Molecular basis of the lipid-induced MucA-MucB dissociation in Pseudomonas aeruginosa

MucA and MucB are critical negative modulators of sigma factor AlgU and regulate the mucoid conversion of Pseudomonas aeruginosa. Previous studies have revealed that lipid signals antagonize MucA-MucB binding. Here we report the crystal structure of MucB in complex with the periplasmic domain of MucA and polyethylene glycol (PEG), which unveiled an intermediate state preceding the MucA-MucB dissociation. Based on the biochemical experiments, the aliphatic side chain with a polar group was found to be of primary importance for inducing MucA cleavage. These results provide evidence that the hydrophobic cavity of MucB is a primary site for sensing lipid molecules and illustrates the detailed control of conformational switching within MucA-MucB in response to lipophilic effectors.

T o adapt to changing environments, Gram-negative bacteria usually use the envelope stress responses (ESR) system to coordinate gene expression 1 . In pathogenic bacteria, ESR controls numerous cellular processes such as, virulence factor production, motility, antibiotic resistance, and bacterial survival 1,2 . In most ESR systems, extra-cytoplasmic function sigma factors (σ ECF ) are crucial transcription elements required for stress signal transmission and amplification 3 . σ ECF is specifically inhibited by the cognate anti-sigma factors, whereas the relief of this inhibition is frequently regulated by a cascade of cleavage reactions named regulated intramembrane proteolysis (RIP) 4 .
AlgU (also known as AlgT or σ 22 ) is the key σ ECF of P. aeruginosa. It is responsible for transduction of the extracellular stimulus that regulates mucoid transition of P. aeruginosa in cystic fibrosis (CF) patients 5 . MucA and MucB are cognate antisigma factors for AlgU activity 5 . In non-mucoid strains, MucB forms a stable complex with MucA that serves as a fine-tune control mechanism that protects MucA from cleavage. It has been proven that misfolded outer-membrane proteins (OMPs) can activate periplasmic protease AlgW and thus initiate the proteolysis of MucA by cleaving its periplasmic domain 6 . Increasing evidences indicate that the major signals for MucA proteolysis require both unfolded OMPs and off-pathway lipopolysaccharides (LPS) induced by extracellular stress 7,8 .
Because of the physiological importance of mucA and mucB, their mutations are commonly associated with a persistent and mucoid phenotype of P. aeruginosa that has been isolated from the lungs of cryptogenic fibrosing alveolitis (CFA) patients 9 .
Recently, a structural studies of MucB, MucA peri -MucB, and MucA cyto -AlgU revealed a redox-sensitive stress-response mechanism in MucB 10 . However, despite intense investigations, there remain many unanswered questions about MucA/MucBassociated RIP signaling. Of interest is determining how signal molecules such as LPS influence the MucA-MucB complex and release the MucA for AlgW degradation. In this study, we provide structural evidence that the hydrophobic core-occupied MucB undergoes dramatic conformational changes in the regions of residues 92-113, 192-216 and 230-237, exposing the cleavage site of MucA to the solvent. We also show that the fatty acid moiety is crucial for inducing the release of MucA from MucB. Using sitedirected mutagenesis and a functional assay, we verified the roles of the critical residues involved in the MucA-MucB interaction and MucB-lipid association, providing experimental support to interpret the mechanism of the mucA/mucB-controlled mucoid phenotype in P. aeruginosa.

Results
MucB exclusively protects MucA peri from AlgW degradation. AlgW, MucA, and MucB are the functional equivalents to DegS, RseA, and RseB from Escherichia coli and all sense and transmit outer-membrane stress through similar mechanisms 7,8,11 . AlgW and DegS both belong to the PDZ-containing serine proteases and possess high sequence identity (42.5%), but different substrate specificities 7 . To evaluate the function and exchangeability of the Muc-and Rse-systems, we overproduced and purified the periplasmic parts of MucB/RseB and AlgW/DegS (without the signal peptide and transmembrane domain) to reconstitute the proteolysis events of the periplasmic regions of MucA (MucA peri : residues 106-194) and RseA (RseA peri : residues 120-216). Both MucB and RseB were tested in two sets of protease cleavage experiments in the presence of a peptide agonist (YVF) 7 . MucB and RseB worked well to prevent cleavage in their parental systems but could not be substituted for each other, suggesting that the requirement for the strict choice of functional elements might lie in the compatibility of MucA/RseA and MucB/RseB (Supplementary Fig. 1A). Next, we systematically investigated the profile of the MucB-controlled AlgW cleavage toward MucA. As Supplementary Fig. 1B shows, the protective effect of MucB on MucA was not persistent and decreased after prolonged incubation. Nonetheless, a certain amount of MucA was intact after 1 h of reaction. By comparison, MucB suppressed AlgW action in a ratio-dependent manner in that full protection of MucA required at least an equal molar equivalent of MucB.
To further investigate the specific recognition between MucA peri and MucB, we measured their binding affinity (K d value of 1.5 µM) by using isothermal titration calorimetry (ITC) and successfully obtained the MucA peri -MucB complex in a 1:1 ratio via a Ni-NTA column and subsequent gel-filtration purification (Supplementary Fig. 2A, B). The modest interaction but moderate selectivity and stability of the MucA-MucB complex might be indicative of the exquisite control over the RIP under diverse stress conditions.
Overall structure of MucA peri -MucB. The MucA peri -MucB was crystallized and the structure was solved by molecular replacement (MR) using the N-terminal domain (NTD: residues 25-208) and C-terminal domain (CTD: residue 209-315) of E. coli RseB as the MucB model. Based on the electron density map generated from the MR, we were able to build residues 145-191 of MucA. The final structure was refined to 1.9 Å, with R work = 0.1729 and R free = 0.1929 (Supplementary Table 1).
The RseA peri -RseB complex (PDB code: 3M4W, the sequence identity between MucB and RseB, MucA and RseA are 30.16% and 34.15%, respectively) revealed that the RseA peri 142-151 is buried in the hydrophobic pocket of RseB, and thus avoids cleavage by DegS 13 . In our MucA peri -MucB complex structure, the N terminal of MucA peri (residues 106-145) is too flexible to be detected in the crystal structure (Fig. 1). Remarkably, the MucA peri -MucB complex structure revealed two extra binding regions. First, MucA 145-158 , corresponding to the structureunsolved RseA 159-168 segment, exhibited an αβ turn conformation and formed antiparallel main-chain hydrogen bonds with the βO of MucB. Second, in contrast with the invisible RseA 190-216 , MucA 181-191 adopted a helical conformation (αII) and bound to MucB-NTD. As was observed for the primary binding in αI, the similar extra binding patterns observed in MucA-MucB may also exist in the Rse system and we speculate that the MucA peri -MucB complex structure represents an intermediate state (inter-state) that is different from the fully protected state (protect-state) presented in the RseA peri -RseB complex 13 .
The conformational gating of MucB and release of MucA. Both the apo and RseA peri -bound forms of RseB have been solved [12][13][14] , and both show that the spatial arrangement of the N-and Cterminal domains is rigid. Such structural rigidity is attributed to the extensive inter-domain interactions ( Supplementary Fig. 3), including either direct hydrogen bonding or hydrophobic interactions represented by a buried inter-domain surface area of 925.9 Å 2 . In MucB, the central residues such as, Y119, R146, Y147, and D247 are structurally conserved with those of RseB. However, the inter-domain interface area is slightly smaller (847.1 Å 2 ) ( Supplementary Fig. 3). Structural superposition of the MucA peri -bound MucB with the RseA peri -bound RseB gave a rmsd (root mean square deviation) value of 1.85 Å for 239 equivalent Cα (Fig. 2a). The major conformational variations are at three loop regions including residues 92-113, 192-216, and 230-237.
By comparing with the recently reported MucB-apo (6IN8) structure 10 , we found that the main conformational variations occurred in the C-terminal domain and the loop region containing residues 92-113 (Fig. 2c). In MucB-apo structure, the loop 92-113 folded into the cavity of MucB, while in MucAcomplexed form, MucB 92-113 shifted outward to generate adequate space for MucA binding (Fig. 2c). Another proline residue P106 was found to be involved in this structural transition. This rearrangement allowed MucA to form αII, and reduced the interactions between the C and N domains of MucB. Therefore, MucA binding initiates the MucB 92-113 rearrangement and MucA αII anchoring, leading MucA N-terminal cleavage site to be exposed to the solvent. This conversion might facilitate the subsequent AlgW coupling.
Interactions between MucA peri and MucB. The periplasmic domain of membrane-spanning anti-sigma factors from different species were aligned together based on the structure superposition between MucA peri and RseA peri (Fig. 3a). The selected homologs were divided into Rse-and Muc-type groups. The sites that were observed or predicted to be involved in hydrogen and charge-charge interactions were aligned as shown in Fig. 3a, b. The primary binding between MucA/RseA αI and the MucB/ RseB C-terminal domain represent a universal determinant for anti-sigma factors coupling because the RseA peri truncation with αI alone still interacts with RseB 12 . Fig. 1 The side-by-side overall stereo structure views of the MucA peri -MucB and RseA peri -RseB complexes. MucAperi-MucB complex (top) and the RseAperi-RseB complex (PDB code: 3M4W, bottom). The MucA peri is colored in orange, the NTD, CTD of MucB are colored in green and blue. A bound PEG molecule is shown as hot-pink sticks (top). Accordingly, the RseA peri is colored in magenta, the NTD, CTD of RseB are colored in cyan and yellow, respectively. These figures were generated by PyMOL (http://www.pymol.org).
To understand the contribution of residues that participate in MucA peri -MucB interactions (Fig. 3c), a variety of MucA mutants were generated and subjected to the His-affinity pulldown experiments (Fig. 3d). First, most of the MucA mutants still retained the ability to form complexes with MucB except for three residues in the αI region (W158, R162, and H170A), suggesting that the extensive interaction network may provide evolvability that underlies the anti-sigma factor recognition specificity. Second, the existence of redundant contacts, especially the fact that the αII region can be deleted without losing the MucA peri -MucB association, also increased the structural flexibility and facilitated the conformational changes that mediated the transition from protect-state to inter-state.

The binding pocket for the lipid-A moiety of LPS in MucB.
MucB and RseB can function as periplasmic LPS sensors, recognizing and interacting with the lipid-A portion 8 . The MucB-NTD forms a U-shape beta-half-barrel and a large hydrophobic cavity with an approximate surface area of 3500 Å 2 , making this domain the most likely to bind to LPS. The MucA peri -MucB complex structure had a polyethylene glycol (PEG) molecule bound, which may be incorporated during crystallization (  10 . In our PEG-bound structure, MucB 92-113 is stable and can be modeled. The circular form of this PEG molecule, although different from the liner form of lipid or PEG molecules in other lipoproteins such as LolB 15 , provides an opportunity to visualize the binding site for the acyl chains of lipid-A. Notably, most hydrophobic residues that line the PEGbinding pocket are conserved among MucB/RseB homologs ( Fig. 3b and 4a), supporting the view that binding of "off-pathway" LPS molecules is a common feature in this family 8 . The limited diameter (13-17 Å) of this concave cavity suggests that MucB may bind a single lipid-A molecule at this site. In a recently reported protein structure in complex with LPS (6S8H) 16 , lipid-A occupies a range of 5-19 Å, which confirms that the MucB cavity is able to accommodate part of the lipid-A moieties. The similar binding regions of PEG in MucA peri -MucB and the RseB 108-117 in RseA peri -RseB (Fig. 2a) indicates that the rearrangements of the MucB-NTD and MucB 92-113 regions may be associated with LPS binding (Fig. 2c).
The L-IIA unit of LPS has been identified as the minimal active fragment in MucA peri -MucB dissociation ( Fig. 4b) 8   MPD; isopropanol; glycerol), disaccharides (maltose, β-D-glucopyranosyl-D-glucose (β-DGDG)) and fatty acids with various chain lengths (C8-C16) were chosen and separately added to the AlgW-mediated MucA peri proteolysis system. After a 30 min incubation at 37°C, the samples were subjected to SDS-PAGE (Fig. 4b). Lipid-A and LPS exhibit positive effects to relief the AlgW cleavage inhibition caused by MucB, whereas boiled LPS under alkali solution which would generated activate fragment L-IIA could induce MucA degradation of the same magnitude as that induced by lipid-A 8 (Fig. 4b). Similar to lipid-A, detergents consisting of a saccharide moiety and a linear alkyl tail exhibited significant effects on MucA peri cleavage, and the detergent with longer acyl chain had more obvious effect ( Supplementary Fig. 5A). Instead, MucA peri remained intact when exposed to PEGs, organic solvents and disaccharides, while the fatty acids showed certain degrees of activation on MucA peri degradation. These data suggest that an alkyl chain with a hydrophilic head group are both functional groups of lipid-A for inducing MucA peri release. This speculation was supported by the subsequent experiment (Supplementary Fig. 5B) in which the lauric acid (C12)-induced MucA peri cleavage process was obviously accelerated by adding disaccharides. In addition, a lower band (around 25 kDa) was observed  under MucB in all cleavage results (Fig. 4b, d; Supplementary  Fig. 5), which was proved to be the cleavage products of MucB by using mass spectrometry (MS) (Supplementary Fig. 6). This specific cleavage on MucB (the RseB is not cleaved by DegS) reflects a novel pattern in MucA/MucB system and is probably a new mechanism for P. aeruginosa to respond to external signal stimuli.
To validate the effects of the alkyl chain, DDM was used as the amphiphilic effector in the following MucA peri degradation experiments. First of all, we performed a competitive protection experiment against DDM-induced MucA peri cleavage in the presence of PEG (Supplementary Fig. 7). In this experiment, the PEG competed with the DDM for the binding site in MucB and the protective effect was observed to be concentration-dependent, suggesting that the PEG and amphiphilic effector binding sites are overlapped. At the same time, the lipid-ELISA assay confirmed the direct and specific binding of lipid-A to MucB but not to MucA peri (Fig. 4c). However, MucB-NTD exhibited lower binding to Lipid-A than that of MucB. We speculated that the CTD was necessary to relief the "plug" effect of loop 92-113 on the lipid-binding pocket. Based on our structural analysis and structural comparison (Fig. 2c), proline residues P106 and P112 are involved in the dynamic behavior of the loop 92-113 . Because of the unique features, proline residues are usually important for the site-specific flexibility in protein structure. Therefore, we introduced Ala-substitutions on the two proline residues in MucB-NTD loop 92-113 to reduce its flexibility and thus to promote lipid-A binding. As we expected, the MucB-NTD-P106A and MucB-NTD-P112A mutants exhibit enhanced binding to lipid-A (Fig. 4c). Furthermore, we carried out site-directed mutagenesis on hydrophobic residues surrounding the bound PEG molecule. However, all of the Ala-substitutions in this region significantly suppressed or even abolished MucB production, indicating that the hydrophobic core is indispensable for its lipoprotein fold. Alternatively, Trp substitutions were introduced at five positions (L31W, F40W, L151W, L159W, and F176W) where the larger hydrophobic side chain would retain the protein stability but sterically hinder the interactions between amphiphilic effectors and MucB. As expected, all five Trp substitutions in MucB retained its protection effects on MucA ( Supplementary  Fig. 8), and the lipid-ELISA assay revealed that all of the five mutants were obviously impaired in their ability to binding lipid-A (Fig. 4c). However, the small alterations of the concave surface may not totally abolish lipoprotein function 17 , accordingly, mutants F40W, L151W, L159W, and F176W still remained the sensing ability of amphiphilic effector (Fig. 4d). Even so, we identified that residue L31 was a "hot spot", in which the Trp substitution significantly decreased the sensitivity of MucB to DDM, this effect is likely attributable to its direct effects on lipidbinding and/or movements in the tunnel.
MucA/MucB variants affect the alginate biosynthesis. MucA and MucB are negative controllers of alginate production and their mutation results in the obvious mucoid phenotype conversion in P. aeruginosa. Based on the structural analysis, we further investigated the MucA mutations that disrupted the MucA peri -MucB interactions (W158A, R162A, and H170A in MucA, Y166A was also included for comparison) or adversely affected the DDM-induced MucA peri degradation (L31W in MucB) in vivo to elucidate their influences on alginate biosynthesis.
None of the tested mutations affected the bacterial growth ( Supplementary Fig. 9). Although both the ΔmucA and ΔmucB strains exhibited enhanced alginate biosynthesis, ΔmucA appeared to have more influence than ΔmucB. This observation highlights the profound role of MucA in suppression of AlgU activity and suggests that MucB is the primary but perhaps not the only controlling parameter for the function of MucA (Fig. 5). As expected, mutations that attenuated the MucA-MucB association (MucAW158A, MucAR162A, and MucAH170A) increased the alginate production to a similar level as that of ΔmucB. In contrast, MucAY166A and MucB-L31W had no impact. In conclusion, the results from the mutagenesis and functional studies were consistent with known biological data and provide a framework to explain many mechanistic observations.

Discussion
MucA and MucB are a pair of negative regulators of the alternative σ 22 factor AlgU and are essential for the membranespanning signaling in P. aeruginosa. This pathway controls alginate production and bacterial infection in response to extracytoplasmic stimuli 11 . RIP-mediated MucA degradation is a key mechanism in which accumulated misfolded protein and lipid signaling act as multiple stressors to initiate the cascading proteolytic cleavages of MucA and thus active AlgU 8 . In this work, the MucA peri -MucB complex structure is in an inter-state revealing conformational changes preceding MucA-MucB dissociation ( Fig. 1 and 2). The conserved primary binding of the MucA αI segment to the MucB C-terminal domain confirms a universal coupling pattern among their homologs in different species. The variable residues involved in the MucA/MucB interactions may provide specificity for their association (Fig. 3). We uncovered the structural basis for the LPS-MucB interaction and proved that this is a regulatory element that promotes MucA-MucB dissociation. These findings are further validated by biochemical and functional analysis ( Fig. 4 and 5). These results give essential mechanistic insights into the process of LPSinduced anti-sigma factor degradation during envelope stress-response signaling in bacteria.
Protease-regulated σ ECF stress-response systems are widely distributed in bacteria and are involve in diverse physiological functions 18 . MucA/MucB and their E. coli homolog RseA/RseB represent two groups of anti-σ ECF factors (Fig. 3b) 12,13 , bearing similar 3D structures and protein-protein binding patterns but exhibiting strict interaction specificities ( Supplementary Fig. 1) 7 . We speculate that the distinct binding selectivity of MucA/MucB compared with that of RseA/RseB is a consequence of the coevolution of anti-σ ECF factors to allow bacteria to establish new stress-signaling pathways. Moreover, the existence of multiple types of anti-σ ECF factors in several bacterial species raises the possibility that selective protein-protein recognition has a role in accurately transmitting different extracellular signals 18 .
Although various interactions have been observed between MucA peri and MucB, the intermediate binding affinity indicates that the complex is in an inter-state ( Supplementary Fig. 2). This is in accordance with the feedback response mechanism used to meet the flexibility and adaptability of bacteria in response to environmental challenges. Since MucA αII does not have a significant role in MucB recognition (Fig. 3d), its insertion into the hydrophobic cavity of MucB may not function to stabilize the complex but instead serve as a "plug" that induces large conformational changes in the MucB 92-113 loop, thus ensuring efficient release of MucA cleavage site.
Because of the high structural similarity with general lipoproteins such as LppX, LolA, and LolB, the NTDs of RseB and MucB have been predicted to be the principal region responsible for binding the lipid signals 8,12,14,15 . Similar to LolB 15 , the MucB-NTD in our MucA peri -MucB structure bound with a PEG molecule in what appears to be the LPS (lipid-A) anchoring site ( Fig. 4a; Supplementary Fig. 4A). The lipid-ELISA results found that the purified NTD had lower binding to lipid-A than wildtype but the Ala-substitutions on P106 or P112 exhibited significant lipid-A binding (Fig. 4c). This result validated the lipidbinding property of NTD and also revealed that the structural integrity of MucB was necessary for lipid-A binding, in which the CTD acts as a stabilizer to restrain the motion of loop 92-113 .
By comparing MucB-NTD with the RseB-NTD and MucBapo-NTD (Fig. 2a, c), we could observe dihedral angle changes in proline residues P106 and P112. The dramatic conformational variations in MucB 92-113 provide clear structural evidence to support the previous speculation that the loop connecting βE and βF serves as a lid to regulate lipophilic compound binding 14 .
Consistently, the lipid-ELISA data (Fig. 4) experimentally support the "plug" mechanism of loop 92-113 to lipid-binding pocket and highlights the contribution of proline residues to this function. In addition, distinct from typical lipoproteins, this lid motion in MucB/RseB also acts as a mechanism to facilitate the release of MucA/RseA.
To characterize the nature of the effector binding at the inner cavity of the MucB-NTD, a number of organic molecules were tested for their ability to induce MucA cleavage. The results indicated that MucB appeared to interact with a variety of amphiphilic compounds implying that this system has the capability to receive even more activating signals. Our findings provide a structural template to interpret the interaction mechanism between the anti-sigma factor and lipid signals. As Fig. 6 shown, MucB-lipid-A binding results in a relatively loose packing structure of MucB and exposes the cleavage site of MucA to the solvent. The conformational change of MucB (loop 92-113 ) forms steric hindrance and induces the MucA αII to serve as a "plug" inserting into the inner cavity of MucB, which lead to the release of MucA from MucB. Moreover, further studies are needed to explore how MucB managed to induce local changes to facilitate access of lipophilic molecules into the hydrophobic center and whether and in what manner lipidinduced MucA-MucB structural rearrangements influence AlgW recognition.
Despite the similarities between RseA/RseB in E. coli and MucA/MucB in P. aeruginosa, it is clear that there are differences between the envelope stress responses in different bacteria because of their diverse niche adaptabilities. The distinctive structural features observed in MucB, such as the C90-C198 disulfide bond (Fig. 2a), would be of interest in future studies exploring this bacterial species-specific signaling pathway. In P. aeruginosa, mutation of mucA is a well-known mechanism for mucoid conversion and usually a major mechanism of mucoid conversion in strains isolated from CF 19 . Understanding the processes of the MucA-RIP system is important to clarify how P. aeruginosa is able to overproduce alginate to exhibit persistence and resistance. The specific MucA-MucB recognition event equips a set of multiple signal-responsive elements in the MucA-RIP system. Most mucA mutants, which commonly result from a truncation leading to a loss of protection by MucB, achieve alginate overproduction at the cost of being insensitive to diverse envelope stresses. We expect that further investigations of the MucA-MucB-modulated signal response in P. aeruginosa will continue to expand our understanding of this topic and aid in the development of therapeutic strategies for treating P. aeruginosa infections in CF.
Bacteria were collected by centrifugation at 4000×g for 15 min and resuspended in 120 mL of lysis buffer consisting of 25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5% glycerol. After sonication, the supernatant was obtained by centrifugation at 15,000 × g for 30 min and then co-incubated with 4 mL Ni-NTA resin (Qiagen) for 1 h. The mixture was washed with lysis buffer complemented with 25 mM imidazole and target protein was eluted with lysis buffer containing 300 mM imidazole. The protein was further purified with size-exclusion chromatography Superdex TM -75 (GE Healthcare), which was pre-equilibrated with solution buffer consisting of 25 mM Tris-HCl (pH 7.5), 150 mM NaCl. Peak fractions were determined by 15% SDS-PAGE gel and stain with Coomassie Brilliant Blue.
Crystallization, data collection, and determination. Crystallization screens were carried out by mixing protein complex with reservoir buffer at 18°C through hanging-drop vapor diffusion method 20 . Crystals were obtained in the solution containing 0.1 M Bis-Tris (pH 6.5), 30% (v/v) Polyethylene glycol monomethyl ether 550 and 0.2 M CaCl 2 . Crystals were soaked in cryo-protectant (reservoir solution supplemented with 20% glycerol) and flash-cooled in liquid nitrogen. Diffraction data were collected on beamline BL18U at the Shanghai Synchrotron Radiation Facility (SSRF), China. All diffraction images were processed with the HKL2000 program package 21 . The structure was determined by molecular replacement using PHENIX package 22 with RseB (PDB code: 2P4B) as a template. The structure model refinement was carried out with PHENIX and COOT 23 . The final refinement statistics for the complex were summarized in Supplementary Table 1.
Construction of P. aeruginosa mucA/mucB gene deletions. A sacB-based twostep allelic exchange strategy was employed to construct full-length mucA and mucB deletions of P. aeruginosa 24 . The upstream and downstream (800 bp) PCR fragments of mucA and mucB were ligated by PCR with the gene-specific primers (Supplementary Table 2). Then target fragments were recombined to the linearized DNA fragment of pEX18Gm with ligation-free cloning system (5× Ligation-Free cloning master Mix, abm). These plasmids were transformed into E. coli S17-1 and then mobilized into P. aeruginosa strains PAO1. Colonies were screened by antibiotic-resistant selection and sucrose-mediated counter-selection 20 . The mucA and mucB single-gene deletion strains were further confirmed by PCR and DNA sequencing.
Construction of supplemented strains and mutants. PCR-amplifed mucA, mucB and site-directed mutagenesis were cloned into the XhoI and EcoRI sites of plasmid pME6032 (Supplementary Table 2) 25 . The recombinant plasmids were transformed into corresponding gene-deleted (ΔmucA or ΔmucB) strains, which were screened by PIA plates complemented with 200 μg/mL tetracycline.
Alginate assay. GDP-mannuronic acid is the precursor in alginate biosynthesis, so the alginate quantification was performed using the uronic acid assay as described previously 26 . Briefly, all strains grew on PIA plates and complementation strains were induced by 0.5 mM IPTG for 36 h 27 . Cell pastes were harvested with buffer A (50 mM Tris-HCl (pH 7.4), 10 mM MgCl 2 ) and the absorbance was determined at 600 nm. Cells were removed by centrifugation at 8000×g for 15 min. The supernatant was added with 15 µg/mL DNase and RNase and then shaked at 37°C for 6 h.
After nuclease digestion, proteinase K was added to a final concentration of 20 µg/mL, and the solution was incubated at 37°C for 18 h in a shaking incubator.
For further purification, the supernatants containing dissolved alginate were placed in dialysis bags (Dialysis Membrane, Standard, RC, 10 kDa cutoff; 24 mm Width; 1.8 mL/cm, Sangon Biotech) and dialyzed in 10 mM Tris-HCl (pH 7.6) at 4°C overnight. A 100 µL dialysis fraction was mixed with 1.0 mL borate-sulfuric acid reagent (100 mM H 3 BO 3 dissolved in concentrated H 2 SO 4 ) and 100 µL carbazole reagent (0.1% in ethanol). The mixture was heated at 55°C for 30 min, and the absorbance was determined at 520 nm 28 .
Alginate was quantified using a standard curve made from brown seaweed (BBI). Briefly, difference concentration (mg/mL) samples of commercial alginate was mixed with borate-sulfuric acid and carbazole reagent. Next, the mixture was heated at 55°C for 30 min, and the absorbance was determined at 520 nm. The numerical values of 520 nm absorbance can be used to made a standard curve. Conversely, accompanied by the same treatment, the quantity of cell alginate was calculated through the numerical value of 520 nm absorbance of cell extractive, at the same time, the OD600 of cell culture was determined. finally, the alginate of cell culture was reported as micrograms of uronic acid per milligram of cell weight.
Isothermal titration calorimetry (ITC). Isothermal titration calorimetry was used to investigate the energetics of biomolecular recognition between MucA peri and MucB 29,30 . All experiments were conducted in solution buffer (25 mM Tris-HCl (pH 7.5), 150 mM NaCl) at 25°C. MucB (20 µM) was placed in the temperaturecontrolled sample cell and titrated with the ligand MucA peri (200 µM), and the total injections were made by stirring speed of 750 rpm for 19 times. Ultimately, the data were analyzed using ORIGIN software 30 .
His-tag pull-down assay. His-tag pull-down assay was performed as described with some modifications 31,32 . First, a PreScission protease cleavage site (LEVLFQ↓GP) was inserted into the site between C terminal of MucB (or MucA peri ) and His-tag, and the His-tag of MucB (or MucA peri ) was removed by HRV 3C protease digestion. Then, 20 µg MucA peri with His-Tag was incubated with excessive MucB (without His-tag) and 20 µL Ni-NTA affinity resin (Qiagen) for 30 min. After centrifuging at 15,000 × g for 3 min, the Ni-NTA resin was washed three times with solution buffer containing 25 mM imidazole. Fractions were eluted with solution buffer containing 300 mM imidazole and determined by 15% SDS-PAGE gel and visualized by Coomassie Brilliant Blue stain.
Lipid-ELISA assay. As reported 34,35 , lipid-A (sigma) were diluted in DMSO an initial amount of 50 μM/well, and the solvent (80 μL) was coated in Nunc-Immuno™ MicroWell™ 96-Well Plates (Thermo Scientific) by overnight incubation at 4°C. Remove the solvent and blocked with a 3% (w/v) BSA in PBS for overnight at 4°C. Thoroughly washed with PBS containing 0.05% Tween-20, and subsequently incubated with twofold serial dilutions (starting at 100 μM) of His-tagged proteins (MucA peri , MucB, and MucB mutants) for overnight at 4°C, the concentration gradients are 0.05, 0.1, 0.2, 0.4, 0.8, 1.6, 3.125, 6.25, 12.5, 25, 50, and 100 μM. The plates were then washed for five times with PBS, incubated with mouse anti-His antibody (Invitrogen) for 2 h, washed again for five times with PBS, incubated with horseradish peroxidase (HRP)-conjugated goat anti-mouse secondary antibody diluted 1:5000 in 1% BSA in PBS, and washed, as a final washing stage, for five times again with PBS. Finally, the bound proteins were detected using TMB-ELISA substrate solution (Horseradish Peroxidase-HRP, beyotime). After incubated for 30 min at room temperature, the reaction was stopped by addition of 1 N HCl. Optical density (OD) was subsequently measured at 450 nm. Each experiment was performed three times, and each point is a mean of three replicates ± SD.
Statistics and reproducibility. All experiments were performed in independent biological triplicate and the results of replicates were consistent. One-way ANOVA analyses was performed using GraphPad Prism 7 (GraphPad, CA, USA). Details of the number of biological replicates are described in the figure legends and Methods. Error bars represent standard deviation. P value of <0.05, which means that there is a significant difference, P value of <0.0001 was considered as extremely significant, which is indicated with ****.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data relevant to this study are supplied in the manuscript and supplementary files or are available from the corresponding author upon request. Atomic coordinates of the refined structures have been deposited in the Protein Data Bank (www.pdb.org) with the PDB code 6JAU.