We collected 12 S. flexneri clinical isolates from various locations between 2017 and 2023. Two international, MSM-related clusters are related to two EpiPulse alerts ( Table 1 ). For all strains, we tested antibiotic resistance against ampicillin, ceftriaxone, ciprofloxacin, gentamicin, azithromycin, trimethoprim–sulfamethoxazole, meropenem, tetracycline, and chloramphenicol. We found all strains to be partially or fully resistant to at least one antibiotic, with ampicillin resistance being the most common, as reported in the literature.

OAg was extracted from the 12 S. flexneri strains and characterized in terms of polysaccharide identity, chain length, quantity, and percentage/position of O-acetyl groups ( Table 2 , Figure 1 ). By growing all strains under similar conditions, the amount of OAg produced ranged from 242 µg of OAg per optical density (OD) for the S. flexneri 2a strain S23BD05840 (S. flexneri 2a) to 2,036 µg of OAg per OD for the S. flexneri 2a isolate S23BD08049 (S. flexneri 2a). Nuclear magnetic resonance (NMR) spectroscopy was used to confirm the OAg structure of all known serotypes and to identify one of the unknown isolates, comparing the anomeric region of the spectra with previously published data ( Supplementary Figures 1 - 4 ) (33). Unknown serotypes were identified as follows: one S. flexneri Y (19096), two S. flexneri 1b [210059 and 13B(76)], and two S. flexneri 3b [I03(75) and 6A(18)]. All isolates except S. flexneri 6 S23BD10088 displayed two main OAg populations with different molecular size distributions: a high-molecular-weight (HMW) population at 48.6–60.5 kDa and a medium-molecular-weight (MMW) population at 13.5–16.1 kDa. Interestingly, the S. flexneri 6 strain exhibited MMW OAg only, with no peak related to the presence of the capsular polysaccharide detected (34). All isolates, with the exception of S. flexneri 4a S23BD10089 (as expected), were O-acetylated, and the details on % and positions of O-acetyl groups are reported in Table 2 ; Supplementary Figures 5 - 8 . Interestingly, all S. flexneri 3b OAgs presented a low percentage of an additional O-acetyl group with respect to the O-acetyl on Rha^I.

Isolates expressing the same OAg [S. flexneri 1b 210059, 13B(76), and S23BD02869; S. flexneri 2a S23BD04932, S23BD05840, and S23BD08049; S. flexneri 3b I03(75), 6A(18), and S23BD06925] showed similar structural features, despite producing different amounts of OAg ( Table 2 , Figure 1 ).

To verify possible structural differences, the S. flexneri 1b and 2a OAgs extracted from the circulating isolates were compared with the homologous OAg displayed on S. flexneri GMMA in the altSonflex1-2–3 vaccine. Interestingly, few differences were observed in terms of both size and O-acetylation levels. In particular, the HMW population was not found in the S. flexneri 1b OAg displayed by GMMA, which showed the MMW peak only, while both HMW and MMW OAgs were present in S. flexneri 2a OAg-GMMA ( Table 2 ). Moreover, even if the total O-acetylation percentage was similar, all the S. flexneri 1b OAgs from the clinical isolates were primarily O-acetylated at position 2 of the Rha^I, while the same OAg on GMMA presented O-acetylation also in position 3 of Rha^III, with a similar ratio between the two positions. All the S. flexneri 2a OAgs were instead O-acetylated in the same positions with similar ratios, but the overall % of O-acetylation (OAc) was different, with the OAg on GMMA being more O-acetylated (147%) compared to clinical strains (37%–49%) ( Table 2 ; Supplementary Figures 5 - 8 ).

To prove the potential effectiveness of altSonflex1-2–3 against real-world strains, we tested the bactericidal activity of immune sera derived from three altSonflex1-2-3-vaccinated preclinical models (mice, rabbits, and rats) via luminescence-based serum bactericidal activity (L-SBA) assay ( Figure 2 ). As a comparison, a pool of sera from non-vaccinated Kenyan adults, who were considered protected from shigellosis upon repeated exposures throughout their lives, were also tested in L-SBA against the panel of isolates, as a comparison.

All sera were bactericidal against all tested S. flexneri 1b strains, confirming vaccine activity against the homologous strain. When tested against S. flexneri 2a strains, all sera were bactericidal against all strains except S23BD05840. However, sera from Kenyan adults were also unable to kill this strain, confirming its resistance. As a possible explanation for this observation, strain S23BD05840 was found to be particularly resistant to complement intrinsic toxicity. Also, OAg characterization showed that this strain produced less OAg compared to the other strains. Since LPS is usually considered to protect bacteria from complement deposition, we may hypothesize that the resistance of this strain may be due to other virulence factors.

Figure 2 also reports bactericidal activity against heterologous strains not included in the vaccine. Importantly, all sera demonstrated bactericidal activity against S. flexneri 3b, 4a, and Y strains. Activity against strain 3b can be explained by a very close structural similarity to serotype 3a present in the vaccine. Indeed, the two OAgs only differ in the presence of an additional Glc on Rha^III in the RU of S. flexneri 3a, compared to S. flexneri 3b. Bactericidal activity is expected against S. flexneri Y, whose OAg comprises a tetrasaccharide shared by all S. flexneri serotypes, with minor O-acetylation. Bactericidal activity against S. flexneri 4a was also observed, albeit with higher variability compared to that of the other strains. Since this was not observed with 4a laboratory strains (data not shown), we hypothesize this could be related to the extent of growth variability of this specific clinical isolate under the conditions used for SBA. All preclinical sera from vaccinated animals displayed lower or no bactericidal activity against the S. flexneri 6 isolate, compared with the sera from Kenyan adults.

For all strains tested, except the S. flexneri 6 strain, titers of preclinical sera were similar to those elicited from human sera. In conclusion, our results suggest that altSonflex1-2–3 could offer protection against a panel of homologous and heterologous antimicrobial-resistant Shigella strains, with a strong potential impact on global health. altSonflex1-2–3 is being tested in Phase 2 clinical trials (NCT06663436 and NCT05073003), and the functionality of sera from vaccinated infants will be tested against the panel of drug-resistant clinical isolates as well to confirm preclinical data. Testing in the Phase 3 clinical trial will confirm the ability of the vaccine to protect in the field.

GlaxoSmithKline (GSK) is committed to the Replacement, Reduction, and Refinement (3Rs) of animal studies. Non-animal models and alternative technologies are part of our strategy and are employed where possible. When animals are required, the application of robust study design principles and peer review minimizes animal use, reduces harm, and improves benefits in studies. Animal studies were performed in the GSK Animal Facility (Siena, Italy) or at Charles River Laboratories (France), in compliance with the relevant guidelines (Italian D.Lgs. n. 26/14 and European directive 2010/63/UE) and the institutional policies of GSK. The animal protocol approved by the Italian Ministry of Health was AWB project No. 643/2021-PR, with approval date 12/08/2021.

Mouse serum was pooled from 10 female CD1 mice vaccinated intramuscularly (IM) with 50 µL of altSonflex1-2–3 at a dose of 6 μg of total OAg (1.5 μg per serotype) with Alhydrogel. Rabbit serum was pooled from eight female New Zealand White rabbits vaccinated IM with 500 µL of altSonflex1-2–3 at a dose of 60 µg of total OAg (15 μg per serotype) with Alhydrogel. Rat serum was pooled from eight Sprague–Dawley rats vaccinated IM with 200 µL of altSonflex1-2–3 at a dose of 24 µg of OAg (6 µg per serotype) with Alhydrogel. All animals were immunized at study days 0 and 28 and were bled on day 42. All sera were stored frozen in 50 µL working aliquots at −80°C until use. All samples tested in SBA were previously heat-inactivated at 56°C for 30 min to remove endogenous complement activity.

The human serum tested was a pool of preimmune sera of 20 Kenyan adults enrolled in the safety cohort of H06_01TP Stage 2 (NCT05073003). All sera were stored frozen in 50 µL working aliquots at −80°C until use. All samples tested in SBA were previously heat-inactivated at 56°C for 30 min to remove endogenous complement activity.

Frozen 20% glycerol stocks were prepared from all S. flexneri strains and stored at −80°C until use. To perform experiments, isolates were inoculated from the glycerol stock into a 5-mL Luria-Bertani (LB) liquid culture in preparation for either large-scale growth for OAg extraction or small-scale growth for SBA.

A 5-mL LB over-day inoculum of each strain was used to seed a 250-mL LB overnight culture. OD_600nm ranged from 1.5 for strain 13B(76) to 8 for strain S23BD05840. OAg was extracted from the bacterial pellet by hydrolysis with 2% acetic acid at 100°C for 4 h and filter-sterilized. The extracted OAgs were purified as previously reported, with slight modifications (35). The samples were subjected to buffer exchange and removal of lower-molecular-mass impurities by ultrafiltration using Amicon Ultra Centrifugal Filter 10 kDa cut-off (Merck, Darmstadt, Germany) against water (3,500 rcf, 10 min, 2 cycles). Next, 25 mM sodium acetate buffer at pH 3.7 was added, reaching a final acetate concentration of 2.5 mM, to precipitate possible protein impurities. After mixing at room temperature (RT) for 30 min, the supernatants were collected through centrifugation and filtered using cation exchange (CEX) chromatography with Sartobind S MA75 filters (Sartorius, Göttingen, Germany) to further remove residual protein impurities. After pH neutralization, the OAg was further purified via anion exchange (AEX) chromatography using a HiTrap Q FF column (Cytiva, Chicago, IL, USA) to remove nucleic acid impurities. All the OAgs eluted at 25 mM NaCl, except the negatively charged S. flexneri 6, which eluted at 200 mM NaCl. The resulting solutions were used to calculate the OAg molecular weight using dextrans as standards, as previously reported (36). A final purification step using Amicon Ultra Centrifugal Filter 10 kDa cut-off against water (3,500 rcf, 10 min, 3 cycles) was performed to remove the lipid A core.

OAgs were extracted from GMMA as previously reported (37) and directly used for characterization.

The OAgs were characterized in terms of O-acetyl ester content, performing a modified microplate procedure of the Hestrin assay (38), and the OAg sugar composition was estimated by High-Performance Anion-Exchange Chromatography with Pulsed Amperometric Detection (HPAEC-PAD) analysis (36). OAg structures were further confirmed by ¹H-NMR analysis and measured using a Bruker Avance III 400 spectrometer at 400 MHz and 298 K ¹H-NMR was also used to determine the position of O-acetyl groups.

OAg amounts were quantified by HPAEC-PAD (36) after the first ultrafiltration step (S23BD04932, S23BD05840, S23BD08049, S23BD02869, S23BD06925, S23BD10088, and S23BD10089 OAgs) or after AEX step [19096, 210059, I03(75), 6A(18), and 13B(76) OAg].

The serum bactericidal assay based on luminescent readout (L-SBA) was conducted as previously described (3, 39, 40). Briefly, sera were serially diluted threefold in Phosphate Buffered Saline (PBS) starting from a 1/50 dilution (25 µL/well). Negative controls without serum were included to monitor non-specific complement killing. Meanwhile, bacteria were cultured to log phase (OD_600nm = 0.20 ± 0.02), diluted in the appropriate buffer to 1 × 10⁶ CFU/mL, mixed with baby rabbit complement (BRC) (Cedarlane, Southern Ontario, Canada, CL 3441-S) ( Supplementary Table 1 ) to a final volume of 100 μL/well, and incubated at 37°C for 3 h. Bacteria were pelleted and resuspended in 100 μL PBS to remove complement-lysed bacteria. Finally, 25 μL of resuspended live bacteria was mixed with 25 μL of BacTiter-Glo (Promega, Madison, Wisconsin G8231) in a white 384-well plate (Greiner Bio-One, Monroe, North Carolina 655904). The reagents lyse live bacteria, which release ATP and enable bacterial quantification. After 5-min incubation on an orbital shaker at 300 rpm in the dark and RT, ATP-induced luminescence was acquired using a luminometer (Synergy HT, Biotek, Swindon, UK), which correlated with the number of living bacteria in the wells.

Relative luminescence values were plotted in Prism (GraphPad, version 9.5.3) and fitted on the logarithmic value of the serum dilutions tested for each serum in a four-parameter logistic regression model; an arbitrary serum concentration of 10⁻¹⁵ was assigned to the well containing no serum (39). To allow correct fitting, the curves were constrained to have a bottom between 0 and the mean of three counts per second corresponding to the lower luminescence detected at T180 in wells where bacteria were killed. The results of the assay are expressed as IC50, representing the dilution of serum able to mediate 50% of killing efficiency with respect to the maximal value. Three independent experiments were performed for each serum on each strain.

The optimal BRC concentration for each strain was established by performing a comprehensive experiment that assessed bacterial growth in a range of increasing BRC percentages in the absence of serum. The selected concentration was determined as the one that most closely resembled the conditions without BRC, wherein the positive impact on bacterial growth was counterbalanced by its inherent toxicity ( Supplementary Table 1 ).