Very long O‐antigen chains of Salmonella Paratyphi A inhibit inflammasome activation and pyroptotic cell death

Abstract Salmonella Paratyphi A (SPtA) remains one of the leading causes of enteric (typhoid) fever. Yet, despite the recent increased rate of isolation from patients in Asia, our understanding of its pathogenesis is incomplete. Here we investigated inflammasome activation in human macrophages infected with SPtA. We found that SPtA induces GSDMD‐mediated pyroptosis via activation of caspase‐1, caspase‐4 and caspase‐8. Although we observed no cell death in the absence of a functional Salmonella pathogenicity island‐1 (SPI‐1) injectisome, HilA‐mediated overexpression of the SPI‐1 regulon enhances pyroptosis. SPtA expresses FepE, an LPS O‐antigen length regulator, which induces the production of very long O‐antigen chains. Using a ΔfepE mutant we established that the very long O‐antigen chains interfere with bacterial interactions with epithelial cells and impair inflammasome‐mediated macrophage cell death. Salmonella Typhimurium (STm) serovar has a lower FepE expression than SPtA, and triggers higher pyroptosis, conversely, increasing FepE expression in STm reduced pyroptosis. These results suggest that differential expression of FepE results in serovar‐specific inflammasome modulation, which mirrors the pro‐ and anti‐inflammatory strategies employed by STm and SPtA, respectively. Our studies point towards distinct mechanisms of virulence of SPtA, whereby it attenuates inflammasome‐mediated detection through the elaboration of very long LPS O‐polysaccharides.

SPtA and STy have independently evolved mechanisms to disseminate systemically and persist in secondary organs (Hiyoshi, Tiffany, Bronner, & Baumler, 2018;Holt et al., 2009). Despite the high degree of genetic relatedness and the plethora of shared pseudogenes, it remains unclear what shared or distinct genomic elements and virulence factors are responsible for enteric fever caused by these serovars (Hiyoshi, Tiffany, et al., 2018;Holt et al., 2009;Johnson, Mylona, & Frankel, 2018;McClelland et al., 2004). For example, the Vi capsule, which is present in STy, but absent from SPtA, has been suggested to mediate immune evasion Wilson et al., 2008;Winter, Winter, Poon, et al., 2014;Winter, Winter, Atluri, et al., 2015). Alternatively, immune evasion by SPtA is thought to be mediated by the production of very long LPS O-antigen chains, a process driven by the polysaccharide copolymerase FepE, which is a pseudogene in STy (Hiyoshi, Wangdi, et al., 2018). Both the Vi capsule and very long O-chains promote the avoidance of the respiratory burst from neutrophils (Hiyoshi, Wangdi, et al., 2018).
Central to Salmonella virulence are two type III secretion systems (T3SS), encoded on Salmonella pathogenicity island 1 (SPI-1) and SPI-2, which secrete effectors that subvert host cell processes (Jennings, Thurston, & Holden, 2017). SPI-1 T3SS mediates invasion into non-phagocytic host cells and is readily recognised by innate immune pathways in macrophages (Srikanth, Mercado-Lubo, Hallstrom, & McCormick, 2011;Wemyss & Pearson, 2019). Following internalisation, Salmonella resides within the Salmonella-containing vacuole (SCV), which is maintained by the activity of the SPI-2 effectors (Jennings et al., 2017). Although the molecular pathogenesis and virulence of SPtA are poorly understood, they are assumed to be similar to those of other Salmonella serovars like STy and the well-studied non-typhoidal Salmonella (NTS) serovar Typhimurium (STm). The latter has frequently been used as a model for typhoidal Salmonella, despite typhoidal and NTS resulting in different disease symptoms in humans, possessing different virulence genes and regulating essential virulence factors in a disparate manner (Gal-Mor, Boyle, & Grassl, 2014;Johnson, Mylona, & Frankel, 2018;McDowell et al., 2019;Sabbagh, Forest, Lepage, Leclerc, & Daigle, 2010).
Macrophages play an important role during infection and systemic dissemination of Salmonella (Dougan & Baker, 2014). They can promote host defence by sensing and responding to infection via inflammasomes, which are multimeric complexes serving as cytosolic caspase-1 activation platforms. Inflammasome activation is initiated upon recognition of bacterial conserved patterns by host proteins such as NLRs (NOD and leucine-rich repeat containing proteins), PYRIN and AIM2. Caspase-1 activation triggers pyroptotic cell death and proteolytic processing and secretion of pro-inflammatory cytokines IL-1β and IL-18, resulting in an inflammatory response (Sanchez-Garrido, Slater, Clements, Shenoy, & Frankel, 2020).
The aim of this study was to determine if and how SPtA activates inflammasomes in human macrophages. We show that the SPI-1 T3SS of SPtA is required for inflammasome-dependent cell death via activation of caspase-1 and caspase-4. Conversely, we found that, by elaborating very long surface O-antigen chains, SPtA dampens pyroptotic cell death. Taken together our data show that SPtA employs a novel stealth infection strategy.

| S. Paratyphi A induces cell death in human macrophages
We infected primary human monocyte-derived macrophages (MDMs) with the reference SPtA strain ATCC 9150 (SPtA 9150) and a clinical isolate originated from a patient with enteric fever in Nepal, SPtA ED199. At 3 h post infection, SPtA ED199 was internalised at significantly higher levels than SPtA 9150 ( Figure 1a). Consistently, compared to SPtA 9150, SPtA ED199 induced two-fold higher cell death, as measured by propidium iodide (PI) uptake (Figure 1b), and about three-fold higher secretion of IL-1β ( Figure 1c).
We next assessed whether these phenotypes are also observed in differentiated human macrophage-like THP1 cells, and furthermore tested four additional Nepalese clinical isolates (Table S1). The four clinical isolates were internalised similarly to SPtA ED199 and trig- To assess the role of caspases in SPtA-induced cell death, we pre-treated THP1 cells with the pan-caspase inhibitor Z-VAD-FMK, which effectively blocked nigericin-induced pyroptosis as a control (- Figure S1A). Z-VAD-FMK treatment abolished cell death during infection with SPtA 9150 and the five clinical isolates (Figure 1h).
To determine if the necrotic cell death seen upon SPtA infection is mediated by inflammatory caspases, we pre-treated THP1 cells with the narrow-spectrum inhibitor Z-YVAD-FMK, which specifically inhibits caspase-1/4. In control experiments Z-YVAD-FMK treatment reduced PI uptake induced by LPS transfection ( Figure S1B). Importantly, Z-YVAD-FMK significantly reduced PI uptake induced by SPtA (Figure 1i), but not to the levels seen with Z-VAD-FMK, pointing towards the involvement of other caspases in SPtA-induced cell death. We therefore additionally pre-treated THP1 cells with the caspase-8 inhibitor Z-IETD-FMK, which reduced staurosporineinduced toxicity in control experiments ( Figure S1C). Caspase-8 inhibition also reduced cell death triggered by all the SPtA strains tested ( Figure S1D). Collectively, these results implicate caspase-1, caspase-4 and caspase-8 in SPtA-induced cell death.
To this end, we knocked down GSDMD expression in THP1 cells using a miRNA30E-based plasmid (GSDMD miR ) (Figure 2d), which was functionally validated by the reduction in pyroptosis after nigericintriggered canonical NLRP3 inflammasome activation ( Figure S2E).
We first investigated the activity and expression of the SPI-1 T3SS in SPtA. As SPtA is internalised by macrophages through phagocytosis As the STm flagellin and T3SS needle and rod can activate the NAIP/NLRC4 inflammasomes (Kortmann et al., 2015;Reyes Ruiz et al., 2017;Yang et al., 2013), we asked whether they play a role in mediating SPtA 9150-induced pyroptosis. We generated NAIP miR THP1 cells, which showed an effective reduction of NAIP protein levels ( Figure S3C). Surprisingly, neither NAIP silencing nor further inhibiting NLRP3 with MCC950 treatment affected caspase-1 activation ( Figure

| Very long O-antigen chains interfere with inflammasome activation
The T3SSs in Shigella and enterohemorrhagic E. coli (EHEC) have been shown to be masked and inhibited by the LPS O-antigen and group 4 capsule, respectively (Caboni et al., 2015;Shifrin et al., 2008;Watson et al., 2019b;West et al., 2005). As FepE mediates the production of very long LPS O-antigen chains in SPtA, we hypothesised that these may interfere with the function of the T3SS and pyroptosis.
To test this, we generated a SPtA 9150 ΔfepE mutant, which reduced the production of very long polymers of O-antigen polysaccharides, without affecting bacterial growth in standard laboratory growth media ( Figure S4A Mean ± SEM from 5 (a, c), 6 (d), or 3 (f, g) independent experiments. * p < .05, ** p < .01, *** p < .001 for the indicated comparisons by Student's t-test (c, d), oneway (a) or two-way ANOVA (f, g) following correction for multiple comparisons; ns, not significant escape later during infection is dependent on the SPI-1 T3SS but is not affected by the very long O-antigen chains ( Figure S5B). These results suggest that basal T3SS activity of WT SPtA is sufficient to mediate rupture of the SCV after bacterial phagocytosis by macrophages.
To further verify the dominant role of caspase-4 in detecting cytosolic ΔfepE bacteria, we quantified ASC assembly into inflammasome specks by using a THP1 cell line expressing RFPtagged ASC (ASC mRFP ) (Figure 4h). WT and ΔfepE bacteria induced ASC-speck formation to comparable levels (Figure 4i and Figure S5C).
We therefore concluded that increased caspase-4 activation by SPtA ΔfepE results in elevated pyroptosis directly, rather than through downstream activation of NLRP3-ASC-caspase-1 inflammasomes, while additional ASC-dependent pathways are also expected to be involved (Figure 4j). Taken together, these results suggest that the very long LPS O-chains limit inflammasome activation and pyroptosis after bacterial escape from vacuoles, which represents a unique mechanism of suppressing caspase-4 activation.

| FepE expression inversely correlates with Salmonella-induced pyroptosis
Although fepE is a pseudogene in STy, it is intact in STm and highly similar to that of SPtA 9150 ( Figure S6). We infected THP1 cells at

| DISCUSSION
Here we showed that SPtA induces caspase-and GSDMD-dependent pyroptotic cell death in human macrophages. Notably, using pharmacological and genetic silencing, we found that SPtA-induced pyroptosis is predominately dependent on caspase-4 activation, with additional contributions made by the NLRP3/caspase-1 inflammasome, caspase-8 and additional ASC-dependent sensors. The very long LPS O-antigen chains expressed by SPtA did not affect macrophage priming, bacterial uptake or escape from the SCV, but they reduced inflammasome activation and pyroptosis.
We have recently showed that infection of macrophages with enteropathogenic Escherichia coli (EPEC) triggers a cell death cascade in which activation of caspase-4 by LPS leads to activation of the NLRP3/ ASC/caspase-1 inflammasome and pyroptosis (Goddard et al., 2019), and caspase-4 is also involved in recognition of STm infection in IFNγprimed macrophages (Fisch, Bando, et al., 2019;Fisch, Clough, et al., 2020;Kutsch et al., 2020;Santos et al., 2020;Wandel et al., 2020). Moreover, we observed that caspase-8, which has been shown to induce GSDMD cleavage (Orning et al., 2018;Sarhan et al., 2018), also contributed to cell death by SPtA. Alternatively   (Hölzer, Schlumberger, Jäckel, & Hensel, 2009), Shigella flexneri (West et al., 2005) and Shigella sonnei (Caboni et al., 2015) T3SS injectisomes to the host lipid bilayer, while the group 4 capsule masks the T3SS in EHEC (Shifrin et al., 2008). Very long O-antigen chains in SPtA have been shown to reduce antibodymediated recognition by the host and respiratory burst in neutrophils (Hiyoshi, Wangdi, et al., 2018), and are also important for STm virulence (Hölzer et al., 2009;Murray, Attridge, & Morona, 2003. Here we show that SPtA very long O-antigen chains restricted host epithelial cell invasion and suppressed SPtA-induced pyroptosis but not vacuolar escape. Therefore, the length of LPS sugar chains might sterically hinder caspase-4-mediated recognition of lipid-A in the cytosol (Shi, Zhao, Wang, Gao, et al., 2014), which appears to act upstream of the major inflammasome pathways activated during SPtA infection (Figure 4j). Loss of very long O-antigen chains rapidly led to increased caspase-4 activation and pyroptosis, but did not increase the percentage of cells containing ASC foci. A host cell can only assemble a single inflammasome "speck," and because WT SPtA can also trigger detectable inflammasome activation (e.g., through a caspase-4-independent canonical inflammasome pathway), our findings suggest that the loss of fepE triggers higher pyroptosis mainly through direct activation of caspase-4, enhanced GSDMD proteolysis, and membrane damage. While STm encodes FepE, it expresses it at much lower levels than SPtA and seems not to exploit this as an immune evasion mechanism. Importantly, we found an inverse correlation between FepE expression and the level of pyroptosis. SPtA and STm express fepE at different levels possibly as a result of their infection strategies in humans. STm induces an extensive inflammatory response in the small intestine which promotes competition with resident microbiota (Stecher et al., 2007), disruption of the intestinal barrier, penetration to the submucosa and neutrophil infiltration (Tükel et al., 2006;Zhang et al., 2003). In contrast, higher expression of FepE in SPtA leads to very long O-antigen chains, which act as an immune evasion mechanism in the intestine (Hiyoshi, Wangdi, et al., 2018), allowing its systemic dissemination. This is analogous to the Vi antigen-mediated escape from immune recognition in S. Typhi infection (Hiyoshi, Tiffany, et al., 2018;Johnson, Mylona, & Frankel, 2018).
O-antigen-mediated evasion of caspase-4 activation may also be used by S. sonnei (Watson et al., 2019a(Watson et al., , 2019b. Whether other bacteria that produce very long O-antigen chains, such as S. Dublin and S. Enteritidis (Murray et al., 2003), Pseudomonas aeruginosa (Kintz, Scarff, DiGiandomenico, & Goldberg, 2008) and S. flexneri (Morona, Daniels, & Van Den Bosch, 2003), may also avoid caspase-4 activation in this manner should be investigated in the future. GBP1 is an interferonγ-stimulated guanylate binding protein which assists in caspase-4 activation by LPS from cytosolic STm (Fisch, Bando, et al., 2019;Fisch, Clough, et al., 2020;Kutsch et al., 2020;Santos et al., 2020;Wandel et al., 2020). Although induction of GBP1 expression is not an essential component for caspase-4 activation, for example during infection with EPEC (Goddard et al., 2019), future work should also investigate whether GBP1 or other GBPs can overcome the reduced inflammasome activation by very long LPS O-antigen chains.
Our results also revealed that clinical SPtA isolates were more virulent than the prototype SPtA 9150 strain in terms of SPI-1 T3SS expression and activity, inflammatory responses and macrophage cytotoxicity. These results are consistent with data generated from an outbreak SPtA isolate, which was more invasive and motile than SPtA 9150 (Gal-Mor, Suez, et al., 2012). It is likely that in vitro adaptation over time has attenuated the prototype strain SPtA 9150, in a manner comparable to STy (Johnson, Ravenhall, et al., 2018). Alternatively, these differences may be the result of polymorphism acquisition in the geographical region where these organisms were isolated, a phenomenon seen with differing STy genotypes (Frankel, Newton, Schoolnik, & Stocker, 1989;Wong et al., 2015). Of note, our data show that SPtA ED199, which has been used in clinical human challenge studies (Dobinson et al., 2017;McCullagh et al., 2015), has similar pathogenicity to the other clinical variants used in this study.
In summary, we show that SPtA induces inflammasomedependent cell death during infection of human macrophages.
This process is dependent on the SPI-1 T3SS and is limited by production of very long O-antigen chains. We propose a model

| Bacterial strains
Bacterial strains used in this study are listed in Table S1. Salmonella were routinely grown in Lysogeny Broth (LB) Lennox at 37 C with shaking at 200 rpm, with the addition of appropriate antibiotics where necessary (kanamycin (50 μg/ml) and spectinomycin (100 μg/ml); see also Table S4). SPI-1 expression was induced by sub-culturing following 1:33 dilution and growth to late exponential phase.

| Generation of bacterial strains and plasmids
Deletion mutants were constructed using the λ-red recombinase system as previously described (Datsenko & Wanner, 2000). Gene deletions were confirmed by sequencing (Eurofins/GATC). pWSK29-Spec (Johnson, Byrne, et al., 2017) vectors were assembled by Gibson assembly according to manufacturer's instructions. Plasmid inserts were confirmed by sequencing with standard M13 primers (Eurofins/ GATC). Primers are listed in Table S2.

| Cell treatments and in vitro infection
For infections, bacterial cultures were washed twice in PBS and diluted to achieve thedesired MOI (MOI of 50 SPtA-infected THP1 cells, MOI 15 for STm-infected THP1 cells and MOI 100 for HeLa cell infections), which was confirmed retrospectively by colony forming unit (CFU) plating. Following infection, cells were centrifuged at 600 g for 10 min to synchronise the infection.

| PI uptake-dependent cell death assays
Cell media were replaced with complete media supplemented with 5 μM Propidium iodide (PI ; Table S4) prior to infection. Cells were infected as described above and fluorescence was measured using a  4.9 | Immunoblotting THP1 cells were infected as above, except prior to infection cells were washed three times with non-supplemented RPMI 1,640 and infections were performed in OptiMEM media supplemented with 1 mM sodium pyruvate. Supernatants were collected at 3 h post infection and precipitated at −20 C in acetone overnight, acetone was aspirated and protein supernatant and lysate samples were prepared as previously described (Eldridge et al., 2017;Goddard et al., 2019). Proteins were separated by SDS-PAGE and transferred to 0.2 μm PVDF membrane using a TransBlot semi-dry electrophoretic transfer machine (BioRad).

| Gentamicin protection assays
Membranes were blocked in 10% milk for 1 h at room temperature and incubated at 4 C overnight with antibodies listed in Table S4. Membranes were incubated with secondary antibodies, before developing with ECL Prime using a BioRad Chemidoc Imager. Quantification of luminescent bands from western blotting was performed using the ImageLab BioRad software after subtracting background based on the UI control and by determining the ratio of the single cleaved band to the sum of the latter plus the pro-form of each protein.
For bacterial SPI-1 protein expression, bacteria were grown to late exponential phase and protein samples were prepared as previously described (Johnson, Byrne, et al., 2017) and transferred onto PVDF membranes as above. Antibodies used are listed in Table S4. 4.10 | Immunofluorescence microscopy ASC mRFP THP1 cells (Goddard et al 2019) plated onto glass coverslips were infected as described above. At 3 h post infection, cells were washed twice in PBS, fixed with 4% paraformaldehyde for 20 min at room temperature and washed thrice with PBS. Cells were quenched in 50 mM NH 4 Cl for 10 min at room temperature and permeabilised with 0.2% Triton X-100 for 4 min. Coverslips were blocked in 2% bovine serum albumin (BSA) for 5 min before staining with the primary antibody in 2% BSA for 45 min at room temperature. Cells were washed and blocked, before adding secondary antibody, Hoechst, and Phalloidin Alexa647 (Table S4) for 30 min at room temperature. The coverslips were washed in PBS before mounting with ProLong Gold antifade reagent and visualised using a Zeiss Axio Observer Z1 microscope at × 40 magnification (Carl Zeiss).
4.11 | Enzyme-linked immunosorbent assay (ELISA) THP1 cells were infected as above for 3 h and supernatants from infected cells were collected. TNF-α or IL-1β were measured using the ELISA kits listed in Table S4 following the manufacturer's guidelines. Sample absorbance was measured using a FLUOstar Omega plate reader (BMG Labtech) at 450 nm, and absorbance at 540 nm was subtracted for well correction.
4.12 | Crude LPS preparation and silver staining 1.5 ml of overnight bacterial cultures were pelleted by centrifugation at 10,000 g and resuspended in 100 μl Laemmli buffer, before boiling at 100 C for 5 min. Proteinase K was added at 1 mg/ml and samples were incubated 60 C for 2 h, prior to addition of 5% β-mercaptoethanol and further incubation for 5 min. Crude LPS samples were run on acrylamide gels, which were then fixed in 10% acetic acid/30% ethanol overnight. Gels were oxidised in oxidative solution (10% acetic acid, 30% ethanol, 1% periodic acid) for 10 min and washed three times for 15 min in water. Gels were stained for 30 min in silver stain solution (0.2 mg/ml silver nitrate) and briefly rinsed in water. Developer solution (10% acetic acid, 30% ethanol, 0.02% formaldehyde) was used to develop the stain before stopping the reaction with 1% acetic acid. Gels were handled at room temperature and incubations were performed with slow agitation.

| RNA-extraction and RT-qPCR
Approximately 6 × 10 8 bacteria were treated with RNAprotect reagent (Table S4) and digested with 15 mg/ml Lysozyme and 2 mg/ml Proteinase K for 20 min according to manufacturer's guidelines. RNA was extracted using the RNease Mini Kit following the manufacturer's instructions. Two microgram of RNA was treated with DNase as per the manufacturer's guidelines, with prolonged incubation time 1 h. Reverse transcription was performed with the MMLV transcriptase following the associated protocol. qPCR was performed using the Power Up SYBR Green master mix on the Applied Biosystems StepOnePlus system. Twenty nanogram of cDNA was used per reaction with primers listed in Table S2 at 0.2 μM final concentration. 16S was used as a reference gene (Table S2).

| Bioinformatics and statistical analysis
Gene sequences were aligned using the ClustalW tool and visualised in JalView. Protein identity was determined using the NCBI blastp suite. No statistical methods were used to determine sample size. All experiments were repeated independently at least three times as indicated in Figure Legends. Cell death assays, gentamicin protection assays, and ELISA were performed with at least two technical repeats for each biological repeat and means from independent experiments were analysed. For immunofluorescence analysis, >100 cells were counted from randomly selected fields, % cells showing events were obtained for each biological repeat and means were compared statistically. When required, data were log-transformed (CFU experiments) or logit-transformed (PI death assays). Normal distribution was tested with the Shapiro-Wilk normality test. Paired two-tailed t-tests, repeated measures one-way or two-way ANOVA were applied to analyse data as indicated in the figure legends. Multiple comparisons were corrected by the Tukey test or the False Discovery Rate (FDR) approach of Benjamini, Krieger and Yekutieli. Statistical significance marked as: *, p < .05; **, p < .01; ***, p < .001; ****, p < .0001. All statistical analyses were performed using GraphPad Prism 8.