Motor-independent retraction of type IV pili is governed by an inherent property of the pilus filament

PilB Does Not Power PilTU-Independent Retraction.

As mentioned above, many T4aP and T4bP are able to retract in the absence of dedicated retraction ATPases, and the mechanism that mediates this PilTU-independent retraction remains unknown. T4cP also lack dedicated retraction motors; however, these pilus systems use a bifunctional motor to power both pilus extension and retraction (36). Therefore, we sought to examine whether PilB powers PilTU-independent retraction of V. cholerae competence pili. Because PilB is required for pilus extension (Fig. 1A), we cannot test this using a ΔpilB mutant. Instead, we sought to determine whether the ATPase activity of PilB was required for PilTU-independent retraction.

In order to test whether PilB powers both extension and retraction via its ATPase activity, we employed the same technique described by Ellison et al. (36) to demonstrate that the ATPase CpaF powers both extension and retraction in Caulobacter crescentus T4cP. For that study, CpaF mutants that exhibit reduced ATPase activity slowed down the speed of both extension and retraction equally, indicating that CpaF likely powers both activities (36). Similarly, if pilB mutants that slow ATPase activity reduce the speed of both extension and PilTU-independent retraction, that would suggest that PilB is also a bifunctional ATPase. Measuring extension and retraction speeds in a ΔpilTU strain, however, is exceedingly difficult, because dynamic pilus events are rare in this hyperpiliated background (20, 37). In order to increase the number of dynamic events, we generated a mutant strain that allowed us to tightly regulate the expression of PilB (ΔpilTU ΔpilB P_BAD-pilB) (37). By inducing PilB expression only during imaging, we greatly increase the number of pilus dynamic events observed (Movies S1–S3).

To test whether PilB is a bifunctional ATPase, we used a pilB mutant (pilB^M391C) that was previously shown to reduce ATPase activity approximately twofold (38) and to correspondingly reduce the competence pilus extension speed approximately twofold (36). Consistent with these prior studies, ΔpilTU ΔpilB P_BAD-pilB^M391C exhibits an average extension speed of ∼72 nm/second, which is ∼1.8-fold slower than the ∼128-nm/second extension speed of ΔpilTU ΔpilB P_BAD-pilB^WT (Fig. 1C) (36). However, retraction speeds in ΔpilTU ΔpilB P_BAD-pilB^M391C were unaltered compared to ΔpilTU ΔpilB P_BAD-pilB^WT with both strains exhibiting an average retraction speed of ∼6 nm/second (Fig. 1C), which suggests that PilB is not a bifunctional motor. Importantly, previous work from our laboratory demonstrated that the ATPases required for polymerization of the other two T4P encoded by V. cholerae (MshE and TcpT) do not contribute to PilTU-independent retraction (20). Together, these results indicate that the retraction observed in the absence of PilTU is likely motor independent. However, this does not reveal the mechanism underlying this motor-independent retraction.

A Forward Genetic Selection Reveals a Role for the Major Pilin, PilA, in Motor-Independent Retraction.

In order to identify the mechanism underlying motor-independent retraction, we performed a forward genetic screen. As discussed above, mutants that exhibit enhanced retraction should also exhibit enhanced levels of natural transformation (Fig. 1B) (20). Thus, we reasoned that a suppressor screen selecting for mutants with enhanced levels of natural transformation might uncover gain-of-function mutations to components involved in motor-independent retraction. To that end, we recursively transformed a population of ΔpilT ΔmutS cells to enrich for mutants with enhanced rates of natural transformation. Using a ΔpilT strain, which exhibits reduced retraction efficiency (Fig. 1B) (20), allowed us to select for mutants with enhanced motor-independent retraction. The deletion of mutS in this background increases the spontaneous mutation rate by disrupting the mismatch repair system (39). To maintain selection for each recursive round of natural transformation, different antibiotic resistance markers were swapped at a neutral locus. This selection was carried out in 11 biologically independent lineages. All lineages exhibited an enhanced transformation frequency following four rounds of selection. Single strains were isolated from each lineage, and targeted sequencing of select pilus genes (pilA, pilB, pilC, pilU, and pilM) was performed to identify potentially causative mutations. From the 11 populations, only four unique missense mutations were identified, all in the major pilin PilA (G34R, V74A, G78S, and G80S). When these pilA mutations were introduced into a clean ΔpilTU genetic background, they all increased transformation frequency ∼100-fold (Fig. 2A), confirming that they were sufficient to promote the enhanced rates of natural transformation observed in the original suppressor mutants.

Open in a separate window

Fig. 2.

A forward genetic screen reveals mutations in PilA that enhance PilTU-independent pilus retraction. (A) Natural transformation assays of the indicated strains. Reactions were incubated with 200 ng of tDNA. All strains, n = 3. (B) Representative images of surface piliation (Top) and aggregation phenotypes (Bottom) for the indicated strains. (Scale bar, 1 µm.) (C) Retraction frequency of the indicated strains measured by epifluorescence time-lapse microscopy of AF488-mal-labeled cells. Data are presented as the percentage of cells within the population that exhibited a retraction event within the 10 min time-lapse. Data are from three independent biological replicates; pilA^parent ΔpilTU, n = 1184; pilA^G34R ΔpilTU, n = 718; pilA^V74A ΔpilTU, n = 709; pilA^G78S ΔpilTU, n = 864; and pilA^G80S ΔpilTU, n = 674. (D) Retraction speeds of the indicated strains measured by epifluorescence time-lapse microscopy of AF488-mal-labeled cells. Data are from three independent biological replicates; pilA^parent ΔpilTU, n = 44; and pilA^G78S ΔpilTU, n = 30. (E) Structural model of PilA based on Phyre2 threading of the PilA sequence (NP_232053) onto PDB 1OQW (alignment coverage 80%, confidence 99.9%, and identity 38%). Residues mutated in our suppressor screen are shown in stick representation and colored magenta. The α1N (residues 1 through 27) and α1C (residues 28 through 53) regions of PilA are also annotated. (F) Predicted model of the V. cholerae competence pilus in which the terminal pilin is shown in teal and other pilins in the fiber are shown in gray. The Phyre2-predicted PilA structure was superimposed onto one chain of the P. aeruginosa PAK pilus (PDB 5VXY) and then the helical properties of the PAK pilus were used to generate a structural model for the PilA fiber. The PilA suppressor mutations are shown as magenta stick representations in the terminal pilin to highlight where they are within the fiber. All bar graphs are shown as the mean ± SD. All box plots represent the median and the upper and lower quartile, while the whiskers demarcate the range. Asterisk(s) directly above bars denote comparisons to the parent strain. Comparisons in A and C were made by one-way ANOVA with Tukey’s post test. The comparison in D was made by Student’s t test. ** = P < 0.01 and *** = P < 0.001.

We also assessed the effect of these pilA suppressor mutations on natural transformation in a background where pilTU was intact. This analysis showed that all mutants transformed at levels similar to the parent strain (SI Appendix, Fig. S1A), suggesting that these alterations to the major pilin did not impact pilus biogenesis or motor-dependent retraction. We directly imaged these pili using our pilus labeling approach. All pilA point mutants produced visible external filaments with and without pilTU, although three of the four strains with pilA^G34R, pilA^V74A, and pilA^G80S mutations produced shorter pili than the pilA^parent strain (Fig. 2B and SI Appendix, Fig. S1B). Even though the ΔpilTU pilA^G34R, ΔpilTU pilA^V74A, and ΔpilTU pilA^G80S strains produced shorter pili, all of these strains aggregate (Fig. 2B), suggesting that they have surface-exposed pili. Moreover, we further confirmed that the fluorescent foci/filaments observed in the ΔpilTU pilA^parent, ΔpilTU pilA^G34R, ΔpilTU pilA^V74A, ΔpilTU pilA^G78S, and ΔpilTU pilA^G80S strains (Fig. 2B) are external pilus filaments by introducing a pilQ deletion into each strain. PilQ is the outer-membrane secretin and is required for the assembly of surface-exposed pilus filaments. In the ΔpilQ background, all mutants lacked foci/visible surface structures (SI Appendix, Fig. S2B). This suggests that the foci and filaments observed when pilQ is intact (Fig. 2B and SI Appendix, Fig. S1B) represent bona fide surface-exposed pili. Additionally, in a ΔpilTU background we used bulky maleimide conjugates (biotin-maleimide + neutravidin) to label pili to sterically obstruct pilus retraction (5, 29). This treatment prevented natural transformation, which further indicates that all of these pilA suppressor mutants still produce external pilus fibers and require pilus retraction to facilitate DNA uptake (SI Appendix, Fig. S2). We also assessed whether these pilA point mutants altered PilA expression levels via Western blot analysis. Although PilA^G80S expression was slightly reduced, the other point mutants had no effect on PilA expression compared to PilA^parent in both the pilTU intact and ΔpilTU backgrounds (SI Appendix, Figs. S1D and S2C). These results indicate that the suppressor mutations in pilA do not induce motor-independent retraction by simply altering PilA expression. Together, these data suggest that the pilA suppressor mutations enhance motor-independent retraction without dramatically altering pilus biogenesis or function.

The increased transformation frequencies observed for the suppressor mutations in a ΔpilTU background could result from increased retraction events or from an increased speed of retraction. To distinguish between these possibilities, we directly observed pilus dynamic activity in strains with labeled pili using time-lapse epifluorescence microscopy. The retraction frequency could be assessed in all mutant backgrounds; however, the speed of pilus retraction could only be accurately measured in the pilA^parent and pilA^G78S strains due to the short length of pilus filaments in the pilA^G34R, pilA^V74A, and pilA^G80S strains. These data revealed that pilus retraction frequency increases ∼10-fold for all of the ΔpilTU pilA suppressor mutants and that ΔpilTU pilA^G78S pili retract with an average speed of ∼12 nm/second which is approximately two times faster than the ∼6 nm/second observed in the ΔpilTU pilA^parent strain (Fig. 2 C and D). We previously found that altering retraction speed even sevenfold has no discernable effect on transformation (20). Thus, the observed increase in the motor-independent retraction frequency likely explains the increase in transformation rates seen in these pilA suppressor mutants. The discrepancy between the observed 10-fold increase in retraction frequency and the 100-fold increase in transformation frequency may be attributed to the fact that many short, dynamic pilus events occur that are not resolved by our microscopy-based approach.

Interestingly, in a background where the pilTU genes are intact, pilA^G78S has no discernable effect on the speed of pilus retraction as both the pilA^parent and pilA^G78S strains retract with an average ∼220 nm/second (SI Appendix, Fig. S1C). Thus, our data indicate that the pilA suppressor mutations only affect pilus dynamics in the ΔpilTU background and did not affect pilus dynamics when pilTU were intact (Fig. 2 A and C and SI Appendix, Fig. S1 A and C). This suggests that these pilA mutations only affect motor-independent pilus retraction and that their effect on pilus dynamics can be overcome by retraction motors.

The Stability of the Pilus Filament Is a Critical Determinant for Motor-Independent Retraction.

In order to understand how these amino acid changes in PilA might increase the frequency of motor-independent retraction, we sought to understand how they might affect the overall structure of the major pilin and the pilus filament. To that end, we generated a model for PilA and for the competence pilus filament based on existing homologous structures. The predicted PilA structure was generated based on the Pseudomonas aeruginosa major pilin structure (Protein Data Bank [PDB] ID 1OQW) using Phyre2 (40, 41). The competence pilus model was generated by superimposing the PilA model onto one of the subunits in the P. aeruginosa pilus structure (PDB 5VXY) and then imposing helical symmetry to generate the multisubunit filament (42). Additionally, the N-terminal half of the extended α-helix, α1N, of PilA was replaced with that of the P. aeruginosa pilin to reproduce the partial melting that this α-helix undergoes when incorporated into the pilus (42, 43). The model of V. cholerae PilA shows the residue positions of the suppressor mutations: Gly34 is located in α1C, the C-terminal half of the primary N-terminal α-helix (α1), and the other positions Val74, Gly78, and Gly80 are located in the αβ-loop that connects α1 with the central β-sheet of the globular domain (Fig. 2E). The pilus filament model revealed that all four suppressor mutation positions that enhance motor-independent retraction (PilA^G34R, PilA^V74A, PilA^G78S, and PilA^G80S) lie at interfaces between pilin subunits (Fig. 2F). When pilins are assembled into a filament, a portion of the α1 of each pilin is melted within a conserved region between residues 14 through 22, presumably to allow α1N to pack within the filament (42, 43). All of the PilA suppressor mutations encode residues that potentially contact α1N of a neighboring pilin subunit including the melted region between residues 14 and 22. Thus, altering pilin contacts in this area might interfere with subunit packing and/or compromise the malleability of the melted region in neighboring subunits, which may destabilize the pilus filament. Additionally, we note that three of the four suppressor mutations in the pilA gene encode substitutions at glycines. Glycine lacks a side chain, which allows greater backbone flexibility and is thus known as a “helix-breaker.” While Val74 is replaced with a smaller alanine, Gly34 is replaced by a bulky positively charged arginine, and Gly78 and Gly80 are both replaced with polar serines. The increase in bulk, polarity, and/or the decrease in flexibility at these sites may disrupt both the backbone conformation at α1C and the αβ-loop as well as the close packing of the subunits, destabilizing the pilus filament.

How might alterations in pilin subunit interactions increase motor-independent retraction in these pilA suppressor mutants? It has previously been suggested that pili can spontaneously depolymerize in the absence of motors (25, 44–46); however, conclusive evidence for this model has been elusive. We hypothesize that a pilus with weaker pilin–pilin interactions would allow newly added subunits at the base of the pilus to diffuse more readily back into the membrane if pilus assembly is stalled, thereby facilitating motor-independent retraction through spontaneous depolymerization (refer to Discussion for a detailed description of this model). A consequence of weaker pilin–pilin interactions may be altered stability of the pilus filament. To test whether the suppressor mutants produce less-stable pili, we purified labeled pili and tested their thermal and chemical stability as well as their protease susceptibility, as previously described (46). Though we were unable to purify the very short PilA^G34R, PilA^V74A, and PilA^G80S filaments in sufficient quantities for these assays, we obtained testable amounts of PilA^parent and PilA^G78S pili from a ΔpilTU background. We found that the PilA^G78S filaments have similar chemical stability to the PilA^parent filaments (SI Appendix, Fig. S3) but are less thermally stable and more susceptible to protease degradation (Fig. 3A). This instability may facilitate spontaneous depolymerization, thereby increasing the frequency of motor-independent retraction.

Open in a separate window

Fig. 3.

The size of residues in α1C may influence motor-independent retraction by altering the stability of the pilus filament. (A) Stability assay comparing thermal stability and protease susceptibility of purified AF488-mal-labeled pilus filaments from ΔpilTU strains expressing the indicated pilA allele. A representative phase-contrast image is shown to highlight that pilus preps are cell free. Fluorescent images show the presence or absence of intact pili under the indicated conditions. (B) Representative images of surface piliation for the indicated ΔpilTU strains after treatment with or without proteinase K. (C) Natural transformation assays of the indicated ΔpilTU strains. Reactions were incubated with 200 ng of tDNA. For ease of mutant generation, the native pilA gene was knocked out, and the indicated pilA allele was expressed at an ectopic location (under the control of an IPTG-inducible P_tac promoter). The amino acid at position 34 is indicated in the bar for each strain. Data are shown as the mean ± SD, and comparisons were made by one-way ANOVA with Tukey’s post test. Asterisk(s) directly above bars denote comparisons to the parent strain. All strains, n = 3. LOD, limit of detection; NS, not significant; and *** = P < 0.001. (D) Representative images of surface piliation for the ΔpilTU strains used in C. All images in panels A, B, and D are representative of three independent biological replicates. (Scale bars, 1 µm.)

Although we could not purify pili from the other pilA point mutants, we sought to test whether the external filaments present on the surface of intact cells are indeed less stable. Since the purified PilA^G78S filaments are more susceptible to proteinase K than the PilA^parent pili (a phenotypic characteristic of less-stable pili), we hypothesized that the surface-exposed filaments present on all of the ΔpilTU pilA suppressor mutants are also less stable and, therefore, more susceptible to protease degradation compared to the ΔpilTU pilA^parent strain. Indeed, when cells with labeled pili were treated with proteinase K, all four pilA suppressor mutants exhibited a complete loss of surface piliation as observed by fluorescence microscopy, while piliation of the pilA^parent was unchanged by this treatment (Fig. 3B). These results are consistent with the protease susceptibility profiles of purified PilA^parent and PilA^G78S filaments and indicate that the pilA^G34R, pilA^V74A, pilA^G78S, and pilA^G80S mutations all reduce the stability of pilus filaments in a similar manner. This supports the hypothesis that reducing filament stability enhances motor-independent retraction. The observed reduction in pilus stability among all of the pilA suppressor mutants is also consistent with the hypothesis that these mutations alter pilin–pilin interactions within the filament.

Insertion of Bulky Residues in α1C of the V. cholerae Competence Pilus Enhances Motor-Independent Retraction.

Since these data suggest that altering pilin–pilin interactions reduces filament stability and enhances motor-independent pilus retraction, we sought to understand the molecular mechanism of filament destabilization. We focused on the PilA^G34R mutation, as it is located in a region of the major pilin, α1C, that is structurally conserved among all major pilins from the type IV filament (T4F) family (19) despite being variable in sequence (SI Appendix, Fig. S4A). The PilA^G34R mutation changes a small, nonpolar amino acid (glycine) to a larger, positively charged amino acid (arginine). To determine whether the size or charge of the arginine side chain is responsible for enhanced motor-independent retraction, we made a panel of mutants with amino acid substitutions at the Gly34 position in PilA. We expected that if the positive charge of the arginine was the critical feature, substitution with lysine but not other amino acids should enhance motor-independent retraction. However, if the size of the arginine is the important feature, substitution with other amino acids that are larger than glycine, like glutamine and phenylalanine, should phenocopy the pilA^G34R mutant (SI Appendix, Fig. S4B lists the volume of amino acid R-groups as a correlate of their size) (47). Transformation assays showed that ΔpilTU pilA^G34K, ΔpilTU pilA^G34Q, and ΔpilTU pilA^G34F all phenocopied ΔpilTU pilA^G34R, indicating that it is the size of the amino acid at position 34 and not its charge that is critical for motor-independent retraction (Fig. 3C). Importantly, these mutations did not alter PilA expression levels (SI Appendix, Fig. S3B). Since substitutions with glutamine (polar) and phenylalanine (nonpolar) both enhanced transformation similarly, this suggests that residue polarity at the Gly34 position in PilA is also not a critical feature for enhancing motor-independent retraction. Each of these mutants also displayed reduced pilus length compared to ΔpilTU pilA^parent, similar to the original ΔpilTU pilA^G34R mutant (Fig. 3D). Changing the Gly34 position to a negatively charged residue (pilA^G34E) abolished transformation (Fig. 3C), most likely due to disruption of pilus biogenesis, which is supported by a lack of observable pilus filaments in this background (Fig. 3D). Substituting glycine with another small amino acid, alanine (pilA^G34A), which is common in α-helices, only had a minor effect on transformation and pilus length (Fig. 3 C and D). Together, these data suggest that amino acids with an R-group volume larger than alanine (SI Appendix, Fig. S4B; henceforth referred to as “bulky” residues) at position 34 in PilA enhance the frequency of motor-independent retraction, which likely occurs through destabilization of the pilus filament.

Residue Bulkiness in α1C Influences Motor-Independent Retraction of the Acinetobacter baylyi Competence Pilus.

We next sought to determine whether the bulkiness at position 34 correlates with the ability of other T4P to retract spontaneously. First, we tested whether bulky residues in α1C promote motor-independent retraction in the naturally retraction motor–dependent competence T4aP of Acinetobacter baylyi. Like V. cholerae PilA, the A. baylyi major pilin ComP has a small amino acid, a serine, at position 34 (SI Appendix, Fig. S4 A and B). A. baylyi competence pili can be studied using the same cysteine knock-in (comP^T123C) and maleimide dye labeling approach described above (48), and we can functionally assess competence pilus dynamic activity through natural transformation assays. Unlike V. cholerae, A. baylyi ΔpilTU mutants do not exhibit any residual natural transformation (Fig. 4A) (20), which is consistent with an absence of motor-independent retraction for this T4aP. We hypothesized that introducing a bulky residue at position 34 of ComP would interfere with pilin–pilin interactions, destabilizing the pilus to promote motor-independent retraction. To that end, we substituted Ser34 of the A. baylyi major pilin ComP with arginine (comP^S34R). The comP^S34R strain produces pilus filaments, indicating that pilus biogenesis is not inhibited (Fig. 4B). Transformation of comP^S34R was ∼1 log lower than the parent when pilTU were intact, suggesting an impact of this major pilin mutation on motor-dependent dynamic activity (Fig. 4A). However, strikingly, transformation was readily detectable in a ΔpilTU comP^S34R mutant, ∼100-fold above the limit of detection in the assay, which is represented by the ΔpilTU comP^parent strain (Fig. 4A). Importantly, this effect was not dependent on the comP^T123C mutation required for pilus labeling (SI Appendix, Fig. S5). These data are consistent with the model in which a bulky residue in α1C enhances spontaneous disassembly during motor-independent retraction.

Open in a separate window

Fig. 4.

Residue bulkiness in α1C influences motor-independent retraction of the A. baylyi competence pilus. (A) Natural transformation assays of the indicated strains. Reactions were incubated with 50 ng of tDNA. All strains, n = 3. Data are shown as the mean ± SD. Asterisk(s) directly above bars denote comparisons to the parent strain. Comparisons were made by one-way ANOVA with Tukey’s post test. LOD, limit of detection; *** = P < 0.001. (B) Representative images of surface piliation for the indicated strains. (Scale bar, 1 µm.)

Reducing Residue Bulkiness in α1C Diminishes Motor-Independent Retraction of the V. cholerae Toxin-Coregulated Pilus.

Next, we tested whether reducing residue bulkiness in α1C diminishes the retraction of the naturally motor-independent T4b toxin-coregulated pilus (TCP) of V. cholerae. TCP has no known retraction motors yet is retractile, and retraction was previously proposed to occur via spontaneous depolymerization (25). TCP dynamic pilus activity can be functionally assessed using a transducing phage (CTXΦ-K_m) that confers kanamycin resistance (49–51). During transduction, CTXΦ-K_m binds to the minor pilin TcpB at the pilus tip (51). TCP retraction is thought to be required for CTXΦ infection, as is the case for many phages that use T4P as receptors (25, 36, 51–54). We formally tested whether CTXΦ-K_m transduction requires TCP retraction by using bulky conjugates (biotin-maleimide + neutravidin) to coat cysteine knock-in pili (tcpA^T22C), which sterically obstructs pilus retraction as previously described (5, 29). Indeed, sterically obstructing pilus retraction inhibits CTXΦ-K_m transduction (SI Appendix, Fig. S6A). Previous work showed that TCP retraction is triggered by incorporation of the minor pilin TcpB into the growing pilus, which presumably stalls pilus extension and initiates TCP retraction (25). All T4P require a glutamate at position 5 of the major pilin for pilus assembly to proceed. This glutamate forms a salt bridge with the N-terminal amine of a neighboring pilin that facilitates subunit docking and neutralizes these charges in the hydrophobic core of the pilus (43). The minor pilin TcpB also requires Glu5 to incorporate into the growing TCP to stall pilus assembly. Thus, a tcpB^E5V mutant cannot initiate pilus retraction (25) and was included as an additional control. As expected, we observed that the tcpB^E5V mutant exhibits dramatically reduced transduction relative to the parental strain (Fig. 5A), as has been observed previously (25, 55).

Open in a separate window

Fig. 5.

Residue bulkiness in α1C helix influences motor-independent retraction in the V. cholerae TCP. (A) Transduction assays of the indicated strains. All strains, n = 3. Data are shown as the mean ± SD. Asterisk(s) directly above bars denote comparisons to the parent strain. Comparisons were made by one-way ANOVA with Tukey’s post test. *** = P < 0.001. (B) Western blot of the indicated strains to detect the major pilin TcpA and RpoA (as a loading and cytoplasmic protein control) in both whole-cell lysates and sheared supernatant samples as indicated. Data are representative of two independent experiments. (C) Representative images of aggregation phenotypes and (D) surface piliation for the indicated strains. (Scale bars in D, 1 µm.) (E) Stability assay comparing chemical stability of purified AF488-mal-labeled pili from the indicated strains. A representative phase-contrast image is shown to highlight that pilus preps are cell free. Fluorescent images show the presence or absence of intact pili under the indicated conditions. All images are representative of three independent biological replicates. (Scale bar, 1 µm.)

To test whether residue bulkiness in the pilus core influences TCP retraction, we targeted residues in TcpA at Thr34 (corresponding to the position of the PilA^G34R suppressor) and at Gln38 (a large residue one helical turn away in α1C) (SI Appendix, Figs. S4 A and B and S7A). These bulky residues were changed to the smallest amino acid, glycine (tcpA^T34G and tcpA^Q38G). We hypothesized that if these bulky residues destabilize the pilus and facilitate retraction, substituting them with smaller residues would reduce TCP retraction frequency and, therefore, reduce phage transduction. Indeed, both the tcpA^T34G and tcpA^Q38G mutants showed a significant reduction in CTXΦ transduction similar to the tcpB^E5V strain (Fig. 5A). A trivial explanation for this phenotype, however, could be that these mutants prevented TCP biogenesis. To test this, we took three approaches to assess the presence of surface-exposed TCP. First, TCP were sheared from the bacterial surface. Western blot analysis for TcpA in sheared-surface fractions indicated that both tcpA^T34G and tcpA^Q38G produced similar amounts of surface-associated pili as the parent and tcpB^E5V mutant (Fig. 5B). Second, surface-associated TCP interact, resulting in aggregation of cells in liquid culture. Aggregation of both tcpA^T34G and tcpA^Q38G were indistinguishable from the parent (Fig. 5C). Third, TCP can be labeled via the same cysteine knock-in and maleimide dye approach described above (tcpA^T22C) (5, 35). We introduced these mutations into the tcpA^T22C background to directly observe surface piliation by epifluorescence microscopy. Importantly, both tcpA^T34G and tcpA^Q38G still exhibit reduced transduction in this background (SI Appendix, Fig. S6B). Consistent with our previous results (Fig. 5 B and C), tcpA^T34G and tcpA^Q38G produce long, external TCP filaments similar in appearance to the parent strain (Fig. 5D). These results indicate that the reduction in motor-independent retraction observed for the tcpA^T34G and tcpA^Q38G mutants is not due to reduced pilus biogenesis and are thus consistent with our hypothesis that smaller residues in α1C may allow for more stable subunit packing, which reduces the frequency of spontaneous retraction.

To test this hypothesis, we purified labeled TCP filaments from the parent, tcpA^T34G, and tcpA^Q38G strains and tested their thermal and chemical stability as well as protease susceptibility as described above. Pilus filaments from all three strains have similar thermal stability and protease susceptibility profiles (SI Appendix, Fig. S6C); however, TcpA^T34G filaments are more resistant to denaturation by 6 M urea compared to the parent and TcpA^Q38G filaments (Fig. 5E). Threonine is only marginally bulkier than glycine (SI Appendix, Fig. S4B), but the tcpA^T34G mutation still enhanced fiber stability and reduced pilus retraction events, which suggests that threonine may be large enough to reduce fiber stability. This is consistent with our hypothesis that small amino acids in α1C of the major pilin allow for stronger pilin–pilin interactions and increase the stability of the pilus, thereby decreasing its propensity to spontaneously retract. In addition to reducing residue bulkiness at these locations, the tcpA^T34G and tcpA^Q38G mutations also modify the polarity at this position (a change from polar to nonpolar amino acids). Thus, these amino acid changes may stabilize TCP by relieving steric stress or by removing the polar hydroxyl group from an otherwise nonpolar environment. Together, these data further support the idea that bulky residues in α1C oriented into the pilus core promote motor-independent retraction.

We thank Nicolas Biais and Stefan Kreida for helpful discussions, Kim D. Seed and Zoe Netter for protocols and strains for transduction assays, Courtney K. Ellison for providing protocols for pilus labeling and purification, and Triana N. Dalia for help with strain construction. This work was supported by Grant R35GM128674 from the NIH (to A.B.D.) and Grant RGPIN-2017-05757 from the Natural Sciences and Engineering Research Council (to L.C.).