Differences in antigenic structure of inactivated poliovaccines made from Sabin live-attenuated and wild-type poliovirus strains: impact on vaccine potency assays.

BACKGROUND
Following the declaration of wild poliovirus type 2 eradication in 2015, the type 2 component was removed from the live-attenuated oral polio vaccine (OPV). This change implies a need to improve global coverage of routine immunization with inactivated polio vaccine (IPV) to ensure type 2 immunity. Several manufacturers use Sabin-OPV strains for IPV production (sIPV) rather than the usual wild-type strains used for conventional IPV (cIPV). However, contrarily to cIPV, potency assays for sIPV have not been standardized, no international references exist and no antigen units have been defined for a sIPV human dose. Thus, sIPV products from different manufacturers cannot be compared and the relationship between antigenicity and immunogenicity of sIPV is not well understood.


METHODS
A collaborative study was conducted in which laboratories used different methods to measure the antigen content of a set of sIPV and cIPV samples with an aim to identify a suitable reference for sIPV products.


RESULTS
The study revealed differences in the reactivity of antibody reagents with cIPV and sIPV products.


CONCLUSIONS
Homologous references are required to measure the antigen content of IPV products consistently. The 1 st WHO International Standard for sIPV was established with new specific Sabin D-Ag Units assigned.

The Global Commission for the Certification of Poliomyelitis Eradication concluded on 20 September 2015 that wild-type 2 poliovirus (WPV2) had been eradicated worldwide, with the last case of polio due to WPV2 globally reported on 24 October 1999 in Aligarh, Uttar Pradesh (India). This announcement was a major step toward the completion of the global efforts to eradicate all 3 WPV serotypes: WPV1, WPV2, and WPV3. WPV3 has not been detected globally since November 2012, in Nigeria, and the only remaining endemic WPV1 strains are now limited to Pakistan and Afghanistan. The declaration of WPV2 eradication was followed by the removal of the type 2 component from the live-attenuated oral polio vaccine (OPV) in April 2016 and resulted in a switch from using trivalent OPV (tOPV), which contained all 3 serotypes, to bivalent OPV (bOPV), which only contains serotypes 1 and 3 [1].
On rare occasion, OPV strains can evolve to virulent strains and result in circulating vaccine-derived polioviruses (cVDPVs), which have been associated with a number of poliomyelitis outbreaks around the world [2]. For this reason, the global eradication of polio requires the cessation of all OPV use. The continued occurrence of polio caused by type 2 VDPV was the reason to implement the switch from tOPV to bOPV in routine immunization programs, even before the remaining strains of wild-type poliovirus are eradicated. The switch from tOPV to bOPV requires the global introduction of IPV, a killed-virus vaccine that is not associated with vaccine-related paralysis, in all routine immunization programs, to maintain immunity levels against type 2 poliovirus. The successful eradication of WPV2 and nearly complete eradication of WPV1 and WPV3 means that laboratory containment of all polioviruses will soon be a requirement; containment for type 2 poliovirus was initiated in April 2016, which means that vaccine production and associated laboratory work requiring handling of live type 2 poliovirus materials should now be conducted at an increased biosafety level [3]. To produce IPV more safely, several manufacturers are producing IPV by using Sabin poliovirus strains, rather than the usual wild-type strains that have been used for conventional IPV (cIPV) for >50 years. As demand for IPV increases, as many as 17 new manufacturers are producing or plan to produce Sabin-strain IPV (sIPV) in the near future [4]. Licensed sIPV products are already being produced and used in China and Japan [5,6].
The potency of IPV is measured in vitro by using a validated enzyme-linked immunosorbent assay (ELISA) with suitable reference preparations and is expressed in D-antigen (D-Ag) units [7,8]. Alternative methods based on surface plasmon resonance have been proposed but need further validation [9,10]. The third World Health Organization (WHO) international standard (IS) for cIPV (National Institute for Biological Standards and Control [NIBSC] code 12/104) was established in 2013 [11]. This IS is used by manufacturers and control laboratories to calibrate laboratory reference reagents. A European Pharmacopoeia biological reference preparation is also commonly used as a reference standard in in vitro potency assays [12]. However, in contrast to cIPV, no reference standards have been established for sIPV, and no requirements in terms of specific D-Ag units required for a human dose have been defined. As most manufacturers use product-specific references, which in most cases have not been calibrated using international references, the equivalence between D-Ag units defined by different manufacturers is difficult to assess. This makes comparison between different sIPV products very difficult, and therefore establishment of batch-release assays suitable for different sIPV products would not be straightforward. For this reason, a collaborative study was conducted to compare different methods for measuring the antigen content of sIPV products, assess the suitability of the cIPV IS 12/104 for sIPV antigen assays, and validate candidate sIPV samples to serve as a WHO IS for sIPV if required. Laboratories were asked to use both their in-house method and a common method with reagents and protocol supplied by the NIBSC.

Preparation of Materials for Collaborative Study
IPV samples for the study were kindly provided by manufacturers. Nine samples were sent to participant laboratories for testing. They included the 2 sIPV candidate standards prepared at the NIBSC, the current IS for cIPV (12/104), another cIPV reference standard (08-143), and 3 sIPV preparations produced by different manufacturers. Full details of the study samples are available in the Supplementary Materials.

Study Design
The primary aim of the study was to characterize the D-Ag content of 2 potential candidate sIPV standards and decide on a recommendation for the establishment of the first IS for sIPV. Fourteen laboratories were invited to participate. Thirteen accepted, of which 7 were manufacturer laboratories and 6 were national control laboratories (Supplementary Materials). Twelve participating laboratories returned data within the requested time frame, and 1 sent data after the final deadline for submission. Laboratories are referred to by a code number that was allocated at random and does not reflect the order of listing. Participants were requested to (1) determine the D-Ag content of the panel of 7 coded trivalent sIPV samples (17001-17007), using the routine in-house ELISA method (Supplementary  Table 1) and a common method with reagents and a protocol provided by the NIBSC (Supplementary Materials); (2) perform 3 independent assays to determine the D-Ag content of the 3 poliovirus serotypes for each study sample against that of cIPV IS 12/104; and (3) test all study samples at the same time for each of the 3 independent determinations, using freshly opened samples for the preparation of dilutions used on any day.

Statistical Analysis
All assay data were analyzed at the NIBSC, using the approach described in the Supplementary Materials. Study samples used as reference standards for the analysis were the current cIPV IS (ie, 12/104), candidate standard 17/130 (coded 17002), or candidate standard 17/160 (coded 17004).

Stability Studies
Samples of the candidate standards were stored at −70°C, −20°C, and 4°C for up to 12 months. Additional samples were placed at 20°C and 37°C for short periods of up to 6 weeks. They were assayed concurrently on 3 separate occasions, using 12/104 as a reference. The potency of the samples was expressed relative to that of the −70°C baseline samples. An experiment to test the potency loss during short-term exposure to increasing temperatures (ie, 15 minutes at 4°C-60°C) was conducted to obtain another indicator of temperature stability.

Validity Criteria for Parallelism of Dose-response Relationships
Using data from all assays performed at laboratories 1-10 that had submitted data when this part of the analysis was completed, slope ratios observed for the coded duplicate samples relative to each other (17002 relative to 17005 for candidate 17/130 and 17004 relative to 17007 for candidate 17/160) were used to define acceptance criteria for parallelism, separately for each poliovirus type. Defining m as max[s, 1/s], where s denotes the slope ratio of the coded duplicates within an assay, nonparametric upper tolerance bounds (with 95% coverage and 95% confidence) for m were determined. The bound and its reciprocal value were used to define acceptable slope ratio ranges for concluding parallelism, giving 0.66-1.51, 0.70-1.43, and 0.65-1.54 for poliovirus types 1, 2 and 3 respectively. These ranges were intended for use in the analysis of data from this study only, to apply consistent criteria to all laboratories and assess the relative performance of the different standards.

Assay Validity
A small number of samples were excluded (around 3%) because they demonstrated nonlinearity in their observed dose-response relationship. No cases of nonparallelism using the criteria described above were observed for 08-143 relative to 12/104. A summary of exclusions due to nonparallelism for the sIPV study samples (coded 17001-17007) is shown in Supplementary Table 2. No consistent improvement in parallelism was observed when using candidate standards instead of 12/104, but this was noted in some cases (eg, for all type 2 poliovirus assays). There were generally fewer exclusions when laboratories used their in-house method than when they used the common method. In addition, there was some evidence that parallelism was better when using 17/160 as a reference versus 17/130 for all methods.

Intralaboratory Variability
Some laboratories did not test all samples concurrently on the same plate, so it was not possible to obtain potency estimates for some vaccines relative to the candidates.
The intralaboratory variability (ie, the between-assay, within-laboratory variability) is shown in Supplementary Table 3 as the average geometric coefficient of variation (GCV) across sIPV study samples for each laboratory and reference. The majority of laboratories (91.8%) achieved between-assay variability of <20%, and 67% achieved a between-assay variability of <10%. There were very few high values, all of which might have been anomalous. Results when using candidate standard 17/160 as a reference were marginally better in this aspect.
Potencies calculated relative to 12/104, 17/130 (coded 17002), or 17/160 (coded 17004) are shown in Supplementary  Figures 1-15. Overall study geometric mean potencies and between-laboratory GCVs are shown in Supplementary Table 4, and side-by-side comparisons of the range of potencies of study samples relative to 12/104 or 17004 obtained in the different laboratories are shown in Figures 1-3. There were some extreme differences between laboratory results when using in-house assays and 12/104 as a reference, particularly with results from laboratories 5 and 12 (which, in several cases, were >10-fold higher than the median result for all other laboratories) but also for a few results from laboratories 7 and 8, although these differences were less extreme ( Supplementary Figures 1-15). For this reason, results from laboratories 5 and 12 have been excluded from the calculations shown in Supplementary Table  4. Remarkably, these differences disappeared in most cases when potencies were estimated using either of the 2 sIPV candidates as a reference ( Supplementary Figures 1-15). No such extreme differences were observed in any laboratory when measuring the potency of sample cIPV 08-143 by using 12/104 as a reference in both in-house assays or common assays. The assays with the common method for D-Ag determination of sIPV study samples from all laboratories were included in the calculations because such large differences between laboratories were not observed. Coded duplicate geometric mean relative potencies of 17/130 (range, 0.935-1.040; Supplementary Table 4) showed very good agreement with their expected value of 1 for all poliovirus types and assay types and low variability (maximum GCV, 21.1%). Values for 17/160 showed even closer agreement (relative potency range, 0.961-1.018) and lower variability (maximum GCV, 11.0%).
A summary of between-laboratory variability based on median GCVs is shown in Supplementary Table 5. Owing to the potential for anomalous results to increase GCV estimates, median absolute deviation values are also shown as  Table 5 and Figure  4). Both measures of variability illustrate that improved interlaboratory agreement for in-house assays is achieved when using either candidate standard, compared with using cIPV IS 12/104 as standard. No such improvement was evident for the common assay, for which all references produced similar results. Agreement between laboratories was even better when the reference and the study sample analyzed were made by the same manufacturer, such as using 17002 (17/ There were some exceptions in which laboratory potency measurements using in-house assays and a sIPV reference differed by more than twice the mean potency estimates from all laboratories. These included the type 1 analysis of sample 17001 in laboratory 6, the type 2 analysis of sample 17006 in laboratory 2, the type 2 analysis of 17003 in laboratory 5, and the type 3 analysis of sample 17003 in laboratory 8. In the first 2 cases, potency estimates were not based on a full data set, because some potency determinations were excluded owing to their failure of validity criteria. Although this number of exceptions is very small and not generalized in any particular laboratory, further standardization of assays might be needed in some of these laboratories.
In summary, there was good agreement between laboratories in D-Ag potency measurements for sIPV study samples when using a homologous sIPV reference, with few exceptions, but neither of the 2 candidates appeared to offer an advantage over the other, based on these data.

Potency Estimates for Candidate Standards, Using Conventional D-Ag Units
A summary of overall potency estimates for 08-143, 17/130, and 07/160 calculated relative to the potency of cIPV IS 12/104 is given in Table 1. As specified above, results of in-house assays from laboratories 5 and 12 have been excluded from the calculations in this table. To mitigate the effects of any other outliers or anomalous results, overall estimates are also shown as robust geometric means and geometric medians. Using robust geometric mean values, potencies of 17/130 were 50, 608, and 450 conventional D-Ag/mL for poliovirus 1, 2, and 3, respectively; for candidate 17/160, potencies were 239, 136, and 237 D-Ag/mL for poliovirus 1, 2, and 3, respectively.

Stability Studies
D-Ag potencies of samples from the candidate standards maintained at −70°C, −20°C, and 4°C are shown in Supplementary Table 6. Results of the analysis of real-time stability at the temperature intended for storage (−70°C) for both sIPV candidates are shown in Figure 5. The results show that potency was maintained at −70°C for candidates 17/130 and 17/160 for 12 and 9 months, respectively. There was a drop in potency of all samples stored at −20°C, with a complete loss in potency of candidate 17/160 (Supplementary  Table 6). There was a loss in the potency of candidate 17/130 after long-term storage at 4°C but not in the potency of candidate 17/160, which remained stable over the 9-month analysis period. Loss of potency following long-term storage at 4°C has been reported for similar concentrated bulk sIPV samples [13]. Samples stored at higher temperatures showed a drop in potency to some degree, particularly at 37°C (Supplementary  Table 7). However, both candidates were stable at 20°C for at least 3 weeks, which is compatible with manipulations at room temperature during the ELISA. An experiment to test loss of potency during short-term exposure to increasing temperatures, which is an indicator of stability, revealed that 17/160 retained its potency at higher temperatures, compared with 17/130, and had a similar profile to cIPV IS 12/104 (Supplementary Fig. 16). Overall, the results indicate that the candidate materials are stable but should be stored at −70°C and carefully handled prior to analysis. A program to measure real-time stability is ongoing by regularly measuring the potency of samples stored at −70°C and 4°C.

DISCUSSION
The antigen potency of a set of 7 sIPV samples was determined using different methods available in as many laboratories. The primary aim of this study was to compare the outcome of different potency assessments, using different antibody reagents and reference standards, and to characterize the D-Ag content of 2 potential candidate sIPVs, to decide on a recommendation for the establishment of the first IS for sIPV. There is currently no WHO IS for sIPV, so manufacturers and national control laboratories must rely on in-house reference materials that have been established independently, and therefore comparison of sIPV products from different manufacturers is not straightforward. The 2 candidate sIPV samples were tested alongside 3 other sIPV products and 2 cIPV reference samples in 13 laboratories from 8 different countries and 3 continents (Europe, Asia, and North America). In-house methods represented a wide variety of assay formats in terms of the antibody reagents used for vaccine capture and detection, the number of replicates and dilutions analyzed per sample, the substrate used to measure the immunochemical reaction, and the statistical analysis used to determine the D-Ag potency of study samples. Good-quality data were received from most participants. The study showed that using cIPV IS 12/104 as a reference to determine the D-Ag content of sIPV products is not the best option because high between-laboratory variability in potency results for sIPV study samples was found when using in-house methods and 12/104 as a reference. This was, however, mostly attributable to few laboratories with potencies very different from the mean values. Agreement between laboratories was better when using 12/104 and a common method across laboratories, which is not surprising. On the contrary, low between-laboratory variability in D-Ag determinations was observed with both in-house and common methods, using 12/104 as a reference to measure the potency of reference sample 08-143, which contains homologous poliovirus strains. This suggests some degree of specificity in the reactivity of the study samples with wild-type and Sabin strain antibody reagents used by the different testing laboratories, as has been shown before [10,[14][15][16][17], and is possibly due to differences in antigen-antibody specificity/avidity between wild-type and live-attenuated poliovirus strains. This was more obvious in a few laboratories, in which large differences in D-Ag potencies of most study samples were obtained when using different reference standards (in-house vs common assays).
Of importance and in agreement with the above observations, there was clear improvement in the agreement between laboratories for D-Ag measurements of sIPV products when using in-house methods in combination with either of the 2 candidate homologous sIPV standards as a reference in the D-Ag ELISA test as compared to those estimated using cIPV IS 12/104. There were also some minor differences in the degree of between-laboratory agreement of measured D-Ag potencies of specific sIPV products when using either of the 2 sIPV candidate references. Interestingly, agreement between laboratories was much better when the reference and the study sample analyzed were made by the same manufacturer. This further indicates a degree of specificity in antigen-antibody reactivity in D-Ag ELISAs, even when using the same poliovirus strain for vaccine production. Possible reasons for this might be that (1) stocks of Sabin viruses used by different manufacturers may contain mutations affecting their antigenic properties, or (2) minor differences in the antigenic structure of some epitopes occur because of differences in the relative concentrations of protein and formaldehyde present in different vaccine products. This phenomenon has also been observed for cIPV products. Very similar results were observed during the collaborative study to establish the third IS for cIPV 12/104, with very low between-laboratory variability observed when measuring candidate samples by using a reference made from the same manufacturer, compared with higher variability when using references made by different manufacturers [11]. It is when such observed differences are major that product-specific references might be necessary, which does not seem to be the case for either cIPV or sIPV products, as shown in this and previous studies concerning cIPV reference standards. Establishing a process for method validation could help further harmonizing D-Ag ELISAs between laboratories, such validating antibody reagents with similar reactivity to different sIPV and cIPV products. However, correlation between in vitro and in vivo potency (and, ideally, a link to immunogenicity potency in human clinical studies) [18,19] should be assessed for different sIPV products, as has been established for cIPV, to demonstrate the equivalence of D-Ag units present in different sIPV products.
In summary, both within-and between-laboratory variability were generally low when using either of the 2 sIPV candidates to determine the D-Ag content of sIPV products. The overall geometric mean potency estimates for the duplicate samples of both candidate standards showed excellent agreement between laboratories and very good consistency between duplicates. There was good agreement between laboratories whichever candidate was chosen as a standard to determine the D-Ag content of sIPV study samples. Stability studies demonstrated that both candidate materials were stable at temperatures used for longterm storage (−70°C) and short-term laboratory manipulation (4°C-20°C). However, both samples were found to be unstable at −20°C, particularly 17/160, for which no D-Ag was detected after 9 months of storage at that temperature. Although we have observed this phenomenon occasionally with cIPV references, we do not have an explanation for this finding. Further real-time stability studies are ongoing. Given the results described above, we conclude that both sIPV candidates, 17/130 and 17/160, are suitable to serve as an IS. Candidate 17/160 was chosen because it showed (marginally) better overall results in terms of assay validity, within-and between-laboratory variability, and thermal stability. Consequently, the WHO Expert Committee on Biological Standardization recently approved candidate 17/160 as the first WHO IS for Sabin-IPV on 1 November 2018. Given the inconsistencies between the antigen properties of cIPV and sIPV, requiring homologous references to measure the antigen content of these products, the potency of sIPV products cannot be measured accurately by using cIPV IS 12/104, and the resulting values will depend on assay conditions. For this reason, it was decided that a new antigen unit, the Sabin D-Ag unit (SDU), specific for sIPV and independent of conventional D-Ag units used for cIPV products, should be defined. This would prevent confusion when using this new sIPV IS, or future sIPV references calibrated with it, to assign antigen potencies to sIPV products. Accordingly, the first WHO IS 17/160 was assigned 100 SDU/mL for each of the 3 poliovirus serotypes.