Salmonella bacteria are zoonotic pathogens of major importance to human and animal health worldwide. According to the European Food Safety Authority and the European Center for Disease Prevention and Control, salmonellosis remains the second most frequently reported gastrointestinal infection in humans after campylobacteriosis. In 2018, Salmonella bacteria were responsible for nearly a third of foodborne disease outbreaks in the EU. Slovakia, Spain and Poland were responsible for 67% of all salmonellosis outbreaks in the EU [45]. Salmonella bacteria are characterized by several virulence factors. An infection can lead, apart from diarrhea, to extraintestinal infections and sepsis.

One of the main features that allows S. Enteritidis bacteria to survive in an extremely unfavorable environment is LPS. Literature data about the involvement of different LPS O-antigen types in complement evasion by S. Typhimurium is ambiguous [29, 30, 34]. Therefore, we examined systematically, using specific deletion mutants, roles of VL-OAg, L-OAg and LMW-OAg LPS in complement evasion by S. Enteritidis. First, we characterized the structural properties of the LPS molecules. The measurement of LPS length and its density on the bacterial surface is difficult due to various reasons. Lipopolysaccharides have high structural variability, several LPS length types exist simultaneously in one bacterial cell, and LPS molecules have a tendency to form aggregates of different sizes and experimentally obtained LPS preparations are heterogeneous.

In order to characterize the average LPS O-antigen length of S. Enteritidis PCM 2817 WT strain and mutants (Fig. 1) a method measuring the average length of LPS molecules present in a given bacterial strain was used (Figs. 3 and S3). The essence of the used method is the analysis of the ratio between the amount of sugar components present in the RUs of the O-specific part (length-dependent component) and in one of the LPS core components (length-independent component). This approach was previously used for E. coli O111 LPS, where the colitose to Kdo ratio was measured colorimetrically [17]. To measure the NeuAc(sialic acid) to Kdo ratio in LPS of Salmonella O48 bacteria GC–MS technique was used [33, 46]. In this study, we determined the amount of the chemical markers of the O-specific part of LPS: rhamnose and tyvelose (Fig. S3A, B), as well as of the components of the LPS core: Kdo and heptose (Fig. S3C) using the peracetylated methylglycoside derivatives for the analysis [47]. Rhamnose is a common component of the O antigen [48, 49] and fimbriae [50, 51]. In turn, tyvelose occurs only in some serovars of Yersinia pseudotuberculosis [52], Y. entomophaga [53], S. Enteritidis and S. Typhi [48]. Tyvelose is also a component of the immunogenic glycoprotein TSL-1 from Trichinella spiralis [54]. The low prevalence of tyvelose as a structural component in bacteria and nematodes makes it an interesting chemical marker for the rapid detection of S. Enteritidis. The method used for the average LPS O-Ag chain length analysis is based on the simultaneous analysis of five sugar markers in one sample: Rha, Tyv, Hep, Kdo and MuAc. This method represents a significant advance over the previously used and described method engaging sialic acid as a marker of LPS length, due to the fact that when Salmonella does not produce NeuAc in the O-antigen, such a method cannot be used [55].

The obtained results of the measurements of the average LPS length (Fig. 3A, B) confirmed that they can be used for comparison between strains. Differences in the results obtained from the measurement of the proportions of sugar markers in the whole bacterial mass (Fig. 3A) and in the isolated LPS preparations (Fig. 3B) for the WT strain and the ΔwzzfepE mutant may be due to the used LPS isolation method. The LPS preparations used for the measurements were isolated using the hot phenol–water method, which is recommended for the isolation of very long and long LPS types [38, 56]. The inclusion in the analysis of muraminic acid also allows to compare the proportion of the total numbers of LPS molecules in the cells between individual strains, assuming that the amount of muramine in the cell is constant (Fig. 3C). The analyzed markers differ significantly in their chemical properties: sensitivity to degradation and susceptibility to disruption of the bonds formed in the LPS molecule. Tyv and Kdo are readily released during the methanolysis reaction. However, their methyl glycosides are sensitive to acid degradation, while efficient release of MuAc from peptidoglycan requires longer methanolysis at a higher concentration of HCl. In addition, ester groups in MuAc and Kdo derivatives are sensitive to elevated pH, which hampers the repetitive neutralization of the mixture after the reaction. The method developed may in the future be the basis for a more detailed analysis of the proportions of different types of LPS (VL-OAg, L-OAg, LMW-OAg) on the bacterial cell surface.

In our previous publication we analyzed 21 Salmonella O48 clinical isolates and showed a high variability in the average length of the O-antigen part of LPS molecules ranging up to two orders of magnitude between extreme values [46]. Subsequent experiments on selected strains have shown that even in the same strain the average length can be gradually increased by the bacteria after stimulation with human serum [33]. The mechanisms for the regulation of O-specific chain length during LPS synthesis are still unclear and there are several hypotheses regarding this phenomenon, involving the regulation by synthesis time (the “molecular clock”) [22], or by the length of emerging O-polysaccharide chain (“molecular ruler”) [57]. Recent studies indicate, that the regulation is effected by the expression ratio between the enzymes: WbdA responsible for the elongation of polysaccharide and WbdD responsible for the termination of the synthesis process [58]. A more detailed description of length control of the O-antigen by Wzz proteins was described previously [20]. Additionally, the effect of environmental factors on length regulation is so far very poorly understood and limited to studies on the effect of bactericidal components of the complement system, composition of medium, temperature and also growth phase [31, 34, 59, 60].

Current knowledge in this area is ambiguous. Palva and Mäkäla showed that the LPS on the surface of S. Typhimurium has a low content of molecules with 2–18 repeating O-specific subunits (LMW-OAg) and a more predominant LPS containing 19–34 repeating subunits (L-OAg) (77%) [61]. It is worth noting that, at the time of publication of the study, the form of VL-OAg LPS and the wzzfepE gene responsible for its formation were not yet known. Despite the presence of bands in the electropherograms indicating very long O-specific chains, the authors do not mention the possibility of the occurrence of LPS with more than 100 repeating O-specific subunits [61]. Numerous studies have shown that the proportions of individual length forms of LPS molecules on the surface of the bacterial cell have a significant impact on the protection of Salmonella against the complement system [19, 29, 30, 34, 62]. Previous studies on this phenomenon were performed using mostly S. Typhimurium. That serotype, although closely related to S. Enteritidis, differ in the composition of its O-antigen structure: in S. Typhimurium the monosaccharide tyvelose, present in S. Enteritidis in the repeating unit, is replaced by abequose. It has been shown previously that seemingly such a small difference as the orientation of the −OH groups in the positions C2 and C4 in one monosaccharide of the O-antigen repeating unit has such a distinct impact on the opsonization capacity of those serovars [63].

Therefore, in this the study, the influence of different S. Enteritidis O-antigen LPS types (Figs. 1, 3) on complement evasion was determined. The analysis of survival in 25% NHS showed that the wild-type strain and ΔwzzfepE mutant were serum resistant (Fig. 4A, B), while ΔwzzST,ΔwzzfepE ΔwzzST and Δwzy were serum sensitive (Fig. 4C–E). The obtained results for the WT strain are in accordance with a previously published study, where the strain showed resistance to 50% NHS [62]. The addition of Ravulizumab (anti-C5 antibody blocking the terminal pathway of complement) in the serum challenge resulted in the survival of the serum sensitive strains (Fig. 4A–E). This proved the involvement of the complement membrane attack complex in the killing of the tested bacteria. The obtained results indicate a small contribution of VL-OAg in the protection of S. Enteritidis bacteria against the lytic effect of the complement system. In contrast, L-OAg clearly protects the bacterial cells from complement mediated lysis.

Literature data initially indicated a greater involvement of the wzzfepE gene, and thus VL-OAg, in protection of S. Typhimurium against complement lysis [29]. However, the results obtained in later experiments showed that it is the wzzST gene, which is essential for S. Typhimurium resistance to human serum [29, 34]. Bravo et al. showed that the survival of S. Enteritidis LPS length mutants in 40% NHS is directly linked with the growth phase. Mutation in the wzzfepE gene was found to lead to serum resistance in the logarithmic growth phase but had no influence on the survival in the stationary phase [34].

C3 activation by the tested S. Enteritidis PCM 2817 strains in 50% NHS showed that mutants with a repertoire of longer O-specific chains on the bacterial cell surface (WT, ΔwzzfepE, ΔwzzST) more effectively activated the C3 protein of the complement system (Fig. 5A). However, we observed differences in complement activation between the entire bacterial cells and isolated LPS molecules. The analysis of that activation by isolated LPS preparations showed that the ΔwzzfepE LPS mutant activated the complement system to a higher degree than the LPS isolated from the WT strain (Fig. 5B). In contrast, C5b-9 deposition on entire bacterial cells was more abundant for the ΔwzzST strain than for the WT strain. This discrepancy might be related to the previously discussed method of LPS isolation (Fig. 6A) [56] or to the difference between the physical state of the LPS molecules on the intact bacterial surface vs. isolated LPS preparation immobilized on polystyrene. The observed discrepancy in the level of activation of the complement system for bacterial cells and isolated LPS molecules could mean that with a shorter O-antigen, there is lowered activation of the complement system. Based on flow cytometry experiments (Fig. 6) it can be concluded that the WT strain and ΔwzzfepE mutant deposit the end products of the complement cascade on their surfaces, but the C5b-9 complex is released from the cell surface without lysing the bacterial cell (Fig. 6B). Similar conclusions were reached by Joiner et al. studying the mechanism of resistance of S. Minnesota smooth and rough strains exposed to human serum. The authors showed that in serum-resistant strains, complement components are deposited on the polysaccharide portion of LPS, which forms a physical barrier separating them from the lipid portion of the cell membrane [28, 64]. It would be useful to know the exact proportions between the individual O-antigen types (VL-OAg, L-OAg, LMW-OAg) of LPS in individual S. Enteritidis PCM 2817 mutants. If mutation in the ΔwzzST gene affects the activation of the complement system to a greater extent (Fig. 6A) than the mutation in the ΔwzzfepE gene, it can be postulated that the appropriate density of LPS with long O-specific chains on the surface of the outer membrane is of key importance in protecting S. Enteritidis against complement lysis. Most likely, the higher density of L-OAg LPS on the ΔwzzfepE mutant is protective against the incorporation of the C5b-9 complex into the bacterial cell membrane, while in the case of the ΔwzzST mutant, the density of VL-OAg LPS is probably not high enough to protect the bacteria from depositing C5b-9 on the outer membrane (Fig. 6). VL-OAg LPS, as shown in the experiments with monosaccharide markers of O-Ag and LPS core (Fig. 3A, B), constitute only a tiny molar part of the total LPS. This implies that VL-OAg molecules are very sparsely distributed on the bacterial surface. In that mutant the L-OAg LPS molecules are most probably replaced by LMW-OAg or LPS with one repeating unit, as the overall number of LPS molecules is constant (Fig. 3C). As it turns out from the SDS PAGE analysis, the molar content of VL-OAg LPS is low, and the lack of L-OAg in wzzST mutant is probably not compensated by VL-OAg LPS. The polysaccharide layer on the outer membrane is thus much more susceptible to the action of serum proteins. Therefore, despite prominent C5b-9 deposition, permeability data suggests that the forming MAC cannot penetrate the membrane (Fig. 6). L-OAg LPS in turn strongly activates the complement alternative pathway (Fig. 5B), which is apparently somehow attenuated by the presence of VL-OAg. In the presence of short and very short LPS molecules this activation capability is very low, although the perforation of the membrane built up of such short molecules is the strongest (Fig. 6B).

In vivo experiments, using laboratory animals is an essential element in determining the degree of pathogenicity of bacterial strains and in understanding interactions between the host and the pathogen. Although the use of rodent models is widely used in the study concerning the pathogenesis of Salmonella infections, mammalian studies are often time-consuming, require an expensive experimental set up and are associated with significant ethical issues. The introduction of invertebrate models allows for a substantial reduction of research costs and minimizing the ethical concerns. Numerous studies have confirmed that the Wax moth G. mellonella is a convenient model to test some aspects of the pathogenicity of bacterial strains [65,66,67,68,69,70,71]. Bender et al. showed that the structure of LPS affects the virulence of S. Typhimurium NCTC 12023. G. mellonella survival after injection with mutant strains lacking the synthesis of VL-OAg or L-OAg was greater, than the survival of larvae after injection with the wild-type strain. In addition, the authors showed that removal of the entire O-specific antigen was associated with the loss of virulence of S. Typhimurium NCTC 12023 in the G. mellonella infection model [70]. Our investigations are in accordance with those observations, showing that the elimination of a particular fraction of O-antigen had an effect of the survival rate of G. mellonella larvae. After challenge with S. Enteritidis PCM 2817 WT and the ΔwzzfepE most of the larval population did not survived (survival 13% and 7%, respectively) (Fig. 7A, CFU/ml 107). The elimination of the L-OAg LPS reduced the pathogenicity to 27% of survival, indicating the importance of long O-antigens as a virulence factor of S. Enteritidis (Fig. 7A, CFU/ml 107). The obtained results confirm the observations of Bender et al. [70] that the structure of LPS may affect the pathogenicity of strains in the infection model of G. mellonella larvae. An interesting observation is that the lack of VL-OAg increases bacterial virulence (increased larval mortality, Fig. 7A). Perhaps the bacteria produce other structures on the surface of the outer membrane that compensate for the lack of VL-OAg. Surprisingly, the survival rate of wax moth larvae subjected to infection with ΔwzzfepE was lower than larvae infected with WT or double mutants. This suggests that VL-LPS can have a moderating role at least in invertebrates (Fig. 7A).

Previous research of other groups concentrated on immunological aspects of LPS length diversity, but the molecular characterization of the O-antigen length was limited to SDS-PAGE analysis with various types of LPS pattern visualization. In our work we have analyzed the length of LPS O-specific polysaccharide using mass spectrometry, therefore we could precisely compare the LPS modal distribution in particular mutants. The elaborated method utilizes tyvelose as an LPS length marker, which has advantages over the previously used method using sialic acid, which could not be used in strains lacking NeuAc [55]. Our method can be also used for quick S. Enteritidis strain identification by detection of Tyv. Previous publications by other groups have mainly studied the S. Typhimurium serotype, which is an predominant serovar in the United States. S. Enteritidis, which is prevalent in Europe and used in the present work, have not yet been studied in this regard. Several publication compare the immunological aspects of reactivity of different strains [28, 64] and different serotypes, where the O-antigen polysaccharide structure is not identical [63]. That approach may introduce variables into the experiment that will not be taken into account. In our publication, we use different mutants of the same Salmonella strain, differing only in the pattern of LPS modal distribution, which allows us to analyze the impact of just that single factor. We also analyze multiple aspects of complement activation, using not only survival in serum but also the more subtle aspects of complement system activation, like C3 activation, specific blocking of C5, C5b9 deposition, bacterial cell permeability and finally preliminary in vivo study on larvae model.

Although the phenomenon of modulation of the length of the lipopolysaccharide molecule probably cannot be used for vaccine construction, the different modal distribution of lipopolysaccharide and the absence of some LPS fractions, especially VL-OAg and L-OAg LPS, may lead to better exposure of some immunogenic structures, such as fimbriae or outer membrane proteins, on the bacterial surface. This in turn may make such bacteria more susceptible to attack by complement or specific antibodies, and vaccines based on the aforementioned structures may be more effective against these bacteria.

To summarize, the obtained data clearly demonstrate that S. Enteritidis bacteria require LPS with a specific modal length to resist complement-mediated lysis and to survive in the G. mellonella infection model. The results indicate a particular contribution of L-OAg in bacterial virulence. The observed inconsistency of the activation potential between isolated LPS and entire bacterial cells is intriguing and requires further study in future.