Intralaboratory evaluation of luminescence based high-throughput Serum Bactericidal Assay (L-SBA) to determine bactericidal activity of human sera against Shigella

Despite the huge decrease in deaths caused by Shigella worldwide in the last decades, shigellosis is still causing over 200,000 deaths every year. No vaccine is currently available, and the morbidity of disease coupled with the rise of antimicrobial resistance renders the introduction of an effective vaccine extremely urgent. Although a clear immune correlate of protection against shigellosis has not been established yet, the demonstration of bactericidal activity of antibodies induced upon vaccination may provide one means of functionality of antibodies induced on protecting against Shigella. The method of choice to evaluate the complement-mediated functional activity of vaccine-induced antibodies is the Serum Bactericidal Assay (SBA). Here we present the development and intra-laboratory characterisation of a high-throughput luminescence-based SBA (L-SBA) method, based on the detection of ATP as a proxy of surviving bacteria, to evaluate the complement-mediated killing of human sera. We demonstrated the high specificity of the assay against homologous strain without any heterologous aspecificity detected against species-related and not species-related strains. We assessed linearity, repeatability and reproducibility of L-SBA on human sera. This work will guide the bactericidal activity assessment of clinical sera raised against S. sonnei. The method has the potential of being applicable with similar performances to determine bactericidal activity of any non-clinical and clinical sera that rely on complement mediated killing. IMPORTANCE Shigella is an important cause of diarrhoea worldwide and antimicrobial resistance is on rise, thus efforts by several groups are ongoing to produce a safe and effective vaccine against shigellosis. Although a clear immune correlate of protection has not been established, demonstration of bactericidal capacity of sera from patients immunised with Shigella vaccines may provide one means of protecting against shigellosis. We have developed and fully characterised a novel high-throughput L-SBA method for evaluation of functionality of antibodies raised against S. sonnei in human sera. This work will allow the clinical testing of human sera raised against GMMA-based and potentially all vaccines producing antibodies than can work via complement mediated manner.


48
Diarrheal diseases, such as shigelloses and salmonelloses, are the second leading cause of death 49 worldwide, resulting in millions of deaths per year, mostly in developing countries (1). Shigella is 50 major cause of sustained endemic bacterial diarrhoea, especially in low and middle-income countries 51 where accessibility to clean water is restricted. Although the improvement of hygienic conditions in 52 the last decade has dramatically reduced the burden of the disease, Shigella is still responsible for 53 more than 200,000 deaths, with a third of them being in young children (1). On top of deaths in 54 endemic countries, enteric diseases are causing diarrhoea to travellers and militaries in developed 55 countries, further increasing the burden and the economic and social impact of them. Therefore, the 56 huge morbidity and mortality of the disease coupled with the rise of antimicrobial resistance (2) 57 render the introduction of a vaccine a priority for public health. Although several approaches have 58 been tried during the years by several groups worldwide, no vaccines are licensed yet. Among the 59 different approaches used to produce Shigella vaccines, many of the candidate vaccines target the 60 serotype-specific O-Antigen (OAg) part of the lipopolysaccharide (LPS), as OAg has been identified 61 as a key antigen recognized by the immune system after natural infection (3). In fact, although 62 multiple immune mechanisms may provide protection against Shigella and are not yet fully 63 elucidated, it is well established that antibodies directed to OAg can fix complement and kill target 64 bacteria in a serotype-specific manner (3,4). Genus Shigella is composed by four subgroups (S. flexneri, S. sonnei, S. dysenteriae, and S. boydii) and each of them, with the exception of S. sonnei, is 66 composed by different serotypes, for a total of over 50 different serotypes based on the structure of 67 the OAg, with relative prevalence of serotypes changing geographically and over time (5). As LPS 68 antibody production can confer protection from homologous serotypes, a multivalent Shigella 69 vaccine is necessary to induce antibodies to LPS OAg from multiple serotypes in order to confer 70 broad protection. 71 Several approaches are currently in development to deliver the O-antigen to the immune system, 72 including whole cell attenuated bacteria (6), vaccines in which the O-antigen are chemically-(7) or 73 bio-conjugated to carrier proteins (8), synthetic vaccine conjugates (9), and GMMA based vaccines 74 (10). GMMA are outer membrane exosomes from Gram-negative bacteria, genetically modified to 75 induce hyperblebbing and to reduce the reactogenic potential of lipid A (11, 12). GMMA are easy 76 and inexpensive to produce, and highly immunogenic (10,(13)(14)(15)(16). The most advanced GMMA based 77 vaccine, 1790GAHB (10) has been tested in phase I and IIa clinical trials, conducted in European 78 (17) and endemic sites (18), and has been demonstrated to be well tolerated, immunogenic, and able 79 to induce a strong anamnestic response after boosting (19).

80
On top of vaccine immunogenicity, traditionally assessed through measurement of serum antibodies 81 via antigen specific ELISA, also the functionality of antibodies raised needs to be documented.

82
Although no correlate of protection has been yet established for Shigella, different approaches to  The working principle of SBA relies on reconstituting in vitro conditions in which antibodies 89 recognize antigen on the surface of target bacterium and bind to exogenous complement, activate the 90 characterisation L-SBA on human sera classical pathway, thus resulting in bacteriolysis and death of the target organism. The major problem 91 with traditional SBA is that it relies on plating and counting the target bacteria. Therefore, 92 conventional SBA has been often considered time-consuming and labor-intensive for screening large 93 datasets and clinical samples (22). However a lot of efforts have been made in order to increase the 94 analytical throughput of the assay, resulting in the development of both conventional (23) and non-  After bacteria lysis, ATP becomes available to trigger a luciferase-mediated reaction, resulting in a 101 measurable luminescence signal. In L-SBA the level of luminescence detected is proportional to the 102 number of living bacteria present in the assay wells, which is inversely proportional to the level of 103 functional antibodies present in the serum (24). Result of the assay is the IC50, the dilution of sera 104 able to kill half of the bacteria present in the assay, thus representing the SBA titer of the sera. We 105 have already demonstrated the possibility to use the L-SBA to determine the bactericidal activity of 106 sera raised against S. sonnei GMMA in pre-clinical models (14). Here we present the further 107 development of this method, showing its full characterisation using human sera, and in particular 108 sera raised against S. sonnei GMMA based vaccine (1790GAHB) as model. We have characterised 109 the method intralaboratory by assessing its specificity, linearity, and precision.

110
The L-SBA assay described here is a useful tool for measuring functional antibodies elicited not only 111 by GMMA based vaccines, but in general to assess Shigella-specific functional antibodies in vitro, 112 and potentially of all vaccines that induce antibodies capable of complement mediated killing, either 113 from preclinical and clinical sera. bridging. Assay conditions developed in pre-clinical studies resulted to be optimal also when using 128 human sera, without detection of prozone effect when assaying sera finally diluted at 1:30 in the 129 assay ( Fig. 1A). A prozone effect is defined for a curve readout vs dilution (in this case luminescence 130 vs serum dilution) a condition in which for the first points tested (the least diluted) the readout value 131 (luminescence) is higher than readout value obtained with points highly diluted. 132 We did an initial assessment of the homoscedasticity of the data produced at the different sera 133 dilutions, performing 12 independent sera dilution series repeated in 6 different plates. The test for 134 equal variance of the data obtained per each sera dilution (per each plate) confirmed the lack of 135 homoscedasticity of the data (Fig. S1). In the 4-Parameter Logistic (PL) fitting of luminescence 136 versus Log transformed sera dilutions, the sum of squared residuals weighted for the inverse of 137 luminescence^2 was minimized. An improved analysis method to directly obtain SBA titres from 138 raw luminescence data was also implemented. The aim was to minimise any raw data manipulation 139 by operator (i.e. not to normalise for the dilution giving the highest luminescence for each dilution 140 characterisation L-SBA on human sera series, and not to individually select this value and do calculations for each sample within each run), 141 thus reducing at minimum risk of errors. The latter is a critical aspect when testing clinical samples 142 to ensure integrity of the data. To further improve the analysis, we also included in the 4PL fitting the   Specificity. Specificity of the assay is the ability of an analytical procedure to determine solely the 182 concentration of the analyte that it intends to measure. In case of S. sonnei 1790GAHB vaccine, the 183 target antigen is considered the LPS OAg. 184 We initially assessed the homologous specificity by pre-incubating homologous S. sonnei purified 185 LPS at different concentrations with test serum prior to perform the L-SBA. The aim was to 186 determine the lowest concentration of LPS able to inhibit ≥ 70% of the IC50. Homologous  (Table 1).

Limit of detection (LoD) and Limit of quantitation (LoQ). Finally, we determined the LoD and 211
LoQ of the assay, representing respectively the lowest SBA titer than can be detected under the assay 212 conditions, and the lowest SBA titer that can be quantified with a suitable precision. To do so 213 NVGH2863 was pre-diluted in PBS to generate a sample with low but detectable SBA titer. These 214 conditions simulated the worst-case scenario possible for the assay, and thus the one expected to give 215 characterisation L-SBA on human sera the highest variability. Twelve independent serial curves were tested and IC50 calculated as reported 216 in Table 2.

217
Limit of detection (LoD) and limit of quantitation (LoQ) of the assay were calculated accordingly to 218 the ICH guideline Q2(R1) (27), by applying the following formulas:  (Table 3). To overcome these limitations, we have recently developed an high-241 throughput SBA method based on luminescence readout (L-SBA), that has been already extensively  The work performed here will allow the analysis of the clinical samples by SBA, confirming if the 248 antibodies elicited by 1790GAHB are able to kill Shigella (i.e. functionality of antibodies induced). 249 We have successfully optimised the fitting of the data, including in data analysis a point mimicking a 250 sera billion times diluted, allowing to establish conditions on which, without any normalisation, SBA 251 titers can be directly obtained from raw luminescence values. The latter represents a crucial aspect to 252 increase the throughput of the assay, but especially to reduce any potential bias due to manipulation 253 of raw data when testing clinical samples.

254
In this work we have characterised L-SBA on human sera, demonstrating that in the working 255 conditions tested, it is able to detect sera having a SBA titer as little as 45, with virtually no upper 256 limit of detection, and to quantify with precision sera with IC50 of 100. Work is currently ongoing to 257 further reduce LoD and LoQ. The assay showed low variability, in particular the repeatibility 258 corresponded to the intermediate precision (CV% of 6.15%). Neither operator nor day of analysis 259 resulted to be significant on the overall variability. Furthermore, L-SBA resulted to be highly specific 260 for the key active ingredient of the vaccine candidate, as by depleting the serum with as little as 0.1 261 µg/mL of homologous LPS, over 95% reduction of IC50 was observed, whereas no depletion was 262 observed when depleting sera with S. flexneri 2a, S. flexneri 3a and Salmonella Typhimurium OAg, 263 with only a marginal SBA titer depletion (28%) observed after incubation with S. flexneri 1b OAg.

264
Linearity of the assay was also assessed resulting to be good within the tested range, with a slight 265 characterisation L-SBA on human sera deflection with more diluted samples, as reported for similar assays (23). In line with that, good 266 fitting of the data was obtained with a second order exponential model (Fig.S4). 267 Using the SBA configuration described here, up to 132 specimens can be tested per day by a single 268 operator, including also a standard serum to validate each plate (12 plates can be assayed per day per 269 operator). The analytical throughput of the described Shigella SBA is superior to that of another 270 high-throughput assay recently described by Nham et al. (23), not only in terms of number of 271 samples that can be assayed by one operator in one day, but also in terms of not having to rely on 272 overnight incubation of plates to enable the bacteria to grow and become colonies ( Table 3). As our 273 assay uses standard reagents and requires only a luminometer to detect ATP, L-SBA can be 274 considered simple enough to be adopted by laboratories around the world. Although inter-laboratory 275 variability was not evaluated in our study, in case this would be observed, results could be 276 normalised with the use of reference serum (23).

277
In conclusion, L-SBA applied to human sera represents an assay fully suitable to perform clinical 278 analysis in high-throughput. Due to its specificity and versatility L-SBA can be applied to determine   The human serum tested was an anti-human S. sonnei IgG standard serum (NVGH2863) that was 306 created by pooling sera from adult subjects immunised with 1790 GAHB in non-endemic European 307 populations (17). NVGH2863 has been already used as standard serum for S. sonnei LPS IgG 308 assessment by ELISA (17-19). Frozen 50 µL working aliquots of the serum were stored at -80°C 309 until use. In setup experiments a standard serum obtained from mice immunised with 1790GAHB 310 (24) was also included.

311
All samples tested in SBA were previously Heat Inactivated (HI) at 56 °C for 30 min to remove 312 endogenous complement activity.

313
Various aliquots of HI NVGH2863 serum have been used and treated as described below to 314 determine different assay parameters. 315 characterisation L-SBA on human sera Samples to assess repeatibility and intermediate precision: each sample consists on the same 316 HI NVGH2863 serum; 12 identical samples were assayed each day by two operators and the assay 317 was repeated in three different days by each of the two operators independently (72 samples in total, 318 36 per operator, 12 on each day).

319
Samples to assess limit of detection and limit of quantitation: HI NVGH2863 was diluted 20 320 times v:v in PBS to generate a sample with low but detectable SBA titer (expected IC50 to be around 321 100). Twelve identical NVGH2863 prediluted serum samples were assayed on the same day by the 322 same operator.

323
Samples to assess linearity: HI NVGH2863 serum was assayed neat or diluted 2, 4, 8, 16, 32-324 fold (v:v) with PBS prior performing the assay; samples were prepared independently by two 325 operators on the same day, with each sample assayed twice by the same operator on the same day (4 326 IC50 obtained for each dilution, 2 IC50 by each operator).

327
Samples to assess specificity: two sets of samples were prepared to assess homologous and 328 heterologous specificity of the assay using HI NVGH2863 serum diluted 1:1 (v:v) in PBS alone or 329 PBS supplemented with different quantity of homologous or heterologous purified polysaccharides.

330
In the first experiment HI NVGH2863 serum was spiked with homologous (S. sonnei) purified LPS 331 at different final concentrations (50, 20, 5, 1, 0.1 µg/mL respectively) and compared with sample 332 spiked 1:1 with PBS alone, incubated overnight (16-18 hours) at 4°C shaking at 200 rpm in an orbital 333 shaker prior being tested. Each spiked sample was assayed in duplicate by the same operator on the 334 same day. The lowest concentration of LPS between the ones tested able to inhibit >70% the IC50 335 was then used in a second experiment to determine the heterologous specificity. In the second 336 experiment HI NVGH2863 serum diluted 1:1 (v:v) in PBS supplemented with S. flexneri 1b, S.         Figure S1. Test for equal variances for homoscedasticity. Variances (x axis) from 12 replicates 545 versus Log sera dilutions (y axis) were plotted for each of 6 different plates tested.