The BR domain of PsrP interacts with extracellular DNA to promote bacterial aggregation; structural insights into pneumococcal biofilm formation

The major human pathogen Streptococcus pneumoniae is a leading cause of disease and death worldwide. Pneumococcal biofilm formation within the nasopharynx leads to long-term colonization and persistence within the host. We have previously demonstrated that the capsular surface-associated pneumococcal serine rich repeat protein (PsrP), key factor for biofilm formation, binds to keratin-10 (KRT10) through its microbial surface component recognizing adhesive matrix molecule (MSCRAMM)-related globular binding region domain (BR187–385). Here, we show that BR187–385 also binds to DNA, as demonstrated by electrophoretic mobility shift assays and size exclusion chromatography. Further, heterologous expression of BR187–378 or the longer BR120–378 construct on the surface of a Gram-positive model host bacterium resulted in the formation of cellular aggregates that was significantly enhanced in the presence of DNA. Crystal structure analyses revealed the formation of BR187–385 homo-dimers via an intermolecular β-sheet, resulting in a positively charged concave surface, shaped to accommodate the acidic helical DNA structure. Furthermore, small angle X-ray scattering and circular dichroism studies indicate that the aggregate-enhancing N-terminal region of BR120–166 adopts an extended, non-globular structure. Altogether, our results suggest that PsrP adheres to extracellular DNA in the biofilm matrix and thus promotes pneumococcal biofilm formation.

allows estimating the degree of flexibility of proteins. If the intensity is normalized by I(0) and s is multiplied by R g the Kratky plot becomes dimensionless and can be used to compare the foldness of proteins independently of the protein size (7). The theoretical curves for the globular protein and the random chain were calculated using the Guinier law I(s)/I(0) = exp[-(sR g ) 2
AFM tips functionalization. AFM tips were functionalized with BR 187-385 and BR* 120-395 residues using ~6 nm long PEG-benzaldehyde linkers as described by Ebner and coworkers (9). Oxide-sharpened microfabricated Si 3  Single-molecule force spectroscopy. SMFS measurements were performed using a Multimode VIII AFM (Bruker, Santa Barbara, CA), at room temperature in buffer solution (20 mM Hepes, 150 mM NaCl, pH 7.5). Functionalized surfaces were attached to a steel sample puck using a small piece of double-face adhesive tape, and the mounted sample was transferred into the AFM liquid cell while avoid dewetting. The cantilevers spring constants were measured by the thermal noise method (Picoforce, Bruker). Adhesion maps were obtained by recording 32 x 32 force distance curves on areas of given size (5 µm x 5 µm) and calculating the adhesion force for each force curve. All curves were recorded using a maximum applied force of 250 pN, a contact time of 500 ms, and constant approach and retraction speeds of 1000 nm s -1 , unless otherwise stated.

Crystallization, data collection and crystal structure determination
The BR 187-378 dimer was concentrated to 20 mg/mL in 20 mM Na-Citrate, 500 mM NaCl, 10% (V/V) glycerol, pH 5.5. Well-diffracting crystals were obtained in 100 mM Trisodiumcitrate dihydrate pH 5.6; 34-44% 2-methyl-4-pentanediole using the sitting drop vapor-diffusion method. Crystals were flash-frozen in liquid nitrogen. X-ray diffraction data was collected at beam line ID23-1 at the synchrotron radiation facility at the ESRF (Grenoble, France). Crystals diffracted to 2.1 Å resolution and were processed using the XDS program package (10) ( Table 2). A molecular replacement search performed using Phaser (11) revealed three molecules of the BR 187-385 monomer (PDB: 3ZGH) within the asymmetric unit. Initial rigid body and restrained refinement rounds were performed in CCP4 refmac, and Coot was used for manual model rebuilding (12,13). The mFo-DFc electron density difference map clearly indicated that residues Q312-G315 had to be re-built such that residues L202-G315 of chain A were linked to residues Y316-S377 of the symmetry mate [X, -Y, -Z + (1 0 0) & {-1 0 0}), symmetry operation in Coot], and vice versa. The same procedure was applied to swap the corresponding regions between chains B and C. The model was thereafter refined using Phenix (14) with individual isotropic ADP factors, TLS refinement and NCS torsion restraints. The final model was refined to R and R free values of 21.1 and 24.4, respectively, and comprised residues L202-S377, residues N203-S377 and residues N203-S376 for chains A, B and C, respectively. The electron density for the structural model of chain A was of higher quality compared to the density for both chains B and C, which was corroborated by higher values for real-space correlation coefficients. This difference in the quality of the electron density map was also true for the crucial loop region around residue S314.

Structural analysis
The simulated annealing composite electron density map displayed in Figure S2 was calculated using autobuild in Phenix (14). Secondary structure analyses were performed using PDBsum (15) and 2Struc (16).

Molecular dynamics simulations
An arrangement similar to the complex obtained from the 'ParaDock' webserver was obtained using the program 'vina' (17), rigidly docking a short DNA fragment of 10 base-pairs onto the BR 187-385 dimer, with few bumps similar to the 'ParaDock' docking complexes.
The best model from 'ParaDock' docking was subjected to molecular dynamics simulations in order to check whether, on a timescale of tens of nanoseconds, the complex would be stable or spontaneously dissociate. The structure of the complex was prepared using the psfgen utility of the software package NAMD (18). The forcefield CHARMM22 (19) with the CMAP correction for backbone torsion angles (20) was used.
The complex was solvated using the software VMD (21). 44 sodium ions were added using the same software in order to neutralize the system. The system was first energy minimized keeping protein atoms fixed by 300 minimization steps and the solvent was simulated for 20 ps in order to let it soak the solute molecule. Then the whole system was energy minimized by 300 minimization steps and molecular dynamics simulation was run at constant temperature and pressure.
The temperature (310 K) control was effected by Langevin dynamics and the pressure (1.012E05 Pa) was controlled using a Langevin piston. Electrostatics was treated by a simple cutoff (14 Å) scheme. The system was simulated for 50 ns. Due to forcefield inaccuracy and simple treatment of electrostatics the molecule of DNA tends to distort at the end of simulation. We analysed snapshots taken at 100 ps intervals along the simulation between 10 and 40 ns.

Surface display of BR on the surface of Staphylococcus carnosus
Cloning. The PCR-amplified PsrP constructs and the gene-synthesized (Eurofins Genomics, Germany) DUF1542 repeat domain of SasC (Uniprot ID: C7BUR8) were cloned into the staphylococcal display vector 'pHis3C' using a sequence and ligation independent cloning (SLIC) method (22,23). The vector comprises two origins of replication, oriE and oriS, as well as ampicillin and chloramphenicol resistance marker genes for vector construction in Escherichia coli XL1 blue (Agilent Technologies, USA) and protein surface display using Staphylococcus carnosus TM300 (23), respectively.
Coding sequences of the protein-display constructs were confirmed by DNA sequencing and are listed in the supplemental information.

Cellular aggregation assay. Transformed S. carnosus cells were cultured in TB medium
with chloramphenicol (10ug/mL) overnight at 37°C at 160 rpm. Cells were adjusted to OD 600nm ~2 and split into two separate samples, one of which was treated with DNaseI (A3778, Applichem, Germany) at a final concentration of 0.6 µM in the presence of 5 mM MgCl 2 . Cells were incubated for another hour at 37°C at 160 rpm before harvesting by centrifugation (3750g, 4°C, 10 minutes) and gently re-suspended in PBS. Resuspended cells were adjusted to OD 600nm ~0.25 in PBS. After a further one-to-eight dilution, a volume of 150 µL was filled into wells of a 96-µ-well plate (Ibidi, Germanu) and covered with 25 µL of silicone oil AR 200 (Sigma Aldrich, USA). After 3h equilibration at RT, images were taken using the ZOE Fluorescent Cell Imager (Biorad, USA) at 20x magnification. Images were analyzed using the software cell profiler (www.cellprofiler.org), the particle parameter values were imported into the R software package for statistical analysis and visualization using density histogram and q-q plots (24,25).        (G) Q-q plots reveal that Scar-BR 187-378 and Scar-BR 120-378 were significantly more aggregated when bacteria were grown in M9 minimal medium with the addition of eDNA (red and blue, respectively), compared to growth in the absence of eDNA (light and dark green, respectively). Scar-DUF was not affected by the addition of DNA, and had a very similar particle size distribution as shown in Figure 5G. For comparison, qq-plots of Scar-BR 120-378 grown in TB without and with DNase treatment (as semi-transparent blue and green lines, respectively, taken from Figure 5G) are included. The q-q plot of Scar-DUF grown in TB is also included (semi-transparent black line), but identical to the q-q plots of Scar-DUF grown in M9 medium.

Movie S1. Molecular dynamics analysis indicate that the BR 185-387 -DNA complex is kept through non-specific interactions over a simulation period of 50 ns
Starting from the docked conformation, the complex appeared to be flexible but stable with the RMSD of the docking interface (atoms N, CA, C for the protein and P for DNA) in the range of 0.3 nm from the starting conformation. This is comparable to the RMSD of protein backbone atoms alone. The relatively large RMSD is mainly due to the relative movements of the monomers with respect to each other, whereas global backbone RMSD, including all backbone atoms, for individual monomers stays below 0.2 nm.
Although the DNA structure tends to be distorted by the protein, many protein-DNA BR* 120-395 3