Characterization of H/D exchange in type 1 pili by proton-detected solid-state NMR and molecular dynamics simulations
Uropathogenic Escherichia coli invades and colonizes hosts by attaching to cells using adhesive pili on the bacterial surface. Although many biophysical techniques have been used to study the structure and mechanical properties of pili, many important details are still unknown. Here we use proton-detected solid-state NMR experiments to investigate solvent accessibility and structural dynamics. Deuterium back-exchange at labile sites of the perdeuterated, fully proton back-exchanged pili was conducted to investigate hydrogen/deuterium (H/D) exchange patterns of backbone amide protons in pre-assembled pili. We found distinct H/D exchange patterns in lateral and axial intermolecular interfaces in pili. Amide protons protected from H/D exchange in pili are mainly located in the core region of the monomeric subunit and in the lateral intermolecular interface, whereas the axial intermolecular interface and the exterior region of pili are highly exposed to H/D exchange. Additionally, we performed molecular dynamics simulations of the type 1 pilus rod and estimated the probability of H/D exchange based on hydrogen bond dynamics. The comparison of the experimental observables and simulation data provides insights into stability and mechanical properties of pili.
KeywordsUropathogenic E. coli Type 1 pili Hydrogen deuterium exchange Solid-state NMR Molecular dynamics simulations
The pilus rod subunit FimA has an incomplete immunoglobulin-like β-sheet fold (pilin body) and is complemented by a protruding N-terminal extension (Nte, termed donor strand) of a neighboring subunit with non-covalent interactions (Choudhury et al. 1999). This donor strand complementation (DSC) is processed by the highly conserved chaperone-usher pathway in the periplasmic space of UPEC with assistance of the chaperone FimC and the usher FimD (Phan et al. 2011; Geibel et al. 2013). Sequential DSCs along pili give rise to a right-handed helical structure (Hahn et al. 2002). While the essential elastic properties of pili originate from their coil-like arrangement of subunits, the stability of pili mainly depends on complementation of conserved P1–P5 residues (Gly8, Val10, Phe12, Gly14, and Val16) in the donor strand (Puorger et al. 2008). Extensibility and stability of pili have presumably evolved to withstand high shear force, preventing the bacteria from being washed away by the flow of bulk urine (Spaulding, C.N. et al. 2018).
The uncoiling process of the pilus rod is typically divided into three distinct regions (I, II, and III) in force versus elongation response. This process has been investigated by force spectroscopic techniques including optical tweezers (Andersson et al. 2008) and atomic force microscopy (Forero et al. 2006). In region I, the entire pilus rod is stretched without significant conformational change. In region II, uncoiling of the pilus rod at a constant force is a consequence of breakages in layer-to-layer interactions (e.g. subunit i and subunit i+3) along the pilus rod. In region III, the pilus rod undergoes conformational changes in its head-to-tail region (e.g. subunit i−1 and subunit i) while the pilus rod is completely stretched out under higher forces than the constant force in region II. The uncoiled pilus rod can be spontaneously coiled and the restoring force can also be divided into the three stages (I, II, and III). These reversible uncoiling and coiling properties of the pilus rod enable UPEC to resist shear force from bulk urine flow.
For structural studies of inherently insoluble and non-crystalline pili, solid-state NMR and cryo-electron microscopy (cryo-EM) are advantageous methods. We have previously determined a high-resolution model of type 1 pili by combining 13C-detected solid-state NMR, scanning transmission electron microscopy, and inferential structure determination (Habenstein et al. 2015). The protomer in the pilus model was estimated from the self-complemented monomeric FimA structure, which was determined by solution NMR (PDB entry 2JTY) (Puorger et al. 2011). Recently, cryo-EM structures of type 1 pili were determined at a resolution of 4.2 Å (Spaulding et al. 2018; Hospenthal et al. 2017). The structure of the other major type of pili in UPEC (P pili) was determined by cryo-EM at a resolution of 3.8 Å (Hospenthal et al. 2016). A comparison between the solid-state NMR structure and the cryo-EM structures of type 1 pili reveals that their quaternary structures are different while monomeric structures of the pilin body and the donor strand are comparable. The solid-state NMR structure exhibited a highly coiled state with a shorter axial rise per subunit and a higher number of subunits per turn compared to the cryo-EM structures. This difference might originate from the different assembly pathways (in vitro for solid-state NMR versus in vivo for cryo-EM) as described by Hospenthal et al. (2017).
Here we use proton-detected fast magic-angle spinning (MAS) solid-state NMR spectroscopy of perdeuterated, fully proton back-exchanged pili to observe 1HN (amide proton), 13C, and 15N backbone resonances. Fast MAS of perdeuterated proteins with subsequent proton back-exchange has become a valuable method to study structure and dynamics of proteins by proton-detected solid-state NMR (Chevelkov et al. 2006; Lamley et al. 2014; Fricke et al. 2017; Najbauer et al. 2019). Proton-detection requires less amount of sample and provides higher sensitivity compared to 13C/15N-detection due to the higher gyromagnetic ratio of 1H. We achieved nearly complete backbone chemical shift assignments by conducting multi-dimensional solid-state NMR experiments, where assignments are missing for three residues at the N-terminus. We also investigated the degree of hydrogen/deuterium (H/D) exchange at labile sites in perdeuterated and fully proton back-exchanged pili. The previous non-equilibrium H/D exchange method by solution NMR (Wagner and Wüthrich 1982) was adapted for solid-state NMR: the degree of H/D exchange in a protonated protein in D2O buffer is measured after a certain incubation time (Gallagher et al. 1992; Whittemore et al. 2005; Wang et al. 2011; Medeiros-Silva et al. 2017; Grohe et al. 2017; Medeiros-Silva et al. 2016). A similar strategy has been applied to investigate the type iii secretion system needle protein (Chevelkov et al. 2017). The H/D exchange rate of amide protons depends on formation and breakage of hydrogen bonds, local conformational distortion, neighboring side-chains, and accessibility of water and hydroxide ions (OH-) to labile sites (Bai et al. 1993; Englander et al. 1996). Therefore, it is interesting to compare the extent of H/D exchange with the dynamic properties of pili in liquid phase (Maity et al. 2003). In addition to the H/D exchange experiments, we performed molecular dynamics (MD) simulations (McAllister and Konermann 2015; Persson and Halle 2015) on type 1 pili. From the simulations we estimated the probability of H/D exchange by counting the number of hydrogen bonds between backbone H-N and water in the absence of hydrogen bonds between the backbone H-N and other residues.
Materials and methods
Uniformly [2H,13C,15N]-labeled recombinant FimA was expressed in E. coli strain BL21 (DE3) using fully deuterated minimal medium with 15N-ammonium chloride and [2H,13C]-glucose as the sole nitrogen and carbon sources, respectively, as described before (Habenstein et al. 2015). The sample perdeuteration is required to dilute the strong network of proton-proton dipolar couplings that otherwise significantly impair spectral resolution. The perdeuterated pili sample was purified under denaturing conditions and kept at 37 °C for three weeks in 10 mM sodium phosphate buffer at pH 7.0 with 0.02 % sodium azide in H2O (non-denaturing 100 % H2O buffer) to produce fully proton back-exchanged pili (protonation at all labile sites), herein referred to as reprotonated pili (Fig. S1). The pili sample was then collected by centrifugation at 48,700×g and the pellet was packed into a 1.9 mm rotor with addition of a few crystals of DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid) and a small amount of D2O (~ 1 μL) for a field lock on the deuterons.
For the preparation of a redeuterated pili sample (Fig. S1), a pre-assembled reprotonated pili sample was washed with non-denaturing 100% D2O buffer [10 mM sodium phosphate at pH 7.0, 0.02 % sodium azide in D2O; prepared as described in the protocol (Fricke et al. 2017)] and incubated at room temperature for 4 h followed by ultracentrifugation at 48,700×g. These washing and pelleting procedures were repeated three more times. This leads to that only solvent accessible labile protons will be exchanged to deuterons in contrast to the initial reprotonation, which was performed before and during assembly of the pili and resulted in complete proton back-exchange of exchangeable sites. The redeuterated pili sample was packed into a 1.9 mm rotor with a few crystals of DSS. The amount of the reprotonated sample and the redeuterated sample in each rotor is approximately equal.
Solid-state NMR spectroscopy
Proton-detected solid-state NMR experiments were performed on 21.2-T and 16.4-T (1H Larmor frequencies of 900 MHz and 700 MHz, respectively) spectrometers (Bruker Biospin, Germany) equipped with four-channel MAS probeheads (1H, 13C, 2H, 15N) for a rotor diameter of 1.9 mm. All the experiments were carried out at an MAS frequency of 40 kHz. The proton DSS signal was set to 0 ppm and used as internal chemical shift reference. The internal sample temperature was monitored by the water proton resonance relative to DSS (Böckmann et al. 2009). Two dimensional (2D) (H)NH and three dimensional (3D) 3D (H)CANH, 3D (H)CONH, 3D (H)COCA(N)H, 3D (H)CA(CO)NH, and 3D (H)CO(CA)NH experiments were conducted for the reprotonated pili on the 21.2-T spectrometer at an internal sample temperature of 22±1 °C. 2D (H)NH and 3D (H)CANH spectra for the reprotonated pili and the redeuterated pili were acquired on the 16.4-T spectrometer at an internal sample temperature of 17±1 °C. Reintroduction of hetero- and homo-nuclear dipolar interactions under MAS was achieved by cross-polarization (CP) (Pines et al. 1972), SPECIFIC-CP (Baldus et al. 1998), and dipolar recoupling enhanced by amplitude modulation (DREAM) (Verel et al. 2001). Decoupling schemes including WALTZ-16 (Shaka et al. 1983) and X inverse-X (XiX) (Ernst et al. 2003) were used during the chemical shift evolution periods. Gaussian pulse cascades (Emsley and Bodenhausen 1992) (Q3 in Bruker nomenclature) were applied for band-selective inversion on Cα or CO. The detailed experimental procedures and pulse sequences of the above experiments are mostly explained in a previously published protocol (Fricke et al. 2017) and experimental parameters are provided in Tables S1–S8. TopSpin 3.2 (Bruker Biospin, Germany) and CcpNmr Analysis 2.2.4 (Vranken, W.F. et al. 2005) were used to process the spectra and to assign chemical shifts, respectively. The 3D (H)CANH spectra of the reprotonated pili and redeuterated pili were recorded on the same 16.4-T spectrometer 5 months after the sample packing of the redeuterated pili in a 1.9 mm rotor. For the purpose of consistency, the 3D (H)CANH spectra of the reprotonated pili and the redeuterated pili were acquired in a row. The peak volumes of assigned peaks were used to compare H/D exchange between redeuterated and reprotonated pili.
All-atom molecular dynamics simulations
The cryo-EM structure of the type 1 pilus rod containing six FimA subunits (PDB entry 5OH0, 4.2-Å resolution) was selected as the starting structure for molecular dynamics (MD) simulations. The missing amino- and carboxyl-terminal residues in the cryo-EM structure (Ala1 and Gln159) were added using PyMOL 1.3 (Schrödinger 2015). Dominant protonation states of all titratable residues at pH 7.0 were assigned. The protein was immersed in 23054 extended simple point charge (SPC/E) water models (Berendsen et al. 1987) in a box, which ensures that each edge of the box is at least 1 nm apart from the protein in the center of the box. In addition to the solvation, 30 sodium ions and 150 mM sodium chloride replaced randomly the same number of water models to neutralize the net charge of the protein and to reach the same salt concentration as in the cryo-EM study. To correct unfavorable geometry and steric clashes, energy minimization of the system was conducted using the steepest descent algorithm with a force constant of 1000 kJ/mol/nm and a maximum step size of 0.1 Å, and the maximum number of iterations was set to 50000 steps. After the energy minimization, the system was subjected to a 500-ps long canonical ensemble (NVT) equilibration at a target temperature of 300 K using a velocity-rescaling thermostat (Bussi et al. 2007) with a temperature coupling time constant of 0.1 ps. During the NVT equilibration, all heavy atoms were position-restrained with a force constant of 1000 kJ/mol/nm to equilibrate solvent (water and ions) around the protein. After the NVT equilibration, a 5-ns long isothermal-isobaric ensemble (NPT) equilibration was performed with a coupling of the isotropic Parrinello-Rahman barostat (Parrinello and Rahman 1981) at a target pressure of 1 atm (coupling time constant: 2 ps) and a target temperature of 300 K (coupling time constant: 0.1 ps). During the 5-ns NPT equilibration, only the heavy atoms of the protein backbone were position-restrained with a force constant of 1000 kJ/mol/nm to relax protein side-chains and to allow their interactions with solvent. After the NPT equilibration, another NPT equilibration of 5 ns was performed without any position restraints to relax all the atoms. The equilibrated system was submitted to a production run of 200 ns (NPT ensemble) without any position restraints. All steps of the energy minimization, the equilibrations, and the production run were repeated ten times with an assignment of random initial velocities for all the atoms, which were drawn from the Maxwell-Boltzmann distribution at 300 K. All simulations were performed using the Amber99SB*-ILDN force field (Best and Hummer 2009; Lindorff-Larsen et al. 2012). The neighboring particle search was carried out with a time interval of 20 fs for the calculation of non-bonded pair forces with a cut-off radius of 1 nm. Periodic boundary conditions (PBCs) were applied over the simulation time to avoid artifacts from the finite size of the simulation box. Long-range electrostatic interactions were treated by the particle mesh Ewald (PME) method (Essmann et al. 1995) with a Fourier grid spacing of 1.2 Å. The Lennard-Jones interactions and short-range Coulomb interactions were evaluated with a cut-off radius of 1 nm. All covalent hydrogen bonds of the protein were constrained using the parallel linear constraint solver (P-LINCS) algorithm (Hess 2008). An integration time step of 4 fs, instead of 2 fs, was allowed by an application of the virtual site approach (Feenstra et al. 1999), which removes degrees of freedom from hydrogen atoms in the protein. Coordinates were saved every 10 ps. All simulations were performed using the Gromacs 4.6.7 software package (Pronk et al. 2013). For the calculation of root mean square deviations (RMSDs) of backbone atoms, two exterior subunits were excluded. Hydrogen bond analysis between water and backbone H-N as well as backbone H-N and other residues, which was used to estimate the probability of H/D exchange, were performed using the Gromacs tool g_hbond. To avoid any artifacts arising from the limited number of subunits (hexamer), the estimated probability of H/D exchange in the two inner subunits surrounded by neighboring subunits were calculated and selected to represent the polymer structure (Fig. S3).
Data visualizations were carried out using matplotlib (Hunter 2007) with Numpy and Scipy (van der Walt et al. 2011). The cryo-EM structure of the type 1 pilus rod was visualized using PyMOL 1.3 (Schrödinger 2015).
The CSPs of backbone amide 1HN and 15N resonances, calculated from chemical shifts of self-complemented monomeric FimA (FimAa, solution NMR) (Puorger et al. 2011) and polymerized FimA (pili, solid-state NMR), indicate which residues are involved in intermolecular interactions resulting from polymerization (Fig. S2). Residues adjacent to Ala51 of subunit i+3 (Gln89-Ser95 of subunit i; Val37 and Val52 of subunit i+3) and residues adjacent to Ala77 of subunit i (Ile78, Ala87, and Leu113 of subunit i; Arg106 and Thr107 of subunit i+3) show the largest CSPs. While the layer-to-layer subunit interface around Ala51 and Ala77 in pili exhibits a different chemical environment compared to self-complemented monomeric FimA (FimAa), the regions related to DSC are relatively similar (Thr4-Asn18, Val28-Gly35, and Gly146-Tyr158).
There are, however, some amide protons that are completely exposed to H/D exchange despite being involved in intramolecular (Gly35, Thr76, Asp79, Ser90, Ala96, Thr107-Ala109, Asn125, Gly127, Thr140, Thr144, Asn149, and Thr153) or intermolecular (Val5, Gly7–Thr9, Glu15, Ser90, Ser95, and Tyr158) contacts. Interestingly, the two α-helices located in the interface of subunits i and i-1 (Ala25–Ser27 and Thr39–Ser41) are completely exposed to H/D exchange (Fig. 6). This may be related to polymorphism and/or slow conformational dynamics in the interface between subunits i and i−1 (Fig. S4b and Table S9).
In the donor strand, the amide protons of residues Val10–Lys13 and Val16-Val17 including Val10, Phe12, and Val16 (P2, P3, and P5, respectively) are highly protected while Gly8 (P1) is exposed to solvent (Fig. 6a). This indicates that the second half of the donor strand is tightly bound as in a β-sheet, whereas the first nine N-terminal residues are exposed to solvent. In the second half of the donor strand, Gly14 (P4) is relatively strongly exposed to H/D exchange. Mutation of the highly conserved Gly14 of P pili to Asn (Gly14Asn) showed impairment of self-polymerization of P pili (Verger et al. 2007). The mutation may hamper binding of the neighboring hydrophobic P residues (Phe12 and Val16) by an increased steric hindrance of Gly14Asn.
The most protected amide proton is at Val23, which is located in the highly protected short β-strand region of subunit i-1 (residues Cys21-Asp24) between the donor strand and the pilin body. This region is adjacent to the other three subunits (subunits i, i−2, i−3 in Fig. 6a) and is complemented with a highly conserved intramolecular disulfide bond between Cys21 and Cys61. Spaulding et al. reported that mutation of Ala22 to Arg resulted in lowering the unwinding force in force response versus elongation from an optical tweezer experiment (Spaulding et al. 2018), no intracellular bacterial communities (IBCs), and reduction of the ability to invade mouse bladder cells. In the FimA-FimC complex structure, the short β-strand (residues Ala22–Ala25) of FimA interacts with FimC (Crespo et al. 2012). Therefore, the H/D exchange protected short β-strand region may contribute to the mechanical properties and function of the type 1 pilus rod.
In contrast, part of the loop region between β-strands C and D (C-D loop: residues Gly75–Gly100) and the loop region between β-strands D and E (D-E loop: residues Asp105–Thr128) exhibit a high level of H/D exchange (Fig. 5a and b). In the pilus rod, these loops are located in the layer-to-layer subunit interface and on the exterior surface (Fig. 6b). The number of layer-to-layer subunit interactions per helical turn is more than three and these interactions are repeating along the axis of the pilus rod. The H/D exchange and structure of the layer-to-layer subunit interface may explain the elongation-competent property of these regions under external shear stress, which distribute the tension along the pilus rod.
We also performed MD simulations to study the dynamics of the pilus rod. Ten 200 ns MD simulations were carried out, resulting in total in a 2 µs trajectory. To check deviations from the starting structure and integrity of MD systems, the root mean square deviations (RMSDs) of the backbone of the inner four pilus rod subunits were calculated (Fig. S7). RMSDs of all ten MD runs converged to less than 2 Å, indicating that systems were well equilibrated over the simulation time. To estimate the probability of H/D exchange for backbone amide protons from the MD simulations, the number of hydrogen bonds between water and backbone H-N in the absence of hydrogen bonds between the backbone H-N and other residues was calculated (red dots in Fig. 5b). Regions with no or relatively few hydrogen bonds between backbone amide protons and water mostly correspond to the slow H/D exchange sites (orange bars in Fig. 5b) with the exception of Ile103 and Tyr137, where large discrepancies were observed. This may be due to the fact that experiments and simulations were performed at different time scales, i.e. the MD simulation time of 2 µs versus the H/D exchange of several hours.
In the present study, we assigned the backbone chemical shifts (residues Thr4-Gln159) of perdeuterated, fully proton back-exchanged pili based on multi-dimensional NMR spectra acquired by proton-detected solid-state NMR under fast MAS at 40 kHz (Table S9). To investigate H/D exchange of type 1 pili, we employed deuterium back-exchange of a perdeuterated, fully proton back-exchanged sample (Gallagher et al. 1992; Whittemore et al. 2005; Wang et al. 2011; Medeiros-Silva et al. 2017; Grohe et al. 2017; Medeiros-Silva et al. 2016; Chevelkov et al. 2017). This approach is different in the earlier steps before the redeuteration from the inversely fractional deuteration (iFD) (Medeiros-Silva et al. 2016). Our pili sample was uniformly [2H,13C,15N]-labeled during expression, proton back-exchanged at all labile sites by purification in 100% H2O buffer under denaturing conditions, followed by pili assembly in non-denaturing 100% H2O buffer. We applied this method to a supramolecular assembly while the iFD method has been used previously for membrane proteins. This H/D exchange study of pre-assembled type 1 pili revealed the distinct pattern of H/D exchange in the head-to-tail and the layer-to-layer regions. The regions that are highly protected from H/D exchange are located in the latter part of the donor strand (residues Val10–Lys13 and residues Val16–Val17) and the short β-strand (residues Cys21–Asp24) between the donor strand and the pilin body. From the mutational study by Spaulding et al., we can infer that the substitution of Ala22 to the basic and bulky residue Arg may influence the highly conserved intramolecular disulfide bond of Cys21–Cys61 and tight helical packing of subunits by steric clashes. In the head-to-tail interface, the two α-helices (Ala25–Ser27 and Thr39–Ser41) are completely exposed to H/D exchange (Fig. 5b). The exposure of the amide protons in the α-helices may be related to polymorphism and/or slow conformational changes as found from peak splittings in the NMR spectra for the head-to-tail interface (Fig. S4). The layer-to-layer subunit interactions are more prone to break based on the fact that most of the interface is composed of loops and is highly exposed to H/D exchange. Here we found that only two backbone amide protons (Ala51 and Ala77) are protected by layer-to-layer intermolecular contacts. The CSPs between the self-complemented monomeric FimA (solution NMR) (Puorger et al. 2011) and pili (solid-state NMR) show that there are significant differences of chemical environments in the vicinity of Ala51 and Ala77 (Fig. S2) that indicate a small layer-to-layer interface in pili and/or their conformational dynamics. In contrast, many amide protons are protected from H/D exchange by head-to-tail intermolecular contacts (Val10–Gly14, Val16–Val17, Val23, Val32, Leu34, Gln36–Val37, Gly146, Ala148, Ala150, Ala152, Phe154, and Val156) (Fig. 6c, d). The backbone-mediated intermolecular interactions in the head-to-tail interface may contribute to the extreme stability of pili in addition to their hydrophobic side-chain interactions.
The H/D exchange results from NMR experiments and the estimated probability of H/D exchange based on hydrogen bond dynamics from MD simulations are generally in good agreement. Nevertheless, it is important to note that MD simulations only provide dynamics of the stiff coiled-state of the pilus rod on a relatively short time-scale. In order to further improve consistency with experimental observables, MD simulations using enhanced sampling techniques may be considered in future studies (Bernardi et al. 2015).
Our integrative approach combining solid-state NMR spectroscopy and MD simulations give atomic resolution insights into mechanical properties and stability of type 1 pili. Investigations of the degree of H/D exchange at deuterium back-exchangeable sites from the perdeuterated, fully proton back-exchanged pili sample reveal the H/D exchange behavior in the context of assembled pili. We found high stability of backbone amides in the head-to-tail interface and the core of monomeric FimA in pili, whereas the layer-to-layer interface and exterior surface areas exhibit low protection against H/D exchange. The different H/D exchange behavior is consistent with resistibility and extensibility of pili under shear force. MD simulations of the pilus rod provided an estimated probability of H/D exchange of backbone H-N by hydrogen bond dynamics. The present study may also give insights into potential binding sites for new types of antibiotics such as coilicide, which makes UPEC prone to detach from host cells under bulk flow of urine (Klinth et al. 2012). A multi-drug strategy combining this approach with adhesin-host receptor inhibitors (Hartmann et al. 2011; Totsika et al. 2013) and pilicide, which blocks the interaction between chaperone-subunit and usher (Pinkner et al. 2006), may be a conceivable alternative to treat patients with urinary tract infection in the future.
We thank Dr. Han Sun, Dr. Tillmann Utesch, and Dr. Frank Eisenmenger (deceased) for valuable discussions. This work was supported by the Leibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP) and the European Research Council (ERC Starting Grant to A.L.).
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
- Hartmann M, Lindhorst TK (2011) The bacterial lectin FimH, a target for drug discovery—carbohydrate inhibitors of type 1 fimbriae-mediated bacterial adhesion. Eur J Org Chem. https://doi.org/10.1002/ejoc.201100407.
- Krogfelt KA, Bergmans H, Klemm P (1990) Direct evidence that the FimH protein is the mannose-specific adhesin of Escherichia coli type-1 fimbriae. Infect Immun 58:1995–1998Google Scholar
- Najbauer EE et al (2019) Probing membrane protein insertion into lipid bilayers by solid-state NMR. ChemPhysChem 20:302–310Google Scholar
- Schrödinger LLC (2015) The PyMOL molecular graphics system, version 1.8.Google Scholar
- Spaulding CN et al (2018) Functional role of the type 1 pilus rod structure in mediating host–pathogen interactions. eLife 7: e31662.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.