The Legionella collagen-like protein employs a unique binding mechanism for the recognition of host glycosaminoglycans

Bacterial adhesion is a fundamental process which enables colonisation of niche environments and is key for infection. However, in Legionella pneumophila, the causative agent of Legionnaires’ disease, these processes are not well understood. The Legionella collagen-like protein (Lcl) is an extracellular peripheral membrane protein that recognises sulphated glycosaminoglycans (GAGs) on the surface of eukaryotic cells, but also stimulates bacterial aggregation in response to divalent cations. Here we report the crystal structure of the Lcl C-terminal domain (Lcl-CTD) and present a model for intact Lcl. Our data reveal that Lcl-CTD forms an unusual dynamic trimer arrangement with a positively charged external surface and a negatively charged solvent exposed internal cavity. Through Molecular Dynamics (MD) simulations, we show how the GAG chondroitin-4-sulphate associates with the Lcl-CTD surface via unique binding modes. Our findings show that Lcl homologs are present across both the Pseudomonadota and Fibrobacterota-Chlorobiota-Bacteroidota phyla and suggest that Lcl may represent a versatile carbohydrate binding mechanism.


Supplementary Table 1 | Domain boundaries of Lcl from L. pneumophila 130b strain
The predicted N-terminal amphipathic helix is underlined with residues predicted to form the hydrophobic face in bold 7 .

Supplementary Fig. 5 |
Circular dichroism (CD) spectra of Lcl.a, CD spectra of Lcl recorded at 10°C with negative peak at 199 nm and maximum peak at 222 nm.b, Thermal denaturation of Lcl recorded between 10°C and 75°C at 199 nm.Two melting temperatures were determined at 38°C and 44°C.Supplementary Fig. 10 | DUF1566/pfam07603 motif within Lcl-CTD.a, Cartoon representation of trimeric Lcl-CTD with the DUF1566 regions coloured wheat, and the non-DUF1566 S1-S4 strands of Lcl-CTD coloured green.b, Electrostatic surface potential representation of the DUF1566 region within the Lcl-CTD trimer, with the Lcl-CTD S1-S4 strands shown as green cartoon.c, Cartoon representation of the isolated DUF1566 trimer from Lcl-CTD.d, Sequence alignment of the consensus DUF1566 motif and the corresponding region within L. pneumophila 130b Lcl.Identical residues are highlighted with an asterisk (*), and residues that are similar are highlighted with a colon (:).Completely conserved positions in the DUF1566 sequence that are also identical in Lcl are coloured red, and those that are not identical are coloured blue.Supplementary Fig. 12 | SAXS analysis of Lcl-CTD mutants.a, Size-exclusion chromatography coupled with SAXS profile of wild-type and engineered Lcl-CTD.b, Experimental scattering curves of trimeric forms of wild-type and engineered Lcl-CTD.c, DAMMIF bead model of monomeric Lcl-CTD R342A superimposed with chain A from Lcl-CTD, with normalized spatial discrepancy (NSD) score of 1.5.d, Monomer Lcl-CTD crystal structure fit to monomeric Lcl-CTD R342A SAXS data (red open circles) with χ 2 of 1.06.Experimental scattering curve of trimeric wild-type Lcl-CTD (black open circles) is shown for comparison.Supplementary Fig. 15 | Docking of C4S dp8 against monomeric Lcl-CTD.Top three HADDOCK clusters are shown.C4S is as sticks and coloured red, while monomeric Lcl-CTD is shown as a green cartoon.Docking was carried out on monomeric Lcl-CTD, but here it is shown superimposed onto the crystal trimer (yellow).The trimeric structure shown in Cluster 1 is equivalent to the HM model used for subsequent MD simulations.Supplementary Fig. 16 | Stability of Lcl-CTD/C4S during Molecular Dynamics simulations.a, Time evolutions of the RMSD of C atoms from the starting structure before energy minimisation for Lcl-CTD in complex with C4S.The HT1 and HT2 plots represent MD simulations starting from different HADDOCK models where one molecule of C4S dp8 was docked against a trimer of Lcl-CTD (3:1 Lcl-CTD:C4S).The HM plot represents a simulation starting from the HADDOCK model used for subsequent analysis.During HADDOCK one molecule of C4S dp8 was docked against a monomer (chain A) of Lcl-CTD (1:1 Lcl-CTD:C4S) and then prior to running the MD simulations it was reconstituted as a trimer (3:1 Lcl-CTD:C4S).b, The RMSD values for all the replicas starting from the HM model.For a and b, RMSD values were calculated after best-fit superimposition of each frame to the reference structure and the highly flexible N-terminal residues (271-277) were not included in the calculation.The initial equilibration stages with positional restraints were omitted from the plot.

Table 6 | Average frequency of occurrence of Lcl-CTD/C4S contacts during MD simulations.
a Residues are sorted by decreasing average frequency.Lysines are highlighted in bold.bThefrequency was calculated as average over all the replicas (production only).Only residues with a frequency ≥ 10% for at least one chain are reported.Supplementary

Table 7 | Average frequency of occurrence of Lcl-CTD/C4S hydrogen bonds during MD simulations.
Residues are sorted by decreasing average frequency.Lysines are highlighted in bold.Hydrogen bonding interactions were detected using Visual Molecular Dynamics 9 with a Donor-Acceptor distance threshold of 3.5 Å and a Hydrogen-Donor-Acceptor angle of 30 o .b The frequency was calculated as average over all the replicas (production only).When more than one donor/acceptor per residue was involved in hydrogen bonding, the frequency values were added together.Only residues with a frequency >= 10% for at least one chain are reported. a

Table 8 | Primers, plasmids and strains used in this study
*Restriction sites are underlined