Development of a Measles and Rubella Multiplex Bead Serological Assay for Assessing Population Immunity

ABSTRACT

Serosurveys are important tools for estimating population immunity and providing immunization activity guidance. The measles and rubella multiplex bead assay (MBA) offers multiple advantages over standard serological assays and was validated by comparison with the enzyme-linked immunosorbent assay (ELISA) and the measles plaque reduction neutralization (PRN) assay. Results from a laboratory-produced purified measles virus whole-virus antigen MBA (MeV WVA_L) correlated better with ELISA and PRN than results from the baculovirus-expressed measles nucleoprotein (N) MBA. Therefore, a commercially produced whole-virus antigen (MeV WVA_C) was evaluated. Serum IgG antibody concentrations correlated significantly with a strong linear relationship between the MeV WVA_C and MeV WVA_L MBAs (R = 0.962 and R² = 0.926). IgG concentrations from the MeV WVA_C MBA showed strong correlation with PRN titers (R = 0.846), with a linear relationship comparable to values obtained with the MeV WVA_L MBA and PRN assay (R² = 0.716 and R² = 0.768, respectively). Receiver operating characteristic (ROC) curve analysis of the MeV WVA_C using PRN titer as the comparator resulted in a seroprotection cutoff of 153 mIU/ml, similar to the established correlate of protection of 120 mIU/ml, with a sensitivity of 98% and a specificity of 83%. IgG concentrations correlated strongly between the rubella WVA MBA and ELISA (R = 0.959 and R² = 0.919). ROC analysis of the rubella MBA using ELISA as the comparator yielded a cutoff of 9.36 IU/ml, similar to the accepted cutoff of 10 IU/ml for seroprotection, with a sensitivity of 99% and a specificity of 100%. These results support use of the MBA for multiantigen serosurveys assessing measles and rubella population immunity.

INTRODUCTION

All six World Health Organization (WHO) regions have established measles elimination goals, and four of the six regions have established rubella elimination goals. Elimination is defined as the absence of endemic transmission of measles virus (MeV) or rubella virus (RuV) for more than 12 months in a country with adequate surveillance capacity. Verification of elimination evaluates multiple lines of evidence, including documentation that endemic transmission has been interrupted for at least 3 years as well as a reliable measurement of population immunity (1–4). The high-quality laboratory surveillance needed to support elimination efforts is provided by the WHO Global Measles and Rubella Laboratory Network (GMRLN) (1, 5, 6).

Population immunity may be documented in several ways, including the frequently used methods of administrative immunization coverage and postcampaign coverage surveys; however, these estimates are often not reliable measures of population immunity, and pockets of susceptible individuals may still exist (7–10). There are numerous examples of large outbreaks resulting from immunity gaps in areas where reported vaccination coverage estimates were high (11–15), and recent outbreaks have resulted in the loss of elimination status in several countries (16–19). Therefore, seroprevalence studies are a useful tool to measure the effectiveness of vaccination programs and to plan vaccination strategies.

The “gold standard” for evaluating protection against measles is the plaque reduction neutralization (PRN) assay, and for rubella, it is the focus reduction assay (20, 21). However, these assays are time-consuming, require highly technical laboratory expertise, and are not conducive to high throughput. The common method to evaluate immunity is the enzyme-linked immunosorbent assay (ELISA). However, when survey objectives have more than one disease of interest, separate ELISAs are required to measure each of the disease-specific antibodies, increasing specimen volume requirements, staff time, and cost (2). Measles and rubella elimination strategies include the use of a combined measles-rubella vaccine, so serosurveys are often designed to estimate seroprotection to both viruses (22).

The measles and rubella regional verification process has increased demand for high-quality seroprevalence data, creating the need for improved serological methods. As an alternative to ELISA, the multiplex bead assay (MBA) offers several advantages, including the ability to simultaneously measure antibodies for multiple antigen targets with a very small volume of serum (1 to 5 μl) (2, 23–25). In addition, the MBA is capable of the high throughput needed to perform large seroprevalence studies while measuring antibody concentrations for up to 500 different antigens depending on the instrument used (2).

This report describes the further development and validation of an MBA that can provide high-quality seroprevalence data for measles and rubella (23). Recombinant measles nucleoprotein (N) and whole-virus lysate were compared as antigens because both have been used in ELISAs. Whole virus lysate was used for the rubella assay as described previously (23). To validate the use of the MBA for estimating measles and rubella seroprotection, MBA results were compared with measurements of measles and rubella humoral immunity by ELISA and measles immunity by PRN.

MATERIALS AND METHODS

Sera.

To evaluate the MBA for measles and rubella, 516 residual human serum samples from 4 studies were used in various analyses as described in Table 1. Sera are described here by source country; Table 1 indicates serum sets and number of sera used per evaluation. Serum samples from the United States included 40 samples for routine measles case-based surveillance; of these, 34 sera were tested on both measles virus whole-virus antigen (MeV WVA) MBAs. U.S. sera also included 49 paired serum samples (for a total of 98 samples) from an early revaccination cohort study (26), in which the immune response following primary vaccination was evaluated in participants; serum samples used for this study include prevaccination and post-second dose measles, mumps, and rubella (MMR) vaccine given 5 to 7 months following the primary dose. Serum samples from Tajikistan included 100 samples that were collected as part of a serological survey in 2010; of these, 80 samples were tested on both MeV WVA MBAs (27). Another 152 paired serum samples (for a total of 304 samples) from Bangladesh that included pre- and post-measles and rubella (MR) vaccination samples were also included (Table 1) (28). Overall, the serum specimens represented 6-month-old infants to adults.

TABLE 1.

Serum sample sets for measles and rubella MBA evaluation

Serum set	Country and serum source	Purpose	Total no. of sera	MBA performed	Other assays performed	Analyses performed	Figure references
1	USA, residual diagnostic samples (40 sera); Tajikistan (100 sera)	Compare the performance of MeV WVA and N MBAs with PRN and ELISA	140	MeV WVAL, MeV N	Measles PRN, measles ELISA (Zeus and Siemens)	Pearson’s correlation, linear regression, ROC	Fig. 1, 2, and 8A and B
2	USA, residual diagnostic samples (34 sera); USA, paired samples, early revaccination cohort study (49 paired; total, 98 sera); Tajikistan (80 sera)	Determine if MeV WVAC had similar performance as the MeV WVAL MBA	212	MeV WVAL, MeV WVAC	Measles PRN, measles ELISA (Zeus and Siemens)	Pearson’s correlation, linear regression	Fig. 3 to 6 and 8C
3	USA, residual diagnostic samples (34 sera); USA, paired samples, early revaccination cohort study (49 paired; total, 98 sera); Tajikistan (80 sera); Bangladesh, paired vaccination cohort study (152 paired; total, 304 sera)	Define performance characteristics of MeV WVAL MBA by ROC analysis	516	MeV WVAL	Measles PRN, measles ELISA (Zeus and Siemens)	ROC	Fig. 8D
4	Bangladesh, paired vaccination cohort study (152 paired; total, 304 sera)	Evaluation of RuV WVA MBA compared with ELISA	304	RuV WVA	Rubella ELISA (Zeus)	Pearson’s correlation, linear regression, ROC	Fig. 7 and 8E

Reference standards.

The international rubella serum standard RUBI1-94 (1,600 IU/ml; National Institute for Biological Standards and Control [NIBSC], Potters Bar, United Kingdom) was used to determine measles and rubella virus concentrations of serum samples by MBA. RUBI-1-94 was previously calibrated to the 2nd international measles serum standard (NIBSC 66/202; 5 IU/ml) and determined to have a concentration of measles virus antibody of 61.2 IU/ml; thus, a single standard was used for evaluation of both measles and rubella virus antibodies in study serum samples. The measles 3rd international serum standard (NIBSC 97/648; 3 IU/ml) was used to generate the standard curve for evaluation of the nucleoprotein (N) MBA. The 2nd international measles serum standard (66/202) was used to determine measles neutralizing antibody concentrations by measles PRN.

MBA.

(i) Antigens and bead coupling procedure. Three measles virus (MeV) antigens and one rubella virus (RuV) antigen were evaluated. A whole-virus lysate of the MeV Edmonston strain produced by the National Institute for Public Health and Environment, the Netherlands (RIVM), was used for the MeV whole-virus antigen laboratory preparation (WVA_L) MBA. The MeV WVA_L was lyophilized in stabilization buffer containing hydrolyzed gelatin as described previously (23). The antigen was resuspended in phosphate-buffered saline (PBS) with 0.1% Tween 20, frozen at −80°C overnight, slowly defrosted the following day at room temperature (20 to 25°C), and then incubated at room temperature for 30 min, followed by sonication for 1 min in a microsonicator before coupling. A total of 150 μg/12.5 × 10⁶ microspheres from a 0.9-μg/μl stock was used to couple MicroPlex (Luminex, Austin, TX) or MagPlex (Luminex) beads as described below.

Five commercial antigens were evaluated using the same coupling and assay procedure as was performed for the MeV WVA_L. Commercial antigen-coupled beads were tested with serial dilutions of the NIBSC RUBI-1-94 standard and compared to the standard curve from the MeV WVA_L MBA. One commercial measles whole-virus lysate antigen, a sucrose gradient-purified heat-inactivated vaccine strain of measles virus propagated in Vero cells (Zeptometrix, Buffalo, NY), gave results similar to the MeV MBA_L standard curve, this antigen was used to prepare an MeV WVA commercial (WVA_C) MBA using MicroPlex or MagPlex beads. The measles virus commercial antigen (Zeptometrix) was dialyzed against 50 ml of PBS for 4 h at room temperature to remove buffer components that were incompatible with the coupling reaction. Protein concentration following dialysis was evaluated by bicinchoninic acid (BCA) assay (Pierce, Grand Island, NY). The coupling concentration described above of 150 μg/12.5 × 10⁶ microspheres was used for WVA_C coupling following the procedure described below.

A recombinant MeV nucleoprotein antigen (Meridian Life Sciences, Memphis, TN) was produced by baculovirus expression in SF9 cells and partially purified by Mono Q column chromatography (GE Healthcare, Piscataway, NJ) and coupled at a concentration of 6 μg/12.5 × 10⁶ SeroMAP (Luminex, Austin, TX) beads in 50 mM 2-(N-morpholino)ethanesulfonic acid (MES)/NaCl buffer (pH 5.0) (29).

Rubella MicroPlex and MagPlex beads were prepared using 30 μg of HPV-77 rubella whole-virus antigen (Meridian Life Sciences, Memphis, TN) and diluted to 0.5 μg/μl in PBS and 30 μg/12.5 × 10⁶ microspheres, similar to the bead activation and coupling procedure as previously described (23).

Measles and rubella virus antigens were coupled to either MicroPlex, MagPlex, or SeroMAP (Luminex, Austin, TX) beads using a standard protocol for the evaluation of the MBA. MicroPlex and SeroMAP beads were previously used with the measles virus antigen formats described above; MagPlex beads were additionally included for evaluation to allow greater instrument versatility. Beads were activated by incubation in 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC; Thermo Fisher Scientific, Waltham, MA) and N-hydroxysulfosuccinimide (sulfo-NHS; Thermo Fisher Scientific) for 20 min at room temperature protected from light and with constant rotation. Antigen preparations as described above were incubated with the activated beads at room temperature for 2 h protected from light and with constant rotation. Measles WVA-coupled beads required a “resting” time in which bead aggregates were broken apart by sonication and vortexing several times a week for several weeks. Even dispersal of the beads was confirmed visually with a hemacytometer. All bead lots were validated with a known serum panel.

(ii) MBA testing procedure. For MicroPlex and SeroMAP assays, filter bottom plates were used (Millipore, Burlington, MA). Filter bottom plates were prewet with 100 μl of PBS. For MagPlex assays, Mylar bottom plates were used (Bio-Rad, Hercules, CA). Two different assay configurations were used as previously optimized for the different measles MBAs. The MicroPlex and MagPlex beads coupled with MeV WVA or RuV WVA were assayed as follows. Serum samples were diluted either 1:200 and 1:4,000 or only 1:400 in serum dilution buffer containing 3% bovine serum albumin (BSA) in PBS with 0.1% Tween 20. NIBSC standard RUBI-1-94 was 3-fold serially diluted in serum dilution buffer starting at a concentration of 1:400. The PBS was removed from filter plates using gentle suction by a vacuum manifold. Microspheres were diluted to a concentration of 2,500 to 4,000 beads/region/25 μl, and 25 μl/well was added to assay plates. Diluted samples, controls, and standard (25 μl) were added to beads in assigned wells and shaken at 600 rpm for 45 min. Unbound antibody was washed away with PBS with 0.5% Tween 20. Washing steps were repeated three times. The anti-human IgG-R-phycoerythrin (PE) detection antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) was used at a 1:200 dilution, and 50 μl was added to each well and incubated with shaking for 30 min. Following a washing, 100 μl of PBS was added to the beads, shaken for 5 min at 600 rpm, and read on a Luminex instrument. Luminex instruments used for the MicroPlex and MagPlex data collection included a BioPlex 100 and a FlexMAP3D (Bio-Rad).

The SeroMAP beads coupled with MeV nucleoprotein were used in assays with sera at 1:400 dilutions in buffer containing casein, BSA, polyvinyl alcohol, and polyvinylpyrrolidone (Sigma-Aldrich Chemical Co., St. Louis, MO) as previously described (30). Bound IgG antibodies were detected using biotinylated mouse anti-human IgG and IgG4 secondary antibodies (Southern Biotech, Birmingham, AL) and streptavidin-linked PE (Thermo Fisher Scientific, Waltham, MA). Data were analyzed on a BioPlex200 instrument (Bio-Rad) (30).

Plaque reduction neutralization assay.

The WHO standard measles PRN assay was described previously (31). Briefly, 1:4 serial dilutions of serum were preincubated for 2.5 h at 36°C with measles virus (Edmonston strain) at a dilution to achieve 35 to 50 PFU in control wells. Following incubation, the serum-virus mixture was added to a Vero cell monolayer, incubated for 1 h at 36°C, overlaid with a final concentration of 2% carboxymethylcellulose in Leibovitz-15 medium (Quality Biologicals, Gaithersburg, MD) supplemented with 10% fetal calf serum (Bio-techne, Minneapolis, MN), penicillin/streptomycin, amphotericin B (Fungizone), glutamine, phenol red (Gibco, Waltham, MA), and sodium bicarbonate (Sigma, St. Louis, MO), and incubated for 5 days at 36°C. Cells were fixed and plaques visualized by crystal violet staining (32). In this study, the second international anti-measles serum standard (66/202; NIBSC) was included to calculate the measles neutralizing antibody concentration in the reciprocal 50% endpoint titer calculated by the Kärber method (33). A dilution of the second international standard was included in each assay, and endpoint titers for the internal standards did not differ by more than 20%. Based on the standard curve of the second international serum standard, PRN titers were expressed in milli-international units per milliliter. A titer of 8 was equivalent to a concentration of 8 mIU/ml. A PRN titer of 120 was considered protective for measles (20) and is equivalent to 120 mIU/ml based on use of the second international standard.

ELISA.

Sera from the United States were tested for measles virus IgG using a commercial ELISA (Zeus Scientific, Branchburg, NJ) according to manufacturer’s protocols. Sera from Tajikistan and Bangladesh were tested for measles virus IgG using a commercial ELISA (Enzygnost IgG; Siemens, Malvern, PA), and optical density (OD) values were converted to milli-international units per milliliter using the method developed by the manufacturer. Serum set 4 (Table 1) was tested for rubella virus IgG using the Zeus Scientific kit as described in a previous publication (28). The rubella ELISA-normalized OD result was converted to antibody concentration based on an NIBSC international standard curve and lot-specific conversion value, and the concentration in milli-international units per milliliter was determined based on the ELISA result.

Statistics.

Statistical analysis was performed using GraphPad Prism version 6 software. Antibody concentration values determined by standard curve analysis in the MBA were lognormal transformed, linear regression was performed, and an R² value was obtained for the curve fit. Pearson’s correlation coefficient (r) and 95% confidence intervals (CI) were determined across comparison groups.

Sensitivity and specificity analysis.

Sensitivity and specificity were determined by receiver operating characteristic (ROC) curves generated in GraphPad Prism version 6. The comparator for measles ROC curve analysis was a PRN result with ≥120 mIU/ml used as the cutoff for seropositivity (20). The comparator for rubella ROC curve analysis was an ELISA result with ≥10 IU/ml used as the cutoff for seropositivity (22).

Ethics/human subjects.

All samples were deidentified, and the study was determined to be “nonhuman subjects research” by Centers for Disease Control and Prevention human subject review.

RESULTS

Comparison of MeV N MBA and MeV WVA_L MBA with PRN assay and ELISA.

Measles MBAs using either an N antigen or WVA were compared with the “gold standard” PRN assay. The performances of the MeV N MBA and the MeV WVA_L MBA were compared with PRN assay using serum set 1, comprised of 140 serum samples from the United States and Tajikistan (Table 1). Pearson’s R value for the MeV WVA_L compared with the PRN was 0.909 (95% CI, 0.875 to 0.935, and P value, <0.0001), indicating a strong correlation between the IgG concentrations obtained with the MeV WVA_L MBA and the PRN titers (Table 2). In contrast, the MeV N MBA compared with PRN had a weaker correlation, resulting in an R value of 0.656 (95% CI, 0.550 to 0.741, and P value, <0.0001) (Table 2). Linear regression of the lognormal-transformed data (Fig. 1) demonstrated a strong linear relationship between the MeV WVA_L MBA and the PRN (R² = 0.827), but the linear relationship was less strong for the MeV N MBA and the PRN (R² = 0.431), supporting the calculated Pearson’s correlation coefficients. The slope of the linear regression indicated that the log concentration (milli-international units per milliliter) of MeV IgG-specific antibodies was close to one for the MeV WVA_L and PRN (slope = 0.962) but that the MeV N MBA generally resulted in lower calculated IgG concentrations than the PRN, as demonstrated by the slope of 0.5003 (Fig. 1).

TABLE 2.

Pearson’s correlation coefficient evaluation for MeV MBA

Serum set	Total no. of sera	Comparison groups	Ra	R2a
1	140	MeV WVAL MBA vs PRN	0.909b	0.827
140	MeV N MBA vs PRN	0.656	0.431
40	MeV WVAL MBA vs ELISA (Zeus)	0.946b	0.895
100	MeV WVAL MBA vs ELISA (Siemens)	0.827b	0.683
40	MeV N MBA vs ELISA (Zeus)	0.938	0.879
100	MeV N MBA vs ELISA (Siemens)	0.786	0.619
2	212	MeV WVAC MBA vs MeV WVAL MBA	0.962	0.926
212	MeV WVAC vs PRN	0.846	0.716
212	MeV WVAL MBA vs PRN	0.877b	0.768
132	MeV WVAC MBA vs ELISA (Zeus)	0.885	0.784
80	MeV WVAC MBA vs ELISA (Siemens)	0.846	0.715
132	MeV WVAL MBA vs ELISA (Zeus)	0.932b	0.869
80	MeV WVAL MBA vs ELISA (Siemens)	0.805b	0.649
132	MeV ELISA (Zeus) vs PRN	0.862	0.743
80	MeV ELISA (Siemens) vs PRN	0.796	0.633
4	304	RuV WVA MBA vs ELISA (Zeus)	0.959	0.919

^aThe R value indicates Pearson’s correlation coefficient between results of comparison groups listed. The R² value depicts the linear fit of that correlation.

^bEvaluation was performed with both the 140-serum sample set and repeated for the 212-serum sample set.

FIG 1.

Linear regression analysis of MBA compared with PRN. A comparison of the MeV WVA_L (A) or MeV N MBA (B) was conducted using serum set 1 (140 serum samples). The results from the MBA were compared with PRN results for the same samples by linear regression. Linear regression was performed on lognormal-transformed concentrations in milli-international units per milliliter determined by standard curve. Linear regression parameter results for the best-fit regression line are noted in each graph.

The MeV WVA_L MBA and MeV N MBA performed similarly to two commercial IgG ELISA kits. For serum set 1, two commercial ELISA kits were used in prior evaluations, which allowed comparison of the MeV WVA_L and MeV N MBAs to both commercial IgG ELISA kits. Samples of U.S. origin (n = 40) were evaluated by the Zeus MeV ELISA, and samples from Tajikistan (n = 100) were evaluated by the Siemens MeV ELISA. In comparing the 40 samples previously tested with the Zeus ELISA to the MeV WVA MBA, the R value was found to be 0.946 (95% CI, 0.898 to 0.972, and P value, <0.0001). The comparison of the 100 samples previously tested by the Siemens ELISA to the MeV WVA MBA yielded an R value of 0.827 (95% CI, 0.752 to 0.880, and P value, <0.0001) (Table 2). The comparison of the same Zeus (n = 40) and Siemens (n = 100) ELISA results with the MeV N MBA yielded values of 0.938 (95% CI, 0.885 to 0.967, and P value, <0.001) and 0.786 (95% CI, 0.698 to 0.851, and P value, <0.0001), respectively (Table 2). Linear regression analysis demonstrated that the MeV WVA_L MBA and MeV N MBA compared similarly to both ELISAs (Fig. 2).

FIG 2.

Linear regression analysis of MBA compared with ELISA. The MeV WVA_L MBA (A and C) and MeV N MBA (B and D) were compared with two commercial MeV ELISA kits, Zeus and Siemens, using serum set 1. Forty samples were compared with the Zeus ELISA, and 100 samples were compared with the Siemens ELISA. Linear regression was performed on lognormal-transformed index standard ratio (ISR) (normalized OD, Zeus) or concentrations in milli-international units per milliliter (Siemens) compared with concentrations in milli-international units per milliliter from MBA determined by standard curve. Linear regression parameter results for the best-fit regression line are noted in each graph.

Identification and evaluation of a commercial source of MeV WVA.

Widespread use of a WVA-based MBA for large-scale seroprevalence studies would be facilitated by having a commercial source of MeV WVA. One whole-virus lysate commercial antigen of the five commercial MeV antigens tested as described in Materials and Methods gave results similar to the MeV WVA_L MBA NIBSC standard curve (data not shown) and was further evaluated as the MeV WVA_C MBA using a panel of 212 serum samples from the United States and Tajikistan (serum set 2 [Table 1]). The MeV WVA_C with the MeV WVA_L MBA performed comparably; the R value was 0.962 (95% CI, 0.951 to 0.971, and P value, <0.0001), the R² was 0.926, and the slope was 1.06 (Table 2 and Fig. 3).

FIG 3.

Linear regression analysis of MeV WVA_C MBA compared with MeV WVA_L MBA. Evaluation of the commercial antigen was performed with serum set 2 (212 serum samples). Serum IgG antibody concentrations determined by the MeV WVA_C MBA were compared with those from the MeV WVA_L MBA by linear regression. Linear regression was performed on lognormal-transformed concentrations in milli-international units per milliliter determined by standard curve. Linear regression parameter results for the best-fit regression line are noted in each graph.

Comparison of MeV WVA_C MBA and MeV WVA_L MBA with PRN and ELISA.

Serum IgG antibody concentrations determined by the MeV WVA_C and MeV WVA_L MBAs were compared with both the PRN assay and ELISA, tested side by side using serum set 2 (n = 212) (Table 1). The MeV WVA_C MBA demonstrated a strong correlation with the PRN assay, with an R value of 0.846 (95% CI, 0.802 to 0.881, and P value, <0.0001) and an R² value of 0.716, similar to what was observed with the MeV WVA_L MBA (Table 2 and Fig. 4). A comparison of the MeV WVA_L MBA had already been conducted on the BioPlex 100 instrument using 140 serum samples from the United States and Tajikistan (serum set 1 [Table 1]), as described above. Because the MeV WVA_C was tested on a FlexMAP3D instrument, the evaluation of the MeV WVA_L MBA was repeated on the FlexMAP3D instrument with serum set 2. The repeat analysis yielded results similar to those of the initial analysis; there was a strong correlation with PRN, with an R value of 0.877 (95% CI, 0.841 to 0.905) and an R² value of 0.768 (Table 2 and Fig. 4). Both the MeV WVA_C MBA and MeV WVA_L MBA performed similarly to two the commercial ELISA kits (Siemens and Zeus). The correlations of the MeV WVA_C MBA with Zeus and Siemens ELISAs were 0.885 (95% CI, 0.841 to 0.918, and P value, <0.0001) and 0.846 (95% CI, 0.767 to 0.899, and P value, <0.0001), respectively; these results were similar to those of the MeV WVA_L: 0.932 (95% CI, 0.905 to 0.952, and P value, <0.0001) and 0.805 (95% CI, 0.712 to 0.899, and P value, <0.0001), respectively (Table 2). The R² values from linear regression analysis for the WVA_C and WVA_L MBAs compared with the Zeus ELISA were 0.784 and 0.869, respectively (Fig. 5A). For comparison with the Siemens ELISA, the R² values were 0.715 and 0.649 for the WVA_C and WVA_L MBAs, respectively (Fig. 5B).

FIG 4.

Linear regression analysis of MeV WVA MBAs compared with PRN assay. Serum IgG antibody concentrations from serum set 2 (212 serum samples) tested with the MeV WVA_L MBA and MeV WVA_C MBA were compared with PRN titers for the same samples by linear regression. Linear regression was performed on lognormal-transformed concentrations in milli-international units per milliliter determined by standard curve. Linear regression parameter results for the best-fit regression line are noted in each graph.

FIG 5.

Linear regression analysis of MeV WVA MBAs compared with ELISA. From serum set 2, 132 samples were analyzed by the Zeus measles ELISA (A) and 80 samples were analyzed by the Siemens measles ELISA (B). The results from the MBA were compared with ELISA results for the same samples by linear regression. Linear regression was performed on lognormal-transformed ISR (normalized OD, Zeus) or concentrations in milli-international units per milliliter (Siemens) compared with concentrations in milli-international units per milliliter from the MBA determined by standard curve. Linear regression parameter results for the best-fit regression line are noted in each graph.

Comparison of ELISA with PRN.

While the goal of this study was to evaluate the MBA, we also compared ELISA results with the PRN titers using the data obtained with serum set 2. The comparison yielded Pearson’s correlation values of 0.862 and 0.796 and R² values for the lognormal-transformed curves of 0.743 and 0.633 for Zeus and Siemens ELISAs, respectively (Table 2 and Fig. 6). These results indicated that the MeV WVA MBAs correlated as well or slightly better with PRN than the ELISA (Fig. 4 and 6).

FIG 6.

Linear regression analysis of ELISA compared with PRN assay. From serum set 2, 132 samples were analyzed by the Zeus measles ELISA (A) and 80 samples were analyzed by the Siemens measles ELISA (B). The results from the ELISA were compared with PRN titers for the same samples by linear regression. Linear regression was performed on log normal transformed ISR (normalized OD, Zeus) or concentrations in milli-international units per milliliter (Siemens) compared with concentrations in milli-international units per milliliter from PRN determined by standard curve. Linear regression parameter results for the best-fit regression line are noted in each graph.

Comparison of RuV WVA MBA with quantitative ELISA.

The rubella MBA was based on the assay described in reference 23. The rubella WVA MBA compared well with the Zeus rubella ELISA as determined by evaluation of 152 paired pre- and postvaccination samples (serum set 4 [Table 1]). The comparison yielded Pearson’s correlation coefficient of 0.959 (95% CI, 0.948 to 0.967, and P value, <0.0001), an R² value of 0.919, and a slope of 1.54 (Table 2 and Fig. 7).

FIG 7.

Linear regression analysis of RuV WVA MBA compared with ELISA. Serum sample set 4 (152 paired samples) was analyzed using the RuV WVA MBA and compared with samples analyzed by the Zeus rubella ELISA. Linear regression was performed on lognormal-transformed concentrations in international units per milliliter (converted optical density) calibrated to a standard curve.

Evaluation of MeV and RuV WVA MBA performance.

ROC analyses were performed for the MeV N and WVA MBAs as well as the RuV WVA MBA and cutoffs determined using the highest sensitivity and specificity achieved in Prism-tabulated results. The ROC analysis for the MeV MBAs was performed using the PRN seroprotective result of 120 mIU/ml as the MBA sample classifier for positivity (20). The ROC analysis of MeV WVA_L on serum set 1 (n = 140 samples [Table 1]) resulted in an area under the curve (AUC) of 0.924, categorizing this assay in the excellent range (34), with a sensitivity of 95% (95% CI, 88 to 98.3) and a specificity of 75% (95% CI, 59.7 to 86.8) at a cutoff of 137 mIU/ml (Table 3 and Fig. 8A). The ROC analysis of the MeV N MBA performed on serum set 1 (n = 140 samples [Table 1]) yielded an AUC of 0.890, with a sensitivity of 88% (95% CI, 79.2 to 93.4) and a specificity of 80% (95% CI, 64.7 to 90.2) at a cutoff of 9.5 mIU/ml (Table 3 and Fig. 8B). The MeV N MBA had a slightly lower false-positivity rate (20%) than the MeV WVA_L MBA (25%); however, the MeV N MBA had a higher false-negativity rate (13%) than the MeV WVA_L MBA (5%) (Table 3).

TABLE 3.

MBA performance using serum set 1 with 2 × 2 contingency tables comparing MeV WVA_L and MeV N MBAs to PRN

^aFPR, false-positivity rate, calculated by FP/FP + TN, where TN is total negative.

^bFNR, false-negativity rate, calculated by FN/FN + TP, where TP is total positive.

FIG 8.

Receiver operating characteristic (ROC) curve analysis of measles and rubella MBAs. (A) ROC analysis of MeV WVA_L MBA using serum set 1 with a total of 140 samples determined by PRN as a classifier. The MeV PRN with a cutoff value of 120 mIU/ml was used as the comparator for analysis, and a PRN titer of >120 mIU/ml was considered positive. (B) ROC analysis of MeV N MBA using serum set 1 determined by PRN as a classifier with positive samples categorized as described above. (C) ROC analysis of MeV WVA_L MBA using serum set 2 with a total of 212 samples determined by PRN as a classifier with positive samples categorized as described above. (D) ROC analysis of MeV WVA_C using serum set 3 with 516 samples determined by PRN as a classifier with positive samples categorized as described above. (E) ROC analysis of RuV WVA MBA determined by quantitative ELISA as a classifier using serum set 4 with 304 samples. ELISA results were converted to international units per milliliter, and a titer of >10 IU/ml was considered positive. ROC analysis results are displayed for each MBA evaluation.

The performance of WVA_C MBA was found to be similar to that of the MeV WVA_L MBA (Table 4) upon evaluation of serum set 2 (Table 1). The false-positivity rate of the WVA_L MBA was found to decrease to 14% when evaluating results obtained with serum set 2 (Table 1), which contained more negative samples than serum set 1 (Table 4). The ROC analysis of the MeV WVA_L using serum set 2 showed an AUC of 0.939, a sensitivity of 96%, and an increased specificity of 86% compared to the ROC analysis performed with serum set 1 (Table 4 and Fig. 8C). The ROC analysis of the MeV WVA_C MBA performed on serum set 3 (Table 1) resulted in an AUC of 0.966, a sensitivity of 98% (95% CI, 95.8 to 99.6), and a specificity of 83% (95% CI, 78.3 to 87.6), with a cutoff of 153 mIU/ml (Fig. 8D).

TABLE 4.

MeV WVA MBA performance using serum set 2 with 2 × 2 contingency tables comparing MeV WVA_L and MeV WVA_C MBAs to PRN

^aFPR, false-positivity rate, calculated by FP/FP + TN, where TN is total negative.

^bFNR, false-negativity rate, calculated by FN/FN + TP, where TP is total positive.

The ROC analysis for the RuV WVA MBA was performed on serum set 4, with 152 paired serum samples (Table 1), using the quantitative ELISA as a classifier and resulted in an AUC of 1, a sensitivity of 99%, and a specificity of 100%, with a cutoff of 9.36 IU/ml (Fig. 8E).

DISCUSSION

Serosurveys are important tools for estimating population immunity and can provide support of elimination goals and guidance for immunization activities. MBAs are increasingly used for serosurveys (2, 24, 25, 30, 35–38), and a measles and rubella WVA MBA has been described previously (23). However, the measles component of that assay used a laboratory preparation of WVA which is not amenable to scale-up. This study identified a commercially available WVA for the MeV MBA and demonstrated a strong correlation of the MeV WVA_C MBA with the gold standard, PRN assay. Both MeV WVA MBAs and the MeV N MBA performed comparably to MeV ELISA, which is most often used for serological surveillance, with the MBA providing the additional benefit of multiplexing antigen targets. Determining a seroprotective titer requires the assay to yield quantitative results. For rubella, the commercially available ELISAs are standardized to the WHO international standard for rubella and most ELISAs can report a quantitative result (22, 39). However, for measles many ELISAs are not calibrated to the WHO international standard for measles, and therefore, serosurveillance studies are unable to report quantitative results (22). The gold standard to determine measles seroprotection remains the PRN assay because it measures neutralizing antibodies. For this reason, PRN was used as the comparator for this study. Of the three MeV MBAs that were evaluated, the two MeV WVA MBAs correlated better to the PRN assay than the MeV N MBA. This lower correlation observed for the MeV N MBA may be because the N antigen, while being the most highly expressed viral protein in MeV-infected cells, is not the target of neutralizing antibodies. The MeV WVA MBAs use purified viral lysate as the antigen, which contains all viral proteins, including the surface glycoproteins, the hemagglutinin (H) and fusion (F) proteins. Since the H and F proteins are the targets of neutralizing antibodies (40), the MeV WVA MBA likely provides a better measurement of seroprotection than the MeV N MBA. However, N antigen-based ELISA and MBAs have provided accurate estimates of population immunity because the levels of neutralizing and nonneutralizing antibodies correlate in most individuals (31, 41, 42).

The GMRLN consists of 704 laboratories providing support for measles surveillance in 191 countries. In addition to case confirmation and virologic surveillance, many GMRLN laboratories also conduct studies of population immunity. For population immunity studies, reference laboratories within GMRLN may wish to replace the IgG ELISA with the multiplex MBA. The expected widespread use of the MBA across the WHO regions will require a consistent and plentiful source of antigen (43–45). The RuV WVA MBA, as previously described (23), already meets this criterion, and this study identified a suitable commercial source for measles antigen. While other similar technologies for evaluating the antibody response to measles and rubella viruses exist, the measles and rubella WVA beads described here allow multiplexing with multiple antigen targets of public health interest and therefore demonstrate an increased flexibility compared to other commercially available methodologies (46). Additionally, the preliminary estimated cost of the multiplexed measles and rubella MBA is equivalent to that of a single antigen ELISA, with marginal additional cost for increased antigen multiplexing, providing a significant cost benefit consideration for large serosurveillance studies.

This study included serum originating from several countries and from a wide age range, with the goal to establish a cutoff for the assay that may be applied broadly across MBA-based serosurveys. ROC analysis demonstrated a high sensitivity and specificity for both the measles and rubella WVA MBAs and resulted in cutoff values similar to the current cutoff values for correlates of protection (20, 21). The measles cutoff value for protection from disease as measured by PRN is >120 mIU/ml; using the second WHO international standard, the MeV WVA MBAs resulted in slightly different cutoff values, ≥137 mIU/ml for the MeV WVA_L MBA and ≥153 mIU/ml for the MeV WVA_C MBA using PRN as a comparator. The closeness of the PRN and MBA cutoff values further demonstrates that similar results may be obtained from assays developed from different whole-virus antigen sources. While the cutoff is specific to each MeV WVA MBA used to measure seroprotection, the range of the calculated cutoffs demonstrates the reduced variability of the MBA compared to the PRN, which is known to vary 2- to 3-fold between runs (47). Additionally, there is a wide interval between the seroprotective cutoff of the MeV WVA MBA and the typical geometric mean titer for a vaccinated population of 500 to 1000 mIU/ml (11, 48), demonstrating that the cutoff value allows adequate separation from the average signal in a vaccinated population regardless of years postvaccination. The cutoff value for the MeV N MBA (9.5 mIU/ml) was far lower than that established by the other WVA-based assays and likely reflected the more limited range of MeV-specific antibodies measured by the MeV N MBA compared with the MeV WVA MBAs. Additionally, the MeV WVA MBA also had a higher accuracy than the MeV N MBA related to the reduced proportion of false-negative results. The RuV WVA MBA also resulted in a cutoff value (9.36 IU/ml) close to the established correlate of protection (10 IU/ml), which is routinely used for antenatal diagnostic testing (22). While the cutoffs for seroprotection in this study were based on correlates of protection determined by PRN for MeV and quantitative ELISA for RuV, the cutoffs used in other serosurveillance studies varied based on the study and type of assay and ranged from 63 to 350 mIU/ml for measles and 4 to 20 IU/ml for rubella (22).

A limitation of the RuV WVA MBA analysis in this study is that the AUC for RuV was limited to 152 pre- and postvaccination samples from infants for which both MBA and ELISA data were available (serum set 4) and may not represent the range of rubella antibody titers in studies covering wider age ranges. However, the RuV MBA results were similar to previously published results for RuV antibody concentrations (23, 39). Further evidence to support that the ROC analysis results for the RuV MBA acceptably estimated assay performance was demonstrated by looking at the difference between the AUC for the MeV WVA_L MBA with serum set 1 (140 serum samples) and serum set 2 (212 serum samples); there was only a slight drop in the calculated AUC for the MeV WVA_L MBA (0.939) of serum set 2 (212 serum samples) compared to serum set 1 (140 serum samples) (0.924) (Fig. 8). Another limitation in evaluating measles and rubella assay specificity results from the scarcity of negative serum samples within populations due to childhood immunization or infection. Sera from two infant vaccination studies were included in this analysis to increase the number of negative samples. When these negative sera are included MeV WVA_C MBA, specificity increases, and when these populations are considered alone to perform ROC analysis (data not shown), the assay specificity increases to 86.5%, with a similar cutoff. However, because the MBA will be used for serosurveillance of children and adults, the reported sensitivity and specificity are not limited to results obtained with infant sera, and the reported cutoff is based on the combined infant and adult data set. The MBAs in this study were performed with a limited number of conjugated bead lots. Further evaluation of the MBA will be necessary to confirm if the cutoff may be applied broadly across bead lots validated to reduce lot-to-lot variability. The MBA likely offers the best approach to compare results between studies, which is highly desirable but currently difficult to perform with ELISA given differences in sample composition, survey method, and cutoff determination (22).

Serosurveys are important tools for estimating population immunity and providing immunization activity guidance (2, 12, 14, 17). The strong performance of the WVA_C MBA for measles and rubella demonstrated its utility in a multiplexed approach to evaluate population immunity and support evaluation of country immunization programs through identification of pockets of susceptible individuals from either missed vaccination opportunities or waning immunity (19, 24, 25, 35). Serological survey data obtained with a well-validated MBA, as described here, may provide valuable evidence to inform national verification committees, identify areas of risk for measles and rubella outbreaks, and guide immunization programs.