calculated for the three validation samples not exceeding 12.2%. The anti-Hib IgG level apparently did not systematically influence the total error as all three samples showed similar total errors. The linearity of the assay was demonstrated by the regression analysis carried out between the found and the expected validation samples' anti-Hib IgG concentrations (Figure 2B) and by the comparison of the dilution-response curves obtained for the assay standard and the samples (Figure 2C). In the first approach, the regression line obtained had a squared correlation coefficient of 1 with a slope close to 1, while the slopes of the samples' dilution–response curves differed by not more than 10% from that of the assay standard. Finally, the analysis of the IgG-depleted serum confirmed the assay's selectivity/specificity (Figure 2D): The mean OD determined for the 1/100-diluted IgG-depleted serum (Table 2) did not differ significantly from that of the blank as shown by Dunnett's multiple comparisons test (p = 0.47). By contrast, there was a significant difference between the mean blank OD and the OD determined for the calibrator with the lowest anti-Hib IgG concentration of 0.94 ng/ml (p < 0.001). The short-term stability study, where the validation sample with the high concentration of 3 μg anti-Hib IgG/ml was kept for up to 72 h at RT on the bench (Supplementary Figure 1A) or frozen and thawed up to four-times (Supplementary Figure 1B), evidenced the good stability of anti-Hib in IgG-depleted serum under the conditions investigated: According to the regression analysis done, 80% of the initial anti-Hib IgG concentration is expected to be maintained after 138 h at RT or after 23 freezing–thawing cycles. All-in-all, the validation data confirmed that the modified anti-Hib IgG ELISA showed acceptable accuracy, precision, linearity and selectivity/specificity.

Anti-Hib IgG stability data in IgG-depleted serum & GGL

The anti-Hib IgG levels of the commercially available IgG-depleted serum spiked with GGL, mimicking a PID serum sample after IgG replacement therapy, were measured every 3 months over a total of 15 months, when the sample was kept at -20°C. The sample measured after 12 months had an anti-Hib IgG concentration of 96.1% ± 1.8% (mean ± SD, n = 6) of the one initially measured, while 87.6 ± 1.2% were found at the end of the 15-month observation period. The regression line between the relative anti-Hib IgG concentrations, expressed as a percent of the initial one, and the storage time (Figure 3A) showed a negative slope of -0.55, which was identified to deviate significantly from zero (p < 0.001). According to this regression line, it can be expected that 80% of the initial anti-Hib IgG concentration will still be present after 36 months under these conditions. The testing design with the measurement of six replicates for each time point applied allowed a further statistical analysis with ANOVA. Thus, we identified a statistically significant difference between the mean anti-Hib IgG concentrations measured at the different time points. Further analysis of these data with Dunnett's multiple comparison test, comparing the mean anti Hib-IgG levels determined at the different time points with that measured at the start of the stability study, revealed only a statistically significant difference (p = 0.007) for the sample kept at -20°C for 15 months (Figure 3B). The anti-Hib IgG levels determined at all other time points did not differ significantly from that determined at the start of the study. Similarly, the anti-Hib levels of the GGL preparation used for spiking, kept at -80°C, were monitored, measuring in that case four replicates per time point. The samples kept for 12 and 15 months at -80°C showed 99.6% ± 4.0% and 92.4% ± 5.2% of the initial anti-Hib IgG concentrations, respectively. The relative concentration versus time regression line (Figure 3C) had a negative slope of -0.1589, which was shown not to deviate significantly from zero (p = 0.233). According to this regression line, 80% of the initial anti-Hib IgG concentration will be present after 126 months storage under these conditions. ANOVA showed significant differences between the mean anti-Hib IgG levels measured for the different time points. Dunnett's multiple comparison test identified a significant difference for the 6-month sample (p = 0.003) compared with the initial value. The sample, however, showed 109.2% of the initial anti-Hib level (Figure 3D). The anti-Hib IgG levels determined at all other time points did not differ significantly from that determined at the start of the study. Overall, an IgG-depleted serum sample spiked with anti-Hib IgG was demonstrated to be stable for at least 15 months at -20°C when the ±20% stability criterion promoted by the EMA guideline was applied.

Assay qualification data for the anti-PCP IgG ELISA

The mean six-point log-log calibration curve, obtained in 18 runs, ranged from 9.4 to 300 ng anti-PCP IgG/ml (Figure 4A). After modification, the assay's sensitivity increased by a factor of 300 compared with the initial ELISA. Using six non-zero assay standards also made the ELISA compliant with the EMA guidance for ligand-binding assay calibration curves [30]. In addition, the higher assay sensitivity allows higher sample dilutions reducing the possible influence of sample matrix components. The reproducibility of the calibration curves was good as evidenced by the RSDs calculated for the mean slopes and y-intercepts, which were 3.0 and -3.8%, respectively. The mean correlation coefficient of 0.9995 with all values higher than 0.9989 and a mean RTE of 8.9% (range 4.3%–12.0%) demonstrated the quality of the log-log fitting approach. Expectedly, all anti-PCP IgG concentrations re-calculated for the calibration curve standards were within the 100 ± 20% range. The results obtained for the high and low control sample (Figure 4B) were normally distributed as shown by the D'Agostino & Pearson test providing p-values of 0.5996 and 0.6374 for the low and high control, respectively (Figure 4C). Table 1 shows the anti-PCP IgG concentrations measured in six independent runs obtained during assay validation. The inter-assay precision was excellent with RSDs lower than 5%, clearly meeting the EMA criterion of 20% for ligand-binding assays (Figure 5A). The total errors were lower than 11.5%. The results of the regression analysis between the found and the nominal anti-PCP IgG concentrations of the validation sample (Figure 5B) and the comparison of the dilution-response curves obtained for the assay standard and the samples (Figure 5C) demonstrated the linearity of the assay. While the squared correlation coefficient of the regression line obtained with the first approach was 1, the slopes of the samples' dilution-response curves differed by not more than 8% from that of the assay standard. This confirmed an adequate parallelism of the dilution-response curves with the standard and excluded any detrimental influence of the sample matrix on the assay. Finally, analysis of the IgG-depleted serum confirmed the assay's selectivity/specificity (Figure 5D): The mean OD determined for the 1/100-diluted IgG-depleted serum (Table 2) did not differ significantly from that of the blank (Dunnett's multiple comparisons test; p = 0.52). By contrast, there was a significant difference between the mean blank OD and the OD determined for the calibrator with the lowest anti-PCP IgG concentration of 9.4 ng/ml (p < 0.001). The short-term stability study for the sample with 100 μg anti-PCP IgG/ml for up to 72 h at RT (Supplementary Figure 2A) or frozen and thawed up to four-times (Supplementary Figure 2B), showed the good stability of anti-PCP in IgG-depleted serum under the conditions investigated: according to the regression analysis, 80% of the initial anti-PCP IgG concentration is expected to be maintained after 283 h at RT or after 18 freezing-thawing cycles. All-in-all, the validation data confirmed that the modified anti-PCP IgG ELISA showed acceptable accuracy, precision, linearity and selectivity/specificity.

Anti-PCP IgG stability data in IgG-depleted serum & GGL

The commercially available IgG-depleted serum, spiked with GGL, was kept at -20°C and measured every 3 months over 15 months. The sample measured after 12 months had an anti-PCP IgG concentration of 101.2% ± 2.6% (mean ± SD, n = 6) of the one initially measured, while 98.6 ± 2.8% were found at the end of the 15-months observation period. The regression line between the relative anti-PCP IgG levels and the storage time (Figure 6A) showed a negative slope of -0.1089, which did not significantly deviate from zero (p = 0.178). According to the regression line, it can be expected that 80% of the initial anti-PCP IgG concentration will still be present after 184 months under these conditions. ANOVA identified a statistically significant difference between the mean anti-PCP IgG concentrations measured at the different time points. Further analysis with Dunnett's multiple comparison test uncovered statistically significant differences between the zero sample and those after 3 months (p = 0.0002) and 9 months (p = 0.0064). These samples, however, showed 93.1% and 95.0% of the initial anti-PCP IgG levels (Figure 6B). Similarly, the anti-PCP levels of the GGL preparation used for spiking and kept at -80°C were monitored. The relative concentration versus time regression line obtained (Figure 6C) had a negative slope of -0.3952, which was shown to deviate significantly from zero (p = 0.0076). According to this regression line, 80% of the initial anti-PCP IgG concentration will be present after 51 months storage under these conditions. ANOVA showed significant differences between the mean anti-PC IgG levels measured for the different time points. Dunnett's multiple comparison test identified significant differences for the samples at 3 months (p = 0.002; 91.4% of initial), 6 months (p = 0.017; 107.6% of initial), and 15 months (p = 0.001; 90.3% of initial) as compared with the initial value (Figure 6D). Overall, an IgG-depleted serum sample spiked with anti-PCP IgG was demonstrated to be stable for at least 15 months at -20°C when the ±20% stability criterion promoted by the EMA guideline was applied.

Discussion

Although ELISAs like the anti-PCP ELISA provide a concentration of binding IgGs that might not necessarily fully agree with the level of functional antibodies, the introduction of specific pre-adsorption steps for the removal of cross-reactive, non-protective IgGs [31] has definitely increased the specificity of the methods. Thus, Wernette et al. [32] provided a detailed protocol for an anti-PCP ELISA in which the detection limit was 10 ng anti-PCP IgG/ml with an inter-assay RSD of about 30%. On the other hand, Whaley et al. [33] and Borgers et al. [34] evaluated the commercially available multiplexed bead-based immunoassay (XMAP^® pneumococcal immunity, Luminex) that, based on the Luminex platform, allows the simultaneous measurement of antibodies against fourteen pneumococcal polysaccharides (PSs). This overcomes the laborious and time-consuming individual measurement of antibodies against each PS contained in multivalent vaccines. Multiplexing was shown to be useful for evaluating the response to vaccination pneumococcal PS-based vaccine. Just recently, Garrido-Jareno et al. showed the feasibility of a surface plasmon resonance (SPR)-based approach for measuring the response to pneumococcal vaccine [35]. The SPR signal demonstrated a clear correlation with the anti-PCP IgG concentrations measured by ELISA and allowed the classification of subjects into those who responded to vaccination and those who did not. Opsonophagocytic assays (OPAs) [36], where the reaction mixture contains bacteria, complement, phagocytes and the test specimen, can be used to measure the functional response of all anti-PCP immunoglobulins, but does not allow the specific anti-PCP IgG measurement achieved, for example, by ELISA. The killing of bacteria is determined by counting bacterial colonies after plating on agar plates or by chemiluminescence of activated phagocytes. Cha et al. [37] reported on the validation of a multiplexed OPA that allowed the simultaneous measurement of eleven pneumococcal serotypes. Historically, before the implementation of ELISA the measurement of anti-Hib IgG was achieved by a radioactive binding assay (RABA) [38].

For our stability study, we used the commercially available anti-Hib IgG and anti-PCP IgG ELISA applied by Schauer et al. [39] for determining the antibody levels in healthy adult blood donors and hospitalized children. While the antigen used for the anti-Hib IgG ELISA was composed of the Hib polyribosyslribitolphosphate oligosaccharide conjugated to human serum albumin, a mixture of 23 capsular PSs, resembling the mixture of the licensed 23-valent vaccine was applied for the anti-PCP IgG ELISA. These polysaccharides represent approximately 80% of the commonly encountered virulent serotypes. A certain immune response, however, is attributable to cell wall polysaccharide antibodies, which confer limited protection against pneumococcal infection [40]. Consequently, an adsorption of such antibodies has been incorporated in this assay. Before the assay validation, we established a six-point calibration curve using based on our ELISA experience [41]. These modified calibration curves provided an obviously increased sensitivity for both ELISAs and could be constructed reproducibly by log-log fitting. The log-log fitting procedure applied seemed to us more convenient than using the 4-parameter fitting for handling the sigmoidal concentration–response curves characteristic for ligand-binding assays. Undoubtedly, the log-log fitting approach provided a narrower operating range than that which would have been obtained by using a 4-parameter fitting, but this is compensated by the easier calculation and direct and more convenient parallelism evaluation between the dilution series of the standard and the samples. Parallelism testing seems to be especially important for the anti-PCP IgG ELISA because serum specimens have been reported that provide dilution-response curves quite different from those obtained with the standard reference serum [32]. Overall, these modifications allowed us to validate both assays according to the EMA guideline and thus to bring them to a quality level required for their use in bioanalysis. In addition, the sensitivity of both assays could be increased substantially by at least a factor of 100.

For the validation, we followed the recommendations of the EMA guideline and other relevant publications [42,43]. The IVIG preparation GGL was added at three concentration levels to commercially available IgG-depleted human serum to obtain the validation samples. The anti-PCP IgG levels ranged from 5 to 100 μg/ml which agreed well with levels of 99.4 ± 2.1 μg/ml and 58.2 ± 13.6 μg/ml measured in IVIG- and SC-treated patients 3 to 5 days after administration [44]. The same samples, containing 0.14 to 2.84 μg anti Hib IgG/ml, were also used for the validation of the anti-Hib IgG ELISA. Although the anti-Hib IgG concentrations were obviously lower than the anti-PCP IgG levels, the anti-Hib IgG levels could nevertheless be deemed representative of clinical samples because their lower levels are caused by the particular assay standardization which is different for the anti-Hib IgG and anti-PCP IgG ELISA. While the standardization of the anti-Hib IgG ELISA was based on the CBER reference preparation, the anti-PCP IgG assay standardization was achieved by using an affinity-purified IgG preparation.

The assay precision was determined as inter-run precision. Intra-run precision was not explicitly addressed because it is covered by the samples' testing procedure that foresees the measurement of a serial dilution series. As the final result is obtained by averaging the concentrations obtained for the individual dilutions, each sample is run in a minimum of two repetitions during its analysis. Furthermore, the data given by the kit's manufacturer showed no remarkable difference between the intra- and inter-assay precision for the anti-PCP ELISA, where mean RSDs of 4.8% and 4.1% were determined for samples with 10.7 to 173 μg anti-PCP/ml. Our mean inter-run precision for the validation sample with similar anti-PCP IgG levels was 3.1%, which is in the same order of magnitude. GGL, showing a ten-times higher anti-PCP level and a different matrix, was measured with an RSD of 3.5%, while the means (n = 18) obtained for the two control samples showed RSDs of 4.8% and 4.3%. The intra- and inter-assay precision RSDs obtained by the kit's manufacturer were 4.7% and 9.5% for samples, with anti-Hib IgG levels ranging from 0.2 to 5.5 μg/ml. Our mean inter-assay precision was 9.1%, which was in line with the data provided. Also, the GGL matrix did not hamper the precise measurement, as demonstrated by the RSDs of 3.9%. The means (n = 18) obtained for the two control samples had RSDs of 6.2% and 7.1%. For the anti-Hib IgG and the anti-PCP IgG ELISA, the RSDs determined for the validation samples ranged from 7.4% to 10.8% and from 1.1% to 4.4%, respectively. Obviously, the anti-PCP ELISA showed a better precision than the anti-Hib ELISA. This is most probably related to the samples' higher anti-PCP IgG concentrations, which ranged from 5 to 100 μg anti-PCP IgG/ml, compared with 0.14 to 2.8 μg anti-Hib IgG/ml. Interestingly, in the concentration range investigated that showed a spread of 20, there was no obvious relation between the analyte level and the precision determined. As stated above, before implementation of the ELISA, RABA was used to measure anti-Hib IgG. In their comparative study, Madore et al. [38] distributed 20 serum samples with anti-Hib IgG levels from 0.1 to 165 μg/ml to eleven laboratories. RABA precision data, expressed as RSDs determined for the inter-laboratory means, ranged from 14.7% to 63.9%, whereas the ELISA data scattered from 6.7% to 73.8%. The authors concluded based on these data that the anti-Hib ELISA can serve as an alternative to the RABA. The precision data obtained here for the anti-Hib IgG ELISA undoubtedly reflect progress in the development of the anti-Hib IgG ELISA method. Moreover, the assay characteristic total error was clearly emphasized by the EMA guideline recommending a total error of not higher than 30% for ligand-binding assays [30]. The data obtained clearly met this criterion because we found total errors ranging from 9.8% to 11.2% and 7.9% to 11.2% for the anti-Hib IgG and anti-PCP IgG ELISA, respectively. The appropriate assay linearity was investigated and confirmed on two levels. Finally, the assays' selectivity/specificity was confirmed by the comparison of the signals obtained for the IgG-depleted serum, the blank and the assay calibrator with the lowest analyte concentration. In our validation we did not include hemolyzed or lipidic plasma/serum samples because due to increased assay sensitivity of the modified ELISAs higher dilutions can be applied for the measurement which will iron out a possible influence of such samples.

The validated assays were then applied for the measurement of the stability samples. At least two approaches are accepted for evaluating the data obtained in such studies: While in the EMA guideline a ± 20% difference to the concentration measured initially indicates stability [30], guidance documents on incurred sample analysis [45] promote a much wider stability-indicating criterion. In particular, the common acceptance criterion for incurred sample analysis using ligand-binding assays requires that the concentrations obtained by the initial and the reanalysis should be within 30% of their mean for at least two third of the samples tested. We here decided to apply the more stringent 100 ± 20% criterion for the stability evaluation. Moreover, we used regression analysis and statistical multiple comparison testing of the data obtained for the stability samples. Thus, the data we obtained for anti-Hib IgG in IgG-depleted serum kept frozen at -20°C predicted that 80% of the initial concentration would be present after a 36-month storage. In particular, the anti-Hib IgG concentrations measured at the different time points for up to 12 months did not differ significantly from that initially determined at the start of the study. The anti-PCP IgG was shown to be even more stable than the anti-Hib IgG because the regression analysis predicted 80% of the initial concentration would be recovered after 184 months. Statistically significant differences were identified between the anti-PCP IgG starting concentration and the concentrations determined after 3 and 9 months. Both anti-bacterial IgGs also showed good stability in the matrix of GGL when kept at -60°C: The regression analysis predicted that 80% of the initial anti-Hib IgG and anti-PCP IgG concentrations would be recovered after 126 and 51 months under these conditions, respectively. With respect to these stability data, it should be mentioned that from the date of manufacture, GGL, formulated as 10% solution in 0.25 M glycine, pH 4.6–5.1, is stable at 2 to 8°C for 3 years or at room temperature (≤25°C) for 24 months. Of course, this stability includes the preservation of the antibody titers against measles and hepatitis B surface antigen that are determined with a neutralization assay and have to meet the US release criteria, that is, ≥0.22-times the antibody level of CBER reference measles immune globulin lot 176 and ≥0.2 IU/ml, respectively, also at the end of the shelf-life.

Nevertheless, we observed a slight decrease in the anti-bacterial IgG levels over time as measured with these particular ELISAs. The stability data obtained for the GGL preparation, kept frozen at -60°C, could advocate for storage at -60°C because the anti-bacterial IgG concentrations measured decreased even less over time. However, these data were obtained at higher anti-bacterial IgG levels and not in the matrix of IgG-depleted serum. Functionally, the specific binding to their targets, in both cases complex but well-defined polysaccharides structures, requires the presence of specific structural areas, so-called paratopes in the antibody. Ma et al. [46] provide a comprehensive overview of the state-of-the-art analytics applied for the structural analysis of antibodies: Circular dichroism has been widely used for studying the solution stability of antibodies and the method has been shown to discriminate between native, unfolded, and aggregated states of antibodies. Similarly, nuclear magnetic resonance spectroscopy has been applied for this purpose, while mass spectrometry after hydrogen/deuterium exchange has been described for determining the influence of excipients on the stability of antibodies. Size exclusion chromatography, differential scanning calorimetry and fluorimetry, and electrophoretic methods complete the analytical standard repertoire available for the antibody characterization. Apart from these biophysical methods, which require the presence of pure antibody at a relatively high concentration so that they will not work properly in serum matrix at dilute antibody levels, measurement of the antibodies' biological activity, i.e., their antigen binding, seems to be the stability-indicating method that provides the most relevant information. Because of the high selectivity/specificity of such binding assays that are based usually on ELISA or surface plasmon resonance spectroscopy (SPR), neither the serum/plasma matrix nor the low antibody concentration are expected to create insurmountable obstacles. It is reasonable to assume that reduced target binding reflects a structural alteration of the antibody binding site. Theoretically, antibody degradation that results in reduced binding to the target antigen can occur via several mechanisms. Proteolysis that can be pH dependent most often affects the proteolytically susceptible hinge region of the antibody and generates antibody fragments. Of note, different IgG subclasses will be affected differently with subclass IgG2 showing a clearly lower susceptibility against proteolytic degradation than IgG1 [47] because of its rigid hinge region [48]. This is interesting because several reports have identified subclass IgG2 antibodies to be mainly associated with the immune response against bacterial carbohydrate antigens, although in general anti-carbohydrate IgGs are not restricted to the IgG2 subclass [49]. Even though some of those fragments will still bind to the immobilized antigen, they will no longer be detected by ELISA because it usually uses an Fc-specific detection system, which requires the antibody to be intact. In cases where apart from the antigen binding interaction with Fc receptors is also required to exert biological activity, this dual binding feature of ELISA fortunately somehow mimics the loss in biological activity. By contrast, binding activity measured by SPR should be less probably be reduced, although the lower mass transfer caused by the binding of antibody fragments could result in different binding curves. Proteolytic antibody degradation probably occurs in serum samples at a higher probability because during serum preparation activated proteases, including thrombin, are generated during serum preparation. Traces of those proteases could remain in the serum preparation and attack susceptible areas of IgGs despite the presence of natural protease inhibitors. Accordingly, if there is a choice for selecting the patients' specimens to be analyzed, preparing citrated or EDTA plasma instead of serum seems to be a better alternative because the plasma preparation avoids the initiation of the clotting process. Antibody inactivation by aggregation is another possibility, although it seems unlikely to occur because the antibody is stabilized by the biological matrix. Repeated freezing–thawing processes have been described to exert stress that could lead to denaturation and associated loss of biological, that is, binding activity. Finally, chemical modification of the amino acid backbone caused by oxidation or deamidation could impact the binding activity of an antibody given that the susceptible amino acids build up an essential part of the binding surface of the antibody. Methionine, serine, and asparagine belong to the amino acids affected most by oxidation and deamidation. Kuzman et al. [50] recently reported on a model for purified monoclonal antibodies that allowed the accurate prediction of 3-year stability based on data generated in accelerated stability studies over 6 months. As such alterations are also expected to proceed in the frozen state, although at a much lower rate, Michaut et al. [51] and Li et al. [52] evaluated the stability of antibodies generated after vaccination by ELISA retesting of frozen serum and whole blood samples, respectively. While the latter authors applied the ± 20% criterion, the acceptance criterion for incurred sample reanalysis defined for ligand-binding assays [53] was applied by Michaut et al. Here, we applied the stricter ± 20% criterion for the stability evaluation throughout the whole work presented.

Finally, the modified and validated anti-Hib and anti-PCP IgG ELISAs presented here, which demonstrated lower limits of quantitation may also be useful for clinical applications related to the PID diagnosis. Thus, the antibody response to Hib-conjugated vaccine and a pneumococcal vaccine was studied in healthy adults and patients with humoral immunodeficiency [54]. Because none of the immunodeficient patients showed antibody responses to the vaccinations, this methodology might be helpful in detecting and confirming humoral immunodeficiency and support the selection of patients for IgG replacement therapy. To apply validated assays with sensitive quantification limits might provide benefits for this purpose.

Conclusion

The data presented confirmed the adequate performance of the two modified commercial ELISAs when evaluated against the acceptance criteria defined for ligand binding assays by the EMA guideline for bioanalytical method validation. Using these validated ELISAs for the stability evaluation of anti-Hib and anti-PCP IgG in human IgG-depleted serum, we were able to demonstrate the adequate stability of both anti-bacterial IgGs when kept frozen at -20°C. These data furthermore verify the common practice in clinical testing, where samples' batches are measured after having kept them frozen at -20°C or at lower temperatures.