Bidirectional pilus processing in the Tad pilus system motor CpaF

Hohl, Michael; Banks, Emma J.; Manley, Max P.; Le, Tung B. K.; Low, Harry H.

doi:10.1038/s41467-024-50280-6

Download PDF

Article
Open access
Published: 05 August 2024

Bidirectional pilus processing in the Tad pilus system motor CpaF

Nature Communications volume 15, Article number: 6635 (2024) Cite this article

3623 Accesses
6 Citations
13 Altmetric
Metrics details

Subjects

Abstract

The bacterial tight adherence pilus system (TadPS) assembles surface pili essential for adhesion and colonisation in many human pathogens. Pilus dynamics are powered by the ATPase CpaF (TadA), which drives extension and retraction cycles in Caulobacter crescentus through an unknown mechanism. Here we use cryogenic electron microscopy and cell-based light microscopy to characterise CpaF mechanism. We show that CpaF assembles into a hexamer with C2 symmetry in different nucleotide states. Nucleotide cycling occurs through an intra-subunit clamp-like mechanism that promotes sequential conformational changes between subunits. Moreover, a comparison of the active sites with different nucleotides bound suggests a mechanism for bidirectional motion. Conserved CpaF residues, predicted to interact with platform proteins CpaG (TadB) and CpaH (TadC), are mutated in vivo to establish their role in pilus processing. Our findings provide a model for how CpaF drives TadPS pilus dynamics and have broad implications for how other ancient type 4 filament family members power pilus assembly.

Conformational changes in the motor ATPase CpaF facilitate a rotary mechanism of Tad pilus assembly

Article Open access 24 April 2025

Assembly mechanism of a Tad secretion system secretin-pilotin complex

Article Open access 13 September 2023

Acinetobacter baylyi regulates type IV pilus synthesis by employing two extension motors and a motor protein inhibitor

Article Open access 18 June 2021

Introduction

Tight adherence pilus system (TadPS) pili are filaments that extend into the external milieu and facilitate the anchoring of bacteria to surfaces¹. By mediating cell adhesion, they represent essential colonisation factors for pathogens such as Aggregatibacter actinomycetemcomitans, Pseudomonas aeruginosa and Vibrio vulnificus^2,3,4. In non-pathogenic Caulobacter crescentus, anchoring to surfaces occurs through dynamic cycles of pilus extension and retraction until surface contact is established^5,6. Pilus retraction orients the cells and stimulates the synthesis of the permanent holdfast. In addition, TadPS pili bundle to form biofilms^4,7, facilitate DNA uptake⁸ and mediate contact-dependent prey killing in predatory social bacteria^9,10. The TadPS therefore has a broad functional repertoire conserved across many Gram-negative and Gram-positive bacteria¹¹. Despite this prevalence amongst the Eubacteria, the TadPS has an intriguing archaeal heritage as part of the wider type 4 filament family with an ancient lineage originating with the last universal common ancestor (LUCA)^11,12. Closest archaeal ancestors include the rotary archaellum¹³, and those systems producing UV-inducible pili (Ups)¹³, archaeal adhesive pili (Aap)¹⁴ and EppA-dependent (Epd) pili¹⁵. For these pilus systems, little is known about the secretion system powering pilus assembly although it is likely they share close mechanistic principles with the TadPS given their common evolutionary heritage¹⁶. Other eubacterial members of the type 4 filament family include the type 4 pilus system¹⁷ that, like the TadPS, assembles retractable surface pili, and the type II secretion system¹⁸ (T2SS), which assembles a dynamic endopilus retained within the periplasmic space.

The typical TadPS constitutes 12-14 genes in a single operon. These include the outer membrane secretin RcpA with its cognate pilotin TadD^19,20 forming a channel through which the assembled Flp1 pilus^21,22 threads and exits the cell. Pilus assembly occurs at the inner membrane and incorporates the most conserved components of the TadPS including the cytoplasmic ATPase TadA²³ and the inner membrane platform proteins TadB and TadC, which share homology to GspF and PilC in the T2SS and type 4 pilus systems, respectively²⁴. These three proteins are likely sufficient to drive pilus assembly given that they constitute the minimal TadPS machinery observed in Gram-positive bacteria⁴ along with the major pilin Flp1 and TadZ²⁵ implicated in TadPS positioning within the cell. Whilst TadA is broadly distributed across α/β/γ/δ-Proteobacteria, Actinobacteria and Firmicutes²⁶, it is most closely related to TrbB/VirB11-type ATPases, which comprise part of the wider family of motor ATPases that energise type IV secretion systems (T4SS)²⁴. Little is known about how TadA operates including its architecture, mode of assembly and nucleotide cycling mechanism.

In contrast, Geobacter metallireducens PilB from the type 4 pilus system is one of the most studied motors. It forms a C2 symmetric hexamer where nucleotide turnover is coupled to a clockwise sequence of symmetric conformational changes dedicated to pilus extension²⁷. PilC is predicted to nestle within the central channel of PilB possibly rotating like a drive shaft and transducing lever-like conformational changes that elevate pilins from the inner membrane into the assembling pilus²⁷. Intriguingly, by adjusting the conformation of the different nucleotide binding pockets and their relative affinity for different nucleotide states, PilB from Thermus thermophilus likely supports counter-clockwise motion²⁸. Other mechanisms for PilB- and PilC-mediated pilus assembly have been theorised including a compression model where pilins are extruded from the inner membrane¹⁷ or actively extracted by a trimeric PilC chuck-like arrangement²⁹. In the TadPS, evolutionary studies show that TadB and TadC arose from the fission of a single GspF/PilC ancestral gene¹¹, so that TadB and TadC likely constitute a single functional platform protein. Given the evolutionary homology between the TadPS, T2SS and type 4 pilus system, it is quite possible that they all share conserved mechanisms for pilus assembly mediated by their respective ATPase motors and platform proteins. How pilus retraction occurs, though, is an open question, as most systems do not have dedicated retraction ATPases. This is consistent with primordial type 4 filament systems functioning with just a single ATPase. Type 4 pilus systems are exceptional though as retraction is coordinated by the dedicated ATPase PilT, which was acquired relatively late in the evolutionary process^12,16. PilT has been extensively crystallised in multiple different conformations and symmetries^30,31,32. However, cryogenic electron microscopy (cryo-EM) studies of G. metallireducens PilT support a functional C2 symmetric form where counter-clockwise motion antagonises PilB extension resulting in pilus retraction³³.

In the C. crescentus TadPS, pilus extension and retraction are mediated by the ATPase CpaF (TadA) alone²⁶ – yet the mechanism for how one motor drives bidirectional pilus cycling remains elusive. It is also unknown whether this motor bifunctionality is conserved across the TadPS family or is distinct to C. crescentus. Similarly, fundamental questions regarding the TadPS pilus assembly mechanism remain given that only the outer membrane secretin and pilotin²⁰, and the pilus²¹, have been structurally characterised from the core apparatus.

Here we use a combination of cryo-EM, light microscopy and cell-based assays to show how CpaF from C. crescentus forms a C2 symmetric hexamer primed to couple nucleotide cycling to a clockwise rotational mechanism consistent with right-handed pilus extension. We use sequence conservation analysis, Alphafold modelling and structure-guided mutagenesis to probe how CpaF may engage the platform proteins CpaG (TadB) and CpaH (TadC) and drive pilus assembly. Finally, by comparison of CpaF structures in different nucleotide states and with other related motors, a mechanism is suggested for how CpaF drives both pilus extension and retraction.

Results

Cryo-EM structure determination of C. crescentus CpaF with AMPPNP

To probe the mechanism by which CpaF powers the TadPS, single particle cryo-EM was undertaken. CpaF from C. crescentus was cloned and purified to homogeneity (Supplementary Fig. 1A). After incubation with the non-hydrolysable ATP analogue AMPPNP and in the presence of Mg²⁺, the sample was vitrified (Supplementary Fig. 1A) using holey grids and a data set collected for high resolution structure determination. Subsequent data processing yielded a map constituting six subunits arranged in a C2 symmetric ring (Supplementary Fig. 1B and 1E, and Supplementary Table 1). Side chain detail was sufficient to build amino acid residues 150–502 relating to the N-terminal domain 2 (NTD2) and C-terminal domain (CTD) for each subunit at 3.8 Å resolution overall (Supplementary Fig. 1C, 1D and 2A). Whilst residues 1–79 termed N-terminal domain 0 (NTD0) were predicted to be disordered and were not observed in the map, residues 80–149 constituted the N-terminal domain 1 (NTD1) and were predicted to form a 3-helix motif. Due to flexibility between the NTD1 and NTD2, map quality was insufficient for NTD1 model building. A focussed refinement strategy zoned on one subunit improved the resolution of the NTD1 map (Supplementary Fig. 2B and 2C) so that an Alphafold³⁴ model of the 3-helix motif could be rigid body fitted and modelled (Supplementary Fig. 1D). Henceforth, this CpaF structure obtained in the presence of AMPPNP is termed CpaF_AMPPNP.

The CpaF monomer has a modular fold similar to FlaI-like archaeal motors

The basic building block (asymmetric unit) of the CpaF oligomer is a trimer that repeats to form an oval-shaped hexamer with C2 symmetry ~128 Å wide and ~80 Å high (Fig. 1A). Within each subunit, the helical NTD1 is connected to the NTD2 by a short linker (Hinge 1) with conserved residues G147 and G149 likely mediating flexibility between these domains (Fig. 1B and Supplementary Fig. 3). The NTD2 comprises a layer of six β-sheets with two helices and connects to the CTD via an eight-residue lysine-rich linker (Hinge 2) that wraps around and forms part of the nucleotide-binding pocket. The CTD has a conserved RecA-like fold³⁵ with central β-sheets surrounded by α-helices on each side. It includes classical motifs such as P-loop, Walker A, Walker B, ASP box and HIS box (Supplementary Fig. 3) with the nucleotide binding pocket cradled between CTD and NTD2 interfaces. Hidden Markov Model analysis shows that the overall fold of the CpaF monomer is like the archaeal motor FlaI in Sulfolobus acidocaldarius, which powers archaellum rotation³⁶, and the FlaI-like ATPase GspE2 in Archaeoglobus fulgidus with unknown function³⁷. Comparison of their structures provides firm evidence for their shared archaeal ancestry based on tertiary fold, consistent with evolutionary sequence analyses (Supplementary Fig. 4)¹¹.

**Fig. 1: Cryo-EM structure of *C. crescentus* CpaF_AMPPNP.**

CpaF_AMPPNP subunits pack in distinct conformations that induce C2 symmetry within the hexamer

To assemble the CpaF hexameric ring, subunits associate laterally around an extensive central cavity with the CTDs forming the ring base, the NTD2s stacked above and the NTD1s protruding at the top. Importantly, the NTD2s are rotationally offset so that the NTD2 of subunit n interfaces with the CTD of subunit n + 1 in a domain swap (Fig. 1A). This structural unit is henceforth termed NTD2/CTD_n+1. Although CpaF was incubated with AMPPNP, this nucleotide was symmetrically bound in only two of the nucleotide-binding pockets within the hexamer. ADP coordinated to a Mg²⁺ ion (ADP/Mg²⁺) was symmetrically bound in the neighbouring subunit, with ADP alone in the remaining pair (Fig. 1A). The ADP is likely derived from AMPPNP hydrolysis given non-hydrolysable ATP analogues are not entirely resistant to hydrolysis³⁸. Mg²⁺ was present in the vitrification buffer indicating a requirement for specific active site geometry to support its recruitment. Comparison of the three subunits in the asymmetric unit showed that whilst the AMPPNP and ADP bound chains were relatively similar with RMSD Cα = 1.2 Å, the ADP/Mg²⁺ chain underwent a significant conformational change where the CTD rotated ~34˚ around Hinge 2. This induced the CTD to swing upwards 18 Å relative to the NTD2 resulting in a clamp-like effect around the nucleotide binding pocket (Fig. 1C and Supplementary Movie 1). Therefore, it is largely the packing mismatch of the ADP/Mg²⁺ bound subunits relative to the other four that generates the C2 symmetry within the hexamer. Given this, it is important to note that the superposition of the three distinct NTD2/CTD_n+1 units within the hexamer shows close similarity with RMSD Cα \(\le\) 1.1 Å (Fig. 1D). This means that the conformational changes observed between subunits within the CpaF hexamer are predominantly mediated by large-scale rigid body shifts between NTD2/CTD_n+1 units, with Hinge 2 providing the necessary flexibility.

The CpaF_AMPPNP hexamer exhibits symmetric open, closed and open’ nucleotide binding pockets that reveal the catalytic cycle

The conformation of the three subunits in the asymmetric unit, and how they interact with their neighbours defines the shape and accessibility of their nucleotide binding pockets. In those subunits where AMPPNP and ADP alone are bound, the active sites are in similar but still distinct conformations termed open and open’, respectively, with the solvent channel leading to the nucleotide-binding pocket sufficient to facilitate the passage of nucleotide (Figs. 1A, 2A and 2B). These open/open’ states are also characterised by a solvent-accessible cleft running between the CTD and the CTD_n+1 of the neighbouring subunit (termed henceforth the CTD/CTD_n+1 cleft). The cleft is broad enough to support the binding of a HEPES moiety recruited during sample purification (Figs. 2A, B). The equivalent cleft in PilT has been shown to bind ethylene glycol³³. In the ADP/Mg²⁺-bound subunit sandwiched between the open/open’ state subunits, the active site is in a closed conformation with the nucleotide-binding pocket largely occluded by the clamped CTD and NTD2. Consequently, the CTD/CTD_n+1 cleft is compressed with no HEPES bound, and Loop 26 between β-sheets 14 and 15 is now directly contacting H13 in CTD_n+1 (Fig. 2A).

**Fig. 2: CpaF_AMPPNP has symmetric open, closed and open’ nucleotide binding pockets that showcase the catalytic cycle.**

Close analysis of the active sites reveals how nucleotide cycling and catalysis are achieved. In the open conformation, the adenine moiety of the bound AMPPNP is secured at its side and back by conserved R431 in β-sheet 13 and non-conserved F256, respectively (Figs. 1B and 2A). K244 is 4.0 Å from the ribose so within range for hydrogen bonding (Fig. 2A and Supplementary Fig. 5). Note that a K244R mutation (in combination with F243L) reduces pilus extension and retraction rate²⁶ possibly by modulating nucleotide access to the binding pocket or impeding its binding. The rest of the nucleotide is cradled by the P-loop including K287 within the conserved consensus sequence coordinated to the γ-phosphate (Fig. 2C and Supplementary Fig. 5). Conserved ASP box E312, Walker B E357, R359 and HIS box H382 are 4-6 Å from the γ-phosphate tip. NTD2 residues R217 and conserved R223 located within β-sheets 4 and 5, respectively, are turned over 12 Å away from the β- and γ-phosphates. Similarly, R347 in Helix 9 from the neighbouring CTD_n+1 is over 12 Å away from the γ-phosphate. Here, R347 and R359 are some of the most conserved residues amongst CpaF homologues (Supplementary Fig. 3) and across other type 4 filament family motors including the archaellum, T2SS and type 4 pilus system (Supplementary Fig. 6). Nucleotide binding is therefore stabilised by a ring of basic charged residues. However, without R217, R223 and R347 engaged with the phosphates, the active site in the open state is not primed for catalysis. In contrast, the neighbouring subunit in the closed conformation has undergone significant rearrangement and is in a catalytically active state (Fig. 2C and Supplementary Fig. 5). Map was observed for the water-activating Mg²⁺ ion coordinated with ASP box E312 and Walker B E357 located 4 Å and 5 Å away, respectively; whilst R217 and R223 are now stabilising the ADP β-phosphate positioned just 3 Å and 4 Å apart, respectively. Similarly, R347 from the neighbouring CTD_n+1 has moved less than 7 Å away from the β-phosphate and is positioned to stabilise the γ-phosphate during the ATP hydrolysis transition state. In the archaeal FlaI, the equivalent residue R326 has been shown to undertake this stabilising role locating within the hydrogen bonding distance of the released phosphate positioned 4.2 Å from the β-phosphate³⁶. HIS box H382 has switched rotamer moving into a position where it may coordinate the γ-phosphate, similar to the role of HIS box H420 in PilB²⁷. In CpaF, a channel connecting the nucleotide-binding pocket to the central cavity of the hexameric ring is sufficiently broad to enable the γ-phosphate to exit the active site, which likely explains its absence. Finally, examination of the neighbouring open’ conformation subunit with ADP alone bound reveals an active site with similarities to the open state (Fig. 2B and Supplementary Fig. 5) with R217, R223 and R347 rotated away from the nucleotide (Fig. 2C). However, of notable difference, the ADP ligand is loosely bound with a relatively disordered active site including unstructured P-loop with no model built from G282 to G286. Importantly, the sidechain of highly conserved P-loop K287, which is critical for nucleotide binding, has partially disengaged from the nucleotide with HIS box residue H382 moved inwards and now just 5 Å from the β-phosphate. Similarly, β-sheets 12 and 13 become partially disordered (no model built for residues 425–431) so that their connecting loop no longer contacts the HEPES moiety in the CTD/CTD_n+1 cleft as in the open conformation (Fig. 2A). This disordered region includes R431, which cannot now effectively secure the adenine moiety of the nucleotide in the active site. Overall, the architecture of the open conformation is consistent with a state where the ADP nucleotide is bound with low affinity and primed for release. In summary, the CpaF_AMPPNP structure reveals an asymmetric unit in three conformations. Importantly, these conformations showcase the key stages in the nucleotide hydrolysis cycle including ATP loading (open state), ATP hydrolysis to ADP with the exit of the γ-phosphate from the active site (closed state), and ADP bound with low affinity primed for release (open’ state).

In the presence of ADP, the CpaF open’ state resembles the CpaF_AMPPNP open state

To understand how different nucleotide states might regulate CpaF assembly and conformation, the structure of CpaF was determined by cryo-EM in the presence of ADP and Mg²⁺. CpaF incubated with ADP yielded our highest quality map at 3.1 Å resolution overall facilitating a build of the NTD2 and CTD with excellent sidechain detail for all subunits (Supplementary Fig. 1E, 2D and 2F, and Supplementary Table 1). The 3-helix bundle of the NTD1 was again poorly ordered indicating domain flexibility relative to NTD2. The NTD1 from the AMPNP-bound model was therefore rigid-body fitted in the unsharpened map and modelled to complete the build. Henceforth, this CpaF structure obtained in the presence of ADP is termed CpaF_ADP.

CpaF_ADP formed a C2 symmetric hexamer with dimensions and subunit packing similar to CpaF_AMPPNP (Fig. 3A). Superposition of the three distinct NTD2/CTD_n+1 units within the CpaF_ADP hexamer showed remarkable alignment with RMSD Cα \(\le\) 0.35 Å (Fig. 3B). Similarly, superposition of the asymmetric units from both CpaF_ADP and CpaF_AMPPNP showed near identical packing between the NTD2/CTD_n+1 units with RMSD Cα = 1.5 Å (Fig. 3C). As observed in CpaF_AMPPNP, subunits and their nucleotide-binding pockets were in symmetric open, closed and open conformations. However, ADP/Mg²⁺ was now bound in each nucleotide-binding pocket (Fig. 3D and Supplementary Fig. 5). Analysis of the active sites revealed an arrangement and geometry of key residues that was similar to CpaF_AMPPNP. Specifically, in the open and open’ states, residues R217 and R223, and R347 from CTD_n+1, were all positioned > 15 Å from the nucleotide and unable to engage it. In the open state, K287, E357, R359 and H382 were positioned like in the CpaF_AMPPNP open state despite the absence of the γ-phosphate. ASP box E312 was coordinated 4 Å from the Mg²⁺. In the closed conformation, R217 and R223 were positioned less than 5 Å and 3 Å away from the β-phosphate, whilst R347 from CTD_n+1 was now just 8 Å away. All these residues were suitably positioned to stabilise the transition state during nucleotide hydrolysis. However, despite these similarities, there were critical differences between the open’ conformations of CpaF_ADP and CpaF_AMPPNP. In CpaF_ADP, the map around the nucleotide binding pocket was now fully ordered with the P-loop well resolved. Crucially, HIS box residue H382 was turned away from the nucleotide with P-loop K287 fully engaging the β-phosphate just like in CpaF_AMPPNP open state. Moreover, superposition of subunits for CpaF_ADP open’ state with CpaF_AMPPNP open state, including their interfaced CTD_n+1, revealed remarkable similarity with RMSD Cα = 1.0 Å (Fig. 2B). These two states therefore share closer similarity than the open and open’ states of CpaF_AMPPNP with an RMSD Cα = 1.4 Å when superposed (Fig. 2B). Broadly, this means that when ADP/Mg²⁺ is bound to CpaF_ADP in the open state (Fig. 3D), then the nucleotide binding pocket of the open’ state subunit resides in a conformation similar to the open state of CpaF_AMPPNP (Fig. 2C and Supplementary Fig. 5). This finding suggests a model where the CpaF_ADP open’ state has the appropriate conformation and geometry of key residues to support AMPPNP (or ATP) binding when ADP/Mg²⁺ is bound in the open state subunit.

**Fig. 3: Cryo-EM structure of CpaF_ADP.**

CpaF assembles as a relaxed C2 symmetric hexamer in the absence of nucleotide

To understand how the absence of nucleotide might regulate CpaF assembly and conformation, the structure of CpaF was determined by cryo-EM without nucleotide addition but in the presence of Mg²⁺. In this case, CpaF was less stable yielding a map resolved to 4.0 Å resolution overall (Supplementary Figs. 1E, 2E and 2G, and Supplementary Table 1). Map quality was sufficient to build complete models of the CTD and NTD2 for all subunits (Supplementary Fig. 2G). Again, the NTD1s were poorly ordered in the map and were built using the NTD1 from CpaF_AMPPNP rigid-body fitted and modelled. Henceforth, this CpaF structure obtained in the absence of nucleotide is termed CpaF_APO.

CpaF_APO formed a C2 symmetric hexamer with subunits arranged in an open, closed and open’ conformation like CpaF_AMPPNP and CpaF_ADP (Supplementary Fig. 7A). No nucleotide or Mg²⁺ was observed in the nucleotide-binding pockets (Supplementary Fig. 7B) although HEPES was bound in the CTD/CTD_n+1 clefts in both open and open’ state subunits. Superposition of the three distinct NTD2/CTD_n+1 units within the CpaF_APO hexamer showed close alignment with RMSD Cα \(\le\) 0.7 Å (Supplementary Fig. 7C). However, comparison of the CpaF_APO and CpaF_ADP asymmetric units showed that whilst similar, they do not pack identically and cannot be considered equivalent to CpaF_ADP superposed on CpaF_AMPPNP. Specifically, Hinge 2 within the open state subunit is straightened so that its NTD2 moves towards the neighbouring NTD2 of the closed subunit (Fig. 3C). Consequently, the nucleotide-binding pocket in the open state subunit differs significantly in comparison to CpaF_AMPPNP and CpaF_ADP as the associated NTD2/CTD_n+1 unit has rotated (Supplementary Fig. 7B). The overall effect is a relaxed and more rounded C2 symmetric hexamer. CpaF_APO is therefore consistent with a model where CpaF preferentially assembles as a C2 symmetric hexamer irrespective of nucleotide state. Critically though, nucleotide binding accentuates the oval-shaped C2 symmetry observed in CpaF_AMPPNP and CpaF_ADP hexamers and switches the conformation of the open state nucleotide binding pocket, with this form likely to be prevalent in vivo given the high nucleotide concentration in the cell.

CpaF Loop 8 and Central Loop likely interact with CpaG and CpaH, and are essential for TadPS function

To investigate CpaF mechanism, CpaF_AMPPNP and CpaF_ADP structures combined with sequence alignments (Supplementary Figs. 3 and 6), were used to select truncations or highly conserved residues (Fig. 4A) for chromosomal mutation. The impact of these mutations was assessed in two cell-based in vivo functional assays. The first was a TadPS pilus-dependent phage infection using ΦCbK³⁹ where TadPS deactivation impairs phage entry, cell lysis and ultimately plaque formation in C. crescentus lawns. However, this assay does not distinguish extension versus retraction mutants as pilus retraction is not essential for ΦCbK infection²⁶. The second fluorescence microscopy-based assay was therefore performed to directly visualise and quantify TadPS surface pili, which are produced by C. crescentus swarmer cells. This assay involves fluorescent labelling of Tad pili by mutating T36C within the major pilin protein PilA (Flp1). This mutation facilitates binding of a thiol-reactive maleimide-AF488 fluorescent dye (which has strong affinity for cysteine residues) to the pili without compromising pilus integrity⁶. Addition of dye to cell culture results in direct labelling of extended cell-surface pili and also the cell body due to retraction of fluorescent pilin subunits into the cell envelope. Pilus extension mutants therefore have no fluorescent pili or cell bodies, whereas pilus retraction mutants have fluorescent pili but dark cell bodies²⁶.

**Fig. 4: Effect of targeted CpaF mutagenesis on TadPS function in vivo.**

Working with the ΦCbK assay first, we verified that the PilA T36C strain functioned like wildtype (Fig. 4B)⁶. CpaF was then deleted (PilA T36C ΔcpaF) resulting in loss of phage infection before complementation (PilA T36C ΔcpaF::cpaF) restored infection levels. This validated the assay and showed that CpaF is essential for TadPS function^6,23,40. All subsequent mutations were constructed in this complemented strain. As a final control, a CpaF Walker B mutant (E357A) was generated resulting in the absence of phage plaques presumably due to loss of CpaF nucleotide hydrolysis. Cellular expression for individual mutants was verified by Western blot (Supplementary Fig. 8A) whilst hexamer formation was verified by purification and negative stain EM analysis (Supplementary Fig. 8B). The CpaF NTD0 (Supplementary Fig. 3) is predicted to be disordered with unknown function. In order to probe its role, NTD0 was truncated (Δaa1-78) and its ability to support phage infection assayed. Wildtype levels of phage plaques were observed indicating no requirement for this motif in pilus extension (Fig. 4B). The NTD0 truncation was expanded to include removal of NTD1 (Δaa1-146), which resulted in loss of phage infection, however, this mutant should be interpreted with caution as in vivo expression was not readily detected by Western blot (Supplementary Fig. 8A) despite purifying and assembling in vitro (Supplementary Fig. 8B). In contrast, the mutation of two conserved glycines (G147A/G149A) positioned within Hinge 1, connecting the NTD1 to NTD2, inhibited phage infection. Therefore, NTD1 flexibility is an essential part of the CpaF mechanism.

During the course of this study, Alphafold-Multimer⁴¹ enabled the modelling of CpaF with CpaG and CpaH (Fig. 5A and Supplementary Fig. 9) showing a heterodimer that bridges two neighbouring CpaF subunits. Specifically, each CpaG or CpaH subunit interacts with Helix 5 conjoined with Loop 8, and the Central Loop of CpaF (Fig. 5B). To test the predicted interface between CpaF and the platform proteins CpaG and CpaH, two mutants were generated. The first mutant D208K resides in Loop 8, which is conjoined to Helix 5 within the NTD2, whereas the second mutant E331K/G332A/S333A/G334A located to the Central Loop within the CTD (Fig. 4A and Supplementary Fig. 3). These residues are highly conserved amongst CpaF homologues and locate to the CpaF surface facing the inner membrane where CpaG and CpaH are positioned¹⁹. For both these mutants phage infection was inhibited. Whilst it should be cautioned that a proportion of this inhibition may be due to lowered CpaF expression in vivo (Supplementary Fig. 8A) and stability during purification in vitro, the inhibition is consistent with the mutated residues playing an essential role in CpaF function, likely by interfacing with the platform proteins CpaG and CpaH.

**Fig. 5: In-silico modelling of CpaF with CpaG/CpaH.**

To further clarify whether any of the ΦCbK infection-negative mutants were competent to assemble but not retract pili, the mutants were visualised by fluorescence microscopy. As expected, fluorescent cells and pili were observed for the PilA T36C strain (Figs. 4C, E) and the CpaF complemented strain. However, for the CpaF Δaa1-78 truncation strain, whilst the number of piliated cells was similar to the PilA T36C strain, the proportion of fluorescent cells was significantly reduced from 34.3 % ± 5.0 in the PilA T36C strain to 19.9 % ± 5.0 (P < 0.0001; Fig. 4E). The CpaF Δaa1-78 truncation strain therefore has a reduced capacity to retract pili. For all other mutants no fluorescent pili or cells were observed confirming a loss of pilus extension rather than a retraction defect (Figs. 4D, E). Collectively, these mutants uncover key residues in CpaF including how the CpaF N-terminus positively regulates pilus retraction. In addition, they highlight the important role of both Loop 8 and the Central Loop in TadPS function likely due to their direct interaction with CpaG and CpaH platform proteins.

Discussion

Here we use CpaF from C. crescentus as a model to probe how one motor might power both pilus extension and retraction in this system and possibly in other bacterial homologues. The CpaF_AMPPNP hexamer has subunits in three symmetric conformations with AMPPNP bound in the open state, ADP/Mg²⁺ in the closed state and ADP in the open’ state. The structure therefore represents three key stages in the nucleotide hydrolysis cycle where i) the open state has high affinity for ATP and the catalytic residues in the active site are disengaged; ii) the closed state is catalytically active and captured in a post-hydrolysis state with high affinity for ADP, catalytic residues engaged around the nucleotide, and the γ-phosphate released into the central cavity; iii) the open’ state has low affinity for ADP, the catalytic residues are disengaged in the active site and the nucleotide is primed for release. The sequence of these states, combined with the C2 symmetry, is consistent with CpaF_AMPPNP geared for nucleotide cycling in a clockwise motion when viewed from the NTD1 side of the hexamer (Fig. 6). In this model, the long axis of the oval-shaped central cavity rotates ~60˚ with each step of the nucleotide hydrolysis cycle and naturally supports the extension of the right-handed C. crescentus PilA pilus²¹.

**Fig. 6: Model for the bidirectional motion of the CpaF motor.**

Conversely, one way to achieve pilus retraction would be to promote counter-clockwise motion in the CpaF motor. This could be achieved if the open’ state switched to have a high affinity for ATP and the open state to have low affinity for ADP thereby reversing the direction of nucleotide-induced conformational changes over the hydrolysis cycle. In support of this nucleotide switch model, analysis of CpaF_ADP open’ state shows how the active site geometry may be configured to closely resemble an ATP binding conformation. The CpaF_ADP open’ state active site is ordered and whilst ADP/Mg²⁺ is bound it shows remarkable similarity in backbone conformation and side chain geometry to the active site of CpaF_AMPPNP open state with AMPPNP bound (Fig. 6A and Supplementary Fig. 5). This is in contrast to CpaF_AMPPNP open’ state which is comparatively disordered with unstructured P-loop (Fig. 2C) and partially unstructured β-sheets 12 and 13 so that R431 no longer secures the nucleotide base (Supplementary Fig. 5). Concurrently, the CpaF_ADP open’ state shows close similarity to the active site of CpaF_ADP open state (Fig. 6A and Supplementary Fig. 5). As the nucleotide binding pockets are solvent accessible in both these states, the CpaF_ADP structure is consistent with the nucleotide switch model where ATP and ADP binding in the open’ and open states, respectively, would potentiate counter-clockwise motion (Fig. 6B). How a nucleotide switch may be regulated to shift between clockwise and counter-clockwise motion is unclear. Some CpaF homologues have variable NTD0 extensions (Supplementary Fig. 3) with sufficient length to reach the nucleotide binding pockets and in principle to allosterically modulate the geometry of active sites or nucleotide access and affinity. In this respect, it is intriguing that Alphafold models N-terminal Helix 0 (aa 57–66) at the mouth of the nucleotide binding pocket where it could modulate or gate the nucleotide binding pocket (Fig. 5A). Any such interaction would be dynamic as density was not observed for H0 in the CpaF maps. As loss of the NTD0 (aa1–78) reduces pilus retraction (Fig. 4E), this motif has the capacity to control or at least contribute to CpaF pilus extension and retraction bifunctionality. Importantly, the role of the NTD0 in modulating CpaF bifunctionality is also supported by comparison studies between CpaF in both C. crescentus and Asticcacaulis biprosthecum where differences in NTD0 sequence were suggested to account for variations in pilus assembly dynamics²⁶.

Many other motors from type 4 filament members have been crystallised including PilB^27,28,30, PilT^31,32,33 and PilF⁴² from the type 4 pilus system, and FlaI³⁶ and GspE³⁷ from archaeal systems. Multiple combinations of active site conformation and symmetries in the presence of different nucleotides were reported showcasing the conformational promiscuity of these motors. However, at least for PilB and PilT, complementary structure determination using cryo-EM has also been undertaken concluding a model where the C2 symmetric hexamer in open/closed/closed states support clockwise motion versus open/closed/open’ states for counter-clockwise motion, respectively³³ (Supplementary Fig. 10). For the unidirectional motor PilB, a nucleotide switch is not readily supported as the active sites in the closed states are not sufficiently solvent accessible to enable nucleotide exchange. Therefore, the earliest type 4 filament member motors may have been similar to CpaF with open/closed/open’ states that could be geared for bidirectional motion. As previously suggested³³, the emergence of unidirectional motors like PilB would then be due to recent evolutionary specialisation¹⁶.

CpaF and the platform proteins CpaG/CpaH share close homology with GspE and GspF in the T2SS, and PilB and PilC in the type 4 pilus system, respectively¹¹. In these secretion systems they form essential interactions^43,44,45,46 with the PilB/PilC complex located in the cytoplasm and inner membrane⁴⁷. Similarly, CpaF is likely located at the base of the TadPS (Fig. 7A) interacting with CpaG and CpaH to spool pilins to or from the pilus through an unknown mechanism. Based on this homology we chose universally conserved residues within Loop 8 and the Central Loop of CpaF to mutate. These were modelled to interact with CpaG and CpaH using Alphafold-Multimer (Fig. 5A). Loop 8 and the Central Loop were likely essential for TadPS function in vivo presumably as the mutations interfere with CpaG and CpaH binding. This data combined with insights from CpaF_AMPPNP and CpaF_ADP suggest two models for how CpaF energises the TadPS based on either rotary or gyratory movements of CpaG and CpaH. Both models depend on CpaF nucleotide cycling, which drives large-scale rearrangements in the packing between NTD2/CTD_n+1 units enabling the C2 axis to rotate around the oval-shaped hexamer (Fig. 6). Concurrently, this motion induces dramatic changes in the surface topography of the hexamer as the Central Loops undergo lever-like conformational changes, shifting through three symmetric elevations spanning ~13 Å (Fig. 5B). A CpaG/CpaH rotary mechanism (Fig. 7B) is best supported if the stoichiometry is one or two CpaG/CpaH heterodimers per CpaF hexamer. As CpaF cycles, the CpaG/CpaH heterodimer sequentially binds to the CpaF subunit pairing for whose surface conformation it has highest affinity. Due to the C2 symmetry, there are two high affinity binding sites per hexamer enabling up to two CpaG/CpaH heterodimers to bind concurrently. This model also predicts that the previously bound site becomes low affinity due to rotation-induced conformational changes. In support of this release mechanism, when the CpaF Central Loop is in the low position it will be blocked from binding CpaG or CpaH due to steric hindrance from the Loop 8 motif above (Fig. 5B). This rotary model differs from that previously suggested for PilB²⁷ as CpaG/CpaH are not predicted to be positioned in the CpaF central cavity and to be mechanically rotated like a drive shaft. A CpaG/CpaH gyratory mechanism (Fig. 7B) is best supported if the stoichiometry is three CpaG/CpaH heterodimers per CpaF hexamer (Supplementary Fig. 9B) as has recently been shown for PulF/GspF, which models as a C3 symmetric trimer like CpaG/CpaH⁴⁸. In this model, CpaF conformational changes are transmitted directly to the three bound CpaG/CpaH heterodimers which do not rotate relative to CpaF. Instead, the rotation of the CpaF C2 symmetry axis induces a complex gyratory effect in the membrane-spanning domains of the CpaG/CpaH heterodimers, which either move independently of each other or as a concerted unit. Ultimately, these conformational changes in CpaG and CpaH are somehow coupled to adding or removing pilins from an assembling or retracting pilus.

**Fig. 7: Models for CpaF engagement and energisation of CpaG and CpaH.**

In conclusion, the comparison of CpaF_AMPPNP with CpaF_ADP suggests a nucleotide switch model for achieving CpaF bidirectional motion where ATP binding in either the open or open’ state stimulates clockwise or counter-clockwise motion, respectively, upon subsequent hydrolysis. However, other mechanisms for how one motor may drive bidirectional rotary motions are feasible with the bidirectional flagellum motor representing one such paradigm. Here, the flagellum is driven by unidirectional rotary motors that deliver both clockwise and counter-clockwise torque simply by engaging different sides of the C-ring rotor in a cog-like mechanism controlled by phosphorylation⁴⁹. Similarly, in the TadPS it cannot be discounted that CpaF has unidirectional motion and other components such as CpaG and CpaH engage different aspects of CpaF structure so as to channel its motion in opposing directions that ultimately drive pilus extension and retraction. This work therefore provides a framework for further studies that characterise the direct interaction of CpaG and CpaH with CpaF and test the models for powering pilus cycling proposed herein.

Methods

Bacterial cultures

Escherichia coli strain MC1061 was used for cloning and CpaF expression. E. coli was grown in LB media with 25 µg/mL chloramphenicol at 37 °C for cloning and propagation of the shuttle vector pINIT. For cloning of the pBXC3H expression constructs and for expression precultures E. coli was cultivated in LB media with 100 μg/mL ampicillin at 37 °C. Caulobacter crescentus NA1000 was cultured in either PYE liquid (2 g/L bactopeptone, 1 g/L yeast extract, 0.3 g/L MgSO₄, and 0.0735 g/L CaCl₂) or on PYE 1.5% agar plates. Where appropriate, spectinomycin (25 μg/mL) and xylose (0.3%) were supplemented to growth media.

Construction of E. coli expression plasmids

All plasmids used or generated in this study are listed in Supplementary Table 3. cpaF (NCBI Gene ID: 7333491) was PCR amplified from C. crescentus NA1000 genomic DNA using primers CpaF_for and CpaF_rev (Supplementary Table 2). The purified PCR product was cloned into the shuttle vector pINIT_cat using FX cloning⁵⁰ resulting in pINIT_cpaF. After verifying the sequence, pINIT_cpaF was subcloned into pBXC3H fusing CpaF C-terminally to a 3 C protease cleavage site and a His₁₀-tag resulting in the plasmid pBXC3H_cpaF. CpaF mutants for in vitro purification were constructed in pBXC3H using FX and Gibson⁵¹ cloning with the primers listed in Supplementary Table 2.

Construction of C. crescentus knock-out and knock-in vectors

Primers NTToligo4020 and NTToligo4021 were used to amplify a 500 bp region upstream of cpaF including the first 180 nucleotides of cpaF, while primers NTToligo4022 and NTToligo4023 amplified a 500 bp region downstream of cpaF including the last 180 nucleotides of cpaF. Primers NTToligo2978 and NTToligo2979 were used to amplify a 500 bp region upstream of pilA including the first 9 nucleotides of pilA, while primers NTToligo2980 and NTToligo2981 amplified a 500 bp region downstream of pilA including the last 9 nucleotides of pilA. Primers NTToligo2978 and NTToligo4018 were used to amplify a 500 bp region upstream of pilA and the first half of pilA with the T36C mutation, while primers NTToligo4019 and NTToligo2981 amplified a 500 bp region downstream of pilA and the second half of pilA. All PCR products were cloned into suicide vector pNPTS138 using Gibson assembly, resulting in the knock-out plasmids pNPTS138::∆cpaF and pNPTS138::∆pilA, and the knock-in plasmid pNPTS138::pilA (T36C). All constructs were verified by Sanger sequencing.

Construction of C. crescentus complementation plasmids

Double-stranded DNA fragments containing cpaF wildtype and mutant genes with a 3’ FLAG tag were chemically synthesised as gBlocks (IDT). These gene fragments were cloned into vector pXYFPC-1⁵² after removing the yfp gene using Gibson assembly⁵¹. Briefly, 2.5 μL of each gBlock fragment and 2.5 μL of NdeI-NheI-cut pXYFPC-1 were added to 5 μL Gibson master mix (NEB) and incubated at 50 °C for 60 minutes. Five μL were used to transform chemically competent E. coli DH5α cells. The resulting pXYFPC-1::cpaF plasmids contained either the wildtype cpaF or mutant genes and were verified by Sanger sequencing.

Construction of C. crescentus strains

All strains generated in this study are listed in Supplementary Table 3. The suicide plasmids pNPTS138::∆cpaF, pNPTS138::∆pilA and pNPTS138::pilA (T36C) were electroporated into C. crescentus NA1000 cells. Double-crossover exconjugants were generated by sucrose counter-selection as described previously⁵³ resulting in the C. crescentus strains NA1000 ∆cpaF, NA1000 ∆pilA and NA1000 pilA (T36C). Markerless gene deletion mutants were identified by PCR. To construct ΔcpaF complementation strains, electrocompetent C. crescentus NA1000 pilA (T36C) ∆cpaF cells were transformed individually with complementation plasmids pXYFPC-1::cpaF containing either the wildtype cpaF or mutant genes to allow for a single integration at the xylX locus. The correct integration was verified by PCR.

PhiCbK phage-sensitivity assays

C. crescentus strains were inoculated into 10 mL of PYE and cultured for 20 hours at 30˚C with orbital shaking at 250 rpm. Cultures were diluted and grown to an OD₆₀₀ = 0.2–0.3. Subsequently, strains were adjusted to an OD₆₀₀ = 0.2 and 100 μL were mixed with 10 mL of soft PYE 0.5% agar and overlaid onto a solid PYE 1.5% agar base. Five μL of ten-fold serial dilutions of phiCbK phage stock were spotted in triplicate onto each plate. Finally, plates were incubated at 30˚C overnight and imaged the next day. Plaque formation indicated sensitivity to phiCbK phage.

Pili labelling and fluorescence microscopy

C. crescentus strains were inoculated into 10 mL PYE and cultured for 20 h at 30˚C with orbital shaking at 250 rpm. Cultures were then diluted and grown to an OD₆₀₀ = 0.2–0.3. Alexa Fluor 488 C5 Maleimide (Invitrogen) was added to 100 μL of C. crescentus culture (final concentration: 25 μg/mL) and incubated at room temperature for 10 minutes. Samples were centrifuged at 2500 × g for one minute and the pellet was carefully resuspended in 100 μL PBS. Cultures were centrifuged again, and the pellet was resuspended in a final volume of 50 μL PBS. Cells were immobilised on 1 % PYE agarose pads and imaged using a Zeiss Axio Observer Z.1 inverted epifluorescence microscope equipped with a sCMOS camera (Hamamatsu Orca FLASH 4) and Zeiss Colibri 7 LED light source. Images were captured with a Zeiss Plan Apochromat 100x/NA 1.4 Ph3 objective lens and a GFP filter set (excitation: 488 nm, emission: 500–549 nm) using an exposure time of 250 ms. Images were acquired using Zen Blue 2.5 (Zeiss) software and subsequently analysed using ImageJ 1.54⁵⁴. The MicrobeJ plug-in for Fiji⁵⁵ was used to detect C. crescentus cells and to measure the background-corrected mean fluorescence intensity within individual cells. The maximum fluorescence intensity for the non-fluorescent wildtype C. crescentus established an autofluorescence baseline; any value above this was categorised as true fluorescence. Each cell was also manually inspected to determine presence or absence of pili and used to calculate the proportion of the cell population that possessed extended pili.

Western blots

C. crescentus strains were inoculated into 10 mL of PYE and cultured for 20 hours at 30˚C with orbital shaking at 250 rpm. Cultures were diluted and grown to OD₆₀₀ = 0.2–0.3 and 5 mL were centrifuged at 1450 × g for 10 minutes. Cell pellets were resuspended in 300 μL lysis buffer (20 mM HEPES pH 7.9, 50 mM KCl, 10% glycerol, protease inhibitor tablet) and sonicated on ice. The cell lysate was centrifuged at 17,000 × g for 15 minutes at 4˚C. Samples of the supernatant were boiled for five minutes in SDS-PAGE loading dye (containing β-mercaptoethanol) before separation by SDS-PAGE together with a coloured protein broad range standard (NEB). Proteins were transferred onto a PVDF membrane using a Trans-Blot Transfer System (BioRad) and the membrane was blocked in TBS, 0.1% Tween-20, 5% milk powder for 1 h. The membrane was then incubated with a 1:5000 dilution of a monoclonal α-FLAG HRP-conjugated antibody (Sigma, A8592) for 1 h. Finally, the membrane was washed five times in TBS, 0.1% Tween-20 for one minute each and then briefly incubated with SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Scientific™). Blots were visualised using an Amersham Imager 600 (GE Healthcare).

Expression and purification of C. crescentus wildtype CpaF

CpaF was purified from E. coli MC1061. Glycerol stocks of E. coli MC1061 harbouring the plasmid pBXC3H_cpaF were used to inoculate a LB broth preculture grown at 37˚C overnight. The preculture was diluted 1:40 (v/v) into 2xYT and grown at 37˚C for two hours and subsequently at 25˚C for one hour to an OD₆₀₀ of ~0.8–1.0 before induction of protein expression with 0.01% L-arabinose. The next day cells were harvested at 9000 × g for 20 minutes at 4˚C. Pellets were resuspended in cold lysis buffer containing 20 mM Tris/HCl pH 8, 200 mM NaCl, 3 mM MgSO₄, DNase I (Sigma) and one tablet of complete EDTA-free protease inhibitor cocktail (Roche). The cells were disrupted by three passages through an Avestin EmulsiFlex-C5 homogeniser at 20–25 kpsi keeping the samples on ice. Unbroken cells were removed by centrifugation at 27,000 × g for 30 minutes at 4 °C. The supernatant was loaded on Ni²⁺-NTA (Qiagen) columns. The column was washed with 200 mM imidazole pH 7.5, 200 mM NaCl, 10% glycerol and 1 mM PMSF. CpaF was eluted with 500 mM imidazole pH 7.5, 200 mM NaCl, 10% glycerol and 1 mM PMSF. To remove the C-terminal His₁₀-tag the elution fraction was incubated with 3 C protease overnight at 4˚C. Finally, CpaF was separated from the cleaved His₁₀-tag and the 3 C protease via size exclusion chromatography using a HiPrep 26/60 Sephacryl S-300 HR column equilibrated in 20 mM HEPES pH7.4, 150 mM NaCl, 3 mM MgSO₄.

Negative stain microscopy of CpaF mutants

CpaF mutants were purified like wildtype except a Superose 6 10/300 GL column was used for the final gel filtration. For negative stain EM, 3.5 μL of protein sample was applied to glow discharged carbon coated 300 mesh copper grids (Agar scientific). Grids were washed with three drops of water and then stained with 2 % uranyl acetate. Images were acquired on a FEI Tecnai TEM 120 keV microscope equipped with a FEI 2 K Eagle camera.

Cryo-EM grid preparation and CpaF data collection

5 mg/mL CpaF was used for vitrification with 8 mM CHAPSO added to all samples to mitigate particle preferential orientation. For nucleotide associated samples, 4 mM ADP or AMPPNP (Sigma) was also added with a total of 5 mM MgSO₄ and a 45 minute incubation at room temperature. Samples were kept on ice until freezing in liquid ethane using glow-discharged 300 mesh holey carbon copper grids (Quantifoil R2/2, Electron Microscopy Sciences). The CpaF-APO dataset was acquired on a 300 kV Titan Krios electron microscope (LonCEM, The Francis Crick Institute, London, UK) equipped with a Gatan K3 detector operated in super-resolution mode with a pixel size of 0.425 Å. 21,498 movies were collected using a total dose of 40 e^-/Å² fractionated over 44 frames with 2.3 second exposure time and defocus range of −0.5 to −2.5 µm. The CpaF-ADP dataset was collected on the 300 kV Titan Krios IV microscope at eBIC (Diamond Light Source, Didcot, UK) equipped with a Gatan K3 detector operating in super-resolution mode with a pixel size of 0.4145 Å. 24,809 movies were collected using a total dose of 40 e^-/Å² fractionated over 40 frames with 1.8 second exposure time and defocus range of −0.5 to −2.1 µm. The CpaF-AMPPNP dataset collection used the 300 kV Titan Krios at LonCEM. The dataset was acquired in super-resolution mode with a pixel size of 0.55 Å. 5108 movies were collected using a total dose of 40 e^-/Å² fractionated over 40 frames with 4.39 second exposure time and defocus range of −0.5 to −2.8 µm.

CpaF cryo-EM data processing

The APO dataset was processed using cryoSPARC v4.0.1⁵⁶ and later using Relion 3.1⁵⁷. Micrographs were aligned using patch motion correction and Fourier cropped to 0.85 Å/pixel. CTF patch estimation was performed, and 19,493 of the best micrographs were selected for further processing. Template picker yielded 2,920,875 particles which was reduced to 471,974 particles after five rounds of 2D classification. The stack was transferred to Relion 3.1 for further 2D classification resulting in 440,053 particles. An initial reference was generated in Relion 3.1 using a stochastic descent gradient algorithm. 3D classification with C1 symmetry was undertaken to determine symmetry. C2 symmetry was subsequently applied to all processing steps. After several rounds of 3D classification, a final stack comprising 51,697 particles was used for iterative rounds of 3D autorefinement incorporating two rounds of CTF refinement and Bayesian polishing. The final resolution of 4.0 Å was based on gold standard Fourier shell correlations (FSC) = 0.143 criterion. For the ADP dataset, MotionCor2 was used to align movies and Fourier crop to 0.829 Å/pixel. CTFFIND-4.2 was used for CTF estimation. Using Relion 3.1, 23,774 micrographs were selected for particle picking resulting in 4,630,409 particles. After four rounds of 2D classification, 560,951 particles were used for iterative rounds of 3D classification. Using a final stack of 114,641 particles, a similar refinement strategy as for CpaF-APO was followed yielding a map at 3.1 Å resolution overall using gold standard FSC = 0.143 criterion. The CpaF-AMPPNP dataset was processed equivalently to CpaF-ADP. MotionCor2 was used to align movies and Fourier crop to 1.1 Å/pixel. CTFFIND-4.2 was used for CTF estimation. Using Relion 3.1, 5,013 micrographs were selected for particle picking resulting in 419,228 particles. After six rounds of 2D classification, 204,843 particles were used for iterative rounds of 3D classification. Using a final stack of 73,200 particles, a similar refinement strategy as for CpaF-APO was followed yielding a map at 3.8 Å resolution overall using gold standard FSC = 0.143 criterion. Map local resolution was estimated using ResMap⁵⁸. CpaF-APO, CpaF-AMPPNP and CpaF-ADP maps were sharpened using Phenix 1.2⁵⁹.

Model building and refinement

CpaF models were built in Coot⁶⁰ for each asymmetric unit from amino acids 150–502. No supporting map was observed for residues 1–79, which were predicted to be disordered. Map relating to the NTD1 (residues 80–149) was poorly resolved due to flexibility in Hinge 1. Particle subtraction masking out a single protomer of the AMPPNP map followed by focussed refinement improved the map quality in the NTD1 so that a three-helix motif could be readily traced with side chain detail limited to bulky residues. An NTD1 model was generated by Alphafold³⁴, a rigid body fitted and modelled using ISOLDE⁶¹. This NTD1 model was then used for rigid body fitting in all other NTD1 map regions and joined with the NTD2/CTD chains. Models for the complete asymmetric unit were real space refined using Phenix 1.2⁵⁹ but with NTD1s defined as rigid bodies. Finally, a C2 operator was used in Phenix⁵⁹ to reconstitute the CpaF hexamer.

Alphafold-Multimer modelling

CpaF, CpaG and CpaH in 6:3:3 stoichiometry were modelled with Alphafold v2.3.2 installed locally using default ‘multimer’ parameters⁴¹. Due to processing limitations, one model was generated with an overall per-residue local distance difference test (pLDDT) = 74.9. The pLDDT scores for individual subunits were retrieved from the B-factor column of the pdb file (Supplementary Fig. 9C). Individual pLDDT scores for CpaF, CpaG and CpaH were 90.0, 80.1 and 82.0, respectively.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The cryo-EM maps generated in this study have been deposited in the Electron Microscopy Data Bank under accession codes EMD-19275 (CpaF-AMPPNP), EMD-19209 (CpaF-ADP) and EMD-19279 (CpaF-APO). The atomic coordinates have been deposited in the Protein Data Bank (PDB) under accession codes 8RKD (CpaF_AMPPNP), 8RJF (CpaF_ADP) and 8RKL (CpaF_APO). Source Data are provided as a Source Data file. Source data are provided in this paper.

References

Kachlany, S. C. et al. Nonspecific adherence by Actinobacillus actinomycetemcomitans requires genes widespread in Bacteria and Archaea. J. Bacteriol. 182, 6169–6176 (2000).
Article CAS PubMed PubMed Central Google Scholar
Duong-Nu, T. M. et al. A stealth adhesion factor contributes to vibrio vulnificus pathogenicity: Flp pili play roles in host invasion, survival in the blood stream and resistance to complement activation. PLoS Pathog. 15, e1007767 (2019).
Bernard, C. S., Bordi, C., Termine, E., Filloux, A. & de Bentzmann, S. Organization and PprB-dependent control of the pseudomonas aeruginosa tad locus, involved in Flp pilus biology. J. Bacteriol. 191, 1961–1973 (2009).
Article CAS PubMed PubMed Central Google Scholar
Tomich, M., Planet, P. J. & Figurski, D. H. The tad locus: postcards from the widespread colonization island. Nat. Rev. Microbiol 5, 363–375 (2007).
Article CAS PubMed Google Scholar
Sangermani, M., Hug, I., Sauter, N., Pfohl, T. & Jenal, U. Tad pili play a dynamic role in caulobacter crescentus surface colonization. mBio 10, e01237 (2019).
Ellison, C. K. et al. Obstruction of pilus retraction stimulates bacterial surface sensing. Science 358, 535–538 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Pu, M. & Rowe-Magnus, D. A. A Tad pilus promotes the establishment and resistance of Vibrio vulnificus biofilms to mechanical clearance. NPJ Biofilms Microbiomes 4, 10 (2018).
Angelov, A. et al. Novel Flp pilus biogenesis-dependent natural transformation. Front Microbiol. 6, 84 (2015).
Article PubMed PubMed Central Google Scholar
Thiery, S., Turowski, P., Berleman, J. E. & Kaimer, C. The predatory soil bacterium Myxococcus xanthus combines a Tad- and an atypical type 3-like protein secretion system to kill bacterial cells. Cell Rep. 40, 111340 (2022).
Article CAS PubMed Google Scholar
Seef, S. et al. A Tad-like apparatus is required for contact-dependent prey killing in predatory social bacteria. Elife 10, e72409 (2021).
Article CAS PubMed PubMed Central Google Scholar
Denise, R., Abby, S. S. & Rocha, E. P. C. Diversification of the type IV filament superfamily into machines for adhesion, protein secretion, DNA uptake, and motility. PLoS Biol. 17, e3000390 (2019).
Article PubMed PubMed Central Google Scholar
Denise, R., Abby, S. S. & Rocha, E. P. C. The evolution of protein secretion systems by co-option and tinkering of cellular machineries. Trends Microbiol 28, 372–386 (2020).
Article CAS PubMed Google Scholar
Albers, S.-V. & Jarrell, K. F. The Archaellum: an update on the unique archaeal motility structure. Trends Microbiol 26, 351–362 (2018).
Article CAS PubMed Google Scholar
Henche, A. L. et al. Structure and function of the adhesive type IV pilus of Sulfolobus acidocaldarius. Environ. Microbiol 14, 3188–3202 (2012).
Article CAS PubMed PubMed Central Google Scholar
Nair, D. B., Uchida, K., Aizawa, S. I. & Jarrell, K. F. Genetic analysis of a type IV pili-like locus in the archaeon Methanococcus maripaludis. Arch. Microbiol 196, 179–191 (2014).
Article CAS PubMed Google Scholar
Beeby, M. Evolution of a family of molecular rube goldberg contraptions. PLoS Biol. 17, e3000405 (2019).
Article CAS PubMed PubMed Central Google Scholar
Craig, L., Forest, K. T. & Maier, B. Type IV pili: dynamics, biophysics and functional consequences. Nat. Rev. Microbiol 17, 429–440 (2019).
Article CAS PubMed Google Scholar
Naskar, S., Hohl, M., Tassinari, M. & Low, H. H. The structure and mechanism of the bacterial type II secretion system. Mol. Microbiol 115, 412–424 (2021).
Article CAS PubMed Google Scholar
Clock, S. A., Planet, P. J., Perez, B. A. & Figurski, D. H. Outer membrane components of the tad (Tight Adherence) secreton of Aggregatibacter actinomycetemcomitans. J. Bacteriol. 190, 980–990 (2008).
Article CAS PubMed Google Scholar
Tassinari, M., Rudzite, M., Filloux, A. & Low, H. H. Assembly mechanism of a Tad secretion system secretin-pilotin complex. Nat. Commun. 14, 5643 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Sonani, R. R. et al. Tad and toxin-coregulated pilus structures reveal unexpected diversity in bacterial type IV pili. Proc. Natl Acad. Sci. 120, e2316668120 (2023).
Article CAS PubMed PubMed Central Google Scholar
Inoue, T. et al. Molecular characterization of low‐molecular‐weight component protein, flp, in Actinobacillus actinomycetemcomitans Fimbriae. Microbiol Immunol. 42, 253–258 (1998).
Article CAS PubMed Google Scholar
Bhattacharjee, M. K., Kachlany, S. C., Fine, D. H. & Figurski, D. H. Nonspecific adherence and fibril biogenesis by Actinobacillus actinomycetemcomitans: TadA protein is an ATPase. J. Bacteriol. 183, 5927–5936 (2001).
Article CAS PubMed PubMed Central Google Scholar
Peabody, C. R. et al. Type II protein secretion and its relationship to bacterial type IV pili and archaeal flagella. Microbiol. (N. Y) 149, 3051–3072 (2003).
CAS Google Scholar
Xu, Q. et al. Structure of the pilus assembly protein TadZ from Eubacterium rectale: implications for polar localization. Mol. Microbiol 83, 712–727 (2012).
Article CAS PubMed PubMed Central Google Scholar
Ellison, C. K. et al. A bifunctional ATPase drives tad pilus extension and retraction. Sci. Adv. 5, eaay2591 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
McCallum, M., Tammam, S., Khan, A., Burrows, L. L. & Howell, P. L. The molecular mechanism of the type IVa pilus motors. Nat. Commun. 8, 15091 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Mancl, J. M., Black, W. P., Robinson, H., Yang, Z. & Schubot, F. D. Crystal structure of a type IV pilus assembly ATPase: insights into the molecular mechanism of PilB from Thermus thermophilus. Structure 24, 1886–1897 (2016).
Article CAS PubMed Google Scholar
Pelicic, V. Mechanism of assembly of type 4 filaments: everything you always wanted to know (but were afraid to ask). Microbiol. (N. Y) 169, 001311 (2023).
CAS Google Scholar
Solanki, V., Kapoor, S. & Thakur, K. G. Structural insights into the mechanism of Type IVa pilus extension and retraction ATPase motors. FEBS J. 285, 3402–3421 (2018).
Article CAS PubMed Google Scholar
Misic, A. M., Satyshur, K. A. & Forest, K. T. P. aeruginosa PilT Structures with and without Nucleotide Reveal a Dynamic Type IV Pilus Retraction Motor. J. Mol. Biol. 400, 1011–1021 (2010).
Article CAS PubMed PubMed Central Google Scholar
Satyshur, K. A. et al. Crystal structures of the pilus retraction motor PilT suggest large domain movements and subunit cooperation drive motility. Structure 15, 363–376 (2007).
Article CAS PubMed PubMed Central Google Scholar
McCallum, M. et al. Multiple conformations facilitate PilT function in the type IV pilus. Nat. Commun. 10, 5198 (2019).
Article ADS PubMed PubMed Central Google Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Story, R. M., Weber, I. T. & Steitz, T. A. The structure of the E. coli recA protein monomer and polymer. Nature 355, 318–325 (1992).
Article ADS CAS PubMed Google Scholar
Reindl, S. et al. Insights into FlaI functions in archaeal motor assembly and motility from structures, conformations, and genetics. Mol. Cell 49, 1069–1082 (2013).
Article CAS PubMed PubMed Central Google Scholar
Yamagata, A. & Tainer, J. A. Hexameric structures of the archaeal secretion ATPase GspE and implications for a universal secretion mechanism. EMBO J. 26, 878–890 (2007).
Article CAS PubMed PubMed Central Google Scholar
Lacabanne, D. et al. ATP analogues for structural investigations: case studies of a DnaB helicase and an ABC transporter. Molecules 25, 5268 (2020).
Article CAS PubMed PubMed Central Google Scholar
Bendis, I. & Shapiro, L. Properties of Caulobacter Ribonucleic Acid Bacteriophage φCb5. J. Virol. 6, 847–854 (1970).
Article CAS PubMed PubMed Central Google Scholar
de Bentzmann, S., Aurouze, M., Ball, G. & Filloux, A. FppA, a Novel Pseudomonas aeruginosa prepilin peptidase involved in assembly of type IVb Pili. J. Bacteriol. 188, 4851–4860 (2006).
Article PubMed PubMed Central Google Scholar
Evans, R. et al. Protein complex prediction with alphafold-multimer. Preprint at bioRxiv https://doi.org/10.1101/2021.10.04.463034 (2022).
Collins, R. et al. Structural cycle of the thermus thermophilus PilF ATPase: the powering of type IVa pilus assembly. Sci. Rep. 8, 14022 (2018).
Article ADS PubMed PubMed Central Google Scholar
Chiang, P., Habash, M. & Burrows, L. L. Disparate subcellular localization patterns of Pseudomonas aeruginosa Type IV Pilus ATPases involved in twitching motility. J. Bacteriol. 187, 829–839 (2005).
Article CAS PubMed PubMed Central Google Scholar
Takhar, H. K., Kemp, K., Kim, M., Howell, P. L. & Burrows, L. L. The platform protein is essential for type IV pilus biogenesis. J. Biol. Chem. 288, 9721–9728 (2013).
Article CAS PubMed PubMed Central Google Scholar
Py, B., Loiseau, L. & Barras, F. An inner membrane platform in the type II secretion machinery of gram-negative bacteria. EMBO Rep. 2, 244–248 (2001).
Article CAS PubMed PubMed Central Google Scholar
Arts, J. et al. Interaction domains in the Pseudomonas aeruginosa type II secretory apparatus component XcpS (GspF). Microbiol. (N. Y) 153, 1582–1592 (2007).
CAS Google Scholar
Chang, Y.-W. et al. Architecture of the type IVa pilus machine. Science (1979) 351, aad2001 (2016).
Google Scholar
Guilvout, I. et al. Membrane platform protein PulF of the Klebsiella type II secretion system forms a trimeric ion channel essential for endopilus assembly and protein secretion. mBio 15, e0142323 (2024).
Article PubMed Google Scholar
Johnson, S. et al. Structural basis of directional switching by the bacterial flagellum. Nat Microbiol. 9, 1282–1292 (2024).
Geertsma, E. R. & Dutzler, R. A versatile and efficient high-throughput cloning tool for structural biology. Biochemistry 50, 3272–3278 (2011).
Article CAS PubMed Google Scholar
Gibson, D. G. et al. Enzymatic assembly of DNA molecules up to several hundred kilobases. Nat. Methods 6, 343–345 (2009).
Article CAS PubMed Google Scholar
Thanbichler, M., Iniesta, A. A. & Shapiro, L. A comprehensive set of plasmids for vanillate- and xylose-inducible gene expression in Caulobacter crescentus. Nucleic Acids Res 35, e137–e137 (2007).
Article PubMed PubMed Central Google Scholar
Skerker, J. M., Prasol, M. S., Perchuk, B. S., Biondi, E. G. & Laub, M. T. Two-component signal transduction pathways regulating growth and cell cycle progression in a bacterium: a system-level analysis. PLoS Biol. 3, e334 (2005).
Article PubMed PubMed Central Google Scholar
Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–682 (2012).
Article CAS PubMed Google Scholar
Ducret, A., Quardokus, E. M. & Brun, Y. V. MicrobeJ, a tool for high throughput bacterial cell detection and quantitative analysis. Nat. Microbiol 1, 16077 (2016).
Article CAS PubMed PubMed Central Google Scholar
Punjani, A., Rubinstein, J. L., Fleet, D. J. & Brubaker, M. A. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat. Methods 14, 290–296 (2017).
Article CAS PubMed Google Scholar
Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure determination in RELION-3. Elife 7, e42166 (2018).
Article PubMed Google Scholar
Kucukelbir, A., Sigworth, F. J. & Tagare, H. D. Quantifying the local resolution of cryo-EM density maps. Nat. Methods 11, 63–65 (2014).
Article CAS PubMed Google Scholar
Liebschner, D. et al. Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr D. Struct. Biol. 75, 861–877 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr D. Biol. Crystallogr 66, 486–501 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Croll, T. I. ISOLDE: a physically realistic environment for model building into low-resolution electron-density maps. Acta Crystallogr D. Struct. Biol. 74, 519–530 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Whitfield, G. B. & Brun, Y. V. The type IVc pilus: just a Tad different. Curr. Opin. Microbiol. 79, 102468 (2024).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

We thank Nora Cronin (LonCEM, The Francis Crick Institute) and Karen Davies (eBIC, Diamond Light Source) for cryo-EM data collection support and Paul Simpson for in-house EM support. We thank Ngat T Tran for some strain constructions for preliminary experiments. This work was funded by a Wellcome Trust Senior Research Fellowship (215553/Z/19/Z) to HL. This work was also supported by a Royal Commission for the Exhibition of 1851 Research Fellowship (E.J.B), a Lister Institute Prise Fellowship (T.B.K.L), the Wellcome Trust Investigator Grant 221776/Z/20/Z (T.B.K.L), and the BBSRC-funded Institute Strategic Programme Harnessing Biosynthesis for Sustainable Food and Health (BB/X01097X/1).

Author information

Authors and Affiliations

Department of Infectious Disease, Imperial College, London, UK
Michael Hohl, Max P. Manley & Harry H. Low
Department of Molecular Microbiology, John Innes Centre, Norwich, UK
Emma J. Banks & Tung B. K. Le

Authors

Michael Hohl
View author publications
Search author on:PubMed Google Scholar
Emma J. Banks
View author publications
Search author on:PubMed Google Scholar
Max P. Manley
View author publications
Search author on:PubMed Google Scholar
Tung B. K. Le
View author publications
Search author on:PubMed Google Scholar
Harry H. Low
View author publications
Search author on:PubMed Google Scholar

Contributions

All authors designed experiments. MH purified proteins and undertook negative stain and cryo-EM studies, including sample preparation, data collection, processing, and model building. HL contributed to data collection and processing. MM contributed to data processing. EJB undertook C. crescentus cell-based assays including strain generation, phage infection assays and fluorescent light microscopy. HL and TL supervised the work. HL wrote the paper with contributions from all authors.

Corresponding author

Correspondence to Harry H. Low.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Julien Bergeron and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Movie 1

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Hohl, M., Banks, E.J., Manley, M.P. et al. Bidirectional pilus processing in the Tad pilus system motor CpaF. Nat Commun 15, 6635 (2024). https://doi.org/10.1038/s41467-024-50280-6

Download citation

Received: 18 January 2024
Accepted: 05 July 2024
Published: 05 August 2024
DOI: https://doi.org/10.1038/s41467-024-50280-6

This article is cited by

The structure of the Tad pilus alignment complex reveals a periplasmic conduit for pilus extension
- Sasha L. Evans
- Iryna Peretiazhko
- Julien R. C. Bergeron
Nature Communications (2025)
Conformational changes in the motor ATPase CpaF facilitate a rotary mechanism of Tad pilus assembly
- Ian Y. Yen
- Gregory B. Whitfield
- P. Lynne Howell
Nature Communications (2025)