LolA and LolB are conserved in Bacteroidota and are crucial for gliding motility and Type IX secretion

Flavobacterium johnsoniae encodes several LolA and LolB proteins

While the presence of LolA homologs has been reported in Bacteroidota, no LolB homologs have been identified in bacteria of this phylum by sequence similarity searches. We started from the hypothesis that Bacteroidota might encode distantly related LolB homologs with low sequence identity, but that would share a similar structure. To identify such homologs, we performed an in silico prediction analysis searching for remote homologs of E. coli LolB in F. johnsoniae (see Methods for more details).

This analysis identified two LolB homologs in F. johnsoniae: Fjoh_1066 (LolB1) and Fjoh_1084 (LolB2). Both candidates are predicted to encode an SPII signal sequence, in accordance with LolB being a lipoprotein. We performed the same analysis to search for E. coli LolA homologs and three proteins were identified: the two LolA already found by sequence similarity and reported in the literature (Fjoh_2111, LolA1, and Fjoh_1085, LolA2)^18,19 and a third one, Fjoh_0605, LolA3. LolA is a periplasmic carrier and thus harbors an SPI signal allowing it to cross the IM via the Sec pathway and reach the periplasm²⁰. All three LolA homologs carry such a signal sequence and are thus likely localized in the periplasm.

While genes Fjoh_2111 (lolA1), Fjoh_0605 (lolA3), and Fjoh_1066 (lolB1) are encoded in different genomic loci, genes Fjoh_1084 (lolB2) and Fjoh_1085 (lolA2) are part of an operon responsible for flexirubin biosynthesis. Flexirubin is thought to be localized in the OM, and thus LolA2 and LolB2 might be involved in the synthesis and/or transport of this pigment^19,21.

Next, we modeled the F. johnsoniae LolA and LolB protein structures and compared their overall secondary and tertiary structures with respect to the crystal structures of their E. coli homologs (Fig. 1). F. johnsoniae LolA and LolB variants retain the main features of the E. coli unclosed β-barrel presenting a convex and a concave side, composed of 11 antiparallel β-strands in which helices (α- and/or 3₁₀-helices) are embedded (Fig. 1), and enclosed by an N-terminal α-helix. Among the LolA proteins, E. coli LolA (LolA_Ec) and LolA1 have the same mainly disordered C-terminal region with one short 3₁₀-helix as well as one extra parallel β-strand, while LolA2 and LolA3 have a shorter and longer C-terminal end, respectively. In addition, the E. coli LolB (LolB_Ec) specific feature, consisting of a protruding loop comprising an exposed hydrophobic amino acid, is conserved in LolB1 and LolB2 (Fig. 1). In LolB_Ec and LolB1 this is a leucine, while in LolB2 a phenylalanine. LolB2 is structurally closer to LolB_Ec than LolB1 which has a long N- and C-terminal disordered tail, as well as a very long loop extension compacted beneath the barrel.

Comparison of crystallized E. coli LolA (PDB entry: 1UA8) and LolB (PDB entry: 1IWM) structures with models of LolA1, LolA2, LolA3, LolB1, and LolB2 homologs from F. johnsoniae. Each protein is shown as a cartoon representation and secondary structure elements are colored as follows: α-helix (purple), 3₁₀-helix (blue), β-sheet (yellow), and coil (gray). On each structure, the N- and C-terminal positions are indicated. In LolA1, the loop extension between α-helix 3 and β-strand 7 is highlighted with a dashed green ellipse. In LolB, LolB1, and LolB2, the upward conserved loop is highlighted with a dashed red circle. In LolB1, the compacted loop extension between α-helix 3 and β-strand 7 is highlighted with a dashed blue ellipse.

The mechanism of transfer of lipoproteins from LolA_Ec to LolB_Ec occurs via a complex formation through electrostatic potential complementarity, forming a contiguous hydrophobic surface, i.e., a tunnel-like structure, composed of both LolA and LolB concave sides. The lipoprotein is transferred due to a higher affinity of its lipid moiety for LolB than LolA due to a hydrophobic gradient between the cavities of the two partners²². LolA_Ec concave side, especially along the edge of the β-barrel opening, is enriched in negatively charged amino acids, interacting with the LolB_Ec convex side, which is mainly positively charged²². With this mechanism in mind, we compared the hydrophobicity (Supplementary Fig 1) and charge state (Supplementary Fig 2) of F. johnsoniae LolA and LolB homologs to highlight differences and, potentially predict LolA–LolB interactions.

An interesting tendency is a distinct enrichment in highly hydrophobic (leucine and isoleucine) amino acids highlighted for all the F. johnsoniae LolA and LolB homologs (Supplementary Table 1). Consequently, these are overall more hydrophobic than LolA_Ec and LolB_Ec, and LolA2 stands out as the most hydrophobic one. Furthermore, LolB1 contains a higher proportion of hydrophobic residues than LolA1, suggesting a favorable hydrophobic gradient for the lipoprotein transfer between LolA1 and LolB1. Although LolB2 has fewer hydrophobic residues than LolA2, its higher content in leucine and their local concentration within the concave side of the β-barrel could nonetheless allow the lipoprotein transfer toward LolB2 (Supplementary Fig 1). The same rationale can be applied to the putative LolA1-LolB1 couple. In the case of LolA3, the side chains pointing toward the inside of the concavity are substituted in tyrosine and isoleucine, increasing the overall polarity of the binding site. Regarding the electrostatic-driven LolAB paired complexes formation, a slight decrease in arginine compensated by a large increase in the lysine content is observed for all the F. johnsoniae proteins (Supplementary Table 1), especially for LolA1 and LolA2, changing the concave side charge from negative to mostly positive (Supplementary Fig 2). LolA3 seems closer to LolA_Ec, with a predominance of aspartate and glutamate residues on the concave side, giving it a global negative charge. Most remarkably, its unique long and disordered C-terminal region bears a high positive charge, due to a dense population of lysine. Furthermore, while the positive character of the LolB convex side is mainly conserved in its F. johnsoniae homologs, a more evocative negative charge is exhibited by the concave side, particularly at the level of the loop extension of LolB1, absent in LolB_Ec and LolB2 (Supplementary Fig 2). Therefore, the mechanism and preferential orientation by which LolA-LolB complexes are formed in F. johnsoniae most likely differ from that of E. coli, and pair selectivity will also arise from subtle modifications in hydrophobicity and charge distribution.

Deletion of lolA1 and lolB1 affects gliding motility and Type IX secretion

Transposon mutagenesis screens to identify gliding motility genes of F. johnsoniae and of another member of the Bacteroidota resulted in non-gliding mutants that had transposon insertions in lolA homologs^18,23. The F. johnsoniae gene identified was Fjoh_2111, which we refer to as lolA1. Since LolA allows lipoproteins to cross the periplasm and localize to the OM, the lack of gliding of the Tn mutants could be due to mislocalization and consequent depletion of lipoproteins in the OM. To confirm the absence of motility reported for a lolA1 mutant and to determine whether deletion of any of the other F. johnsoniae LolA and LolB proteins could have a similar effect, we assessed the gliding motility of the lolA1, lolA2, lolA3, lolB1, and lolB2 mutants on plates. As expected, lolA1 deletion resulted in non-spreading colonies, and, a lack of motility comparable to that of a gldJ non-gliding mutant was also observed when lolB1 was deleted²⁴ (Fig. 2a and Supplementary Movie 1). GldJ is an OM lipoprotein required for both gliding and T9 secretion^24,25. Deletion of lolA2, lolA3, and lolB2 did not affect gliding. Co-deletion of both lolA1 and lolB1 resulted in the same phenotype as the single lolA1 mutant. Plasmid-borne complementation of lolA1 and lolB1 deletion with lolA1 and lolB1 fully restored gliding motility (Fig. 2a).

Gliding motility on MM agar plates after 48 h (a). Starch degradation by T9-secreted amylases on LB starch plates after 24 h (b) and amylase activity of cell culture supernatants (c). Data in (c) is represented as means of n = 3 independent biological replicates ± standard deviation, and statistical analysis was a one-way ANOVA followed by Dunnett’s multiple comparison test (F (22, 46) = 47.47). Asterisks indicate the following: (****): p < 0.0001.

Since gliding motility and T9 secretion are highly interlinked and both require several OM lipoproteins for their activity^8,26, we next determined the impact of lolA and lolB homologs deletion on T9SS activity. This secretion system is unique to Bacteroidota and is involved in the secretion of various proteins mainly involved in adhesion and nutrient acquisition, such as α-amylases that allow F. johnsoniae to hydrolyze and feed on starch²⁵. To monitor Type 9 secretion, we measured the starch degradation activity of secreted amylases on starch-containing plates and on cell culture supernatants. Starch degradation was significantly reduced in the lolA1, lolB1, and lolA1lolB1 mutant strains compared to the WT and like that of the gldJ mutant (Fig. 2b, c). Deletion of lolA2, lolA3, and lolB2 did not affect amylase secretion (Fig. 2b, c). Plasmid-borne complementation of lolA1 and lolB1 strains fully restored amylase secretion (Fig. 2b, c). The reduced amylase activity observed for the lolA1 and lolB1 mutants could not be ascribed to the lack of motility of these strains as a gldJ-548 mutant, which is non-motile but has an active T9SS⁸, degraded starch to the same extent as the WT strain (Fig. 2b, c).

Overall, these data indicate that LolA1 and LolB1 are crucial for both gliding motility and T9 secretion and reinforce our initial hypothesis that these two proteins might belong to the same pathway, namely the transport of a subset of lipoproteins to the OM. Given their pivotal role in F. johnsoniae, we focused on further characterizing the function of these two proteins.

Deletion of lolA1 and lolB1 affects outer membrane proteome composition

The absence of gliding and T9 secretion could be due to mislocalization of OM lipoproteins crucial for these pathways in the absence of LolA1 and LolB1. To test this hypothesis, we determined the impact of lolA1 and lolB1 deletion on the OM protein composition of bacteria grown in CYE-rich medium. We purified the OM of WT, lolA1, and lolB1 strains and identified proteins by mass spectrometry. Overall, the absence of lolA1 or lolB1 respectively affected 609 and 171 proteins in a significant manner (fold change ≥ 1.5; significance ≥ 20) (Fig. 3a, Supplementary Data 1 and 2). We sorted the proteins according to their localization: cytoplasmic (no SP), integral OM and soluble periplasmic proteins (SPI), periplasmic-facing lipoproteins (SPII), and surface-exposed lipoproteins (SPII-LES). The OM proteome of the WT and lolB1 mutant was enriched in membrane proteins and lipoproteins while, that of the lolA1 mutant contained many cytoplasmic proteins, accounting for 66% of significant entries (402 out of 609) hinting at possible intracellular protein leakage as also suggested by the extracellular debris and cell material observed by electron microscopy for this mutant (Fig. 3a, Supplementary Data 2 and Supplementary Fig. 3). We clustered proteins harboring an SP by predicted function and mapped those involved in polysaccharide utilization to their respective(s) polysaccharide utilization loci (PULs)²⁷.

Proteins identified by mass spectrometry whose abundance significantly differs (FC ≥ 1.5, significance ≥ 20) in the OM fraction of the lolA1 and lolB1 deleted strains compared to the WT clustered by predicted localization (SignalP server). SPI signal peptide I (OMP or periplasmic), SPII periplasmic-facing lipoproteins, SPII-LES surface-exposed lipoproteins (a). Proteins with an SP whose abundance is affected in the OM fraction of the lolA1 and lolB1 mutants compared to the WT clustered by functional class (EggNOG) (b). Gliding and T9SS-specific proteins whose abundance is altered in lolA1 and/or lolB1 OM compared to the WT (c).

In both mutants, the most affected known cellular functions were polysaccharide utilization (2 proteins increased (+) and 45 decreased (−) in lolA1, +6/−14 in lolB1) and cell wall/membrane/envelope biogenesis (+5/−21 in lolA1, +5/−12 in lolB1). In addition, several gliding/T9 secretion-specific proteins were affected (+3/−9 in lolA1, +3/−6 in lolB1). Overall, 207 SPI/SPII proteins were impacted by the deletion of lolA1 and 106 by the deletion of lolB1 (Fig. 3b). Among these, the abundance of 59 proteins was decreased in both mutants and among them several gliding/T9 secretion-related proteins (Fig. 3c). GldK and GldJ, two lipoproteins essential for gliding motility and T9 secretion²⁸ were much less abundant in the OM of the lolA1 and lolB1 mutants than in the WT strain (GldK FC in lolA1 = 0.48 and 0.08 in lolB1; GldJ FC in lolA1 = 0.05 and 0.22 in lolB1). GldN, another critical component of the gliding and T9 types of machinery and the physical partner of GldK^29,30 was also less abundant (FC = 0.53 in lolA1 and 0.33 in lolB1). One SprF-like protein (Fjoh_3951), implicated in T9 secretion²⁸, was also less abundant in the OM of both mutants (FC = 0.16 in lolA1 and lolB1). We also noticed a few proteins potentially involved in envelope biogenesis whose amount in the OM was altered in the two mutants, namely a peptidoglycan-binding LysM protein (FC = 3.12 in lolA1 and 1.86 in lolB1), a BamB-like protein (FC = 0.44 in lolA1 and 0.57 in lolB1), a polysaccharide export protein Wza (FC = 0.27 in lolA1 and 0.23 in lolB1), and a peptidoglycan hydrolase (FC = 0.18 in lolA1 and 0.28 in lolB1).

Among the 148 proteins specifically affected by the deletion of lolA1, we found other components of the gliding/T9 secretion types of machinery (Fig. 3c). Among them, SprT (FC = 0.41), SprF (FC = 0.37), and SprD (FC = 0.05) are all required for gliding motility^31,32. The absence of lolA1 resulted in the decrease of numerous OM proteins dedicated to polysaccharide utilization and inorganic ion transport (Fig. 3b) of which most were SusC-like proteins or TonB-dependent receptors, i.e., proteins with a β-barrel domain.

Within the 47 proteins specifically affected by the deletion of lolB1, we also found gliding/T9 secretion-related proteins (Fig. 3c). LolA1 amount was also increased in the lolB1 mutant (FC = 2.04). In contrast with the data specific to lolA1, we did not observe an obvious pattern in the data specific to lolB1. The loss of lolB1, as for lolA1, decreased the number of PUL-encoded proteins in the OM but to a much lower extent (Fig. 3b).

We did not observe a general mislocalization of OM lipoproteins upon deletion of lolA1 or lolB1, as one could have expected. Nevertheless, the proteomics data confirm and provide an explanation for the loss of gliding motility and T9 secretion that we observed for both mutants. Although the deletion of lolA1 or lolB1 alters the OM protein composition in different ways, bacteria lacking either of these two proteins fail to properly localize a subset of proteins and lipoproteins, including core components of gliding motility/T9 secretion such as GldN, GldK, and GldJ. While the deletion of lolA1 was more consequential for the OM of F. johnsoniae than the deletion of lolB1, the most impacted cellular functions (polysaccharide utilization and cell wall/membrane/envelope biogenesis) were in common. Considering that the abundance of several OM lipoproteins was altered in both mutants, our data suggest that lolA1 and lolB1 participate in the same pathway, i.e., the transport of lipoproteins to the OM and, in particular, lipoproteins involved in gliding motility and T9 secretion. Yet, lolA1 deletion clearly determined a bigger perturbation of the OM proteome. Among the IM and cytoplasmic proteins identified in the proteomic analysis, the majority were more abundant in the lolA1 (323 out of 402) and lolB1 (38 out of 66) mutants OM compared to the WT, thus pointing to some envelope defects of these mutants, probably more pronounced in the absence of LolA1 (Supplementary Data 2).

Deletion of lolA1 and lolB1 affects growth and cell morphology

The OM proteome composition alterations observed prompted us to test whether the lack of LolA1 and LolB1 would affect bacterial fitness. To this aim we monitored the growth of these mutants in two media, casitone yeast extract (CYE), the rich medium normally used to grow F. johnsoniae and motility medium (MM) a poor medium that stimulates gliding motility³³.

As shown in Fig. 4a, all mutants grew like the WT in CYE. In contrast, in MM, the deletion of lolA1 severely impacted growth while the deletion of lolB1 did not (Fig. 4a). In the absence of both LolA1 and LolB1 growth was comparable to that of the single lolA1 mutant. Plasmid-borne complementation of lolA1 and lolA1lolB1 mutants with lolA1 and lolA1 and lolB1 fully restored growth to the WT levels (Fig. 4a). Deletion of any of the other LolA and LolB proteins did not affect growth in any condition tested (Fig. 4a).

Growth curves of WT and mutant strains in CYE and MM liquid media. Data are represented as means of n = 3 independent biological replicates ± standard deviation (a). Phase-contrast microscopy images of bacteria grown for 16 h in CYE and MM media (bar = 5 µm) (b).

Next, we assessed whether the mutations had any morphological effect on bacteria. We observed by optic and transmission electron microscopy cells grown for 16 h in CYE medium and found that lolA1 bacteria presented some aberrant shapes, some resembling the shape of a lollipop, while others being completely round (Fig. 4b and Supplementary Fig. 3). The same abnormal morphologies, along with more cell aggregation, were also observed in the lolA1lolB1 double mutant, while lolB1 mutants resembled the WT strain. Both lolA1 and lolA1lolB1 mutants displayed detached OM and release of intracellular material as a result of severe envelope defects (Supplementary Fig. 3). The aberrant cell morphology phenotype of the lolA1 and lolA1lolB1 strains was enhanced when bacteria were grown in MM and the formation of abnormal cells was also observed for the lolB1 mutant in this medium (Fig. 4b). Lack of any of the other LolA and LolB proteins did not affect cell morphology in either CYE or MM (Supplementary Fig. 4). Complementation of lolA1, lolB1, and lolA1lolB1 with plasmid-borne lolA1, lolB1, and lolA1 and lolB1 fully restored the cell morphology to WT levels (Fig. 4b).

Overall, while growth in the MM medium was affected only by the deletion of lolA1, the deletion of lolA1 and lolB1 affected cell morphology, in particular in MM for this latter mutant. Co-deletion of lolA1 and lolB1 determined an enhancement in abnormal cell morphology in both CYE and MM (Fig. 4b and Supplementary Fig. 3).

Deletion of lolA1 and to a minor extent of lolB1 affects envelope integrity

The observed growth and morphological phenotypes of the lolA1 mutant when grown in MM hinted to a possible membrane instability. A main difference in composition between these two media is the absence of magnesium in the form of MgSO₄ in MM. Divalent cations and, in particular, magnesium, are known to stabilize LPS–LPS interactions³⁴. In bacteria with OM alterations due to protein and/or lipid shifts, magnesium has been shown to compensate for these phenotypes. We thus tested whether the growth defect and morphology changes observed for the lolA1 and lolB1 mutants could be rescued by magnesium supplementation in MM. As shown in Fig. 5, the addition of 8 mM MgSO₄ (as in CYE) to MM restored growth (Fig. 5a) of the lolA1 and lolA1lolB1 mutants, thus suggesting an alteration of the OM of these mutants. Concerning cell morphology, the addition of magnesium alleviated abnormal cell morphologies of the mutants but did not totally restore them to WT levels (Fig. 5b).

Growth curves of WT and mutant strains in MM liquid medium without and with 8 mM MgSO₄. Data are represented as means of n = 3 independent biological replicates ± standard deviation (a). Phase-contrast microscopy images of bacteria grown in liquid MM medium without and with 8 mM MgSO₄ (bar = 5 µm) (b). Growth after 24 h in CYE liquid medium in the presence of different OM perturbing agents (SDS, Polymyxin B, and EDTA) (c). Data in c are represented as means of n = 3 biological independent replicates ± standard deviation, and statistical analysis was a one-way ANOVA followed by Tukey’s multiple comparison test (SDS 0.001% F (3, 8) = 6.602; SDS 0.005% F (3, 8) = 440.9; Polymyxin B 2 µg/ml F (3, 8) = 43.21; Polymyxin B 20 µg/ml F (3, 8) = 540.5; Polymyxin B 40 µg/ml F (3, 8) = 669.9; EDTA 0.2 mM F (3. 8) = 4.37; EDTA 1 mM F (3, 8) = 99.03). Asterisks indicate the following: (*): p < 0.0332, (**): p < 0.0021, (***): p < 0.0002, (****): p < 0.0001, (ns): p > 0.05).

We also tested whether lolA1 and/or lolB1 deletion might affect the sensitivity of the mutants to several OM perturbing agents, namely SDS, Polymyxin B, and EDTA. We observed that while deletion of lolA1 and/or lolB1 did not affect growth at low SDS concentrations, the lolA1 mutant showed significant growth impairment at a higher concentration (Fig. 5c). The lolB1 strain also showed reduced growth although to a lesser extent than the lolA1 mutant (Fig. 5c). The lolA1lolB1 double mutant showed a phenotype like the lolA1 mutant. Polymyxin B affected the growth of the three mutants in a similar concentration-dependent fashion (Fig. 5c). In contrast, EDTA affected the growth of the lolA1 and lolA1lolB1 mutants, but not of the lolB1 mutant at the tested concentrations (Fig. 5c).

At an EDTA concentration of 1 mM, the lolA1lolB1 double mutant showed an increased sensitivity compared to that of the lolA1 mutant, meaning that the deletion of lolB1 in a lolA1 mutant background increases the sensitivity of bacteria to this compound.

Taken together, these results suggest that the deletion of lolA1 affects OM stability, as the lolA1 and lolA1lolB1 mutants showed acute sensitivity to all tested stresses. On the other hand, lolB1 single deletion seems to have a lower impact on OM stability, as this mutant was less sensitive than the lolA1 mutant to SDS and Polymyxin B and was resistant to EDTA. The increased sensitivity of the lolA1lolB1 double mutant to EDTA compared to the lolA1 single mutant supports our observations of cell morphology (Fig. 4b) as well as the idea that the deletion of both lolA1 and lolB1 is more detrimental to F. johnsoniae than their single deletion.

The protruding loop and the C-terminal region of LolB1 are crucial for its function

In E. coli LolB, the hydrophobic protruding loop between β-strands 3 and 4 is critical for lipoprotein insertion³⁵. A highly conserved residue, Leu68, found at the tip of the loop has been shown to be crucial for LolB activity. When this leucine is replaced with a polar residue or deleted, LolB can no longer efficiently insert lipoproteins into the OM^13,35. A loop is also present in LolB1 with a leucine (L74) at its tip (Fig. 6a). To see whether the loop and this residue are crucial for F. johnsoniae LoLB1 function, we tested whether the expression of LolB1 where Leu74 is replaced by the polar glutamate (L74E) or where the loop is deleted (residues 73–76) could complement the gliding and T9SS phenotypes of the lolB1 mutant. As shown in Fig. 6b, c, the LolB1_L74E variant only partially restored gliding motility whereas amylase secretion was restored to WT levels. In contrast, deletion of the loop completely abolished gliding and secretion.

I-TASSER model of LolB1 with the protruding loop (in red) with the conserved L74 (in cyan) and the C-terminal domain (in blue) (a). Gliding on MM agar plates after 48 h and starch degradation on LB starch plates after 24 h of lolB1 mutants expressing: WT LolB1, LolB1_L74E, LolB1_∆73-76, LolB1_∆223-243 and LolB1_C17G (b). Amylase activity of the supernatant of WT and lolB1 mutants expressing: WT LolB1, LolB1_L74E, LolB1_∆73-76, LolB1_∆223-243 and LolB1_C17G (c). Data in c are represented as means of n = 3 independent biological replicates ± standard deviation, and statistical analysis was a one-way ANOVA followed by Dunnett’s multiple comparison test (F (22, 46) = 47.47). Asterisks indicate the following: (****): p < 0.0001. Detection by Western blot of C-terminally His-tagged: WT LolB1, LolB1_L74E, LolB1_∆73-76 and LolB1_∆223-243. GroEL protein levels are shown and serve as a loading control (d).

These data suggest that the loop and the conserved Leu are important for LolB1 function, as in E. coli LolB. Partial complementation of the LolB1_L74E protein could not be ascribed to a different protein expression level of this mutation since the protein amount of WT LolB1 and LolB1_L74E variants was comparable (Fig. 6d). In contrast, deletion of residues 73–76 of LolB1 (LolB1_∆73-76) resulted in a significant reduction of protein amount, as shown by the weak signal in the western-blot and by MS detection (Fig. 6d and Supplementary Table 2), thus suggesting that this mutation affects protein stability or induces degradation of this non-functional LolB variant.

A main difference between E. coli LolB and LolB1 is the presence of a C-terminal domain that could not be folded by modeling software (Fig. 6a). We wondered whether this domain could be important for LolB1 function and thus deleted it (lolB1Δ223–243). This mutant LolB1 could only partially complement gliding while amylase secretion was like WT levels (Fig. 6b, c). As for LolB1_∆73-76, deletion of the C-terminal domain determined a significant reduction in the protein levels, as shown by the weak signal in the western blot and low spectral count detected by MS (Fig. 6d and Supplementary Table 2). However, the evidence that deletion of the C-terminal tail of LolB1 still allows protein secretion despite determining the strongest reduction in protein levels (Supplementary Table 2), suggests that this deletion affects protein stability but not completely protein function as observed for the LolB1 73–76 residues deletion which abolished both gliding and secretion.

In E. coli, LolB does not require its lipid anchor to correctly insert lipoproteins in the OM³⁶. To see if this was also the case for LolB1, we generated a periplasmic soluble form of LolB1, by introducing a C17G mutation in the signal peptide of LolB1 (LolB1_C17G), thus replacing the lipidated cysteine and generating an SPI. In trans expression of LolB1_C17G fully complemented lolB1 deletion in F. johnsoniae as shown by the recovered gliding motility and amylase secretion (Fig. 6b, c). As for E. coli LolB, lipidation does not seem to be crucial for F. johnsoniae LolB1 function.

Overall, these data support the evidence that LolB1 shares common features with E. coli LolB, as the loop, while having some compositional differences, and the C-terminal domain, both crucial for its function.

E. coli and F. johnsoniae LolA and LolB are not interchangeable

In silico structure comparison and protein characterization show high structural similarity between F. johnsoniae LolA1 and LolB1 and E. coli LolA and LolB respectively (Fig. 1) while highlighting some differences regarding hydrophobicity and charge distribution (Supplementary Figs. 1 and 2 and Supplementary Table 1). We thus wondered whether LolA and LolB from E. coli might complement the deletion of lolA1 and lolB1 in F. johnsoniae. In trans expression of LolA_Ec and LolB_Ec did not restore gliding nor T9 secretion of the lolA1 and lolB1 mutants (Fig. 7a–c). Similarly, co-expression of LolA_Ec and LolB_Ec did not restore gliding nor secretion of the lolA1lolB1 double mutant (Fig. 7a–c).

Gliding motility on MM agar plates after 48 h (a) and starch degradation on LB starch plates after 24 h (b). Amylase activity of cell culture supernatants (c). Data in c are represented as means of n = 3 independent biological replicates ± standard deviation, and statistical analysis was a one-way ANOVA followed by Dunnett’s multiple comparison test (F (22, 46) = 47.47). Asterisks indicate the following: (****): p < 0.0001.

Next, we tested whether F. johnsoniae LolA1 and LolB1 could complement the deletion of lolA and lolB in E. coli. Since lolA and lolB are essential genes in E. coli, we first expressed the F. johnsoniae proteins LolA1 and LolB1 individually or together (LolA1–LolB1) in the E. coli MG1655 WT strain and then attempted to delete lolA or lolB. While deletion of lolA and lolB could be achieved when LolA_Ec and LolB_Ec were expressed in trans (Supplementary Fig. 5), we could not obtain any mutant when the F. johnsoniae proteins were expressed. Similar results were obtained when LolA2 and/or LolB2 proteins were expressed in E. coli. Lack of complementation by E. coli LolA and LolB of F. johnsoniae mutants and by F. johnsoniae LolA1 and LolB1 of E. coli mutants could not be ascribed to lack of expression of these proteins since all of them could be detected by mass spectrometry analysis (Supplementary Table 3). In conclusion, despite structural and physico-chemical similarities, E. coli and F. johnsoniae LolA and LolB proteins are not interchangeable.

LolA and LolB are conserved among Bacteroidota

We next explored if the F. johnsoniae LolA and LolB proteins were conserved in bacteria of the phylum Bacteroidota. To this aim, we performed an in silico sequence similarity search for homologs of the five F. johnsoniae LolA and LolB proteins in several Bacteroidota species. This analysis identified LolA and LolB homologs in all the analyzed species (Fig. 8 and Supplementary Table 4). In addition, while we found LolA1 and LolB1 homologs in all species, LolA2, LolB2, and LolA3 homologs seem to be more restricted. Among the species carrying only LolA1 and LolB1 homologs, there is C. canimorsus, which belongs to the family Flavobacteriaceae and is a normal oral commensal of dogs and cats which can cause severe infections in humans upon contact with these animals³⁷. In C. canimorsus, LolA and LolB homologs, Ccan_16490 (LolA) and Ccan_17050 (LolB), are essential since attempts to delete their encoding genes failed. Regarding lolA2 and lolB2, both genes were always found to be encoded in loci dedicated to the synthesis and transport of a flexirubin-like pigments¹⁹, suggesting that LolA2 and LolB2 are specific to these processes.

Homologs of F. johnsoniae LolA and LolB proteins in several Bacteroidota species identified by DELTA Blast⁶². Phylogenetic tree based on NCBI taxonomy generated via phyloT⁶⁷.

To see whether the F. johnsoniae LolA1 and LolB1 and C. canimorsus LolA and LolB homologs share the same function in these bacteria, we first tested whether expression of C. canimorsus LolA (Ccan_16490) and LolB (Ccan_17050) could complement the deletion of lolA1 and lolB1 in F. johnsoniae respectively. Complementation with C. canimorsus LolA or LolB fully restored the ability to glide and secrete via the T9SS of the F. johnsoniae lolA1 and lolB1 mutants (Fig. 9a, b) as well as the growth and cell morphology phenotypes in MM (Fig. 9c, d).

Gliding motility on MM agar plates after 48 h, starch degradation by T9-secreted amylases on LB starch plates after 24 h (a), and amylase activity of cell culture supernatants (b) of lolA1 and lolB1 F. johnsoniae mutants complemented with C. canimorsus lolA (lolA_Cc) and lolB (lolB_Cc). Data in b are represented as means of n = 3 independent biological replicates ± standard deviation, and statistical analysis was a one-way ANOVA followed by Dunnett’s multiple comparison test (F (22, 46) = 47.47) (ns indicates p > 0.05). Growth in MM liquid medium (c) and phase-contrast microscopy images of bacteria grown in liquid MM medium for 16 h (d) of lolA1 and lolB1 F. johnsoniae mutants complemented with C. canimorsus lolA (lolA_Cc) and lolB (lolB_Cc) (bar = 5 µm). Data in c are represented as means of n = 3 independent biological replicates ± standard deviation. Growth of serial dilution spots of C. canimorsus lolA and lolB mutants expressing F. johnsoniae lolA1 or lolB1 in trans on SB plates (e).

We next tested whether the expression of F. johnsoniae LolA1 and LolB1 could bypass the lethality of the deletion of lolA and lolB of C. canimorsus. To test complementation with LolA1, we performed a plasmid exchange in the C. canimorsus lolA (Ccan_16490) mutant strain expressing plasmid-borne C. canimorsus lolA with the vector encoding lolA1 (see “Methods” for additional details). To test LolB1 complementation, we first expressed LolB1 in trans in the C. canimorsus WT strain and then deleted lolB (Ccan_17050). Both C. canimorsus lolA and lolB deletion mutants were viable when LolA1 or LolB1 were expressed, thus confirming that the F. johnsoniae and C. canimorsus proteins share similar functions (Fig. 9e). However, while complementation of lolB deletion in C. canimorsus with lolB1 did not result in any growth defect on rich medium, expression of lolA1 only partially complemented lolA deletion (Fig. 9e). Overall, these results suggest that homologs of F. johnsoniae LolA1 and LolB1 share the same function in C. canimorsus and that the same could be true in Bacteroidota in general.

Simultaneous deletion of all F. johnsoniae LolA and LolB homologs is not lethal and does not affect surface lipoprotein localization

Next, to see whether F. johnsoniae could tolerate the lack of all its LolA and LolB homologs, we generated a lolA1 lolA2 lolA3 lolB1 lolB2 quintuple mutant by sequentially deleting all genes. This indicates that the bacterium can perform its vital functions even in the absence of any LolA and LolB.

Finally, we tested whether surface-exposed lipoproteins are correctly localized in the absence of all LolA and LolB homologs in F. johnsoniae. To this end, we monitored surface lipoprotein exposure using a reporter lipoprotein: the sialidase (SiaC) from C. canimorsus. SiaC is an OM lipoprotein facing the periplasm where it is responsible for the cleavage of sialic acid from eukaryotic glycoproteins³⁸. The addition of the lipoprotein export signal (LES) to SiaC (LES-SiaC) was shown by our group to be sufficient to localize it at the surface of C. canimorsus^16,17.

We thus expressed SiaC and LES-SiaC in the F. johnsoniae WT and in the lolA1 lolA2 lolA3 lolB1 lolB2 quintuple mutant strain and verified their localization by immunofluorescence microscopy using anti-SiaC antibodies and secondary fluorescently labeled antibodies (Fig. 10). As expected, SiaC WT was not exposed at the bacterial surface of both strains while we found LES-SiaC surface exposed in both WT and LolAB lacking strains (Fig. 10), thus suggesting that surface-exposed lipoproteins correctly localize at the cell surface in the absence of any LolA and LolB proteins and thus that the pathway responsible for surface-lipoproteins localization is independent of LolA and LolB homologs of F. johnsoniae.

Immunofluorescence microscopy images of F. johnsoniae WT and lolA1 lolA2 lolA3 lolB1 lolB2 mutant bacteria expressing sialidase (SiaC) or LES-sialidase (LES-SiaC) labeled with anti-SiaC serum (scale bar = 5 µm).