Intracellular Shigella remodels its LPS to dampen the innate immune recognition and evade inflammasome activation

Analysis of LPSs from Acellular and Intracellular Shigella.

During the invasive process, Shigella multiplies within the cytoplasm of the epithelial cells with a doubling time close to that of bacteria grown in laboratory medium (30). To acquire insights about eventual LPS modification during the residence of bacteria within this cell population, we developed an experimental protocol designed to recover shigellae from infected HeLa cells. At 3 h after infection, HeLa cells containing bacteria were processed to purify the iLPS. iLPS was extracted and analyzed by MALDI-TOF MS and compared with that derived by an equivalent number of bacteria grown under conventional conditions. In preliminary experiments, Shigella was cultured in cell medium, DMEM, to allow a comparison between iLPS and bacteria grown under the same medium as the host cells. However, Shigella cultured under these conditions was submitted to cell lysis making unrealistic to collect bacteria for LPS studies. Therefore, we abandoned this approach and performed an LPS analysis of bacteria grown under standard conditions, i.e., in trypticase soy broth (TSB) (31).

iLPS (in contrast to aLPS) did not exhibit dramatic modifications in the core oligosaccharide or the O-antigen. In fact, the core oligosaccharide region of iLPS possessed the same chemical structure already described by our group (31), the only difference being a phosphate group replacing a 2-amino-ethyl phosphate group linked to the heptose I (Fig. 1, species Oligosaccharide). The O-chain portion was characterized through NMR spectroscopy, and was constituted by the same O-repeating unit already found in other Shigella strains (Fig. S1 and Table S1) (32). On the contrary, the lipid A domain of iLPS exhibited a dramatic change in the acylation pattern (Fig. 1B); in fact, it was highly underacylated and mainly characterized by a very high percentage of tri- and tetraacylated lipid A species and a low amount of pentaacylated species, and only trace amounts of the canonical hexaacylated lipid A were present (Fig. 1B). It is worth noting that this information was achieved directly on the intact LPS molecule; in fact, the lipid A MS spectrum was directly obtained on the intact LPS upon in source fragmentation obtained in our optimized conditions (33); in this way, we can exclude any chemical change in lipid A as a result of chemical treatment. In aLPS, the canonical hexaacylated lipid A species was highly prevalent (Fig. 1B), whereas underacylated species were present in minor amounts. All relevant details of the purification procedures are reported in SI Materials and Methods.

The absence of contaminants during the purification steps was assessed through several analyses as GLC-MS (SI Materials and Methods and Fig. S2), showing the presence of sugar residues exclusively ascribable to the LPS moiety.

Activation of NF-κB by LPS Extracted from Acellular and Intracellular Shigellae Correlates with the Degree of Lipid A Acylation.

To ascertain the absence of contamination in the purified LPS, iLPS and aLPS were analyzed in various biological assays (as reported in Fig. S3). The LPS were first analyzed for the presence of bacterial lipoproteins (BLPs; Fig. S3A) and PGN (Fig. S3 B and C) in HEK 293 cells expressing TLR2 (i.e., BLPs) or hNod1 or hNod2 (i.e., PGN) (34). Then, LPSs were assessed in bone marrow-derived macrophages (BMDMs) in the presence or absence of Oxidized 1-palmitoyl-2-arachidonoyl-sn-glycero-3-phosphocholine (OxPAPC)C (40 μg/mL; Fig. S3D), which inhibits the signaling induced by the bacterial LPS (35).

Changes in the degree of lipid A acylation has a dramatic effect on the binding of this molecule to MD-2 and TLR4. The hexaacylated lipid A species has optimal inflammatory activity. In fact, in the MD-2 pocket, the presence of six acyl chains allows the phosphate groups of glucosamine to interact with positively charged residues of TLR4, thus promoting the dimerization and activation of the receptor complex (36).

Compared with the lipid A with six acyl chains, the lipid A with five lipid chains is ∼100-fold less active, and those with four lipid chains completely lack agonistic activity (37, 38).

Therefore, the biological activity of iLPS was assessed in vitro in the model of HEK 293 cells, which stably express human CD14, MD-2, and TLR4. HEK 293 human (h) TLR4 were exposed to different LPS concentrations (1 and 10 ng/mL). NF-κB activation was measured through the assessment of luciferase activity; il-8 mRNA and release was evaluated via quantitative real-time PCR (qPCR) and ELISA, respectively (SI Materials and Methods). Consistent with its chemical composition, iLPS induced lower NF-κB activation (P < 0.01; Fig. 2A, Upper) and il-8 mRNA production and release than did aLPS (Fig. 2A, Left). Several natural or synthetic hypoacylated lipid A structures have been reported to exhibit a high potential as antagonists for endotoxically active LPS (7). We assessed whether iLPS could interfere with TLR4-mediated signaling induced by hexaacylated lipid A. HEK 293 hTLR4 cells were preincubated 1 h with different amounts (1, 10, and 100 ng/mL) of iLPS or aLPS and then exposed to 100 ng/mL of hexaacyl Escherichia coli LPS for 4 h.

iLPS significantly antagonized hexaacyl LPS-dependent TLR4-mediated NF-κB activation at low concentrations (P < 0.001 at 1 and 10 ng/mL; Fig. 2B). The absence of antagonistic effect at high iLPS doses could like be explained by the increase in the iLPS blend of hexaacylated lipid A species in addition to tri- and tetraacylated forms. No antagonistic effect was observed with aLPS (Fig. 2B, Right).

LPS of Intracellular Shigella Modulates Cytokine Release in BMDMs.

We next proceeded to compare the ability of the two LPS forms to activate BMDMs for cytokine production (TNF-α, IL-1β, IL-6, KC, CXCL-10, CCL-5, IL-18, IFN-β, and IL-10). BMDMs were treated with 1 and 10 ng/mL of the two LPS forms as described earlier, and cytokine release was recorded at two time points, 6 h and 18 h (Fig. 3). Hexaacylated E. coli LPS was used in parallel as a control. First, for all three LPS forms (iLPS, aLPS, and E. coli LPS), we observed a low release of IL-1β and a barely detectable release of IL-18, especially at 6 h of stimulation. By comparing the two forms of Shigella LPS (aLPS and iLPS), we found that, at both time points, stimulation with iLPS (especially with 10 ng/mL) caused a dramatic reduction of the yield of cytokine release with respect to the corresponding amounts of aLPS. The only exception was CCL-5 (Fig. 3): the amount of this cytokine was the same with 10 ng/mL of all three LPS forms at 18 h.

It must be stressed that TNF-α, which is considered the canonical readout measuring the proinflammatory potency of LPS, was undetectable with 1 ng/mL and barely detectable with 10 ng/mL of iLPS, with the latter values being significantly lower (P < 0.001) with respect to aLPS.

Because IFN-β production was undetectable (Fig. S4 A and B, Left), we proceeded to analyze the mRNA levels for this cytokine. The exposure of 1 ng/mL of all forms of LPS did not result in the expression of ifn-β mRNA; with 10 ng/mL, the expression of IFN-β was similar with aLPS and E. coli LPS, but reduced with iLPS (Fig. S4C).

MyD88 and TRIF Pathways Are Engaged by both Forms of Shigella LPS.

TLR4-mediated LPS signaling engages two pathways, one mediated by the adaptor protein MyD88 leading to the production of proinflammatory cytokines, and the other involving the proteins Mal (also called Tirap), TRIF, and TRAM allowing the expression of genes encoding type 1 of IFN (IFN-1) and IFN-associated genes (39). The first signaling pathway is activated by the TLR4 plasma membrane and is transmitted through the adaptor molecule MyD88. Then, TLR4 is internalized in the endosomal network in which a second signal pathway is triggered through TRIF and TRAM (40). The TRIF pathway leads to the activation of the transcription factor IFN regulatory factor-3 (IRF-3) (41), which regulates expression of IFN-1 and associated genes.

We therefore assessed whether Shigella LPS could engage both signaling pathways and whether the two LPS forms could selectively activate MyD88 or the TRIF-mediated circuit.

iLPS and the aLPS were used to stimulate BMDMs from myd88- and trif-defective mice by applying the same experimental scheme described earlier. In myd88^−/− BMDMs, the production of TNF-α was totally abrogated after 6 h and 18 h exposure with all LPS forms tested (Shigella iLPS and aLPS and E. coli LPS; Fig. 4). In trif-defective macrophages, TNF-α was still produced, although with a significant reduction with respect to all conditions applied in WT BMDMs. However, the relative proportions and differences among the three LPS forms observed in WT BMDMs were maintained in trif^−/− cells. The low amount of IL-1β observed following LPS stimulation in WT macrophages was abolished and reduced (at 18 h) in myd88^−/− and trif^−/− cells, respectively (Fig. 4). IL-6 yield was suppressed in both myd88^−/− and trif^−/− macrophages under all experimental conditions applied (Fig. 4). Likewise, KC was null in myd88^−/− macrophages at both time points and reduced in trif^−/− BMDMs (Fig. 4).

CXCL-10 has been described as being under the control of IRF-3 (42). The CXCL-10 release was similar in WT and myd88^−/− macrophages at 18 h (Fig. 4). In trif-defective cells, the amount of CXCL-10 was reduced with all LPS tested compared with the corresponding data in WT cells. Production of CCL-5 was similar in WT and myd88^−/− cells at both time points, whereas, in trif-defective cells, it was significantly reduced. Low mRNA for ifn-β was noticed particularly in trif^−/− cells (Fig. S4C). In summary, both forms of LPS (iLPS and aLPS) seem to signal mainly via the MyD88 pathway with the involvement of TRIF-mediated signaling.

iLPS and aLPS Differently Influence the Macrophage Response to S. flexneri Infection.

The interaction of S. flexneri with macrophages is followed by the death of infected cells. It has been reported that this process mainly depends on a mechanism of pyroptosis involving the formation of the NLRC4-mediated inflammasome, caspase-1 activation, and IL-1β release (23). Macrophage death and subsequent release of IL-1β are hallmarks of Shigella infections and play a central role in eliciting the inflammatory reaction in the gut.

The inflammasome assembly requires a priming signal, often referred as “signal 1,” via TLRs or intracellular receptors, which is necessary to up-regulate the expression of certain inflammasome receptors and the substrate pro–IL-1β. Then, a “signal 2” can promote a complex formation (43) and events downstream of the inflammasome activation. Recently, various reports have proposed that two inflammasome platforms could coexist in infected cells. One, usually called “death complex,” leads to cell death whereby activation of caspase-1 does not necessarily involve its autoproteolytic cleavage. The other governs IL-1β production and release and requires caspase-1 processing (43). Given the relevance of these processes in the context of shigellosis, we decided to analyze whether and to what extent macrophage stimulation with iLPS or aLPS could affect macrophage death and IL-1β release induced by Shigella.

In preliminary experiments, C57BL/6 BMDMs were first stimulated with different amounts of LPS, as described earlier, and then infected with S. flexneri. We observed that priming the macrophage with 1 ng/mL of LPS (of both Shigella LPS forms and E. coli LPS) did not change responses upon Shigella infection compared with those obtained with untreated and infected cells. Consequently, we performed all of the following experiments mainly by using the optimal LPS concentration of 10 ng/mL. We analyzed parameters associated with the inflammasome in BMDMs stimulated with iLPS, aLPS, or E. coli LPS for 4 h and infected with M90T [multiplicity of infection (MOI) of 10] over 3 h.

First, we evaluated the rate of cell death and cytotoxicity through the analysis of propidium iodide (PI) staining and lactate dehydrogenase (LDH) release (Fig. 5A). Caspase-1 activation and caspase-1 maturation were analyzed through the fluorescent labeled inhibitor of caspases (FLICA) and Western blot approaches, respectively (Fig. 5B). No difference was observed in cell death, LDH release, and caspase-1 activation/recruitment in infected BMDMs that were previously unstimulated or stimulated with aLPS or iLPS. However, in cell lysates, the processing of caspase-1 was significantly reduced upon iLPS treatment with respect to aLPS, whereas the cleaved caspase-1 fragment (p10) was weakly detectable when the M90T-infected macrophages were not previously exposed to LPS (Fig. 5B). Likewise, in cell supernatants, the cleaved fragment of caspase-1 was weakly detectable following pretreatment of the M90T-infected macrophages with iLPS. These results seem to suggest that cell death is a process independent of LPS priming and involving caspase-1 recruitment, and that full cleavage of caspase-1 is not necessary to this process, in accordance with the hypothesis of the death complex.

We proceeded to analyze the production of IL-1β under the conditions described earlier (stimulation plus infection). Stimulation with 10 ng/mL of iLPS produced a release of a significantly lower amount of IL-1β with respect to 10 ng/mL of aLPS (P < 0.001) and E. coli LPS (Fig. 5C, Left). IL-18 secretion (P < 0.001) also followed the trend observed for IL-1 β (Fig. 5C, Right).

To verify whether this result reflected a general limited ability of iLPS to stimulate M90T-infected macrophages to produce inflammatory cytokines, we also measured the release of TNF-α, IL-6, and KC. The trend observed for IL-1β was partially maintained for IL-6 but not for TNF-α and KC, for which the values obtained with iLPS were similar to those recorded with aLPS and E. coli LPS (Fig. S5). As observed in the experiments with LPS stimulation, both LPS-mediated pathways, MyD88 and TRIF, were involved in the signaling leading to the production of IL-1β following LPS stimulation and M90T infection (Fig. S6).

In cell lysates of C57BL/6, the low IL-1β release observed with 10 ng/mL of iLPS was associated with a reduced production of il-1β mRNA (Fig. 5C, Lower Left), also reflected in a level of mature IL-1β that was lower than that produced by aLPS (Fig. 5C, Lower Right).

NLRC4/Ipaf has been reported as the NLR protein mediating the inflammasome assembly in macrophages infected with Shigella (23). In accordance with these data, the secretion of IL-1β and IL-18 in nlrc4 defective BMDMs were abrogated under all conditions tested (Fig. S7).

In summarizing the results, it appears that treatment of iLPS did not affect the cell death rate of BMDMs with respect to the data obtained with aLPS or with untreated M90T-infected macrophages. Caspase-1 recruitment was similar under all conditions, whereas maturation was reduced when BMDMs were treated with iLPS compared with aLPS. Likewise, stimulation with 10 ng/mL of iLPS strongly reduced the expression, processing, and release of IL-1β with respect to aLPS.

iLPS Influences the Neutrophil Response.

Neutrophils play a pivotal, although somewhat complex, role during Shigella infections. Neutrophils are able to efficiently ingest and kill the bacterium, yet, paradoxically, the inflammation that is triggered by PMNs is also necessary for Shigella to colonize successfully. To determine how iLPS may influence neutrophil responses, we compared the ability of iLPS and aLPS to prime neutrophils. Human neutrophils were isolated and primed for 30 min with Shigella iLPS, aLPS, or LPS purified from S. typhimurium (Fig. 6). Neutrophils were then stimulated with 100 nM N-Formyl-Met-Leu-Phe (fMLF) to produce an oxidative burst. Compared with aLPS or S. typhimurium LPS, neutrophils primed with iLPS induced significantly lower levels of reactive oxygen species (ROS) production at all concentrations tested. In contrast, ROS production in neutrophils was similar after aLPS and S. typhimurium LPS priming. These data suggest that hypoacylation of Shigella LPS can reduce priming of NADPH oxidase in human neutrophils, giving the bacterium an additional mechanism to control inflammation.

Bacterial Strains and Growth Conditions.

The S. flexneri 5a strains used are M90T streptomycin-resistant, and the derivative M90T transformed with the plasmid pILL₁₁₀₁ encoding an afimbrial adhesin of uropathogenic E. coli (Afa E) (66) M90T ΔmsbB1msbB2 has been described by D’Hauteville et al. (28). Bacteria were grown in TSB (BBL; Becton Dickinson) and maintained on agar (TSA). Streptomycin and spectinomycin were added to cultures at 100 mg/mL.

Recovery of LPS from Intracellular Bacteria and Preparation of Bacterial Cultures for LPS Isolation.

To collect and purify the LPS from intracellular Shigella, 2 g bacteria were necessary. To maximize the collection of intracellular bacteria, HeLa cells seeded in 120-mm-diameter plates were infected with S. flexneri M90T pILL₁₁₀₁ (66) at an MOI of 50. After 30 min of incubation at room temperature and an additional 30 min at 37 °C, the acellular bacteria were removed by extensive washing with PBS solution. A total of 10 mL of fresh medium containing gentamicin (60 μg/mL) was added to each plate, and infected cells were incubated at 37 °C for 2 h. At this time, after three washes with PBS solution, the cell monolayers were detached with trypsin and then subjected to sonication and enzymatic hydrolysis to recover intracellular LPS (SI Materials and Methods). Approximately 400–500 infections in 120-mm-diameter plates provided 20 g of infected HeLa cells, from which ∼2 g of intracellular bacteria were recovered. To isolate LPS of bacterial grown in TSB, M90T pILL₁₁₀₁ and M90T ΔmsbB1msbB2 pILL₁₁₀₁ were grown in in a shake flask culture of 4 L under constant aeration at 37 °C for 12 h. A preculture in the same medium was used to inoculate the flask. These conditions resulted in approximately 0.5 g⋅L⁻¹ of dried cells. The culture was checked for purity at the end of the cell cycle. Bacterial culture was then centrifuged and briefly washed in PBS solution, and the pellet dried, yielding approximately 2 g of dried bacteria for LPS isolation and purification (SI Materials and Methods). The full experimental scheme (including the LPS analysis) was repeated twice.

HEK 293 hTLR4/CD14/MD2 Cell Culture, Transfection, and Stimulation.

Stably transfected HEK 293 hTLR4/MD2-CD14 cell lines (InvivoGen) were seeded into 96-well plates at the concentration of 3 × 10⁵ cells per milliliter. For NF-κB studies, cells were transfected with Firefly luciferase reporter constructs, pGL3.ELAM.tk, and Renilla luciferase reporter plasmid, pRLTK, as published previously (34). HEK 293 hTLR4/MD2/CD14 were exposed to different concentrations of Shigella iLPS or aLPS or E. coli LPS (LPS-EB ultrapure; InvivoGen; 1 and 10 ng/mL) and stimulation was prolonged for 4 h or 18 h. For the competition assays, the cells were primed with Shigella iLPS (1, 10, and 100 ng/mL) for 1 h and then restimulated with 100 ng/mL of E. coli LPS (LPS-EB ultrapure; InvivoGen) for 4 h. To assess the absence of contamination (Fig. S3) in LPS preparations, HEK 293 hTLR2 and HEK 293 expressing hNod1 or hNod2 (all cell lines from InvivoGen) were stimulated (hTLR2) and transfected (hNod1 or Nod2) (34) with Shigella and E. coli LPS. Sonicated E. coli PGN (1 μg/mL) and L-Ala-γ-D-Glu-mDAP (TriDAP) (1 μg/mL) and MDP (1 μg/mL) were used as controls in HEK hNod1 and Nod2 studies, whereas Pam3CSK4 (InvivoGen) was used in hTLR2.

NF-κB–dependent luciferase activity was measured by using the Dual-Luciferase Reporter Assay System (Promega) as reported previously (34), and il-8 mRNA was analyzed through qPCR. IL-8 production was quantified after 18 h of stimulation by ELISA.

BMDM Culture and Infection and Stimulation Assays.

All animals were on pure C57BL/6 background and were maintained in a specific pathogen-free animal facility; all experiments were performed in accordance with the guidelines established in the Principles of Laboratory Animal Care (directive 86/609/EEC) and approved by the Italian Ministry of Health. BMDMs were derived from bone marrow cells collected from 5-wk-old WT C57BL/6 (Charles River Laboratories) (67), myd88^−/−, trif^−/−, ipaf (nlrc4)^−/−, or tlr2^−/− female mice (gifts from M. Rescigno, FIRC Institute of Molecular Oncology Foundation-European Istitute of Oncology Campus, Milan, Italy; F. Granucci, Università degli Studi di Milano–Bicocca, Milan, Italy; Mathias Chamaillard, Université Lille Nord de France–Institut Pasteur de Lille, Lille, France; and Anna Zumsteg, Max Planck Institute for Infection Biology).

Neutrophil Isolation and ROS Measurements.

All human subjects are informed on admission that tissue samples may be used for research purposes, through an appropriate informed consent form approved by the Ethics Committee of the Max Planck Institute (Berlin). Human neutrophils were isolated from blood obtained from the blood bank in a protocol approved by the ethics committee of the Charité Hospital (Berlin, Germany). Neutrophils were purified by Histopaque/Percoll (68). After isolation, neutrophils were resuspended in Hanks balanced salt solution (Gibco) containing cations supplemented with 5% FCS and seeded at 7.5 × 10⁴ neutrophils per well in a 96-well plate and allowed to rest for 30 min. Neutrophils were then primed for 30 min with aLPS, iLPS, or S. typhimurium S-form TLR-grade (Enzo Life Sciences). After priming, neutrophils were stimulated with 100 nM fMLF (Sigma). ROS formation was measured over time by chemiluminescence (69) by using 50 μM luminal and 1.2 U/mL HRP (Calbiochem). Chemiluminescence was detected by using a Victor Light 1420 counter (Perkin-Elmer), and data are displayed as relative light units.

Cell Death/Cytotoxicity Studies and Caspase Activity.

Cell death/cytotoxicity were assessed by using PI staining (Apoptosis Detection kit; BD Pharmingen) and LDH release (Cytotox96 Cytotoxicity Assay; Promega), according to the manufacturers’ instructions. Caspase-1 activity was evaluated through carboxyfluorescein FLICA Apoptosis Detection kit caspase assay (Immunochemistry Technologies). Cell death parameters and caspase-1 activity were analyzed by using a flow cytometric analysis on a FACSCalibur cytometer (Becton Dickinson). Data acquisition (10⁴ events for each sample) was performed by using CellQuest software (Becton Dickinson). Analysis was performed with FlowJo software (TreeStar). LDH release was quantified on a LT4000 ELISA reader at a wavelength of 450 nm (Labtech International).

Western Blot Analysis.

Total protein extracts obtained through cell lysis and Western blot procedures were carried out as described (27). Protein extraction from cell culture supernatants was performed as reported (70). Monoclonal antibody to caspase-1 (IMG-50 28; Imgenex/Histo-Line Laboratories) was used to recognize full-length caspase. Caspase-1 and IL-1β processing was determined by using rabbit anti–mouse caspase-1 p10 antibody (sc514; Santa Cruz Biotechnology) and goat anti-mouse IL-1β antibody (401-NA; R&D Systems), respectively; mouse monoclonal anti-hsp70 (SMC-164 C/D; Stress Marq Biosciences) was used to ensure equal loading.

Statistical Analysis.

Data are reported as means ± SD, and the numbers of independent experiments are indicated in the legends of each of the figures. Statistical calculations and tests were performed by using the Student t test. A P value of 0.05 was considered statistically significant. A P value of 0.001 was considered extremely significant.