Interrogating the genus Yersinia to define the rules of lipopolysaccharide lipid A structure associated with pathogenicity

26 Pathogen recognition by the immune system relies on germline-encoded pathogen recognition 27 receptors which identify conserved pathogen-associated molecular patterns (PAMPs) such as the 28 lipid A section of the lipopolysaccharide (LPS). The assumption that pathogens and mammalian-29 associated bacteria remodel their lipid A PAMP because of host-microbe co-evolution is a long 30 held-belief of microbial pathogenesis. We set out to test this fundamental principle by interrogating 31 a Gram-negative genus presenting evidence of evolutionary events linked to the acquisition of 32 essential virulence traits, resulting in pathogenic and non-pathogenic species. The genus Yersinia 33 fulfil these requirements; the acquisition of the pYV virulence plasmid is one of the evolutionary 34 events associated with virulence. At 37°C, only pathogenic Yersinia switch to deacylated lipid A, a 35 modification that diminishes TLR4/MD-2 recognition and reduces inflammation. An engineered 36 chimeric pathogenic Yersinia strain expressing the non-pathogenic lipid A profile efficiently 37 engages with TLR4, demonstrating it is sufficient to switch the acylation pattern to modify the 38 recognition by TLR4 and subsequent activation of inflammation. The lipid As of pathogenic and 39 non-pathogenic species are modified with aminoarabinose and palmitate; therefore, only the 40 reduced acylation of the lipid A PAMP is a trait associated with virulence. The decorations of lipid 41 A do not alter TLR4 engagement but confer resistance to antimicrobial peptides. The chimeric 42 pathogenic Yersinia strain expressing the non-pathogenic lipid A profile allows to ascertain whether 43 the switch in the lipid A PAMP affects virulence. This strain showed enhanced motility due to an 44 upregulation of the flhDC master regulator, and impaired cellular invasion through downregulation 45 of rovA, a key invasion regulator. The expression and function of pYV-encoded virulence factors 46 Yops and YadA were not affected. Nonetheless, the chimeric strain was attenuated in vivo, 47 demonstrating that virulence factors cannot overcome a switch in the lipid A PAMP associated with 48 pathogenicity. 49 50 51 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 3

52 The recognition of pathogens through the actions of a set of germline encoded innate immune 53 receptors is one of the foundations of immunology. These receptors, known as pattern recognition 54 receptors (PRRs), detect conserved molecules common to large number of microbes to launch an 55 antimicrobial programme [1]. These microbial molecules are distinguishable from “self”, important 56 for the viability and fitness of the pathogen in different environments and, therefore, they are stable 57 [2]. These molecules contain so-called pathogen-associated molecular patterns (PAMPs) recognized 58 by the PRRs [2]. Well-characterized molecules are the lipid A section of the lipopolysaccharide 59 (LPS), the flagellin of the bacterial flagella, and bacterial nucleic acids. 60 The limits of the universal rule of PRR recognition of PAMPs were challenged when it was 61 recognized that pathogens and members of the microbiome may modify their PAMPs to affect the 62 activation of PRRs [3-5]. This has been particularly studied in the case of the LPS lipid A 63 recognized through the PRRs To ll-like receptor 4/myeloid differentiation factor 2 (TLR4/MD-2) 64 complex, and inflammatory caspases, caspase-11 in mouse and caspases-4/5 in humans [6-10]. LPS 65 receptors best sense a bisphosphorylated diglucosamine to which are attached six saturated fatty 66 acyl chains with lengths of 12 or 14 (occasionally 16) carbons, the so-called “hexa-acyl lipid As” 67 [11]. Variations of this structure by changing the number and length of fatty acids, the 68 phosphorylation status as well as by adding some lipid A decorations, phosphoethanolamine and 69 aminoarabinose may result in impaired PRR recognition [12]. Therefore, these changes of the lipid 70 A PAMP are thought to represent adaptations of microbes to the challenge imposed by the innate 71 immune system as a result of the co-evolution between host and microbes. However sound this 72 notion, much of the research has focused only on pathogens, and there is extensive evidence 73 showing that environmental conditions influence the lipid A structure. It cannot then be rigorously 74 ruled out that these lipid A changes found in pathogens may simply reflect bacterial adaptation to 75 changing environments and, therefore, are shared by non-pathogenic Gram-negative bacteria. 76 Supporting this notion, deep-sea bacteria express lipid A with decorations found in pathogenic 77 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 4 bacteria [13]. Therefore, the question remains whether indeed there is a distinct lipid A PAMP 78 expressed by pathogens shaped by host-pathogen co-evolution. 79 To test this principle rigorously, it is necessary to interrogate bacteria from the same genus as they 80 should produce the same lipid A. Within the genus, there should be evidence of evolutionary events 81 linked to the acquisition of essential virulence traits, resulting in, at least, two well defined distinct 82 groups, one encompassing pathogenic species and another including non-pathogenic species lacking 83 almost all the virulence factors expressed by the pathogenic group. The comparison between these 84 two groups of species should help to delineate features of the lipid A PAMP of pathogens if they do 85 exist. 86 Careful consideration of Gram-negative bacterial diversity reveals that the genus Yersinia meets all 87 the requirements to define the lipid A PAMP associated with virulence. The genus includes 26 88 species of which only Y. pestis , and the enteropathogenic Y. pseudotuberculosis and Y. 89 enterocolitica are mammalian (including humans) pathogens [14, 15]. The remaining known 90 species are commonly found in soil and aquatic environments and are non-pathogenic in mammals 91 [16]. Evolution studies of the genus Yersinia r evealed that the most prominent event during the 92 emergence of pathogenesis is the acquisition of the pYV virulence plasmid, encoding the Ysc type 93 III secretion system (T3SS) crucial to blunt cell-intrinsic immunity [14, 15, 17]. 94 In this study, by examining pathogenic and non-pathogenic species of the genus Yersinia, we 95 demonstrate that indeed there is a lipid A PAMP associated with virulence. This lipid A PAMP is 96 characterized by reduced acylation. In contrast, the decorations of the lipid A PAMP with 97 aminoarabinose and palmitate, though useful to survive within the mammalian host, are not unique 98 traits of the pathogenic lipid A PAMP. We provide molecular mechanistic insights into the 99 enzymology of the lipid A PAMP. Finally, our results demonstrate the negative effect on virulence 100 of switching the lipid A PAMP expressed by pathogens. 101 102 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 5

103 Variations in the lipid A PAMP expressed by pathogenic and non-pathogenic Yersinia spp. 104 In Y. pestis, and the highly virulent enteropathogenic Y. pseudotuberculosis and Y. enterocolitica 105 phylogoup 2 (serotype O:8) (hereafter YeO8), growth temperature regulates the lipid A acylation 106 pattern being predominantly hexa-acylated and hepta-acylated at 21 oC and tetra-acylated at 37 oC 107 [18-22]. Only in bacteria grown at 21oC, the lipid A is decorated with aminoarabinose and palmitate 108 [18, 20, 23-25]. To investigate whether this lipid A PAMP is found in other pathogenic Yersiniae, 109 we determined the lipid A structure of two additional enteroptahogenic Y. enterocolitica strains 110 from phylogroup 3 (serotype O3, hereafter YeO3) and phylogroup 5 (serotype O:9, hereafter YeO9) 111 (Table S2). Phylogroup 3 is the most frequent cause of human yersiniosis [26]. We extracted lipid A 112 and predicted structural composition by MALDI-TOF mass spectrometry (MS) in the negative ion 113 mode. Mass spectral data were used to predict the number of acyl changes, the loss of a phosphate 114 from the di-glucosamine sugar backbone (mono-phosphorylation), and the decorations of the lipid 115 A with aminoarabinose, palmitate, or phosphoethanolamine. When grown at 21oC, YeO3 and YeO9 116 lipid As contained primary ionizable peaks m/z 1,797 and 1,824 (Fig 1A and Fig 1B), which are 117 predicted to be hexa-acylated lipid A species (Table 1). m/z 1,824 corresponds to a structure 118 containing two glucosamine residues, two phosphate groups, four 3-OH-C14, one C12 (laurate), 119 and one C16:1 (palmitoleate) (Table 1). m/z 1,797 contains C14 (myristate) instead of C16:1. Peak 120 m/z 1,414 corresponds to a tetra-acylated lipid A (Table 1); we have demonstrated that this species 121 is caused by the activity of the LpxR deacylase that removes the 3 ′ -acyloxyacyl residue of the lipid 122 A [20]. YO3 and YeO9 also encode the lpxR deacylase (Table S1). Two more peaks of m/z 1,954 123 and 2,063 were detected and they correspond to the modification of the species m/z1824 with 124 aminoarabinose (m/z 131) and palmitate ( m/z 239), respectively (Fig 1A and Fig 1B). YeO3 and 125 YeO9 encode the pmr/arn operon responsible for the decoration with aminoarabinose, and the pagP 126 acyltransferase responsible for the addition of palmitate to produce hepta-acylated lipid A (Table 127 S1) [23]. These lipid A species are also found in the lipid A of YeO8 grown at 21 oC (Table 2) [20, 128 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 6 21, 23]. When grown at 37oC, YeO3 and YeO9 lipid As contained hexa-acylated lipid A, m/z 1,797, 129 and a tetra-acylated lipid A, m/z1,388, corresponding to a species with two glucosamine residues, 130 two phosphate groups, three 3-OH-C14 units, and one C14 (Fig 1C and Fig 1D). These lipid A 131 species are also found in the lipid A of YeO8 grown at 37 oC [20] (Table 2). m/z 1,388 peak is 132 consequence of the action of LpxR [20]. No lipid A modifications were detected (Fig 1C and Fig 133 1D). Altogether, it can be concluded that the lipid A PAMP of pathogenic Yersinia spp is 134 characterized by its reduced acylation at 37oC and by the temperature-dependent decoration of lipid 135 A with aminoarabinose and palmitate. 136 To ascertain the lipid A PAMP expressed by non-pathogenic yersiniae , we analysed the lipid A 137 from five strains belonging to Y. enterocolitica phylogroup 1A (Table S2). This phylogroup 138 includes non-pathogenic Y. enteroclitica strains isolated from a variety of environmental sources 139 and they do not harbour the pYV virulence plasmid [17, 27]. Strikingly, these strains expressed a 140 different lipid A than the pathogenic Y. enterocolitica strains (Table 2). When grown at 21 oC, we 141 detected peaks m/z 1,797 and 1,824 (Fig 1E and Table 1) and a new species m/z 1,768. The latter 142 corresponds to a hexa-acylated lipid A containing two glucosamine residues, two phosphate groups, 143 four 3-OH-C14, and two C12 (laurate) (Table 1). This lipid A species was modified with 144 aminoarabinose, m/z 1,899, and palmitate, m/z 2,007 (Fig 1E and Table 1). When grown at 37 oC, 145 these non-pathogenic strains expressed hexa-acylated lipid As m/z 1768 and m/z 1,797, and no 146 decorations were detected (Fig 1F and Table 2). No tetra-acyl lipid A species were detected at any 147 growth temperature. These results suggest that the reduced acylation of the lipid A at 37oC could be 148 a trait of the lipid A PAMP from pathogenic strains whereas this is not the case for the decorations 149 of the lipid A with aminoarabinose and palmitate. 150 We next extracted lipid As from twelve non-pathogenic Yersinia spp, Y. aldovae, Y. bercovieri, Y. 151 mollaretii, Y. frederiksenii, Y.intermedia, Y. kristensenii, Y. rhodei, Y. ruckeri, Y. similis, Y. 152 massilensis, Y. pekkaneniii, and Y. nurmii (Table S2). The lipid As extracted from these species 153 grown at 21oC were similar to those produced by Y. enterocolitica 1A strains (Fig 1G and Table 2). 154 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 7 At 37oC, we did not detect any tetra-acylated lipid A species (Fig 1H and Table 2), and the peaks 155 detected corresponded to hexa-acylated lipid As m/z 1,768 and 1,797 (Table 1). Only for Y. aldovae, 156 Y. bercovieri, Y. mollaretii, Y. frederiksenii, Y. intermedia, and Y. kristensenii lipid As did we detect 157 a peak m/z 2,007, consistent with the addition of palmitate to the hexa-acylated species m/z 1,768 158 (Fig 1H, Table 1 and Table 2). 159 Altogether, these results demonstrate that the switch to deacylated lipid As at 37 oC only occurs in 160 pathogenic Yersiniae. The temperature-dependent variation in the modifications of the lipid A 161 observed in pathogenic Yersiniae is also found in non-pathogenic Yersiniae (this work) [18, 20, 23-162 25]. Therefore, the modifications of the lipid A with aminoarabinose and palmitate cannot be 163 considered a trait uniquely found in the lipid A PAMP of pathogenic bacteria in contrast to the 164 reduced acylation. 165 Enzymology of Yersinia spp lipid A PAMP acylation. 166 One conspicuous difference between the lipid As of the pathogenic and non-pathogenic Yersiniae is 167 the absence of tetra-acylated species in the latter. Bioinformatic analysis revealed the absence of the 168 deacyase lpxR in the genomes of the non-pathogenic Yersiniae (Table S1), explaining the lack of 169 LpxR-dependent tetra-acylated lipid As m/z 1,388 and 1,414 in the lipid As of these species. We 170 questioned whether LpxR will be sufficient to produce deacylated lipid As in non-pathogenic 171 Yersiniae. YeO8 lpxR was introduced at a single copy in Y. aldovae genome using the Tn7 delivery 172 system [28]. When grown at 21 oC, the lipid A expressed by this chimeric non-pathogenic strain, Y. 173 aldovae::LpxRYeO8, contained peaks m/z 1,414 and 1,388 consistent with the deacylation of the 174 hexa-acylated species m/z 1,797 and m/z 1,824 (Fig 2A). At 37 oC, we detected lipid A species m/z 175 1,360 and m/z 1,388 resulting from the de acylation of the hexa-acylated species m/z 1.768 and 176 1,797, respectively (Fig 2B). These results demonstrate that the lack of lipid A deacylation is not an 177 intrinsic property of non-pathogenic Yersiniae lipid As. 178 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 8 Another conspicuous difference is the presence of the hexa-acylated species m/z 1,768 only in the 179 lipid As from non-pathogenic yersiniae (Table 2). In pathogenic Yersiniae, the late acyltransferase 180 LpxM transfers C12 to the 3 ′ -R-3-hydroxymyristoyl group, whereas LpxL and LpxP transfer C14 181 and C16:1, respectively to the 2 ′ -R-3-hydroxymyristoyl group [21, 22]. In silico analysis revealed 182 the genomes of non-pathogenic Yersiniae encoded three late acyltransferases, lpxM, lpxL and lpxP 183 (Table S1) with more than 90% sequence identity to the pathogenic late-acyltransferases. We then 184 sought to characterize the substrate specificity of these enzymes in vivo by expressing them in a 185 surrogate host. This approach has been used previously to characterize the function of lipid A 186 enzymes in the context of the bacterial membrane [29, 30]. Y. aldovae lpxM was cloned and 187 expressed in E. coli lpxM mutant strain BN2 [31]. As expected, lipid A extracted from BN2 188 contained only one species m/z 1,588 corresponding to two glucosamine residues, two phosphate 189 groups, four 3-OH-C14, and one C12 (added by the E. coli acyltransferase LpxL) (Fig 2C). BN2 190 complemented with Y. aldovae lpxM contained species m/z 1,768 consistent with the addition of 191 C12 to the lipid A (Fig 2D), establishing that LpxM from non-pathogenic Yersiniae esterifies the 192 lipid A with C12 likewise LpxM from pathogenic Yersiniae [21, 22]. Similar results were obtained 193 expressing lpxM from Y. bercovieri, and Y. intermedia (Fig S1A and Fig S1B). 194 To confirm whether LpxP from non-pathogenic Yersiniae transfers C16:1 to 2 ′ -R-3-195 hydroxymyristoyl group like LpxP from pathogenic Yersiniae [21, 22], we complemented YeO8 196 lpxP mutant, strain YeO8-Δ lpxPGB [21], with lpxP from Y. aldovae. lpxP from the non-pathogenic 197 strain restored the species m/z 1,824 in YeO8 lpxP mutant (Fig 2E and Fig 2F); demonstrating that 198 LpxP from non-pathogenic Yersiniae transfer C16:1 to the lipid A. Y. bercovieri and Y. intermedia 199 lpxP also complemented YeO8-Δ lpxPGB (Fig S1C and Fig S1D). 200 These enzymatic activities raise the question which is the late acyltransferase esterifying the 2′ -R-3-201 hydroxymyristoyl group with C12 to give m/z 1,768 in non-pathogenic Yersiniae. It can be 202 postulated that either LpxL from non-pathogenic Yersiniae has a relaxed specificity and can transfer 203 C12 and C14 to the 2 ′ -R-3-hydroxymyristoyl group, or non-pathogenic Yersiniae encode a hitherto 204 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 9 unknown late acyltransferase. Ruling out the latter, analysis of the genomes of non-pathogenic 205 Yersiniae did not reveal the presence of another putative late lipid A acyltransferase. To determine 206 whether LpxL from non-pathogenic Yersiniae has a relaxed specificity, we determined the activity 207 of LpxL in the background of E. coli lpxL mutant, strain BN1- Δ lpxL [30]. This strain produces a 208 penta-acylated lipid A species, m/z 1,641 (Fig 2G) containing two glucosamine residues, two 209 phosphate groups, four 3-OH-C14, and one C14 (added by E. coli LpxM) [30]. The lipid A 210 produced by BN1-Δ lpxL harbouring LpxL from YeO8 contained species m/z 1,824, consistent with 211 the addition of C14, confirming previous findings (Fig S1E). [21]. Notably, BN1- Δ lpxL expressing 212 LpxL from Y. aldovae produced lipid As containing species m/z 1,797, consistent with the addition 213 of C12, and m/z 1,824, consistent with the addition of C14 (Fig 2H). Similar results were observed 214 when lpxL from Y. bercovieri, and Y. mollaretii were expressed in BN1 Δ lpxL (Fig S1F and Fig 215 S1G). Altogether, these results establish that LpxL from non-pathogenic yersiniae esterifies the 216 lipid A 2 ′ -R-3-hydroxymyristoyl group with C12 and C14, demonstrating a relaxed acyl chain 217 length selectivity of lipid A late acyltransferases. 218 Molecular basis of non-pathogenic Yersinia spp LpxL relaxed specificity. 219 To elucidate the molecular basis of the relaxed specificity for C12 and C14 observed in non-220 pathogenic LpxL, we used template-based homology modeling to predict the structural models for 221 LpxL from YeO8, Y. aldovae , and E. coli. The models were based on the crystal structure of 222 Acinetobacter baumannii LpxM (PDB ID 5KN7) [32], which showed up as a relevant hit (E-value 223 of 3e-12) when searching PDB with the YeO8 LpxL sequence as bait. LpxM, LpxL and glycerol-3-224 phosphate acyltransferases (GPAT) all belong to the lysophospholipid acyltransferases (LPLATs) 225 superfamily, and share the GPAT family motifs (cyan in Fig S2). The crystal structure was used to 226 identify the active site (corresponding to the LpxM site with bound n-dodecyl- β -D-maltoside) with 227 an entrance to a deep hydrophobic tunnel (corresponding to the LpxM tunnel with a glycerol bound) 228 in the studied LpxL proteins. Due to its hydrophobic nature, this tunnel was predicted to serve as 229 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 10 the binding site for an acyl chain (Fig 3). At the bottom of the tunnel, YeO8 LpxL has a leucine 230 (L146), which corresponds to a phenylalanine in Y. aldovae (F142) and E. coli (F143) LpxL. This 231 residue likely functions as a hydrocarbon ruler and confers the ability to precisely measure and 232 incorporate hydrocarbon chains of a specific length [33, 34], similarly to residues in enzymes from 233 the GPAT family. Homology models of E. coli , Y. aldovae and YeO8 LpxL with C12 and C14 234 docked to the hydrophobic tunnel predict a stable complex between E. coli LpxL and C12, whereas 235 the binding of C14 is not hampered by the side chain of F143 restricting the tunnel size (Fig 3). 236 Moreover, F143 is unlikely to swing away to make space for C14 due to the more polar 237 environment at the bottom of the tunnel, which is unfavourable for the hydrophobic side chain of 238 F143 (Fig S3). Also Y. aldovae LpxL is predicted to bind C12, with F142 stabilizing the interaction 239 with the acyl chain. However, Y. aldovae LpxL F142 may be able to swing away from the tunnel 240 cavity to provide a stable interaction site also with C14, since the surrounding amino acids L112, 241 I141 and L253 are small hydrophobic residues, which can interact with both F142 and the carbon 242 tail of C14. On the other hand, YeO8 LpxL L146 has a smaller side chain than phenylalanine, 243 thereby facilitating and stabilizing the binding of C14, but being unable to form interactions with 244 the shorter C12 that would place the acyl chain in the correct position for the catalysis to take place. 245 Altogether, these findings suggest that F143 ( E. coli LpxL), F142 ( Y. aldovae LpxL) and L146 246 (YeO8 LpxL) act as hydrocarbon rulers. Furthermore, the alignment of Yersinia LpxL homologs 247 showed a high conservation rate among the residues lining the hydrophobic tunnel, while also 248 highlighting that the hydrocarbon ruler residue is conservatively a phenylalanine in non-pathogenic 249 species and a leucine in pathogenic species (Fig S2). 250 To test these predictions, we mutated the putative hydrocarbon rule of YeO8 LpxL from leucine to 251 phenylalanine and assessed the specificity of the enzyme in vivo. When expressed in BN1- Δ lpxL 252 strain, the lipid A contained species m/z 1,797 and m/z 1,824 (Fig 3E), validating that L146 is the 253 hydrocarbon ruler of YeO8 LpxL and that its replacement by phenylalanine is sufficient to relax the 254 acyl chain specificity of the enzyme. 255 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 11 Molecular dynamics simulations of TLR42/MD-22 with the lipid A PAMP. 256 We next investigated the ability of mammalian LPS receptors to interact with the lipid A PAMP 257 produced by pathogenic and non-pathogenic Yersiniae . Although TLR4 is often considered the 258 receptor for LPS, MD-2, a co-receptor that physically interacts with the extracellular domain of 259 TLR4, is principally responsible for LPS binding [35]. In the inactive state, TLR4 exists as a 260 heterodimer with the co-receptor MD-2. The binding of LPS to TLR4/MD-2 creates a dimerization 261 interface that facilitates the formation of the active complex consisting of a “dimer of dimers” 262 (TLR42/MD-22) [35]. LPS interacts with a large hydrophobic pocket in MD-2 and directly bridges 263 the two components of the multimer. The F126 loop of MD-2 is particularly important in stabilizing 264 the active TLR42/MD-22 complex, undergoing a structural shift upon LPS binding [35]. 265 To gain insights into the potential of the different lipid A molecular species as TLR4/MD-2 266 agonists, we used explicitly solvated atomic-resolution molecular dynamics (MD) simulations to 267 assess the interaction between the lipid A molecular species and the TLR4 2/MD-22 complex, and 268 with the F126 residues within the MD-2 loop encompassing 120 to 129 residues. We performed six 269 sets of independent 1,000 ns simulations, modelled using the X-ray structure of the hexa-acylated 270 LPS-bound-(TLR4/MD-2)2 dimer in its “activated state” (Fig 4A). Each system thus corresponded 271 to the dimeric TLR4/MD-2 complex with two lipid A species, one embedded within each of MD-2 272 monomers. As a control, we tested E. coli lipid A, known to strongly activate the TLR4/MD-2 273 complex and, therefore, a benchmark molecule to determine activation. Of the Yersinia lipid A 274 species, we assessed the hexa-acylated lipid A species m/z 1,797 and m/z 1,824 found in pathogenic 275 and non-pathogenic Yersiniae, the hexa-acylated species m/z 1,768 found only in non-pathogenic 276 Yerisiniae, the hepta-acuylated lipid A species m/z 2,007 found in pathogenic and non-pathogenic 277 Yersiniae, and the tetra-acylated lipid A species m/z 1,388 present only in pathogenic Yersiniae at 278 37oC (Table 1 and Table 2). We reasoned that these lipid As would allow us to determine the effect 279 of lipid A acylation on the interaction with the TLR4/MD-2 complex and to investigate the effect of 280 the decoration with palmitate. Calculation of the root-mean-square deviation (RMSD) of the TLR4 281 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 12 ectodomains relative to their starting structures over the last 200 ns of simulations revealed that all 282 systems studied, apart from the lipid A species m/z 1388, retained the near-experimental “activated 283 state” structure (Fig 4B). Lipid A m/z 1.388 amongst all studied species has only four lipid tails in 284 comparison to six or seven held by other species. Hence, the buried area between the MD-2:lipid A 285 complex and the TLR4 dimer as well as total number of contacts between lipid A and (TLR4/MD-286 2)2 were the lowest in the case of m/z 1388 (Fig 4C and Fig 4D). We next looked at the loop of MD-287 2 containing F126 which is essential in maintaining the active TLR4 2/MD-22 complex and overlaid 288 the F126 side chain every 200 ns using two aligned trajectories composed of MD-2:lipid A. In 289 systems containing E. coli lipid A or m/z 2,007, the F126 side chain retained its orientation facing 290 the hydrophobic lipid tails throughout the entire trajectory for both monomers (Fig 4E). In contrast, 291 lipid A species m/z 1,768, m/z 1,797 and m/z 1,824 retained the F126 side chain orientation in only 292 one of the MD-2 monomers, while m/z 1388 did not show a structurally well-defined side chain 293 conformation (Fig 4E). The loop 120-129 containing this residue was also the least stable in the 294 case of m/z 1388, whereas no differences were found between any of other lipid A species and E. 295 coli (Fig 4F). Altogether, these simulations uncover a reduced interaction of lipid A species m/z 296 1,388 with the TLR4/MD-2 complex, resulting in loss of receptor stability. 297 Ligand-induced conformational changes that bring the C-terminal regions of TLRs together are 298 important for the activation of downstream signalling. We then hypothesized that the loss of 299 receptor stability in m/z 1,388 systems may lead to separation of the membrane-proximal C-terminal 300 regions of TLR4. To test this, we performed principal component analysis (PCA) for the TLR4 C α 301 atoms, in order to filter the “noise” from the trajectory and identify the dominant collective motions 302 of the receptor chains. In simulations of the active TLR4 2/MD-22 complex E. coli lipid A revealed 303 no significant “separating” motions of TLR4 chains, consistent with stabilization of the complex 304 when bound to agonist (Fig 4I). Similar results were obtained testing lipid A species m/z 2,007, m/z 305 1,797 and m/z 1,824 (Fig 4I). In contrast, in the case of m/z 1,388 we detected vertical sliding 306 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 13 motion between the two chains (Fig 4I). This pattern of motion is consistent with a loss of 307 proximity of the membrane-proximal C-terminal regions and inactivation of the receptor complex. 308 Collectively, the MD simulations demonstrate that the hexa-acylated and hepta-acylated lipid A 309 species found in pathogenic and non-pathogenic Yersiniae lead to stabilization of the TLR42/MD-22 310 receptor complex in a similar way as the benchmark E. coli lipid A and, therefore, it is expected that 311 they are agonists of TLR4. On the contrary, the presence of the tetra-acylated lipid A species m/z 312 1,388 found only in pathogenic Yersiniae grown at 37 oC within the active TLR4 2/MD-22 receptor 313 complex leads to destabilization of the complex, supporting the notion that this lipid A species is a 314 poor agonist of TLR4. 315 TLR4 recognition of the lipid A PAMP. 316 We next sought to provide experimental evidence to support the results obtained by MD simulations 317 suggesting differences in the activation of TLR4 between hexa-acylated and hepta-acylated lipid A 318 species on the one hand, and the tetra-acylated ones on the other hand. Ligand-dependent 319 dimerization of TLR4/MD-2 heterodimers results in endocytosis of the active TLR4 2/MD-22 320 complex [36-38]. Therefore, the loss of TLR4 from the plasma membrane represents the most 321 proximal event in the initiation of TLR4 pathway activation. This event can be monitored by flow 322 cytometry and represents a sensitive quantitative assay to assess the interaction of the lipid A 323 PAMP with endogenous PRRs in the natural cellular context (the macrophage surface [36-38]. Live 324 E. coli expressing hexa-acylated lipid A interacts efficiently with TLR4 [36-38]; and, therefore, it 325 was used as a benchmark to delineate the engagement of different Yersinia strains expressing 326 different lipid A PAMPs with TLR4 as we did for the MD simulations. Pathogenic, YeO8 and 327 YeO3 strains grown at 21 oC behaved similarly to E. coli, in that they triggered the loss of TLR4 328 (Fig 5A). Of note, these lipid As present modifications with aminoarabbnose and palmitate (Table 1 329 and Table 2), suggesting that the decorations of the lipid A PAMP had no effect on the 330 internalization of TLR4. These infections were done with the pYV-cured strains to avoid any effect 331 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 14 due to the activation of the YsC T3SS upon contact with macrophages [39-41]. Non-pathogenic 332 strains, Y. aldovae, Y. nurmii and Y. enterocolitica 1A 0902, grown at 21oC also induced the loss of 333 TLR4 from the membrane (Fig 5A). These lipid A s are also decorated with aminoarabinose and 334 palmitate (Table 1 and Table 2), providing further evidence that the decorations do not affect the 335 engagement with TLR4. In sharp contrast, infection with pathogenic strains grown at 37 oC did not 336

in the internalization of the receptor (Fig 5A) whereas non-pathogenic strains engaged with 337 TLR4 (Fig 5A). This is so even in the case of Y. aldovae lipid A that it is decorated with palmitate 338 (Table 1 and Table 2). No significant differences were observed on the engagement with TLR4 339 between non-pathogenic strains grown at 21oC and at 37oC (p > 0.05 for each comparison; Fig 5A). 340 Altogether, this analysis established that the deacylated lipid A PAMP expressed by pathogenic 341 strains at 37 oC limits engagement with TLR4. The data also support the notion that the lipid A 342 decorations with aminoarabinose and palmitate do not affect engagement with TLR4. 343 We next asked whether changes in the lipid A PAMP expressed by pathogenic Yersiniae would be 344 sufficient to affect the interaction with TLR4. To address this question, we constructed a pathogenic 345 chimeric strain expressing the lipid A of non-pathogenic Yersiniae. First, we constructed a double 346 lpxR and lpxL mutant in the YeO8 background, and then we introduced lpxL from Y. aldovae at the 347 single copy in the chromosome using the Tn7 delivery system to generate strain YeO8NPL. We 348 extracted the lipid A from YeO8NPL and analysed its structure by MALDI-TOF. Figure S2 shows 349 that when YeO8NLP was grown at 21 oC (Fig S4A), the lipid A contained species m/z 1,768, m/z 350 1,797 and m/z 1,824 corresponding to hexa-acylated lipid As, and m/z 2,063 corresponding to an 351 hepta-acylated lipid A. Species m/z 1768 and m/z 1,1824 are modified with aminoarabinose, m/z 352 1,899 and m/z 1954. When grown at 37oC, lipid A from YeO8NPL contained species m/z 17668 and 353 m/z 1,797 (Fig S4B). No tetra-acylated lipid A species were found at any temperature. Altogether, 354 these results demonstrate that YeO8NPL expressed the same lipid A as non-pathogenic Yersiniae. 355 Infection with YeO8NPL, grown either at 21 oC or at 37 oC, induced the loss of TLR4 from the 356 membrane (Fig 5B), and the levels were significantly lower than those found in cells infected with 357 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 15 YeO8 (Fig 5B). No differences were observed in TLR4 engagement between YeO8NPL grown 358 either at 21 oC or at 37 oC (p>0.05; Fig 5B). The fact that the levels of TLR4 in cells infected with 359 the lpxR mutant grown either at 21 oC or at 37 oC were significantly lower than those found in cells 360 infected with YeO8 (Fig 5B) further demonstrates the notion that increased acylation of the lipid A 361 PAMP increases the engagement with TLR4. 362 We reasoned the differences in engagement with TLR4 between Yersiniae expressing different lipid 363 A PAMPs should translate to differences in TLR4-dependent inflammatory responses. When we 364 challenged macrophages with bacteria grown at 21 oC, YeO8NPL induced higher levels of TNF α 365 than the wild-type strain (Fig 5C). Similar results were observed when macrophages were 366 challenged with bacteria grown at 37oC (Fig 5C) although the levels were lower than those found in 367 cells infected with bacteria grown at 21 oC (Fig 5). This was dependent on the well-known anti-368 inflammatory action of the pYV-encoded Ysc T3SS [42, 43] because the TNF α levels induced by 369 wild-type bacteria cured of the pYV were higher than those induced by bacteria harbouring the 370 virulence plasmid. This was not the case for YO8NPL grown at 21oC because we did not detect any 371 significant difference in the TNF α levels induced by pYV positive and cured bacteria (Fig 5C). 372 Nevertheless, at 21 and 37 oC YeO8NLP cured of the pYV induced higher TNF α levels than the 373 wild-type strains cured of the plasmid (Fig 5C). We next isolated LPS to evaluate whether purified 374 LPS behaved similarly to whole bacteria. LPS purified from YeO8NPL grown at 37 oC induced 375 more TNFα than LPS isolated from YeO8 and the lpxR mutant (Fig 5D), the levels induced by the 376 latter were higher than those induced by the wild-type LPS (Fig 5D). 377 Altogether, these results demonstrate that the lipid A acylation pattern dictates the engagement with 378 TLR4 and the activation of inflammatory responses. The lipid A PAMP from non-pathogenic 379 bacteria engages efficiently with TLR4 and evokes a higher inflammatory response than that of 380 pathogenic bacteria. Switching the lipid A PAMP expressed by pathogenic bacteria to a non-381 pathogenic PAMP is sufficient to modify the recognition by TLR4 and subsequent activation of 382 inflammation. 383 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 16 Interaction of the lipid A PAMP with antimicrobial peptides. 384 Likewise PRRs, antimicrobial peptides interact with the lipid A PAMP [44, 45]. The decorations of 385 the lipid A counteract this binding and, therefore, they protect Gram-negative pathogens against 386 antimicrobial peptides [46]. The temperature-dependent variations in the modifications of the lipid 387 A from pathogenic and non-pathogenic Yersiniae led us to test the susceptibility to antimicrobial 388 peptides of the Yersinia strain. We and others have used polymyxin B and magainin II to probe the 389 contribution of the modification of the lipid A with aminoarabinose and palmitate, respectively, to 390 bacterial susceptibility to antimicrobial peptides [23, 47-49]. When bacteria were challenged with 391 polymyxin B, we observed that bacteria grown at 21 0C were significantly more resistant than those 392 grown at 37 oC (Fig 6A). These results are consistent with the detection of lipid A modified with 393 aminoarabinose in the lipid As produced by yersiniae grown only at 21 oC. We and others have 394 previously shown that the modification of the lipid A with aminoarabinose mediates resistance to 395 polymyxin B in Yersiniae [23-25]. No significant differences were found between pathogenic and 396 non-pathogenic strains at any growth temperature (Fig 6A). YeO8, YeO3 and YeO9 were also more 397 susceptible to magainin II when grown at 37 oC than at 21 oC (Fig 6B). This is consistent with the 398 lack of lipid A modification with palmitate at 37 oC (Fig 1). Previous work demonstrated that the 399 lipid A modification with palmitate mediates resistance to magainin II [23, 48, 49]. In the case of 400 the non-pathogenic strains, the susceptibility to magainin II differentiated two groups. Those strains 401 more susceptible at 37 oC than 21 oC like the pathogenic strains, and those showing no significant 402 differences in the susceptibility between temperatures (Fig 6B). The latter group includes Y. 403 bercovieri, Y. mollaretii, Y. frederiksenii, Y. intermedia, Y. aldovae, and Y. kristensenii; these strains 404 present the modification of the lipid A with palmitate at both growth temperatures (Table S2). 405 Altogether, these results indicate that the resistance to antimicrobial peptides is not a unique trait of 406 pathogenic bacteria and, therefore, the modifications of the lipid A PAMP leading to resistance 407 cannot be strictly considered a result of the host-pathogen co-evolution to evade their antimicrobial 408 action. 409 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 17 Effect of the lipid A PAMP on the function of Yersinia virulence factors. 410 The ability to switch the lipid A PAMP of pathogenic Yersinia to a non-pathogenic lipid A PAMP 411 offered the unique opportunity to ascertain the effect of the lipid A PAMP from non-pathogenic 412 yersiniae on the expression and function of Yersinia virulence factors. We investigated the flagellar 413 regulon, invasin (Inv), responsible for the for the invasion of the host, and the anti-host Ysc T3SS, 414 encoded in the pYV virulence plasmid. 415 (i) Flagellar regulon and the lipid A PAMP. 416 The flagellar regulon regulates several virulence genes and, therefore, contributes to Y. 417 enterocolitica pathogenesis [50]. Moreover, motility is important for the invasion of cells [51]. In 418 vitro, YeO8 is motile when grown at 21oC but not at 37oC [52, 53]. To examine the influence of the 419 non-pathogenic lipid A PAMP on the flagellar regulon, we first quantified the migration of the 420 strains in motility medium (1% tryptone-0.3% agar plates). YeO8NPL was more motile than the 421 wild-type strain (Fig 7A and 7B). In contrast, and corroborating previous results [20], the lpxR 422 mutant was less motile than the wild-type (Fig 7B), suggesting that the increased acylation of 423 YeO8NPL is not responsible for the enhanced motility. On the other hand, the expression of Y. 424 aldovae lpxL in the background of the YeO8 lpxL mutant was sufficient to increase the motility (Fig 425 7B). YeO8 lpxL did not affect the motility of the lpxL mutant (Fig 7B). We have previously shown 426 that absence of lpxL does not affect the motility of YeO8 [21]. These results indicate that the 427 expression of Y. aldovae lpxL increases the motility of pathogenic yersiniae. 428 Yersinia motility is related to the levels of flagellins which, in turn, are regulated by the expression 429 of flhDC, the flagellum master regulatory operon [52]. We then hypothesized that the expression of 430 flhDC could be higher in YeO8NPL due to the expression of Y. aldovae LpxL. To test this 431 hypothesis, the flhDC::lucFF transcriptional fusion [53] was introduced into the chromosome of the 432 strains and the luciferase activity was determined. Validating our hypothesis, luminescence was 433 higher in the YeO8NPL and YeO8 lpxL mutant complemented with Y. aldovae lpxL than in the wild 434 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 18 type strain (Fig 7C). No differences were observed in the luciferase levels between the wild-type 435 strain and the lpxL mutant (Fig 7C). Further supporting previous results [20], luminiscence was 436 lower in the lpxR mutant than in the wild type, and there were no differences in luminescence 437 between the YeO8, the lpxL mutant, and the complemented strain expression YeO8 lpxL (Fig 7C). 438 To corroborate further that the flagellar regulon is upregulated in YeO8NPL, we next assessed the 439 expression of the phospholipase yplA whose expression is regulated by flhDC [50, 54]. The 440 luciferase activity of the transcriptional fusion yplA::lucFF was higher in the YeO8NPL and YeO8 441 lpxL mutant complemented with Y. aldovae lpxL backgrounds than in the wild type, the lpxL mutant 442 and the mutant complemented with YeO8 lpxL (Fig 7D). No significant differences were observed 443 between the later strains (Fig 7D). In contrast, the luciferase levels were lower in the lpxR mutant 444 than in the wild-type strain (Fig 7D). 445 Altogether, these results demonstrate that the flagellar regulon is upregulated in pathogenic 446 yersiniae expressing the non-pathogenic yersiniae lipid A PAMP triggered by the expression of Y. 447 aldovae lpxL. 448 (ii) Invasin and the lipid A PAMP. 449 Inv is an outer membrane protein responsible for the invasion of the host [55, 56]. We then asked 450 whether the invasion of cells is affected by the lipid A PAMP using a gentamicin protection assay. 451 The number of intracellular bacteria in HeLa cells was 80% lower in cells infected with YeO8NPL 452 and the lpxR mutant (Fig 7E) whereas no differences were found between cells infected with the 453 wild-type strain and the lpxL mutant complemented with Y. aldovae lpxL or YeO8 lpxL (Fig 7E). 454 The differences in invasion are not due to differences in attachment to the cells because the number 455 of bacteria attached to HeLa cells was not significantly different between cells infected with the 456 wild type, YeO8NPL or the lpxR mutant (Fig 7F). These differences in the invasion of cells led us 457 to assess the expression of inv . We constructed an inv::lucFF transcriptional fusion which was 458 introduced into the different strains, and inv expression was monitored as luciferase activity. 459 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 19 Luminescence was lower in the YeO8NPL and the lpxR mutant backgrounds than in the wild-type 460 strains and the lpxL mutant complemented with YeO8 lpxL or Y. aldovae lpxL (Fig 7G), suggesting 461 that the reduced expression of inv mediates the differences in cell invasion between strains. 462 RovA regulates the expression of inv [57, 58]. Therefore, we sought to determine whether the low 463 inv expression correlates with the expression of rovA. A rovA::lucFF transcriptional fusion [21] 464 was introduced in to the strains and the luciferase activity determined. The expression of rovA was 465 also downregulated in YeO8NPL and the lpxR mutant backgrounds (Fig 7H). No differences were 466 found between the wild type and the lpxL mutant complemented with YeO8 lpxL or Y. aldovae lpxL 467 (Fig 7H). Collectively, this evidence establishes that the increase acylation of the lipid A in YeO8 468 expressing the non-pathogenic lipid A PAMP results in a decrease expression of inv and its positive 469 regulator rovA with a concomitant decrease in the invasion of cells. 470 (iii) pYV-encoded virulence factors and the lipid A PAMP. 471 The Ysc T3SS, encoded in the pYV virulence plasmids, is required for Yersinia virulence and it 472 secretes a set of protein effectors called Yops that enable Y. enterocolitica to multiply 473 extracellularly in lymphoid tissues [15]. Analysis of Yop secretion showed no differences between 474 the wild-type, the lpxR mutant, and YeO8NPL (Fig 7I). The disruption of the cytoskeleton upon 475 injection of YopE to cells is one of the most sensitive read-outs to assess the activity of the Ysc 476 T3SS [59]. YeO8, YeO8NPL and the lpxR mutant induced similar disruption and condensation of 477 the actin microfilament structure whereas this was not the case in cells infected with the yopE 478 mutant (Fig S5A and Fig S5B). 479 YadA is another pYV-encoded factor mediating bacterial adhesion, bacterial binding to proteins of 480 the extracellular matrix and complement resistance [60, 61]. We analysed the expression of the 481 YadA trimeric form by SDS-PAGE followed by Coomasie blue staining. Figure 7J shows that 482 YeO8, YeO8NPL and the lpxR mutant produced the same amount of YadA. To investigate YadA 483 functionally, we assessed the binding of YadA expressing bacteria to collagen using as negative 484 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 20 control YeO8 cured of the pYV. We observed no differences in collagen binding between YeO8 485 and YeO8NPL (Fig S5C and Fig S5D) whereas, as expected, the strain cured of the pYV did not 486 bind to collagen (Fig S5C and Fig S5D). We have demonstrated that the lpxR mutant binds to 487 collagen as well as YeO8 [20]. 488 Taken together, these results suggest that expression of the non-pathogenic lipid A PAMP does not 489 affect the production and function of pYV-encoded virulence factors. 490 Effect of the lipid A PAMP on Yersinia virulence. 491 To establish the effect on virulence of switching the lipid A PAMP, BALB/c mice were infected 492 orogastrically, and 2 days postinfection, mice were euthanized, and the. numbers of bacteria present 493 in the Peyer's patches were determined by plating. All strains colonized the Peyer’s patches (Fig 494 7K). However, the loads of the lpxR mutant were seven times lower than those of the wild-type 495 strain (Fig 7K). The double mutant lacking lpxR and lpxL was even more impaired than the lpxR 496 mutant to colonize the Peyer’s patches; the bacterial loads were eleven and two times lower than 497 those of the wild type and lpxR mutant, respectively (Fig 7K). The bacterial loads of YeO8NPL 498 were not significantly different than those of the double mutant (p> 0.05) and six times lower than 499 that of the wild-type strain, demonstrating that a switch from a pathogenic to a non-pathogenic lipid 500 A PAMP is enough to attenuate virulence. 501 502

503 This study was designed to address one of the fundamental questions of the pattern recognition 504 concept: is there a PAMP associated with pathogenicity shaped by the challenge imposed by the 505 innate immune system? By probing the lipid A PAMP, a key molecular pattern expressed by Gram-506 negative bacteria, we drew three conclusions. First, we demonstrated that the reduced acylation of 507 the lipid A PAMP is an evolved adaptation to evade recognition by mammalian LPS receptors, 508 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 21 limiting the activation of inflammatory protective responses and, therefore, contributing to 509 virulence. Second, the decorations of the lipid A PAMP do not represent a unique adaptation to a 510 mammalian ecosystem and are shared by pathogenic and non-pathogenic bacteria. And third, 511 virulence factors cannot overcome a switch in the lipid A PAMP associated with pathogenicity, 512 resulting in decrease virulence. 513 By using MD simulations and a highly sensitive flow cytometry assay to detect endogenous TLR4, 514 we have demonstrated that the reduced acylation of the lipid A PAMP results in reduced interaction 515 with TLR4 with a concomitant decrease in inflammation. Our work provides a conceptual 516 framework to explain why so many pathogens including Pseudomonas aeruginosa, Helicobacter 517 pylori, Francisella turalensis, Porphyromonas gingivalis, as well as members of the microbiome 518 such as Prevotella intermedia and Bacteroides spp, express lipid As with reduced acylation [62-75]. 519 This is also in agreement with evidence demonstrating that the mammalian host environment 520 induces a reduction in the lipid A acylation state not observed in bacteria culture in vitro as it has 521 been demonstrated for Klebsiella pneumoniae [76]. Noteworthy, a common feature shared by all 522 these different lipid As is the comparative lower activation of the innate immune system compare to 523 lipid As similar to the so-called canonical “hexa-acyl lipid A”. This correlates with a body of work 524 probing diverse lipid As generated by bacterial glycoengineering in which those lipid As with 525 reduced acylation induce less inflammation [77, 78]. 526 It is interesting to note that the hepta-acylation of the lipid A found in non-pathogenic strains when 527 grown at 370C, for example Y. aldovae, is not associated with reduced engagement with TLR4 and 528 a decrease inflammation as our MD simulations and experimental evidence establishes. This is in 529 sharp contrast to previous reports suggesting that PagP-controlled hepta-acylation results in limited 530 TLR4 activation [79-82] . However, our results are consistent with observations made probing 531 Acinetobacter baumannii, Pseudomonas aeruginosa, and E. coli expressing hepta-acylated lipid A 532 [78, 83, 84]. Likewise our work, these studies assessed LPS-mediated inflammatory responses 533 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 22 interrogating live bacteria instead of testing purified LPS, highlighting the importance of 534 interrogating the lipid A PAMP in the physiological context of the outer membrane. 535 Our results emphasize the role of the Ysc T3SS to overcome inflammation induced by the lipid A 536 PAMP. However, and despite its efficiency, the pYV-encoded T3SS cannot control the 537 inflammation induced by the non-pathogenic lipid PAMP. Overall, these findings support the 538 notion that the low inflammatory response associated to infections by pathogenic Yersiniae is 539 dependent on the reduced activation of the LPS receptor complex induced by the pathogenic lipid 540 PAMP couple to the anti-inflammatory action of the Ysc T3SS. 541 There is a wealth of evidence demonstrating that the decorations of the lipid A PAMP contribute to 542 the virulence of Gram-negative pathogens [46]. Our results do not challenge these findings; 543 however, they demonstrate that this is not a unique trait of pathogenic bacteria. The decorations of 544 the lipid A fortify the outer membrane, acting as a defence mechanism against antimicrobial 545 peptides and other toxins [46]. Undoubtedly, these lipid A structural changes are also useful for 546 bacteria living in the environment where they could be also exposed to these agents produced by 547 other microbes such as Streptomycetes , plants, and aquatic and marine organisms, explaining why 548 the lipid A decorations are not unique to bacteria colonizing or infecting mammalian hosts. In the 549 Yersinia context, it was previously believed that the temperature-dependent expression of these 550 lipid A decorations represented an adaptation to the environment found in the mammalian gut and 551 lung mucosae [23, 24]. Here, we have shown that this is not the case because non-pathogenic 552 species also display this trait. In pathogenic strains, the temperature-dependent regulation of the loci 553 controlling lipid A modifications is explained by H-NS-dependent negative regulation alleviated by 554 RovA [23]. Interestingly, non-pathogenic Yersiniae do not encode rovA; however, they do encode 555 slyA which has been shown to alleviate the negative regulation exerted by H-NS in other Gram-556 negative pathogens [85-87]. We posit the SlyA-H-NS circuit controls the expression of the loci 557 controlling the lipid A decorations in non-pathogenic Yersiniae. Future studies are warranted to 558 validate this notion. 559 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 23 One striking finding of our study is the dramatic effect on virulence of switching the lipid A PAMP. 560 Increasing the acylation pattern, as occurs in the lpxR mutant, already resulted in an attenuation of 561 virulence. Remarkably, expressing the non-pathogenic lipid A PAMP triggered an even further 562 attenuation. This finding cannot be explained by lack of function of the Ysc T3SS because we did 563 not observe any differences in the production and function of the pYV-encoded Yops and YadA 564 between the wild-type strain, YeO8NPL and the lpxR mutant. At least two non-exclusive 565 explanations may account for the low-level colonization of Peyer's patches: (i) switching the lipid A 566 PAMP results in a decrease penetration of the intestinal epithelium, and (ii) YeO8NPL is cleared by 567 a host defence mechanism present in the tissue. In support of the former, inv expression was lower 568 in YeO8NPL than in the wild type and the mutant was impaired in its ability to invade epithelial 569 cells. However, the reduced inv expression cannot explain fully the low bacterial loads in the 570 Peyer’s patches of mice infected with YeO8NPL because no differences in inv expression were 571 found between the lpxR mutant and YeO8NPL. Supporting the latter, the heightened inflammation 572 triggered by YeO8NPL upon TLR4 recognition of the lipid A PAMP will result undoubtedly in 573 vivo in the activation of host protective proinflammatory responses. This agrees with previous 574 studies illustrating the crucial role of TLR4 signalling to control Yersinia infections [88-90]. 575 576

577 Ethics. 578 Balb/c mice were purchased from Charles River. Mice were age and sex-matched and used 579 between 8-12 weeks of age. The experiments involving mice were approved by the Queen’s 580 University Belfast’s Ethics Committee and conducted in accordance with the UK Home Office 581 regulations (project licence PPL2910) issued by the UK Home Office. Animals were randomized 582 for interventions but researches processing the samples and analysing the data were aware which 583 intervention group corresponded to which cohort of animals. 584 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 24 Bacteria strains and growth conditions. 585 Bacterial strains are described in Table S2. Bacteria were grown overnight in 5 mL of Miller’s 586 Luria Broth (LB) medium (Melford) at 37°C or 21°C on an orbital shaker (180 rpm). Overnight 587 bacterial cultures were refreshed 1/10 into a new tube containing 4.5 mL of fresh LB. After 3 h at 588 37°C, bacteria were pelleted (2,500x g, 20 min, 22°C), and resuspended in PBS to an OD 540 of 0.8 589 (corresponding to 4x10 8 CFUs/ml). Where appropriate, antibiotics were added to the growth 590 medium at the following concentration: ampicillin (Amp), 50 μ g/ml for E. coli and Yersinia spp, 591 and 100 μ g/ml for Yersinia spp in agar plates; kanamycin (Km), 50 μ g/ml for E. coli and Yersinia 592 spp, and 100 μ g/ml for Yersinia spp in agar plates; gentamycin (Gm), 100 μ g/ml; chloramphenicol 593 (Cm), 25 μ g/ml; tetracycline (Tet), 12.5 μ g/ml; streptomycin (Str), 100 μ g/ml; trimethroprim 594 (Tmp), 100 μ g/ml. 595 To cure the pYV plasmid from pathogenic Yersinia, bacteria were grown at 37°C in Congo Red 596 Magnesium oxalate agar plates [91]. Colony size and lack of uptake of Congo Red were used to 597 detect loss of the virulence plasmid. This was further confirmed by testing the YadA-dependent 598 autoagglutination ability [92]. 599 Isolation of lipid A and MALDI-TOF analysis. 600 Bacterial cultures were grown overnight at 21 ºC or 37 ºC in a 5 ml of LB. On the next day, the 601 cultures were used to inoculate 1 litre Erlenmeyer containing 250 ml of LB broth (cultured at the 602 same temperature and agitation). After 24 hours, the bacterial cells were harvested by centrifugation 603 (2,000 x g, 15 minutes, room temperature), washed twice in phosphate buffer (PB, 1.5 g of 604 Na2HPO4, 0.2 g of KH2PO4). The final pellet was resuspended in 10 ml of PB and placed at -80ºC. 605 The cell pellet was lyophilized. 606 The lipid A extraction was carried out by an ammonium hydroxide / isobutyric acid method [21, 607 93]. Briefly, 10 mg of the lyophilized bacterial was resuspended in 400 μ l of a solution of butiric 608 acid (98% pure Sigma-Aldrich) / 1M ammonium hydroxide (5:3, v/v), and incubated during 2 hours 609 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 25 at 100ºC with vortexing every 15 minutes. Then, the hydrolysed product was cooled down (5 610 minutes incubation in ice), and centrifuged (2,000 x g, 15 minutes, room temperature). The 611 supernatant was then transferred to a new microcentrifuge tube and the same volume of water was 612 added (1:1, v/v). Next, the samples were frozen at -80ºC and lyophilized. After, the samples were 613 washed in 400 μ l of methanol twice and centrifuged (2,000 x g, 15 minutes, room temperature). The 614 final pellet was solubilized in a mixture of chlorophorm/methanol/water (3:1.5:0.25, v:v:v), as a 615 final concentration 1mg/ml. Each sample was treated with traces of the ion-exchange resin (Dowex 616 50W-X8; H+). 1 μ l of the lipid A suspension were spotted on a MALDI-TOF target, dried (5 min at 617 room temp), and 1 μ l of matrix was added on top. The 2,5-Dihydroxybenzoic acid (Bruker) was 618 used as a matrix dissolved in (1:2) acetonitrile-0.1% trifluoroacetic acid (Following manufacturer´s 619 instructions). The lipid A analysis was performed by mass spectrometry using a Bruker autoflex 620 speed TOF/TOF mass spectrometer (Bruker Daltonics Inc.) in negative reflective mode. Each 621 spectrum was obtained as an average of at least 1000 shots (ion-accelerating voltage set at 40kV). 622 Two controls were used for calibration: Peptide Calibration Standard II (Bruker), a mixture of 623 known peptide sizes, and lipid A extracted from the E.coli strain MG1655 grown in LB at 37 ºC. 624 The resulting spectra interpretation was based on previous studies showing that a mass/charge (m/z) 625 above 1000 is proportional to the corresponding lipid A specie. 626 Construction Y. aldovae strain encoding lpxR. 627 We used the Tn7 single copy-based system [28]. The Tn 7 transposon integrates at the site-specific 628 attTn7, located downstream of the conserved glmS gene, which encodes an essential glucosamine-629 fructose-6-phosphate aminotransferase [28]. lpxR from YeO8 was amplified by PCR using the 630 primers listed in Table S3 and the following cycling conditions: 30 x cycles of: 95ºC for 45 sec, 631 56ºC for 45 sec, 72ºC for 1.30 min, using 50 ng of genomic DNA, 0.2 mM dNTPs, 10 μ M of 632 primers, and 0.2 μ l of Pfu DNA polymerase in a final volume of 50 μ l). The PCR product was gel-633 purified (Qiagen Gel extraction kit), phosphorylated, and cloned into PvuII-digested, Antarctic 634 phosphatase treated, and gel-purified, pUC18R6KTmini-Tn7TKm to give pUC18R6KTmini-635 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 26 Tn7TKmLpxR. This vector was mobilized into Y. aldovae by triparentral conjugation with the help 636 of a conjugative E. coli β 2613 carrying the plasmid pTNS2, encoding the Tn7 transposase. The 637 colonies grown were thereafter screened for resistance to kanamycin and susceptibility to 638 ampicillin. Correct integration of the Tn7 transposon was confirmed by PCR using the primers 639 YeO8_GlmS_Tn7comF and Tn7com_R, and Tn7com_F, YeO8_PstS_Tn7comR (Table S3). 640 Additionally, the presence of the lpxR gene was PCR-confirmed using YeO8_lpxR_F, 641 YeO8_lpxR_R primers. 642 Complementation of E. coli strains. 643 E. coli BN1 is an lpxT, eptA and pagP mutant, producing hexa-acylated lipid A [31]. We previously 644 constructed a lpxL mutant of BN1 [30]. This strain was complemented with lpxL from YeO8, Y. 645 aldovae, Y. mollaretti , and Y. bercovieri . The genes were amplified by PCR and cloned into 646 pGEMT-Easy to give pGEMT lpxLYeO8, pGEMT lpxLY. aldovae , pGEMT lpxLY. mollaretti , and 647 pGEMTlpxLY. bercovieri. The plasmids were introduced into E. coli BN1 Δ lpxL by electroporation. E. 648 coli cells were made competent utilizing a microcentrifuge-based procedure (**Choi 2006). 649 E. coli BN2 is an lpxT, eptA , pagP, lpxM mutant producing penta-acylated lipid A [31]. This strain 650 was complemented with lpxM from Y. aldovae, Y. intermedia , and Y. bercovieri. The genes were 651 amplified by PCR and cloned into pGEMT-Easy to give pGEMT lpxMY. aldovae , pGEMT lpxMY. 652 intermedia, and pGEMT lpxMY. bercovieri . The plasmids were introduced into E. coli BN2 by 653 electroporation. 654 Complemention of YeO8 lpxP mutant. 655 YeO8-Δ lpxPGB mutant [21] was complemented using a Tn7-based system. lpxP genes from Y. 656 aldovae, Y. bercovieri and Y. intermedia were amplified by PCR, gel-purified (Qiagen Gel 657 extraction kit), phosphorylated, and cloned into PvuII-digested, Antarctic phosphatase-treated, and 658 gel-purified, pUC18R6KTmini-Tn7TCm [28] to give pUC18R6KTmini-Tn7TCm lpxPY. aldovae , 659 pUC18R6KTmini-Tn7TCmlpxPY. bercovieri , and pUC18R6KTmini-Tn7TCm lpxPY. intermedia . These 660 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 27 vectors were mobilized into the YeO8 lpxP mutant by triparental conjugation, and the correct 661 integration of the Tn7 transposon at the attTn7 site confirmed by PCR. The presence of the 662 complementing genes was PCR-confirmed. 663 Site-directed mutagenesis of YeO8 lpxL 664 The site-directed mutagenesis of the lpxL gene from YeO8 was performed by PCR, as previously 665 described [94] using as template plasmid pGEMT lpxLYeO8. For that, the primers pairs (Table S3) 666 were designed with a mismatch in the amino acid leucine, position L146. The obtained PCR 667 products were gel purified, phosphorylated with T4 polynucleotide kinase, ligated, and digested 668 with DpnI to break down any remaining template plasmid. The ligated PCR-product was 669 transformed into chemically competent E. coli C600 to obtain pGEMTlpxLYeO8Leu146. Plasmid DNA 670 was isolated from transformants and the lpxL gene was sequenced to confirm the generated 671 mutations and to ensure that no other changes were introduced. pGEMTlpxL YeO8Leu146 was 672 introduced into E. coli BN1-Δ lpxL by electroporation. 673 Construction YeO8NPL, YeO8 strain producing a non-pathogenic lipid A. 674 To construct a YeO8 strain producing the lipid A of non-pathogenic yersiniae, we first constructed 675 a double lpxR-lpxL mutant which was then complemented with Y. aldovae lpxL. 676 To construct a lpxL mutant, two sets of primers were used to amplify 800-900 bp regions flanking 677 the lpxL gene. Gel-purified LpxLUp and LpxLDOWN fragments were annealed at their overlapping 678 BamHI sequence in a PCR machine (8 cycles of 95ºC for 45 sec, 55ºC for 90 sec, 72ºC for 90 sec). 679 The reaction included Takara Ex Taq DNA Polymerase, and 0.2 mM dNTPs. The annealed 680 fragment was amplified by PCR using primers YeO8LpxLF1 and YeO8DOWNR1, gel-purified and 681 cloned into pGEMT-Easy to obtain pGEMTΔ lpxL. A kanamycin resistance cassette flanked by FRT 682 recombination sites was obtained as a BamHI fragment from pGEMTFRTKm [49] and it was 683 cloned into BamHI-digested pGEMT Δ lpxL to generate pGEMT Δ lpxLKm. The 3.5 kb lpxL::Km 684 allele was obtained by NotI restriction digestion of pGEMT Δ lpxLKm, gel-purified and cloned into 685 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 28 NotI-digested, Antarctic phosphatase-treated pJTOOL1 [95] to obtain pJTOOL Δ lpxLKm. 686 pJTOOL1 is a suicide vector that carries the defective pir-negative origin of replication of R6K, the 687 RK2 origin of transfer, and an Cm resistance marker [95]. It also carries the sacBR genes that 688 mediate sucrose sensitivity as a positive selection marker for the excision of the vector after double 689 crossover [95]. pJTOOL Δ lpxLKm was introduced into the diaminopimelate (DAP) auxotrophic 690 E. coli donor strain β 2163 [96] and mobilised into the lpxR mutant of YeO8 [20] via conjugation. 691 Bacteria were diluted and aliquots spread on Yersinia selective agar medium plates (Oxoid) 692 supplemented with Cm. Bacteria from 5 individual colonies were pooled and allowed to grow in LB 693 without any antibiotic overnight at RT. Bacterial cultures were serially diluted and aliquots spread 694 in LB without NaCl containing 10% sucrose and plates were incubated at room temperature. The 695 recombinants that survived 10% sucrose were checked for their antibiotic resistance. The 696 appropriate replacement of the wild-type alleles by the mutant ones was confirmed by PCR. The 697 kanamycin cassette was excised by Flp-mediated recombination using plasmid pFLP2Tp [20] [97]. 698 The generated mutant was named YeO8-Δ lpxRΔ lpxL. 699 Y. aldovae lpxL was amplified by PCR, gel-purified, phosphorylated and cloned into PvuII-700 digested, Antarctic phosphatase treated, and gel-purified, pUC18R6KTmini-Tn7TKm to give 701 pUC18R6KTmini-Tn7TKmlpxLY. aldovae . The plasmid was mobilized into YeO8- Δ lpxRΔ lpxL by 702 triparental mating to obtain YeO8NPL. The correct integration of the Tn7 transposon was assessed 703 by PCR. 704 Autoagglutination analysis. 705 YadA is a pYV-encoded outer membrane protein mediating agglutination when bacteria are grown 706 in minimal medium [92]. Therefore, the presence or absence of pYV in pathogenic Yersinia can be 707 determined using an autoagglutination test. 1 ml of RPMI1640 media without pH indicator (Gibco) 708 was inoculated with half colony taken from a fresh LB agar plate grown at room temperature for 48 709 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 29 h. The culture was grown at 37 ºC, without agitation for 20 h, and the autoagglutination was 710 visually determined. 711 Molecular dynamics simulations 712 All-atom simulations were performed using the GROMACS 2018.3 simulation package [98] 713 utilizing the CHARMM36m force field with the TIP3P water model [99]. The experimental 714 structure (PDB: 3FXI4) of the (TLR4:MD-2)2 heterotetramer bound to E. coli ReLPS was used. 715 Protein charges were assigned according to neutral pH with charged termini. Lipid A structures and 716 topologies: i) m/z 1768, ii) m/z 2007, iii) m/z 1797, iv) m/z 1388 and v) m/z 1824 were built using 717 CHARMM-GUI ligand builder & modeler [100], while vi) E. coli lipid A was built using 718 CHARMM-GUI membrane builder. In total, six simulation systems were setup of (TLR4:MD-2) 719 complexes bound to a given lipid A type. Each lipid molecule headgroup was aligned to the 720 experimentally solved ReLPS lipid A headgroup. Each dimeric construct was placed in a 721 dodecahedron box with ~16 nm box edge. Approximately 90,000 TIP3P water molecules were 722 added to the box and 150 mM NaCl salt whilst neutralizing the overall system charge. Energy 723 minimization was performed using the steepest descent minimization algorithm with a 0.01 nm 724 energy step size. The system was equilibrated in a step wise fashion: i) 1 ns with position restraints 725 on protein alpha carbons and lipid headgroups in the NVT ensemble; ii) 20 ns with position 726 restraints on protein alpha carbons and lipid headgroups in the NPT ensemble; iii) 50 ns with 727 position restraints on protein alpha carbons in the NPT ensemble. All position-restrained 728 simulations were run with a force constant of 1,000 kJ mol-1 nm-2. Unrestrained production runs 729 were set to 1000 ns in the NPT ensemble. A temperature of 310 K was maintained using the 730 velocity rescaling thermostat with an additional stochastic term using a time constant of 1 ps. 731 Pressure was maintained semi-isotropically at 1 atm using the Parrinello-Rahman barostat [101] and 732 a time constant of 5 ps. All bonds which involved hydrogens were constrained using the LINCS 733 algorithm [102]. Equations of motion were integrated using the leap-frog algorithm with a time step 734 of 2 fs. Long-range electrostatic interactions were described using the particle mesh Ewald method 735 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 30 (PME). The short-range electrostatics reals pace cut-off was 1.2 nm and the short-range van der 736 Waals cut-off was also 1.2 nm. Periodic boundaries conditions were applied in all directions. 737 Simulations were performed on: i) an in-house Linux cluster composed of 8 nodes containing 2 738 GPUs (Nvidia GeForce RTX 2080 Ti) and 24 CPUs (Intel® Xeon® Gold 5118 CPU @ 2.3 GHz) 739 each as well as on ii) the National Supercomputing Center (https://www.nscc.sg) using 4 nodes 740 containing 128 cores and 256 logical cores (AMD EPYC 7713 @ 2.0 GHz) each. 741 All simulations snapshots were generated using VMD [103]. The root mean square deviation 742 (RMSD) of the TLR4 dimer backbone was calculated with respect to the experimental structure 743 after alignment onto the same set of atoms. The RMSD of the loop encompassing backbone atoms 744 of residues 120-129 was calculated with respect to the experimental structure after alignment onto 745 each MD-2 monomer backbone. The (TLR4:MD-2) 2-lipid A contacts were calculated based on a 746 0.4 nm cutoff distance. The buried solvent accessible surface area (SASA) between two MD-2-lipid 747 and TLR4 dimer was calculated as a sum of MD2-lipid and TLR4 SASA before subtracting 748 (TLR4:MD-2)2–lipid SASA and dividing by two. The RMSDs, protein-lipid hydrogen bonds, 749 protein-lipid contacts and SASA values were calculated as a mean over two monomers and 750 averaged over last 200 ns of trajectory. Principal component analysis was performed on TLR4 751 dimer alpha carbons. Porcupine plots were generated using extreme conformation of the most 752 dominant principal component (PC1) which corresponded to ~30% of the total variance captured. 753 Modelling LpxL-lipid interactions. 754 YeO8 LpxL sequence was used as bait with the Basic Local Alignment Search Tool (BLAST) at 755 NCBI ( http://blast.ncbi.nlm.nih.gov/) to search the non-redundant protein sequence database for 756 similar sequences in the Yersinia genus, as well as in E. coli (YeO8: A1JN09; Y. enterocolitica O3: 757 gi|491299878; Y. enterocolitica O9: gi|325666359; Y. enterocolitica 1A: gi|571261928; Y. 758 bercovieri: gi|238715289; Y. mollaretii : gi|238720151; Y. frederiksenii : gi|238722579; Y. 759 intermedia: gi|238728988; Y. aldovae : gi|238704192; Y. kristensenii : gi|238699514; Y. ruckerii : 760 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 31 gi|238707399; Y. rohdei : gi|238712349; Y. similis : gi|588288876; Y. nurmii : gi|902505539; Y. 761 pekkanenii: gi|902541928; Y. massiliensis : gi|518039378; E. coli : P0ACV0). Protein Data Bank 762 (PBD) was searched for crystal structures with bound ligands, which could serve as template for 763 modeling. The only significant hit was LpxM from A. baumannii in complex with n-dodecyl- β -D-764 maltoside and glycerol at 1.99 Å resolution and an E-value of 3e-12 (PDB ID 5KN7) [104], which 765 belongs to the same LPLATs superfamily as LpxL. Since we focused on the conserved hydrophobic 766 lipid binding site of LpxL and the crystal structures deposited in the PDB in general agree much 767 more closely with experimental data where the predicted AlphaFold models and deposited 768 structures differ [105], we did not use ligand free AlphaFold models in this study aiming to predict 769 C12 and C14 binding to LpxL. MALIGN [106] in the BODIL modeling environment [107] was 770 used to align the LpxL sequences from the Yersinia genus and E.coli, after which the LpxM 771 sequence was added to the alignment using prealigned sequences. For modeling, all sequences 772 except YeO8 LpxL, Y. aldovae LpxL or E. coli LpxL and the A. baumannii LpxM template were 773 deleted from the alignment. A set of 10 models was created with MODELLER [108] and the model 774 with the lowest value of the MODELLER objective function was analyzed and compared with the 775 crystal structure of LpxM. The rotamers of the hydrophobic tunnel were chosen as close to the 776 rotamers in LpxM as possible using Maestro. C12 and C14 were sketched in Maestro and 777 minimized and prepared for docking with LigPrep and force field OPLS4. Different ionization was 778 allowed within a pH range of 7.0 ± 2.0. All LpxL models were prepared with the Schrödinger 779 protein preparation wizard and minimized using the OPLS4 force field, before generating a grid 780 centered around the residues in the hydrophobic tunnel. The minimized and prepared acyl chains 781 were docked with the Glide program with standard precision (SP) mode, resulting in 10 poses per 782 prepared acyl chain and protein model. The resulting poses were ranked based on the Glide emodel 783 score and the five best scored complexes of each prepared acyl chain and protein model were 784 analyzed visually. The complex with the best pose was further minimized with OPLS4 force field, 785 using water as solvent, method PRCG and maximum iterations 2500. PyMOL was used to prepare 786 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 32 pictures of the final complex models and ESPript ( https://espript.ibcp.fr) [109] to prepare the 787 alignment picture. 788 Cell culture. 789 Immortalised BMDM (iBMDM) cells (BEI Resources, NIAID, NIH: Macrophage Cell Line 790 Derived from Wild Type Mice, NR /i6 9456) were grown in Dulbecco's modified Eagle's medium 791 (DMEM; Gibco 41965) supplemented with 10% heat /i6 inactivated foetal calf serum (FCS), 792 100 U/ml penicillin and 0.1 mg/ml streptomycin (Gibco) at 37°C in a humidified 5% CO 2 793 incubator. 794 Carcinoma human alveolar basal epithelial cells (A549, ATCC CCL-185), and human carcinoma of 795 Henrietta Lacks (HeLa, ATCC CCL-2), were maintained in RPMI 1640 (Gibco) tissue culture 796 medium supplemented with 1% v/v HEPES, 10% v/v heat inactivated fetal bovine serum (FBS) and 797 1% v/v antibiotics (penicillin and streptomycin) at 37 ºC, in a humidified 5% CO2 incubator. 798 Cells were routinely tested for Mycoplasma contamination. 799 Cell surface expression of TLR4 by flow cytometry. 800 The cell surface expression of TLR4 was determined as previously described [13]. Overnight 801 cultures of E. coli MG1655, grown at 37oC in LB, and Yersinia strains. grown in LB at 21 oC, in an 802 orbital shaker (180 rpm) were diluted 1:10 in LB medium and grown at 21 oC or 37 oC, Yersinia 803 strains, or at 37 oC, E. coli, for 3.5 h. Bacteria were collected by centrifugation (2500×g, 20 min, 804 24°C) and resuspended to an OD 600 of 1.0 in PBS (approximately 5 x 10 8 CFU/ml). 0.4x10 6 805 iBMDMs were seeded in complete DMEM 24-hours prior to lifting with PBS/EDTA and 806 resuspension in complete DMEM. The following day, 0.5x10 6 iBMDMs were lifted and 807 resuspended in 5 ml tubes in 2 ml of complete DMEM and infected with live bacteria at a 808 multiplicity of infection of 50 bacteria per cell. Following a 20-minute incubation at 37°C, 5% 809 CO2, tubes were placed on ice, centrifuged at 400 x g (4°C), washed with 2 ml of cold PBS, 810 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 33 resuspended in 2 ml of ice-cold PBS and stained with antibodies to TLR4 (Biolegend, 117606, 811 1:200 dilution) for 20 min at 4°C. Cells were then washed with 2 ml ice-cold PBS and centrifuged 812 400 xg for 5 min, before resuspension in 200 μ l PBS. TLR4 monomer surface expression was 813 measured using PE laser channel on the FACs Canto II and analysed using FlowJo version 10.7.2. 814 Data analysis was performed using Prism. 815 LPS purification 816 LPSs from Yersinia strains were extracted and purified as previously described using the hot 817 phenol-water method [110]. Bacteria were cultured overnight in 5 ml LB at 21 oC. Then, the entire 818 culture was used to inoculate a 2 l flask, containing 1 l of LB broth, which was incubated for 24 819 hours in agitation at 37 ºC. On the next day, an aliquot of the cultures was plated onto LB agar to 820 check for any contamination. Then, cells were collected by centrifugation (6,000 x g, for 20 minutes 821 at 4ºC). The bacterial pellet was washed twice in PBS, weight and resuspended in deionized water 822 (250 ml of water per 100 g of pellet). Then, the same volume of 65ºC pre-warmed liquid phenol 823 was added, and stirred at 65ºC in a water bath for 20 minutes. The mixture was then incubated 824 overnight at 4ºC following a rapid cool down on ice. Later, samples were centrifuged using a high-825 speed centrifuge and special glass tubes (30 ml COREX tubes). The aqueous upper phases were 826 combined and transferred into a new 250 ml glass flask. 5 volumes of methanol were added per 827 volume of sample (5:1), and sodium acetate-saturated methanol to a final concentration of 1 % 828 (v/v). Sample was incubated at -20ºC overnight. On the next day, samples were centrifuged (6,000 829 x g, 15 minutes, 4ºC). The pellets were resuspended in 20 ml of water, transferred to a dialysis tube 830 and incubated in a tank containing deionized water during 2 days at 4ºC, changing the water after 831 24 h. The content of the dialysis tubes was transferred to a glass flask and lyophilized. 832 Crude LPS preparations were dispersed (10 mg/ml) in 0.8% NaCl-0.05% NaN3-0.1 M Tris-HCl (pH 833 7), digested with nucleases (50 μ g/ml; 1 h 37 ºC during with agitation) and proteinase K (50 μ g/ml; 834 3 hours at 55ºC and overnight at room temperature). The proteinase K treatment was repeated twice. 835 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 34 LPSs were sedimented by ultracentrifugation (6 h, 100,000 × g), and freeze-dried. Protein 836 contamination was tested using a BCA assay following the manufacturer’s recommendations. 837 (Thermofisher). LPSs were repurified using phenol reextraction in the presence of deoxycholate to 838 eliminate lipoprotein contaminants [111]. 839 Macrophage culture and TNFα ELISA. 840 iBMDMs were seeded 24 h before infection in 24-well plate at a density of 0.5 x 10 8 cells per well 841 in complete medium. Before infection, cells were washed with PBS and suspended in complete 842 medium without antibiotics. Bacteria were grown in 5 /i6 ml LB at 37°C or 21 oC, harvested at 843 exponential phase (2,500/i6 ×/i6 g, 20/i6 min), and adjusted to an OD 600 of 1.0 in PBS. Infections were 844 performed using a multiplicity of infection of 25 bacteria per cell. Infections were performed in the 845 same medium used to maintain the cell line without antibiotics and incubated at 37°C in a 846 humidified 5% CO 2 incubator. Cells were centrifuged (200 x g, 5 min) to synchronize infection. 847 After 30 min, cells were washed with PBS and incubated in 100 μ g/ml of Gm in complete medium 848 without antibiotics for 3 hours. Supernatants from infected cells were collected and spun down at 849 12,000/i6 ×/i6 g for 5/i6 min to remove any debris and stored at -80 oC. TNFα in the supernatants was 850 determined using a Murine TNF- α standard 3,3 ′ ,5,5′ -tetramethylbenzidine (TMB) enzyme-linked 851 immunosorbent assay (ELISA) development kit (PeproTech, catalog number 900-T54). 852 To challenged iBMDMs with LPSs, cells were seeded the day before treatment in 96-well plates at 853 a density of 0.5x10 5 cells per well in complete medium. The day of the experiment, cells were 854 washed with PBS and resuspended in complete medium. Cels were challenged with 10 ng/ml of 855 repurified LPSs in complete medium with antibiotics. After 1 h, supernatants were collected, spun 856 down, and stored at -80 oC. TNF α was determined using an ELISA (PeproTech, catalog number 857 900-T54). 858 Antimicrobial peptides susceptibility. 859 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 35 To assay Yersinia strains for resistance to polymyxin B and magainin II, we used a modified 860 version of the sensitivity assay described by Llobet et al [49]. Briefly, each strain was grown in LB 861 until an OD600 of 0.3, washed once in PBS and diluted in liquid testing media (1% v/v Tryptone soy 862 broth, 10% v/v 100 mM phosphate buffer [pH 6.5], 2% v/v 5 M NaCl) to an approximate 863 concentration of 4 × 10 4 CFU/ml. Twenty /i6 five microlitres of each diluted strain was then mixed 864 with 5 μ l of polymyxin B (10 μ g/ml) or magainin II (10 μ g/ml) and incubated at 37°C for 1 h. 865 Fifteen microlitres of the suspension was thereafter spread onto LB agar and incubated overnight at 866 21°C or 37 oC. Colony counts were determined and results were expressed as percentages of the 867 colony count of bacteria not exposed to antibacterial agents. All experiments were done with 868 duplicate samples on at least three independent occasions. 869 Reporter strains. 870 Two reporter plasmids were constructed to assess the expressions of inv, and ylpA . The promoter 871 regions of inv , and ylpA were amplified by PCR using Phusion High-Fidelity DNA Polymerase 872 (NEB), YeO8 genomic DNA as template and the primers shown in Table S3, which contain an 873 EcoRI site at their 5’. The PCR products were gel-purified, digested with EcoRI and cloned into 874 EcoRI-SmaI-digested, Antarctic Phosphatase-treated pGPL01Tp suicide vector [23] and then 875 transformed into E. coli GT115 cells to obtain pGPLTp inv, and pGPLTpylpA. pGPL01Tp contains 876 a promoterless firefly luciferase lucFF gene and an R6K origin of replication and can be mobilized 877 by conjugation [23]. Correct insertion of the amplicons was verified by restriction digestions with 878 EcoRI and SmaI. Vectors were introduced into E. coli β 2163, and then mobilised into Y. 879 enterocolitca strains by conjugation. Cultures were then serially diluted and checked for Tmp 880 resistance by plating on LB Tmp agar at 21°C. Correct insertion of the vectors into the chromosome 881 was confirmed by PCR using the relevant lucFF_check and promoter sequence primers (Table S3). 882 Luciferase assays. 883 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 36 Overnight cultures of the reporter strain in LB at 210C, were refreshed 1:10 and grown for 3 h in LB 884 at 21°C in agitation (180 rpm). The cells were then pelleted, washed once in sterile PB buffer (1M 885 disodium hydrogen orthophosphate, 1.5 mM potassium dihydrogen orthophosphate, pH 7.0) and 886 adjusted to an OD 600 of 1.0. One hundred microlitres of each suspension were added to an equal 887 volume of luciferase assay reagent (1 mM d /i6 luciferin [Synchem] in 100 mM sodium citrate buffer 888 pH 5.0), vortexed for 5 s and then immediately measured for luminescence (expressed as relative 889 light units [RLU]) using a GloMax 20/20 Luminometer (Promega). All strains were tested in 890 quadruplicate from three independent cultures. 891 Motility assay. 892 Phenotypic assays for swimming motility were initiated by stabbing 2 µl of an overnight culture at 893 the centre of agar plates containing 0.3% agar and 1% tryptone [52]. Plates were analyzed after 24 h 894 of incubation at 22 oC and the diameters of the halos migrated by the strain from the inoculation 895 point were compared. Experiments were run in triplicate in three independent occasions. 896 Adhesion and invasion of epithelial cells. 897 Overnight cultures of bacteria in LB at 21oC, were refreshed 1:10 in 5-ml LB and grown at 21 oC 898 for 3 h. Bacteria were collected by centrifugation (3,220 x g, 20 min at room temperature) and 899 resuspended in PBS to an OD 600 of 1.0. HeLa cells were seeded in 24-well plates at a density of 900 5/i6 ×/i6 104 in complete RPMI medium containing antibiotics. 2 days after seeding, cells were starved 901 for 16 h before infection using RPMI 1640 medium (Gibco 21875) supplemented only with 902 10/i6 mM HEPES. Before infection cells were washed once with PBS and infected with a 903 multiplicity of infection of 25 to 1. To determine the adhesion of Yersinia to cells, after 30 min 904 infection, monolayers were washed seven times with PBS, and cells were lysed in 300 μ l 0.5% 905 saponin in PBS. The lysates were serilally diluted in PBS and plated onto LB agar plates incubated 906 at 21oC. To determine the invasion of cells, after a 30 min infection, monolayers were washed twice 907 with PBS and then incubated for an additional 90 min in medium containing Gm (100 µg/ml) to kill 908 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 37 extracellular bacteria. This treatment was long enough to kill all extracellular bacteria. After this 909 period, cells were washed three times with PBS and lysed for 5 min in 300 μ l 0.5% saponin in PBS 910 and bacteria were plated. Adhesion and invasion are expressed as CFUs per monolayer. 911 Experiments were carried out in triplicate on three independent occasions. 912 Analysis of Yops secretion. 913 Overnight cultures of Y. enterocolitica strains were diluted 1 ∶ 50 into 25 ml of TSB supplemented 914 with 20 mM MgCl 2 and 20 mM sodium oxalate in a 100-ml flask. Cultures were incubated with 915 aeration at 21°C for 2.5 h, and then transferred at 37°C for 3 h. The optical density at 540 nm of the 916 culture was measured and the bacterial cells were collected by centrifugation at 1500× g for 30 min. 917 Ammonium sulphate (final concentration 47.5% w/v) was used to precipitate proteins from 20 ml of 918 the supernatant. After overnight incubation at 4°C, proteins were collected by centrifugation 919 (3000×g, 30 min, 4°C) and washed twice with 1.5 ml of water. Dried protein pellets were 920 resuspended in 50 to 80 µl of sample buffer and normalized according to the cell count. Samples 921 were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using 922 12% polyacrylamide gels and proteins visualized by Coomassie brilliant blue staining. 923 Actin disruption by Yersinia infection. 924 A549 cells were seeded on 13 mm circular coverslips in 24-well tissue culture plates to 70% 925 confluence. Cells were serum starved 16 h before infection. Overnight cultures of Y. enterocolitica 926 strains grown at 21°C were diluted 1 ∶ 10 into 5 ml of LB and grown with aeration at 21°C for 1.5 h 927 and then 1 h at 37°C. Bacteria were pelleted, washed once with PBS and resuspended to an 928 OD600/i6 = of 1 in PBS. Cells were infected with this suspension at a multiplicity of infection of 25∶ 1. 929 After 1 h incubation, the coverslips were washed three times with PBS and then cells fixed with 930 3.7% PFA in PBS pH 7.4 for 20 min at room temperature. PFA fixed cells were incubated with PBS 931 containing 0.1% saponin, 10% horse serum, Hoechst 33342 (1 ∶ 2500), and Orange-Red-phalloidin 932 (1∶ 100) (Invitrogen) for 30 min in a wet dark chamber. Finally, coverslips were washed twice in 933 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 38 0.1% saponin in PBS, once in PBS and once in H 2O, mounted on Prolong gold antifade oil 934 (Molecular Probes P36930). Coverslips were visualised on the Leica SP8 confocal microscope. 935 Experiments were carried out in duplicate in three independent occasions. 936 Analysis YadA production. 937 Bacteria were grown overnight in 2 ml RPMI 1640 medium lacking phenol red at 37°C without 938 shaking. The OD 540 of the culture was measured and CFUs were determined by plating serial 939 dilutions. Bacteria from 1-ml aliquot were recovered by centrifugation (16 000×g, 10 min, 4°C) and 940 resuspended in 200 µl of SDS-sample buffer. Samples were incubated for 4 h at 37°C and kept 941 frozen at −20°C. Samples were analyzed by SDS-PAGE using 12% polyacrylamide gels and 942 proteins visualized by Coomassie brilliant blue staining. Samples were normalized according to the 943 cell count and they were not boiled before loading the gel. 944 Binding assay to collagen-coated slides. 945 Overnight cultures of Y. enterocolitica strains grown at 37°C were diluted 1∶ 10 into 5 ml of LB and 946 grown with aeration at 37°C for 2.5 h. bacteria were pelleted, washed once with PBS and 947 resuspended to an OD540 of 0.3 in PBS. 948 13 mm circular coverslips in 24-well tissue culture plates were coated overnight at 4°C with 10 949 µg/ml human collagen type IV (Sigma) in PBS (final volume 100 µl). Coverslips were washed three 950 times with TBS and later they were blocked for 1 h at 4°C with 2% BSA in TBS. Finally, coverslips 951 were washed three times and were incubated at 37°C with 100 µl of the bacterial suspension. After 952 1 h incubation, the coverslips were washed three times with PBS and then bacteria fixed with 3.7% 953 paraformaldehyde (PFA) in PBS pH 7.4 for 20 min at room temperature. PFA fixed cells were 954 incubated with PBS containing 0.1% saponin, 10% horse serum and Hoechst 33342 (1 ∶ 25000) for 955 30 min in a wet dark chamber. Finally, coverslips were washed twice in 0.1% saponin in PBS, once 956 in PBS and once in H 2O, mounted on Aqua Poly/Mount (Polysciences). Fluorescence images were 957 captured using the ×100 objective lens on a Leica DM5500 microscope equipped with the 958 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 39 appropriate filter sets. Acquired images were analysed using the LAS imaging software (Leica). 959 Bacteria were counted in images from three randomly selected fields of view. 960 In vivo experiments. 961 8- to 12-week-old Balb/c mice (Charles River) of both sexes were infected by oral gavage with ∼ 1 962 × 10 9 Y. enterocolitica strains in 100 μ l PBS. 72 h post infection, mice were euthanized using a 963 Schedule 1 method according to UK Home Office approved protocols. Peyer’s patches (3-5 per 964 mouse) were immersed in 1 ml sterile PBS on ice and processed for quantitative bacterial culture 965 immediately. Samples were homogenised with a Precellys Evolution tissue homogenizer (Bertin 966 Instruments), using 1.4 mm ceramic (zirconium oxide) beads at 4,500 rpm for 7 cycles of 10 s, with 967 a 10-s pause between each cycle. Homogenates were serially diluted in sterile PBS and plated onto 968 Yersinia selective agar medium plates (Oxoid), and the colonies were enumerated after incubation 969 at 21°C. Data are expressed as CFUs per gr of tissue. 970 Statistical analysis. 971 Statistical analyses were performed using one-way analysis of variance (ANOVA) with the 972 Dunnett’s multiple comparisons test, the one-tailed t test, or, when the requirements were not met, 973 the Mann-Whitney U test. Normality and equal variance assumptions were tested with the 974 Kolmogorov-Smirnov test and the Brown-Forsythe test, respectively. Results are given as means ± 975 SD. A P value of <0.05 was considered statistically significant. All analyses were performed using 976 GraphPad Prism for Windows (version 10.4.1) software. 977

978 We are grateful members of Bengoechea lab for helpful discussions. We also thank the 979 bioi nfor matics ( J .V. L e ht on en) , a nd s t r uct ur al biolog y ( F I N S t r u c t ) i nfr astr u c t ur e su p p o rt from 980 Biocen ter Finla n d a n d CSC IT Cen t e r for S c ien c e f or compu tat ion al i nfra s tru c t ur e s u pp o rt at t he 981 Str uc t ur al B i oi nfor m ati c s Laborato ry, A/i36 bo Akademi University. The funders had no role in study 982 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 40 design, data collection and analysis, decision to publish, or preparation of the manuscript. This work 983 has been funded by Medical Research Council (MRC, MR/V032496/1) and Biotechnology and 984 Biological Sciences Research Council (BBSRC, BB/P006078/1) grants to J.AB. P.J.B. and J.K.M. 985 would like to acknowledge BII core funds. The computational work for this article was partially 986 performed on resources of the National Supercomputing Centre, Singapore ( https://www.nscc.sg). 987 This work was funded by the Sigrid Juselius Foundation to T.A.S. and K.M.D. 988

989 1. Medzhit ov R, Janew ay Jr C A. Inn ate im munity : impac t on the ada p tive immun e r e spon se . Cur ren t 990 opin ion in i mmunology . 1997 ;9(1 ):4 -9. 991 2. Jane way Jr CA, Medzh it o v R . Inna te immun e rec ognition . Ann ual r evie w of immu nol ogy. 992 200 2;20(1) :197-21 6. 993 3. Vanc e RE , Isb erg RR, P or t noy DA. Pa t tern s o f path ogen e s i s: di sc r im in ation of pat hogenic an d 994 nonpa th ogenic mic r obe s by the in na te immune s y stem . Ce ll h os t & m icro be. 2 00 9;6(1) :10-21 . 995 4. Stei mle A, Aute nrie th IB, F r ic k J-S . S tr uc t ur e and func t io n: Lipi d A modi fica tio ns i n comme ns al s an d 996 pathog en s. In tern ation al J ou rna l o f Medi c al Mic r o biology . 2016; 306(5 ):290 -301 . 997 5. Ander s en- N i ssen E, S mith K D , S t r ob e KL, Barr et t SLR, Co ok son BT, Log an SM , e t al. Evas ion o f Tol l -998 lik e r e cep tor 5 by fla g ella te d ba c teria . P rocee dings o f t he na tional ac ademy of s c ienc e s. 999 200 5;102(26):92 47 -52. 1000 6. Pa r k BS , L e e J- O. R e cogni t i on o f lipop o l ys a cc harid e pat t e rn by TL R4 c omplex es. Ex perime ntal & 1001 mole cula r m edici ne . 2013 ;45(12 ): e66- e. 1002 7. Vig anò E, Dia mo nd CE, S pr eafic o R, Bala c hande r A , Sobot a RM, Mor tell aro A . Huma n caspa s e-4 an d 1003 c as p a se-5 reg ula te th e one - st ep no n-ca nonic al in fla mma s o m e activ ati on in mono cyte s. N ature 1004 c ommunic ation s . 2015;6 (1) :8 761. 1005 8. Ca s son C N, Yu J, R e ye s V M, Ta schuk FO, Ya dav A, Cop enhav er AM , e t al. H uma n ca s p as e -4 me diat e s 1006 nonc anonic al in flamma som e activ a tion ag ains t gram -neg a t i ve b a cte rial pa tho gens . P roc ee ding s of t h e 1007 Natio nal Acad e my of S cienc e s . 2015;1 12(21):6688 -9 3. 1008 9 . K a y a g a k i N , W o n g M T , S t o w e IB , R a m a n i S R , G o n z a l e z L C , A k a s h i - T a k a m u r a S , e t a l . N o n c a n o n i c a l 1009 infla mma some ac tivation by int rac ellul a r L PS i ndepend e nt o f TLR4 . Sc ience . 2013; 34 1(6151):124 6 -9. 1010 10. Hajja r AM, Ern s t RK, Tsa i JH, Wil son C B, Mill er S I. Human Toll-lik e r ec ep tor 4 rec ogniz e s ho s t-1011 speci fic L PS mod i fication s . Na tur e immun ology. 2002;3(4) :3 54-9 . doi: 10.1038 /ni 77 7. 1012 11. Se ydel U , Oik a wa M, F uka se K, Ku s um oto S , B rand enburg K . In tr i n s ic c onf or mation of lip id A i s 1013 re spon s i ble fo r ag oni st i c and an ta gon is tic a ctivi t y . Eur J Bioc h em. 2 000;26 7(10) :3032-9. doi : 1014 10. 1046/j.1432-10 33 .2000 .01326.x . Pub Med PM ID: 1080640 3. 1015 12. Si mps on BW , Trent MS. Pu shing t he env elo pe: LPS modi fica ti on s a nd t he ir con se quenc e s . N at Rev 1016 Mic r ob iol. 201 9;17 (7) :403-16 . Epub 20 19/ 0 5/ 3 1. d oi: 10 .1038 /s 4 1579 -019 -020 1-x. Pub M e d PM ID : 1017 311 42822; P ubMed Cen tral P MCID: P M C P M C 6913091 . 1018 1 3 . G a u t h i e r A E , C h a n d l e r C E , P o l i V , G a r d n e r F M , T e k i a u A , S m i t h R , e t a l . D e e p - s e a m i c r o b e s a s t o o l s 1019 to r e fi n e th e rul es of innat e im mune pat te rn rec ogni tion . Sc i Im munol. 2021;6 ( 5 7 ). doi : 1020 10. 1126/ s c iimmunol .ab e053 1. PubMe d P M ID : 3 3712473 ; P ubMed Ce n t ral PM CI D: P MC PM C836704 8. 1021 14. McNal ly A, Th om s on NR , Re u t e r S , Wr e n BW . 'Add, st i r a nd red uce ': Yer sinia spp . as mod e l bac t e ri a 1022 for p a t h og en evolu tion . Na t Re v Mic r ob iol. 2016;14 (3) :17 7-90 . E pub 20 16/02/ 1 6 . doi : 1023 10. 1038/nrmicro .2015 .29 . PubM e d PM ID: 268760 35. 1024 15. Se abaug h JA, Ande rson D M. P a thogenic ity an d virule nce of Y er s ini a . Vi r ul enc e. 1025 202 4;15(1) :2316 439. 1026 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 41 16. Su lakv elidze A. Y er s in i ae o ther t han Y. ente roco litic a, Y. ps eudo tube r c ulo si s , a nd Y. pes t i s: t he 1027 ig nor e d s pec ie s . Microb e s and In fe ction . 2 000;2 (5):497 -5 13. d oi: ht t p s: // d oi .or g / 10 .1016 /S128 6-1028 457 9(00)00311 - 7 . 1029 17. Reut er S , Conn or TR, Barqui st L , Wal k er D, Feltw ell T , Harri s SR, et al. Pa rallel ind epen d en t 1030 ev olution of p a t ho g enic ity within the g enu s Ye rsinia . Proc e eding s o f th e Na t i o nal Acad e my of S cie nc e s. 1031 201 4;111(18):67 68 -73. 1032 18. Rebe il R, Ern st RK, Gow en B B, Mil ler SI, H i nn ebu sch B J. V a r i a tion in lipid A s truct ure in t he 1033 pathog enic yer sinia e. Molec ul ar mic r ob iology . 2004 ; 52(5 ):1363 -73. doi : 1 0.1111 /j.136 5-1034 295 8.2004.0 4059.x . 1035 19. Kaw ahara K, T suka no H, W atan abe H , Lindn er B , Ma t s u ur a M. Modi fic atio n o f the struc t u r e a nd 1036 ac tivity o f l ipid A in Ye r s i nia pe s ti s l ipo poly s ac cha r id e by grow t h t empera tur e. In fecti on an d immunity . 1037 200 2;70(8) :4092 - 8 . 1038 20. Rei ne s M , L lobe t E, Dahl strom KM, Per ez -G u tie rrez C, Llomp a r t CM, Torr eca bot a N , et al . 1039 Dec iphe ring the ac yla t i on pa tt e r n o f Ye rsinia e nte r oc oli tic a lipid A. PL oS pa t h og ens . 2012;8(10 ): e 1002978. 1040 doi: 10.1371 / jou rna l. ppat .1002 978; 10. 1 371/journal .ppa t .1 002978. 1041 21. Per ez -Gu t ie rrez C , L lobe t E, Llompa rt C M, Rei ne s M, B engo ech ea JA . Rol e o f l i pid A ac yla t i on in 1042 Ye r si nia en te roc olitica v irulenc e . In fec ti on and immu nity. 2010 ;78 ( 6 ) :2 768 -81. doi: 10.1 128 /IA I.01417 -09 ; 1043 10. 1128/IAI .01417 -0 9. 1044 22. Rebe il R, E rn st RK, Jarr ett CO , Ad ams KN, Mille r S I, Hinn ebu sch B J . Ch arac teriz ation o f la te 1045 ac yltran sf era se g ene s of Y ersini a pe s ti s and their rol e in tempe ratu re -dep ende n t li pid A variation . Journ al 1046 of Ba ct eriol ogy . 2006 ;188(4) :1381-8 . doi: 10. 1128/JB .188.4 .1 381-1388. 2006 . 1047 2 3 . R e in es M , Ll o be t E , L l o m p ar t CM , Mo r a nt a D , P er ez - G ut i er r e z C , B eng o e ch ea JA . M o l e cu l a r b as is o f 1048 Ye r si nia ent eroc oli tica t empe ra t ur e- d ep end ent r e si s tance to antimic r o b ia l peptide s . Journa l of 1049 Bac te riology . 2012;1 94(12 ):3173 -8 8. doi : 10 .1128/JB .00308 - 1 2 ; 10.112 8/JB.00 308 -12. 1050 24. Anisimov A P , De nto vskaya S V, Tita rev a GM, Bakht ee va IV, S haik hu t d inova RZ, B ala khonov S V, et al. 1051 Intr as peci e s and temp era ture -d epend e nt va r i a tion s in su s c ep tibility of Yer s i ni a pe stis to th e b ac t e r ic idal 1052 ac tion of s erum and to polymy xin B. In fec tion an d immuni t y . 2 00 5;73(11) :732 4-31. doi : 1053 10. 1128/IAI .73 .11. 7324 -7331 .2005. 1054 25. Marce au M, Se bba ne F, Ewa nn F , C ol lyn F, Lindne r B, C ampo s MA, et al . Th e pmrF poly myxi n-1055 re si s tan c e operon o f Y ersini a p s e ud otub erculo si s is upre gulat ed b y the PhoP - Pho Q two -c omponen t sy ste m 1056 but no t by Pmr A- PmrB, and i s no t r eq ui r e d for virul enc e . Mi crobi olog y (Re adi ng, Eng land). 2004 ;1 50( Pt 1057 12):3 947-57 . doi : 10.10 99 / mi c.0.27 426 -0 . 1058 26. Fredrik sson -Ahom aa M, S t ol l e A, Ko rke ala H. M ol ecula r epid emi ology of Y ersi nia ent eroc oli tica 1059 infe cti on s. FEMS Immunol og y & Medic al Mic r obi ology . 2006;47 ( 3 ):315 -29. 1060 27. Hall M, Cha tt aw ay MA, Reu te r S, Sav in C, S trauch E, Ca r ni el E, et al. U s e of who le-g enu s genom e 1061 seq uenc e da ta to devel op a multilocu s seq uenc e t y ping too l that acc ur at ely iden t i fi e s Yer sinia i sola te s t o 1062 the spec ie s a nd su bs pecie s l evel s . Jou rna l o f clini cal microbiol ogy. 2015;53 (1):35 -4 2. 1063 2 8 . C h o i K H , G ay n o r J B , W h it e K G , L op ez C , B o s io C M , K a r k h off -S ch w e i z e r R R, e t al. A T n 7 - b a s e d b r o ad-1064 range bac te r ia l cl oning a n d ex pr e ssion syste m. N ature meth od s. 2005;2 (6 ) :4 43 -8. doi: nmeth765 [pi i ]. 1065 29. Barth olome w TL , Kidd T J , S a P e s soa J, C onde Alv arez R, B engoe ch ea JA. 2 - Hyd r ox ylatio n o f 1066 Aci neto bac t e r b auma nnii Lipid A Con tr i bu t e s t o Vi rulenc e . Inf ect Immun. 2019;87 (4) . Ep ub 20 19/ 0 2 / 13 . 1067 doi: 10.1128 /I AI.00 066-19 . P ub Med PMID: 30745 327; Pub Me d Cen tral PMC I D : P MCP MC64 34125. 1068 30. Mill s G, Dumig an A, Ki dd T, Hobl ey L, B eng oeche a J A . Iden tific a tion and Charac terizatio n of Tw o 1069 Kle bs i e lla p ne umoni ae l pxL Lipid A La te Ac yltran sfe ra se s and Th eir Role in Vi rule nce . Infe ction an d 1070 immu nity. 201 7;85(9) :10.11 28 / IAI . 00068 -17. P rint 2017 S ep . doi : e00068 -17 [pii] . 1071 31. Needh a m BD , C a r roll S M, Gil e s DK, G e orgi ou G, W hi tel ey M, Tr ent M S . Mo d ula ting t h e inna t e 1072 immu ne re spon se by c ombina tori al en gi neering o f en do t ox in. P roce edi ng s of the Na tion al Acad emy o f 1073 Sc ienc e s o f th e Unite d S t a te s of America . 2013;11 0 (4) :1464-9. doi : 10.1073 / pna s .121 8080110 ; 1074 10. 1073/pna s .121808 0110. 1075 32. Dova la D, R ath CM, H u Q, Saw yer WS, S hia S, Elli ng R A, e t al . S tructu re -guide d e nz y mology of t h e 1076 lip id A ac yltran s fe r a s e L pxM r e veal s a d ual ac tivity mec hani sm. Proc eedi ngs o f the N a tion al Acad emy of 1077 Sc ienc e s. 2016;1 13(41 ):E606 4 -E71. 1078 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 42 33. Wyc koff TJ , Li n S, Co tt er R J , Dot son GD, Rae t z CR . Hydroca r b o n r ul ers in U DP -N- ac etylgl ucos amine 1079 ac yltran sf era se s. J ournal o f Biol og ical Ch emis tr y . 1998;2 73(49 ):3236 9 -72. 1080 34. Ahn VE, L o EI, E ng el CK, Che n L, Hwa ng PM, Kay LE, e t a l . A hydroc a r bon rule r me as u re s palmi t a te 1081 in th e e n z yma tic acy lation o f e n dotoxi n. Th e EM BO Jou rnal. 2004 ;2 3(15) :2931-41 - 4 1. doi : 1082 http s : / /doi. org/10 .1038 /sj.emb oj. 7600320 . 1083 35. Pa r k BS , Song DH , Kim HM, Choi BS , L ee H , Le e JO. The struc tural b a si s o f li popolysac ch aride 1084 rec ognition by the TL R4- MD-2 c omplex . Natu re. 2009 ;458 (7242):1 191 -5. Epu b 2009/03/03. doi : 1085 10. 1038/na ture07830 . Pub Med P MID: 1 92 52480. 1086 36. Za noni I, O s t un i R, M arek L R, Ba rre si S , B a r ba la t R, Ba rton GM, et al . C D14 cont rol s t h e LPS - i nduce d 1087 end ocyto si s o f Toll-li ke r ec ep tor 4. Cell . 2011;14 7(4) :868-80 . 1088 37. Ta n Y, Zan oni I, Cull en TW, Goodman AL , Kagan JC. Me cha nism s o f Toll-lik e rec e p t or 4 en docyto si s 1089 reve al a c ommon im mune -eva s ion s tr ategy u sed b y path oge nic a nd comm e nsa l bac te r ia . Immunity. 1090 201 5;43(5) :909-22 . 1091 38. Aka s hi S, Sai toh S -i, W a ka bay a shi Y, Ki ku chi T, T akamu r a N , N agai Y, e t al . L i popolys ac charide 1092 inter ac tion with c ell sur face Tol l -like re ce ptor 4- MD -2: hi gher a ffini ty t han t h a t w ith MD -2 or C D14. Th e 1093 Journa l of expe rime nta l medi cine . 2003; 198 ( 7 ):1035 -42. 1094 39. Durand EA, Ma ldon ado -Ar ocho F J, Ca stil lo C , Wal sh RL , Me cs a s J . T he pre s en c e of pro fe ssion al 1095 pha gocy t e s dicta te s th e n umber o f ho st c ell s t arge ted f or Y op tr an slocati on during infecti on. C ellul ar 1096 mic robiology . 2010 ;12(8 ):1064 -82 . 1097 40. Marke ton MM, DeP aolo RW, DeB ord K L, Jabri B, S chne e wind O . Pla gu e bac t e r ia targe t immune 1098 c ells du ring i nfec tion. Sci e nce. 2005;309 (5 741):1739-41 . 1099 41. Köbe r le M , K lei n- Günt her A, S chütz M, F r itz M, B er c ht old S, To lo s a E, e t al . Yer sinia en teroc olitic a 1100 targe t s cells o f t h e inna te and ada p t i ve immune sys tem by injec tion of Yops i n a m ous e in fe ctio n model. 1101 PL oS pa t h o gen s. 2009;5 ( 8 ): e100055 1. 1102 42. Thi efe s A, Wol f A , Do errie A, A G r a ssl G, Ma tsumo to K, Aut enri eth I, et al. Th e Ye rsini a 1103 ente roc olitic a e ffe cto r Y opP in hibit s ho st ce ll s i gnalli ng by ina ctiva t in g t h e pr o te i n k inas e TAK1 in th e IL -1 1104 signa lling pa thway . EMBO re port s. 2006 ; 7(8):838 -44. 1105 4 3 . Z h ou H , M on a c k DM , K a y ag a ki N , W er t z I, Y i n J , W o lf B , e t a l . Ye rs in i a v ir u le n c e f a c t o r Y o pJ a c ts a s a 1106 deu biquiti na se t o inh ibit N F -κB ac tivatio n. The Jou rnal o f expe r i men tal med icin e . 2005;20 2(10):1327 -32 . 1107 4 4 . B r o g d e n K A . A n t i m i c r o b i a l p e p t i d e s : p o r e f o r m e r s o r m e t a b o l i c i n h i b i t o r s i n b a c t e r i a ? N a t u r e 1108 revi ews mic r obi ology. 2005;3(3 ):2 38 -50. 1109 45. Nika ido H . M ole cular b a sis of ba c teri al outer m embran e perme ab ility r evisi ted. Microbio logy an d 1110 mole cula r b iology review s : MMBR. 2003 ;67 ( 4 ):5 93-656 . 1111 46. Needh a m BD, Tren t MS . Fo rti fying the ba r rie r: t h e impa ct o f l ipid A r e mod e lling on b acterial 1112 pathog en e si s . Natu re r evi ewsMic r obiolo gy. 20 13;11(7 ):467 -81 . doi : 10.1 038/nr m icro3047 ; 1113 10. 1038/nrmicro3047 . 1114 47. Kidd TJ, Mill s G , S a -P e ssoa J , Dumiga n A, Fra nk CG , In su a JL , e t a l . A Kle b s ie ll a p neumo nia e 1115 antib io tic r e si sta nce mec hani sm tha t subdue s ho s t de fenc e s a nd p r omo te s v irulenc e . E M BO m olecul a r 1116 medi cine . 2017;9 :430 -47. d oi : e2016 073 36 [pi i] . 1117 48. Guo L, Lim KB, Poduje C M, Da niel M, G unn JS, H ack et t M, e t a l. L ipid A a cyl ation and b acte rial 1118 re si s tan c e a gain st ver teb rat e antimi cr obi al peptid e s. C ell. 1998;95 ( 2 ): 189-98 . 1119 49. L lobet E, Ca mpo s MA, Gimenez P, Mora nta D, Be ng oeche a JA. A n alysi s o f the ne tworks con trollin g 1120 the antimic ro bial -pep tide - d e pend ent in duc t i on of Kleb siell a pn eumonia e v irul e nce f ac to r s . In fectio n an d 1121 immu nity. 201 1;79(9) :3718 - 3 2 . doi: 1 0.1 128/IA I.0522 6 -11; 10. 1128 /IA I.05226 -11 . 1122 5 0 . Y o un g G M , S ch m i e l D H , M i l le r V L. A n e w pa t h w a y f or t h e s e cr et i o n of v i r u l en ce fac t o rs b y b a ct e r i a : 1123 the fla gell a r ex por t ap para tu s fun ctio ns a s a pro t e in - s ec re t i on s y stem . P ro c eeding s o f th e N ation al 1124 Aca demy of Sci enc e s of t he U ni ted St at e s o f America . 1999;96 ( 1 1 ):6456 -61. 1125 5 1 . Y o u n g G M , B a d g e r J L , M i l l e r V L . M o t i l i t y i s r e q u i r e d t o i n i t i a t e h o s t c e l l i n v a s i o n b y Y e r s i n i a 1126 ente roc olitic a . In fec tio n and im munity . 20 00;68(7) :432 3-6 . 1127 5 2 . Y o u n g G M , S m i t h M J , M i n n i c h S A , M i l l e r V L . T h e Y e r s i n i a e n t e r o c o l i t i c a m o t i l i t y m a s t e r r e g u l a t o r y 1128 ope r on , flh D C, i s r equir ed for fla gelli n produc tion , s w imming motility , a n d swa r min g mot i lity. Jou r n al o f 1129 Bac te riology . 1999;1 81(9 ):2823 -33 . 1130 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 43 53. Beng oech e a JA, Na jden ski H , Sk ur ni k M. Lip opolysac charid e O antig en st atu s of Ye rsini a 1131 e n t er o co l i t i ca O: 8 is es s en t ia l f o r vi r u le n ce a nd a bse n c e of O ant ig e n a f fec ts t he e xp r e s s i o n of ot h e r 1132 Ye r si nia vi r ul ence fac tor s. Mol ecu la r mic r ob iology . 20 04;52(2 ):451 - 6 9 . do i: 1 0.1111 /j.136 5-1133 295 8.2004.0 3987.x . 1134 54. Sc hmiel DH , Y oung G M, M ille r VL . The Ye rs i ni a en te roc olitica pho spho lipa se g e ne ypl A is par t o f t h e 1135 flag ella r reg ulon . Jou rnal o f Bact eriol og y. 20 00;182(8):23 14 -20. 1136 55. Pep e JC , Mil le r VL. Ye r sinia en te rocolit ic a i nvasin: a primary rol e i n th e i niti a tion o f inf ection . 1137 Proc e eding s of the Nation a l Aca demy o f Scie nce s of the Uni ted S tat e s of Am erica. 1993;90(1 4):6473 -7. 1138 56. Mill er V, Fal kow S . Ev idence for t w o ge n etic lo ci in Yer s i nia ente roc oli tica tha t ca n promote inva s io n 1139 of e pi t h elia l cell s . In f e c tion a nd immunit y . 1988;5 6(5):1242 -8. 1140 57. Rev ell P A, M ill er VL . A chr omos omally e ncod ed re gulat or i s r e quir ed for e xpre s sion of th e Ye rsini a 1141 ente roc olitic a inv gene a nd f or virulen ce . Mole cula r m icrobi olog y. 20 00;35(3 ):677 - 8 5. 1142 58. Nage l G , L ahrz A, D er sch P. Env ironment al c on tr o l o f inva sin ex pre ssion in Y e rsini a 1143 pse udotub erc ulo s i s i s me di at ed by regu la t i on o f Rov A, a t ran scrip tional ac tivat or of the S lyA / Ho r f amily. 1144 Molec ular mi cro biolog y. 200 1;41(6 ):124 9-69 . 1145 59. Rosqvi st R, F orsb erg A, Wol f - Wa tz H. Intr ace llula r ta rge t i ng o f the Ye rs i n ia YopE cy tot ox in i n 1146 mam malian cell s i nd uce s ac tin mic r o fi la ment di srup tion. Inf ec tion a nd immunity. 1991;59(1 2) :4562-9 . 1147 6 0 . S k u r n i k M , T o i v a n e n P . L c r F i s t h e t e m p e r a t u r e - r e g u l a t e d a c t i v a t o r o f t h e y a d A g e n e o f Y e r s i n i a 1148 ente roc olitic a and Y ersinia p se udo tuberc ulosi s . Jou rnal o f Ba ct eriolog y. 199 2;174 ( 6):204 7-51 . 1149 61. El Tah ir Y, Skurnik M. YadA, the mu lt i fa ceted Ye r sinia adh esi n. In tern atio nal journal o f me dical 1150 mic robiology : IJM M. 2001 ;291(3 ):209 -1 8 . doi: 10 .1078 /1438-42 21 -00119. 1151 62. Hof s t ae d t e r CE , Chandl er CE , Met CM , Gille spie JJ, Ha rro J M, G oo dlet t D R , e t al. Dive r g ent 1152 Pseudomonas aeruginosa Lpx O enz ymes p e r f orm s it e -s peci fic lipid A 2-hy dr ox yla t i o n. mBio . 1153 202 4;15(2) :e028 23-23 . doi : do i:10.1 128 / mbio.0282 3-23. 1154 63. Erns t RK, Yi EC, G uo L, L im KB, Burns JL, Hac kett M, e t al. S peci fic lipopo lys acc har ide found in cy s tic 1155 fibro si s ai rway Pse udomon a s ae r ug in osa . Sc ienc e ( N e w York, N Y ). 1999;28 6(5444 ):1561 -5 . 1156 64. Erns t RK, Moskow it z SM, E merson JC , Kraig GM , Adam s KN, Harve y MD, e t al. U ni que l ipid a 1157 modi f i cati on s in P seudo mona s ae r ug ino sa i s ola t e d fr om th e a irway s o f p a t i ent s wi th cy s tic fibr o sis. Th e 1158 Journa l of in fecti ou s di sea s es . 2007 ;196( 7):1088-9 2 . doi: JID3825 4 [pii ]. 1159 65. Tran AX, St ead CM, Tr ent M S . R em ode ling of Hel ico ba ct er py lori li p opoly sac charide . Jou rnal o f 1160 end otox in re search . 2005;11(3 ):1 61 -6. 1161 66. Stea d C M , Be a s le y A, Co tt er RJ, Tre nt MS. D eci pheri ng the U nu sual Ac y lation P at tern o f 1162 Helicobacter pylori Lipid A. J our n al of Ba c teriol ogy . 2008;1 90(21) :7012 - 2 1 . doi: d oi :10.112 8/ jb .00667 -08. 1163 67. Sh af fer SA, H arvey MD, Go odle tt DR, Ernst RK. S tructu ral he te roge n eity a nd env ir o nmen tally 1164 regul at ed remo deling of Franc i s e ll a tu la ren s i s s ub spec i es no vici da lipid A c ha ra c t e r iz ed by t andem ma s s 1165 spec trome try. Journ al of the Ame rica n S oc iety fo r Ma s s Spec trome try. 2007;18(6 ):1 080-92 . 1166 68. Phi llip s NJ , S chil ling B, McLe ndon M K, Apic ella MA, G i b s on BW . Nove l mo dific a t io n o f lipid A of 1167 Fran ci sella tul ar en s i s . In f e c tion a nd imm unity . 2004;7 2(9 ):5340 -8. 1168 69. Vinog r a d ov E, P e r ry MB, C onla n JW. St r uctural ana ly si s o f F r a nci s ella t u lar en si s l i popolys ac charide . 1169 Europe an journ al of bi och emi st ry . 2002; 269 ( 2 4):6112 -8. 1170 70. Coa t s S R, Jon e s JW , Do CT, Bra ham PH, Bainbridge B W, To TT, et al . H u man T oll-lik e rec ep tor 4 1171 re spon s e s to P . g ingiv alis a r e reg ula ted by lipid A 1 -and 4EiLP -pho sp hat as e ac t i viti es . C ellula r mic r ob iol ogy . 1172 200 9;11(11) :158 7-99. 1173 71. Weint raub A, Zä h r in ger U , W OLLE N WEB ER HW, S ey del U , Rie t s c hel ET . S t r uc tura l charac teriza t io n 1174 of th e lipid A c ompon ent o f Bac ter oide s fragili s st r ain N CTC 934 3 lipopoly s a ccha r i de. Eur opean jo u r na l o f 1175 bioc hemis try. 198 9;183 (2) :425-31 . 1176 72. Berez ow AB, E r ns t RK, Coa t s SR, Braha m PH, Karimi-N ase r L M, Da rveau RP . The s tr uc tu r a lly simil ar, 1177 pen t a -a cyl a ted lipop oly s ac ch a r i de s o f P orphyromona s gingi vali s and Bac tero ide s elic i t s trikingl y dif fer en t 1178 inna te immune r e s pon s es . Microbi a l pat hoge ne s i s . 2 009;47 (2) :68 - 7 7 . 1179 73. Vat ane n T, K o stic AD, d’He nnez el E, S ilja nder H, F ranzo s a E A, Ya ssour M, e t al . Vari atio n in 1180 mic robiome L PS i mmunogeni city cont rib ute s t o a utoimmuni ty in human s. Ce ll. 20 1 6;165(4) :842 - 5 3 . 1181 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 44 74. Hashimo to M, A sai Y, Tama i R, J in no T , Uma tani K, O ga wa T. Chemic a l s tr u c ture an d 1182 immu nobiologi cal activ i ty of li pid A f r om Pr evotel l a int ermedi a AT CC 25 611 lip opoly sac charid e. FE BS 1183 le t t er s. 2003;5 43 (1 -3):98 -102. 1184 7 5 . D i L o r e n z o F , P a l m i g i a n o A , S i l i p o A , D e s a k i Y , G a r o z z o D , L a n z e t t a R , e t a l . T h e s t r u c t u r e o f t h e 1185 lip ooligo sac charid e f rom Xant homona s oryz a e pv. O ry z a e: t he ca u s a l a gen t o f the ba c teri al le af bli gh t i n 1186 rice . C arbohy dra te Re s e arch. 20 16;427:3 8 -43. 1187 76. L lobet E, Ma rtin ez - Moli ne r V, Moran ta D, D ahl st r om KM, Reg uei ro V , T omas A, et al . Dec iph ering 1188 tis s u e- i nduced Kleb si ella pne umonia e lip id A structu re . Proc e eding s o f th e Na tion al Acade my of S ci enc e s o f 1189 the Uni ted S t at es o f Am er i ca. 2015;112 (4 6):E6369-78 . doi: 10 .1073 /pna s .15088 20112 [doi ] . 1190 7 7 . G r e g g K A , H a r b e r t s E , G a r d n e r F M , P e l l e t i e r M R , C a y a t t e C , Y u L , e t a l . R a t i o n a l l y D e s i g n e d T L R 4 1191 L igands f o r Vacc ine Adjuvan t Di s c ove ry. mBio. 2017; 8(3) . Epu b 201 70509. doi: 10.112 8/ mBi o.00492 -17 . 1192 PubMe d PM ID: 284874 29; PubM ed C ent r a l PM CI D: PMC PM C5424 205. 1193 78. Needh a m BD , C a r roll S M, Gil e s DK, G e orgi ou G, W hi tel ey M, Tr ent M S . Mo d ula ting t h e inna t e 1194 immu ne re s p on se by c ombinat orial engi nee r i ng o f e ndo tox in. P roc N a t l Acad S ci U S A. 2013 ;110 (4) :1464-9 . 1195 Epu b 2013010 7. doi: 10.1 073 /pna s. 12 180801 10. PubM ed PM ID : 232 97218; PubMe d Ce ntra l PM CI D: 1196 PMC PM C35 57076. 1197 79. Kaw asaki K, Ern s t RK , Mill e r SI. D eacy lation an d palmitoy la tion o f lipi d A by S almone lla e out e r 1198 memb r a ne enz y me s modu la te h o s t s i gn aling throug h T oll -like rec e pto r 4. Journ al of endo toxi n re s earch . 1199 200 4;10(6) :439-44 . 1200 80. Muroi M, Ohni shi T, T anamot o K-i. MD -2 , a novel acc e s s ory mo lecul e , is invol ved in s pec i e s- s p ec ific 1201 ac tion s of S al monell a lipid A. Inf ection a n d immuni t y . 2002 ;70(7) :3546-50 . 1202 81. Muroi M, Tana mo t o K-i. T he poly s a cc h aride por tion play s a n indi spen sa ble role in Sal monell a 1203 lip opolys ac charide -i nduc ed ac tiva tion of NF-κB thr ough human To ll -like r e c epto r 4. In fe ct ion a n d immunity. 1204 200 2;70(11) :604 3-7. 1205 8 2 . T a n a m o t o K - i , A z u m i S . S a l m o n e l l a - t y p e h e p t a a c y l a t e d l i p i d A i s i n a c t i v e a n d a c t s a s a n a n t a g o n i s t 1206 of li popoly sac cha ride acti on on human li ne c ells . The J ournal o f Imm unol ogy . 2000;1 64(6) :3149-56 . 1207 83. Boll JM , Tuck er AT, Kle in D R, Bel tran AM, Brodbel t J S , Davie s B W, e t a l. R einf orc in g L ipid A Ac ylation 1208 on the C ell Sur fac e o f Aci ne tobac ter b a umann ii Prom ote s Ca tionic Antimi crobi al Pep tid e Re si st ance a n d 1209 De sicc ati on S ur v ival . mBio. 2 015;6 (3) :e0 04 78-15. doi : 10 .112 8/ mBi o .00478-15 [d oi]. 1210 84. Se nG u p t a S, Hi ttle L E, Ern st RK , U ria rte S M , Mit c hell T C. A Ps eudomon a s ae rug inosa h ept a- ac yla t e d 1211 lip id A varian t a s s oc iat ed w ith cys tic fib r osi s sel ec tively activ a te s hu man n eut rop hils. Jou r n al o f Leuk ocy t e 1212 Biol ogy. 2 016;100 ( 5 ) :1 047-59 . do i: 10 .11 89/ j lb.4V MA0316 -101 R. 1213 85. Nava r r e W W, Hal s ey TA, Walth er s D, Fry e J, Mc Cl ellan d M, Pot ter JL, et al. Co-re gula tion o f 1214 S al m o n e l la e nt e r i ca ge n es r e qu i r e d fo r v i r u l e nc e a n d r es is t an c e t o a n t i m i cr ob i a l p e pt i d es b y S ly A and 1215 PhoP /Pho Q. Molec ula r mic robiolog y. 20 05 ;56(2) :492-508 . doi: 10 .1111 /j.1365 -2 958.200 5.04553. x. 1216 86. So ng H, Kong W, We ath er spoo n N , Q i n G, T yler W , Tu rk J, e t al. Modula tion of the regula t o r y 1217 ac tivity of bact erial two -c omponen t s y st ems by Sl yA. The J ou rna l of biol og ical c hemi stry. 1218 200 8;283(42):28 158 -68. d oi : 10.10 74 / jb c.M8010 58200; 1 0.1074/jbc . M80105820 0. 1219 87. Kong W, Weath e rspoo n N, Shi Y. Mol ec ular mecha ni s m fo r e st a bli shm ent o f sign al- depe nd en t 1220 regul ati on in the Pho P/ Pho Q s y stem . Jou rnal of Bi olog ical Ch emis tr y . 2008;2 83 (24 ):16612 -21 . 1221 88. Ros adini C V, Zan o ni I, Odend a ll C, Gr een ER, Pac z o sa MK, Ph ilip NH, et al . A Singl e Bac t e r ia l Immun e 1222 Ev asion S t ra teg y Di sman tle s B oth My D 88 a nd TRIF S ignali ng Pa t h ways Dow n stream of TL R4. C ell Ho st 1223 Mic r ob e . 2015;1 8 (6):682 -93. doi : 10.101 6/j.c hom.2015 .11 .006. PubM ed PMID: 2 66519 44; PubM ed Ce n t ral 1224 PMC ID : PM CP MC46 85476 . 1225 89. Hajja r AM, E r n st RK, F ortuno ES, 3rd, Bra s field AS, Ya m CS, New lon LA , e t al . Hum ani z ed TLR4/ MD - 2 1226 mic e revea l LPS rec ognition di ff ere ntia ll y impa cts su scep tibil ity to Yer s ini a pe st i s and Sa lmonell a ent erica . 1227 PL oS pa t h o gen s. 2012;8 ( 1 0): e10029 63. d oi: 10.1371 / jour nal .p pat .10029 63; 10.1 3 71/journal .ppa t.10 02963. 1228 90. Montminy S W , Kha n N , Mc Gra th S, W a l kow icz MJ, Sha r p F , C o nlon JE, e t a l. Vi r ul ence fac to rs of 1229 Ye r si nia pe sti s ar e ove r c ome by a s t rong lipopol y s a cc harid e r e s pon s e. Na tur e immu nology . 1230 200 6;7(10) :1066 - 7 3. d oi: 1 0.1038 /ni138 6. 1231 91. Rile y G , Toma S . De tec t i on of pa thoge nic Yersinia e nte rocol i tica by using Con go red-magn e s i u m 1232 ox alate a ga r medi um. J ournal o f C li nica l Microbio logy. 1989;27 (1 ) :2 13 -4. 1233 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 45 92. Sk urnik M, Bolin I, Heikk inen H, P iha S, Wol f- W a tz H. Vi rule nce pla s mid -a s socia t e d 1234 autoa ggluti na t i on in Ye rsinia s pp . Journa l of Bac t eriology . 1984;1 58 (3) :1033-6 . 1235 9 3 . E l H a m i d i A , T i r s o a g a A , N o v i k o v A , H u s s e i n A , C a r o f f M . M i c r o e x t r a c t i o n o f b a c t e r i a l l i p i d A : e a s y 1236 and rapid me thod for m a ss sp ec t r omet ric c haract eriza tion . Jou r na l o f lipi d r e s earch. 200 5 ;46(8) :1773-8 . 1237 doi: 10.1194 / jlr .D5000 14 -JLR200. 1238 94. Byrap pa S, G a vin DK, G upt a KC . A highly ef ficien t proce dur e fo r s it e- sp eci fic m utage n es i s of ful l -1239 le ngth pla smid s u sing Ven t D NA polym era se. Ge nome r e sea rch. 1995;5 (4):404 - 7 . 1240 95. va n Aart s en J J, Rajak umar K . An op tim ized meth od for s ui cide vec t or-ba s ed a lle lic exc hang e in 1241 Kle bs i e lla pn eumon ia e. Journ al of micro biolo gica l me thod s. 201 1;86(3) :313-9 . doi : 1242 10. 1016/j.mimet .20 11.06.00 6 [doi ]. 1243 9 6 . D e m a r r e G , G u e r o u t A M , M a t s u m o t o - M a s h i m o C , R o w e - M a g n u s D A , M a r l i e r e P , M a z e l D . A n e w 1244 fami ly o f m obiliz able suicid e p la smid s b as ed on bro ad ho st r ange R388 pla s mi d (Inc W ) and R P4 pl a s mid 1245 (Inc P alpha ) conj ugative mac hine rie s a nd thei r cogna te Esche ric hia coli h os t strain s . R e sea r c h i n 1246 mic robiology . 2005 ;156(2 ):245 -55 . doi: S 0923-250 8(04)00256 -6 [ p ii] . 1247 97. Hoang TT , Ka r k hof f - S chw eize r RR, Ku tchma AJ , S chw eize r HP . A br oad -h os t-ran ge Flp -FRT 1248 rec ombina tion sys t e m for s i te -s pec ific e x cision of ch romo somally -l oca te d D NA s eque nce s : applic a tion for 1249 isola tion o f unmark ed P s e ud omona s ae r ugi nos a mu tant s. G e ne . 1998; 212 (1) :77- 86. 1250 98. Van De r Spo e l D, L indahl E , H e ss B, G ro enhof G , Ma rk AE, B er end sen H J. G R OM ACS: f ast, fle xible , 1251 and fre e. J Compu t Ch e m. 2005;2 6(16 ):1 701-18. doi: 10 . 1002/jcc .20291 . P ubMed PMID: 16 211538. 1252 99. Huang J, Rau sche r S, Nawroc ki G , Ra n T, Feig M, de Gro ot BL , e t al . CHAR MM3 6m: an improve d 1253 force fi e ld fo r fol ded and int rin sica lly di s order ed p rot ein s. N a t M eth od s. 2017 ;1 4 (1):71 -3. Epub 2 0161107. 1254 doi: 10.1038 / nm e t h .4 067. Pu bMe d PMID: 278196 58; Pub Med Cen tr a l PMC I D : P MCP MC519 9616. 1255 100 . Kim S, Le e J, Jo S , B rook s CL, 3rd , L ee H S , Im W. C H AR MM -GUI liga nd r ead e r and model e r fo r 1256 CHARM M for c e f ield g ener ation o f s ma ll molec ule s. J Compu t Che m. 2017;3 8(21 ): 1879-86 . Epub 2 0170511. 1257 doi: 10.1002 / jc c.24 829. Pub Me d PM ID : 284976 16; PubMed Cen tr a l PM CI D: P MC P MC5488 718. 1258 101 . Pa r ri nel lo M , R ahman A. P oly mor p hic t ran si tion s in s in gle crys t a l s : A new m olec ular dynamic s 1259

. Journa l o f Applie d Phys ic s. 1 981;52(12) :7 182-90 . doi : 10.10 63 / 1. 328693. 1260 102 . Es smann U, Pe rer a L, Berko witz ML , D a rde n T, Le e H , P eder sen L G. A smo oth pa rticl e me sh Ewal d 1261

. The Jou rnal o f Ch emic al Ph y s i cs . 1995;1 03(19) :8577-9 3 . doi: 1 0.1063 /1.4 70 117. 1262 103 . Humphrey W , Dalk e A, Sc hult en K. V MD : visu al mo l ecul ar d ynamic s. J M ol Grap h. 1 996;14 (1) :33-8 , 1263 27-8. doi : 10. 1016/0263 -785 5(96 )00018- 5. PubM ed PM ID : 87 44570. 1264 104 . Dova la D, R ath CM, H u Q, Saw yer WS, S hia S, Elli ng R A, e t al . S tructu re -guide d e nz y mology of t h e 1265 lip id A ac yltran s fe r a s e Lpx M r e ve al s a dual a ctivity mecha nism . P roc Natl Ac ad Sc i U S A. 1266 201 6;113(41):E 6064 -E71. E pu b 201609 28 . doi: 10.1073 / pna s .16 1074611 3. P ubMed PMID: 2 7681620 ; 1267 PubMe d Cen tral PM CID: P MC PM C506 8295. 1268 105 . Te r w illig er T C , Lieb sc hn er D, C roll TI, W i l liams C J, M c Coy A J, Po on BK, e t al . Alp h aFold p redi cti on s 1269 are val ua ble hypo the se s and ac c eler at e but do no t r eplac e e xperimen t a l s t r uc t u r e d et ermina tion. Na t 1270 Metho ds . 2024; 21 (1):110 -6. E pub 20231 1 30. doi: 10 .1038 / s 4 1592-023 -02 087-4 . PubMe d PM ID: 3 8036854 ; 1271 PubMe d Cen tral PM CID: P MC PM C107 76388. 1272 106 . John son MS, M ay AC , Rodio nov MA, O v erington J P. D i scrimi na tion o f c om mon p r ot ein fold s: 1273 app lica t i on o f pro tein s t ruc ture to se qu ence /s t ruc tu r e c omp ari son s . Me thod s E nz y mol. 1996;2 66:575 -98. 1274 doi: 10.1016 /s0076 -687 9(96 )66036 - 4 . Pu bMed PM ID : 87 43707. 1275 107 . L ehtonen J V , Still D J, Ran ta ne n V V, Ekhol m J, Bjorklund D, If tikha r Z, et a l . B ODI L: a m ol ecul ar 1276 mode ling env ir o nmen t for s tr uctur e- f unctio n a naly si s a nd dr ug d es ig n . J Compu t Aid ed Mol De s. 1277 200 4;18(6) :401-19 . doi: 10 .1007 / s 10 822- 004-37 52-4 . PubM ed P MID : 156630 01. 1278 108 . Sa li A, Bl und ell TL . C ompara tive p rot ein modell ing by s a ti sfac ti on o f s pati al r e s t rai n ts. Jou r na l of 1279 Molec ular Bi o logy. 1993;234 (3 ):779 -815. d oi: S00 22-2836(83 )7 1626-8 [pii ] . 1280 109 . Robe rt X, Gouet P . D ec iphering k ey fea t ure s i n p r otei n st r u ctu re s wi t h the new EN Dscri pt se r v e r . 1281 Nucle ic Ac id s R e s. 201 4;42 (Web S erve r i s sue ): W 3 20-4 . Epub 2 0140421. doi : 10.1 09 3/ na r/gk u316. Pu bMed 1282 PMID: 247 53421; P ubMed Cen tral PM CID: P MC P MC4 086106. 1283 110 . Beng oech e a JA, Diaz R, Mo riyo n I. Oute r mem bra ne di f fere nce s b etwe e n pathog enic an d 1284 env ironmen tal Yer s in ia ent eroc oli tica b iogroup s prob ed with hy dropho bic per meant s and polyc a tionic 1285 pep t id e s. In fec tion a nd immuni t y . 1996 ; 64(12):4891 -9. 1286 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 46 111 . Hirs c h feld M, M a Y, W ei s JH, Vog el S N , Wei s J J . Cut ting e dg e: r epuri fica t i o n o f l i popolys ac charide 1287 el iminat es sign aling th rough both hu man and murine toll -li ke rec epto r 2. J ou r na l o f immu nology 1288 (Bal timo r e , Md: 1950) . 2000;165 ( 2 ) : 618 - 22. doi : ji_v 165n2p61 8 [pii ]. 1289 1290 1291 1292 1293 1294 1295 1296 FIGURE LEGENDS 1297 Figure 1. Characterization of Yersiniae lipid A. 1298 Negative ion MALDI/i6 TOF mass spectrometry spectra of lipid A purified from (A) YeO3 grown at 1299 210C, (B) YeO9 grown at 21 0C, (C) YeO3 grown at 37 0C, (D) YeO9 grown at 37 0C, (E) Ye1A 1300 strain IP266118 grown at 210C, (F) Ye1A strain IP266118 grown at 37 0C, (G) Y. aldovae grown at 1301 210C, and (H) Y. aldovae grown at 37 0C.. Data represent the mass /i6 to/i6 charge (m/z) ratios of each 1302 lipid A species detected and are representative of three extractions. 1303 Figure 2. Structural characterization of Yersiniae lipid A enzymes. 1304 Negative ion MALDI /i6 TOF mass spectrometry spectra of lipid A purified from (A) Y. aldovae 1305 expressing lpxR from YeO8 grown at 21 0C, (B) Y. aldovae expressing lpxR from YeO8 grown at 1306 370C, (C) YeO3 grown at 37 0C, (C) E. coli BN2, (D) E. coli BN2 complemented with lpxM from 1307 Y. aldovae, (E) YeO8 lpxP mutant grown at 21 0C, (F) YeO8 lpxP mutant complemented with lpxP 1308 from Y. aldovae grown at 21 0C, (G) E. coli BN1 lpxL mutant, (H) E. coli BN1 lpxL mutant 1309 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 47 complemented with lpxL from Y. aldovae. Data represent the mass /i6 to/i6 charge (m/z) ratios of each 1310 lipid A species detected and are representative of three extractions. 1311 Figure 3. Defining Yersiniae LpxL hydrophobic tunel 1312 A deep hydrophobic tunnel is formed in all LpxL models and may serve as the binding site for an 1313 acyl chain. The surface of the tunnel is color coded by hydrophobicity in PyMOL, where gray 1314 shows charged residues and red hydrophobic residues. In all models, the amino acids contributing to 1315 the tunnel are mostly hydrophobic. (A) E. coli LpxL with C12 (cyan sticks). (B) YeO8 LpxL with 1316 C14 (yellow sticks), (C) Y. aldovae LpxL with C12 (cyan sticks), (D) Y. aldovae LpxL with C14 1317 (yellow sticks). (E) Negative ion MALDI/i6 TOF mass spectrometry spectra of lipid A purified from 1318 E. coli BN1 lpxL mutant complemented with lpxL from YeO8 in which L146 is mutated to 1319 phenylalanine. Data represent the mass/i6 to/i6 charge (m/z) ratios of each lipid A species detected and 1320 are representative of three extractions. 1321 Figure 4. Molecular dynamics simulations of TLR4 2/MD-22 heterodimer with lipid A species . 1322 (A) Initial snapshot of the (TLR4/MD-2)2 complex bound to m/z 1768 (left). Final snapshots of each 1323 one side of the (TLR4/MD-2) 2 complex bound to m/z 1768 (right). Labelled basic amino acid 1324 residues observed to make significant hydrogen-bonding interactions with m/z 1768 are shown in 1325 licorice format, coloured in blue. (B) Root mean square deviation (RMSD) of TLR4 dimer 1326 backbone atoms. (C) Buried solvent accessible surface area (SASA) between MD-2-lipid A species 1327 and TLR4. (D) Contacts between lipid A species and (TLR4/MD-2) 2. (E) Overlaid every 200 ns 1328 simulation snapshots of F126 side chain and its orientation with respect to the MD-2 :lipid A 1329 species. The pseudo trajectory was made of two aligned MD-2:lipid monomer frames for each 1330 system. The F126 side chain is shown in licorice representation and coloured according to 1331 simulation time (red-white-blue for 0 to 2,000 ns). (F) RMSD of MD-2 backbone loop region after 1332 alignment on the entire MD-2 backbone. (G) Porcupine plot of TLR4 dimer alpha carbons 1333 corresponding to the most dominant motion (PCA1). The colours corresponds to distances as 1334 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 48 labelled inset. In panel (A) and (E) protein is shown in cartoon representation: TLR monomers in 1335 violet and red, MD-2 dimers shown in grey. Lipid A species and side chains are shown in licorice 1336 representation (cyan – carbon, red – oxygen, blue – nitrogen, brown – phosphorus). All values in 1337 panels (B)-(D) and (F) were averaged over last 200 ns and across both dimers. 1338 Figure 5. Engagement of Yersinia lipid As with TLR4 and subsequent inflammation. 1339 (A) Surface expression of TLR4 as measured by mean fluorescence intensity (MFI) on iBMDMs 1340 non-infected (ni) or exposed to live E. coli , pathogenic yersiniae, YeO8 and YeO3, and non-1341 pathogenic yersiniae, 1A strain 0902, Y. aldovae, Y. nurmii , grown at 21 0C (denoted as 21) and 1342 370C (denoted as 37). (B) ) Surface expression of TLR4 as measured by MFI on iBMDMs non-1343 infected (ni) or exposed to live YeO8, chimeric YeO8 strain expressing a non-pathogenic lipid A, 1344 strain YeO8NPL, and YeO8 lpxR mutant (Δ lpxR, strain YeO8- Δ lpxR) grown at 21 0C (denoted as 1345 21) and 37 0C (denoted as 37). (C) TNF α secretion by infected iBMDMs with YeO8 ,and 1346 YeO8NPL.“c” denotes bacteria without the virulence plasmid. Strains were grown at 21°C (denoted 1347 as 21) and 37°C (denoted as 37). (D) TNF α secretion by iBMDMs challenge with 10 ng/ml of 1348 repurified LPS from YeO8, YeO8NPL, and YeO8 lpxR mutant (Δ lpxR, strain YeO8-Δ lpxR) grown 1349 at 370C. 1350 Data are presented as mean ± SD (n /i6 =/i6 3). In panels A and B, **** P/i6≤/i6 0.0001; ns, P/i6 >/i6 0.05 1351 for the comparisons against non-infected cells, and # P /i6≤/i6 0.0001 for the comparisons against 1352 YeO8 grown at 37 oC using One /i6 way ANOVA with Bonferroni contrast for multiple comparisons 1353 test. In panels C and D, **** P/i6≤/i6 0.0001; ns, P/i6 >/i6 0.05 for the indicated comparisons using 1354 One/i6 way ANOVA Dunnett’s multiple comparisons test. 1355 Figure 6. Susceptibility of Yersiniae strains to antimicrobial peptides. 1356 (A) Susceptibility of pathogenic and non-pathogenic Yersiniae strains grown at 21oC or 37oC to 1357 polymyxin B (10 µg/ml). (B) Susceptibility of pathogenic and non-pathogenic Yersiniae strains 1358 grown at 21oC or 37oC to magainin II (10 µg/ml). The data are presented as mean ± SD (n/i6 =/i6 3). 1359 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 49 Figure 7. Effect of the lipid A PAMP on Yersinia virulence. 1360 (A) Motility assay with YeO8 and YeO8NPL in a semisolid agar plate (0.3% agar and 1% 1361 tryptone). Plates were incubated at 22°C for 24 h. (B) Quantification of the diffusion diameter of 1362 YeO8, and YeO8 lpxR mutant (Δ lpxR, strain YeO8-Δ lpxR), YeO8 lpxL mutant complemented with 1363 YeO8 lpxL ( Δ lpxL::lpxLYeO8) and YeO8 lpxL mutant complemented with Y.aldovae lpxL 1364 (Δ lpxL::lpxLYa,). (C) Analysis of flhDC expression by YeO8, YeO8NPL, YeO8 lpxL mutant 1365 complemented with YeO8 lpxL ( Δ lpxL::lpxLYeO8), YeO8 lpxL mutant complemented with 1366 Y.aldovae lpxL ( Δ lpxL::lpxLYa,)YeO8 lpxR mutant ( Δ lpxR, strain YeO8- Δ lpxR), and YeO8 lpxL 1367 mutant ( Δ lpxL, strain YeO8- Δ lpxL) carrying the transcriptional fusion flhDC::lucFF grown at 1368 21°C. (D) Analysis of yplA expression by YeO8, YeO8NPL, YeO8 lpxL mutant complemented 1369 with YeO8 lpxL ( Δ lpxL::lpxLYeO8), YeO8 lpxL mutant complemented with Y.aldovae lpxL 1370 (Δ lpxL::lpxLYa,), YeO8 lpxR mutant (Δ lpxR, strain YeO8- Δ lpxR), and YeO8 lpxL mutant ( Δ lpxL, 1371 strain YeO8-Δ lpxL) carrying the transcriptional fusion yplA::lucFF grown at 21°C.(E) Invasion of 1372 HeLa cells by YeO8, YeO8NPL, YeO8 lpxL mutant complemented with YeO8 lpxL 1373 (Δ lpxL::lpxLYeO8), YeO8 lpxL mutant complemented with Y.aldovae lpxL ( Δ lpxL::lpxLYa,), YeO8 1374 lpxR mutant ( Δ lpxR, strain YeO8- Δ lpxR). Invasion assays were done in triplicate without 1375 centrifugation (n/i6 =/i6 3).(F) Adhesion to HeLa cells by YeO8, YeO8NPL, and YeO8 lpxR mutant 1376 (Δ lpxR, strain YeO8- Δ lpxR). Adhesion assays were done in triplicate without centrifugation 1377 (n/i6 =/i6 3) (G) Analysis of inv expression by YeO8, YeO8NPL, YeO8 lpxL mutant complemented 1378 with YeO8 lpxL ( Δ lpxL::lpxLYeO8), YeO8 lpxL mutant complemented with Y.aldovae lpxL 1379 (Δ lpxL::lpxLYa,), and YeO8 lpxR mutant ( Δ lpxR, strain YeO8- Δ lpxR) carrying the transcriptional 1380 fusion inv::lucFF grown at 21°C. (H) Analysis of rovA expression by YeO8, YeO8NPL, YeO8 1381 lpxL mutant complemented with YeO8 lpxL (Δ lpxL::lpxLYeO8), YeO8 lpxL mutant complemented 1382 with Y.aldovae lpxL (Δ lpxL::lpxLYa,), and YeO8 lpxR mutant (Δ lpxR, strain YeO8-Δ lpxR) carrying 1383 the transcriptional fusion rovA::lucFF grown at 21°C. (I) SDS-PAGE (the acrylamide concentration 1384 was 4% in the stacking gel and 12% in the separation one) and Coomasie brilliant blue staining of 1385 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 50 proteins from the supernatants of Ca 2+- deprived cultures from YeO8, YO8NPL and YeO8 lpxR 1386 mutant (Δ lpxR, strain YeO8- Δ lpxR). Result is representative of three independent experiments. (J) 1387 SDS-PAGE (the acrylamide concentration was 4% in the stacking gel and 10% in the separation 1388 one) followed by Coomasie brilliant blue staining of cell extracts from YeO8, YO8NPL and YeO8 1389 lpxR mutant (Δ lpxR, strain YeO8- Δ lpxR) grown in RPMI 1640 at 37°C. YadA protein is marked. 1390

is representative of three independent experiments. (K) CFUs per gram of Peyer’s patches 1391 from male and female mice infected with YeO8, YeO8NPL, YeO8 lpxR mutant ( Δ lpxR, strain 1392 YeO8-Δ lpxR) and YeO8 lpxR-lpxL double mutant (Δ lpxR Δ lpxL, strain YeO8-Δ lpxR- Δ lpxL). 1393 Data are presented as mean ± SD (n /i6 =/i6 3). In all panels, **** P/i6≤/i6 0.0001; *** P/i6≤/i6 0.001; 1394 **P/i6≤/i6 0.01; ns, P /i6 >/i6 0.05 for the comparisons against non-infected cells, and # P/i6≤/i6 0.0001 1395 for the comparisons against YeO8 using One /i6 way ANOVA with Dunnett’s multiple comparisons 1396 test. 1397 1398 SUPPLEMENTARY FIGURES 1399 Figure S1. Structural characterization of Yersiniae lipid A enzymes. 1400 Negative ion MALDI /i6 TOF mass spectrometry spectra of lipid A purified from (A) E. coli BN2 1401 complemented with lpxM from Y. bercovieri, (B) E. coli BN2 complemented with lpxM from Y. 1402 intermedia, (C) YeO8 lpxP mutant complemented with lpxP from Y. bercovieri grown at 210C, (D) 1403 YeO8 lpxP mutant complemented with lpxP from Y. intermedia grown at 21 0C, (E) E. coli BN1 1404 lpxL mutant complemented with lpxL from YeO8, (F) E. coli BN1 lpxL mutant complemented with 1405 lpxL from Y. bercovieri, (G) E. coli BN1 lpxL mutant complemented with lpxL from Y. molaretii. 1406 Data represent the mass /i6 to/i6 charge (m/z ) ratios of each lipid A species detected and are 1407 representative of three extractions. 1408 Figure S2. Alignment of Yersiniae LpxL homologs. 1409 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 51 The residues are numbered according to YeO8 LpxL. Conserved residues are highlighted in green 1410 and residues contributing to the hydrophobic tunnel are marked with black triangles under the 1411 alignment. The leucine (pathogenic Yersinia) to phenylalanine (non-pathogenic Yersinia) mutation, 1412 which enables binding of C12 and C14 to LpxL in non-pathogenic strains, is highlighted in pink. 1413 The residues that affect the conformation of this phenylalanine in E. coli and makes E. coli LpxL 1414 selective for only C12 are highlighted in yellow. There are four conserved motifs shared with the 1415 GPAT family (highlighted in cyan); HX4D/E (motif 1), a conserved arginine (motif 2), a conserved 1416 negatively charged residue (motif 3), and a conserved proline (motif 4). 1417 Figure S3. Modelling the lipid binding site in LpxL. 1418 (A) E. coli LpxL with C12 (orange sticks) bound in the hydrophobic tunnel. The suggested 1419 hydrocarbon ruler, F143 (pink sticks), is located in the bottom of the tunnel and interacts with the 1420 carbon tail of C12 to form a stable complex. F143 is unlikely to swing away to make space for C14 1421 due to the more polar environment created by Q142. (B) Comparison of C12 (orange sticks) and 1422 C14 (light pink sticks) in the hydrophobic tunnel of E. coli LpxL. The suggested hydrocarbon ruler, 1423 F143 (pink sticks), restricts the tunnel space, causing C14 to stick out into the main groove and a 1424 productive complex for catalysis is unlikely to be formed. (C) YeO8 LpxL with C14 (light pink 1425 sticks) bound in the hydrophobic tunnel. L146 (pink sticks) is the suggested hydrocarbon ruler and 1426 its smaller side chain creates space for C14 to bind, while making it unable to form proper 1427 interactions with the shorter C12 for a stable complex. (D) YeO8 and Y. aldovae LpxL with C12 1428 (orange sticks) and C14 (light pink sticks) bound in the hydrophobic tunnel. The amino acid 1429 numbering is according to Y. aldovae and the only difference in the tunnel is the hydrocarbon ruler 1430 (pink sticks); L146 for YeO8 LpxL and F142 for Y . aldovae LpxL. F142 in Y. aldovae LpxL can 1431 probably swing away to create more space in the tunnel and provide a stable interaction site for 1432 C14, since the surrounding amino acids L 112, I141 and L253 are small hydrophobic residues, 1433 which can interact with both F142 and the carbon tail of C14. 1434 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint 52 Figure S4. Structural characterization of the lipid A produced by YeO8NPL. 1435 Negative ion MALDI /i6 TOF mass spectrometry spectra of lipid A purified from (A) YeO8NPL 1436 grown at 21 oC, (B) YeO8NPL grown at 37 oC. Data represent the mass /i6 to/i6 charge (m/z) ratios of 1437 each lipid A species detected and are representative of three extractions. 1438 Figure S5. Effect of the lipid A PAMP on Yersinia virulence. 1439 (A) Actin disruption by Yersinia infection. A549 cells were non-infected (Control), and infected 1440 with YeO8, YeO8 yopE mutant (Δ yopE, strain YeO8-Δ yopE), YeO8NPL, and YeO8 lpxR mutant 1441 (Δ lpxR, strain YeO8-Δ lpxR). After fixing and permeabilization of cells actin was stained with 1442 Orange-red-phalloidin (1∶ 100) and cells were analyzed by fluorescence microscopy. Result is 1443 representative of three independent experiments. (B) Quantification of the percentage of A549 cells 1444 showing actin disruption. Cells were non-infected (Control), and infected with YeO8, YeO8 yopE 1445 mutant (Δ yopE, strain YeO8-Δ yopE), YeO8NPL, and YeO8 lpxR mutant (Δ lpxR, strain YeO8-1446 Δ lpxR). The number of cells analysed are indicated on top of each of the bars. (C) Assessment of 1447 binding of YeO8, YeO8NPL, and YeO8 cured of the pYV virulence plasmids to collagen-coated 1448 coverslips. (D) Quantification of the number of bacteria bound to collagen-coated coverslips. 1449 Data are presented as mean ± SD (n/i6 =/i6 3). In all panels, ****P/i6≤/i6 0.0001; ns, P/i6 >/i6 0.05 for the 1450 comparisons against the results of YeO8 using One /i6 way ANOVA with Dunnett’s multiple 1451 comparisons test. 1452 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 3, 2025. ; https://doi.org/10.1101/2025.03.03.641127doi: bioRxiv preprint