Type IV Pilus Alignment Subcomplex Proteins PilN and PilO Form Homo- and Heterodimers in Vivo*

Pseudomonas aeruginosa is a leading cause of hospital-acquired infections and is resistant to many antibiotics. Type IV pili (T4P) are among the key virulence factors used by P. aeruginosa for host cell attachment, biofilm formation, and twitching motility, making this system a promising target for novel therapeutics. Point mutations in the conserved PilMNOP alignment subcomplex were previously shown to have distinct effects on assembly and disassembly of T4P, suggesting that it may function in a dynamic manner. We introduced mutations encoding Cys substitutions into pilN and/or pilO on the chromosome to maintain normal stoichiometry and expression levels and captured covalent PilNO heterodimers, as well as PilN and PilO homodimers, in vivo. Most covalent PilN or PilO homodimers had minimal functional impact in P. aeruginosa, suggesting that homodimers are a physiologically relevant state. However, certain covalent homo- or heterodimers eliminated twitching motility, suggesting that specific PilNO configurations are essential for T4P function. These data were verified using soluble N-terminal truncated fragments of PilN and PilO Cys mutants, which purified as a mixture of homo- and heterodimers at volumes consistent with a tetramer. Deletion of genes encoding alignment subcomplex components, PilM or PilP, but not other T4P components, including the motor ATPases PilB or PilT, blocked in vivo formation of disulfide-bonded PilNO heterodimers, suggesting that both PilM and PilP influence the heterodimer interface. Combined, our data suggest that T4P function depends on rearrangements at PilN and PilO interfaces.

Many bacteria, including Pseudomonas aeruginosa, use type IV pili (T4P) 3 for surface attachment/adhesion, biofilm formation, and twitching motility (1)(2)(3)(4)(5)(6). T4P are divided into T4aP and T4bP subgroups, based on differences in pilin size and organization of the assembly machinery (6,7). Here we focus on T4aP, henceforth referred to as T4P. T4P are long, thin, fibrous appendages, which extend from the bacteria, attach to a surface, and are retracted back into the cell, winching the cell forward. This flagellum-independent form of locomotion is called twitching motility. Four subcomplexes make up a T4P machine in P. aeruginosa as follows: the outer membrane (OM) secretin (PilQ) and its pilotin (PilF) (8 -12); the inner membrane (IM) platform protein (PilC) and cytoplasmic motor proteins (ATPases PilBTU) (13,14); the helical pilus fiber, composed mainly of the major pilin subunit (PilA) plus minor pilins (PilVWXE and FimU) and adhesin PilY1 (5,15,16); and the alignment subcomplex proteins (PilMNOP) that span from the cytoplasm to the OM secretin (17)(18)(19)(20). PilN and PilO are bitopic IM proteins with similar predicted secondary structures. These proteins are essential for T4P function and connect the cytoplasmic alignment subcomplex component, PilM, with the IM-associated lipoprotein PilP that interacts with the N0 domain of the secretin monomer PilQ (19,21). Proposed functions of the PilMNOP subcomplex include aligning the IM motor with the OM secretin, concentrating pilin subunits at the site of extension and retraction, and possibly transducing signals via conformational changes from the cytoplasmic ATPases to generate "open" or "closed" states of the gated secretin (19,20).
The T4P and type II secretion (T2S) systems have analogous components and similar architecture (22,23). Whereas the T4P system extends and retracts an elongated pilus, the T2S system specializes in secreting protein substrates from the periplasm into the environment, using a short piston-like "pseudopilus" (23)(24)(25). Although the systems have different outputs, their underlying functional mechanisms, which remain to be deciphered, are thought to be comparable. Orthologous components of the alignment subcomplexes in the two systems (PilMN/GspL, PilO/GspM, and PilP/GspC) have limited sequence identity, but atomic resolution structures revealed a high degree of structural similarity, and all three components form equimolar complexes with a ratio of 1:1:1 (18, 20, 26 -30). PilN-and PilO-like proteins in the T2S system (GspL and GspM, respectively) form both homo-and heterodimers, although it is not yet clear whether the T4P system is similar (26,27,(31)(32)(33)(34). Previous work in P. aeruginosa showed that PilP fails to interact with a PilO homodimer and requires both PilN and PilO for stability (17,20). Although PilO crystallized as a homodimer (18), complementation of a pilO mutant with a plasmid-encoded copy of the gene led to formation of predominantly PilO homodimers at the cost of PilN stability, and thus T4P function (17). In Thermus thermophilus, purified T4P

Selection of Residues for Cys Substitution-Our
PilO homodimer structure (Protein Data Bank code 2RJZ) (18) was used to identify candidate residues predicted to be within appropriate distances for disulfide bond formation (Fig. 1A, top). In the PilO model, two relevant interfaces were identified as follows: the core, comprising mainly contacts between the ␣4-helix and ␤4-strand; and the coiled coils, between the ␣1and ␣2-helices (18). In the PilO ⌬68 model, the coiled-coils are truncated and folded back, because they are not anchored in the IM. Because of the potentially artificial nature of that particular interaction interface, we focused on residues in the core region previously shown to be important for function (42). Residues in the ␣4-helix and ␤4-strand of the PilO homodimer model with an estimated distance between ␤-carbons of less than 8 Å were selected for mutagenesis (Fig. 1A, bottom) (37,43,44).
All attempts to purify soluble P. aeruginosa PilN for structural studies, or to solve an x-ray crystal structure of the PilNO heterodimer, have so far been unsuccessful. The structure of a periplasmic portion of PilN from T. thermophilus is available (29); however, due to the low primary sequence identity with PilN from P. aeruginosa, as well as a different arrangement of secondary structure elements (Thermus PilN has an extra ␣-helix inserted in the first ␣␤␤ motif), we were unable to generate a high confidence model using that structural template. P. aeruginosa PilN is predicted to have a structural organization similar to PilO, and thus we used the truncated P. aeruginosa PilO homodimer structure (18) to generate a high confidence Phyre 2 (45) model of a PilN ⌬57 monomer, and we used the PilO homodimer interface to model a PilNO heterodimer using the same core-core interface (Fig. 1B, top) (42). This model was used to design a series of single Cys substitutions in the ␣4-helix and ␤4-strand of either PilN or PilO (Fig. 1B, bottom). The PilNO model has been partially validated in previous work (42), but for this study, we included a negative control Cys pair (PilN R142C /PilO A158C ) at a C␤ distance of over 12 Å, considered too far for disulfide bond formation (37,44).
To avoid issues arising from non-stoichiometric expression, the mutations encoding the Cys substitutions of interest were introduced at the native pilMNOPQ locus of P. aeruginosa PAK by homologous recombination. Screening of the sequence-verified mutants for effects on expression and stability of each of the inner membrane alignment subcomplex proteins revealed no effects of the Cys substitutions on PilMNOP levels relative to wild type ( Fig. 2A).
Formation of Cys Cross-linked PilNO and PilOO Dimers in Vivo-Whole cell lysates of all mutants were analyzed in parallel in reducing and non-reducing conditions, followed by immunoblotting with anti-PilN or PilO antibodies. Of the PilNO Cys substitutions, only the PilN R142C /PilO D175C variant formed significant levels of cross-linked dimers in vivo (Fig. 2, B and C). Two bands appeared on the anti-PilN blot (Fig. 2B) 2C) at a size consistent with that of PilO homodimers (45.6 kDa). Thus, at least one of each set of selected residue pairs formed the predicted linkage (Fig. 2D). This is the first evidence in any T4P system for PilN homodimerization under physiological conditions.
To further investigate PilN homodimerization, we leveraged a previously described PilN ⌬44 /PilO ⌬51_His co-expression construct (20) that encodes truncated forms of both proteins. These proteins lack their short cytoplasmic N termini and transmembrane segments and encompass a larger portion of the periplasmic domains than the construct used to obtain the aforementioned PilO ⌬68 homodimer structure. A PilN Cys  Fig. 3, A and B). In this case, nearly all PilN monomers were involved in covalent interactions in non-reducing conditions (Fig. 3, A and B) when compared with PilO. This suggests that PilO homodimers might occupy multiple orientations, some disulfide-bonded and others not, implying in the latter case that the PilO D175C mutation was too far or incorrectly oriented for disulfide bond formation. These data confirm the observation that PilN and PilO form homodimers and heterodimers under physiological conditions in P. aeruginosa.
Because the untagged covalently linked PilN R142C homodimers purified with C-terminally His-tagged PilO, we realized that PilN homodimers must interact with PilO homodimers via an interface other than the one shown in Fig. 1B; the most likely candidate for a second interface was the coiled coils, below the core domains. Using size exclusion chromatography, we confirmed that PilN ⌬44 R142C /PilO ⌬51_His (as well as unmodified PilN ⌬44 /PilO ⌬51_His and other Cys mutants) eluted from an S75 size exclusion column at a volume similar to the conalbumin standard (ϳ75 kDa) (Fig. 3C), consistent with a PilNO tetramer (70.4 kDa). Thus, PilNO can simultaneously form homo-and heterodimers via two distinct interfaces.

FIGURE 1. Locations of residues chosen for Cys substitution mutagenesis in PilN and PilO.
A schematic diagram representation mapping the N and C termini, secondary structure elements, and structural domains of a PilO ⌬68 dimer (2RJZ (18)) (A) or a PilN ⌬57 (blue) and PilO ⌬68 (gray) heterodimer using a Phyre 2 model (45) (B) of PilN generated using the PilO ⌬68 homodimer as template. The core-core domains consist of PilN ␣-helices 3-4 and ␤-strands 1-4 and PilO ␣-helices 3-4 and ␤-strands 1-5. Inset shows zoomed view of interaction interface with residue substitutions labeled accordingly. Bottom, tables indicate the distance between the C␤ of the substituted Cys residues (in angstroms) and schematic depiction of residue locations.

Disulfide Bond Formation between PilN and/or PilO Can
Disrupt T4P Function-To assess the functional consequences arising from formation of covalently linked dimers in vivo, the ability of the Cys mutant strains to assemble surface-exposed T4P was assessed (Fig. 4A). The PilN R142C / PilO D175C and PilO A158C/I178C mutants had very low levels of surface pili, and piliation of the PilO Y154C/I178C mutant was comparable with negative controls. Both the PilO S166C/D175C and PilO V161C/D175C mutants had a moderate reduction in surface pili, possibly related to their shared D175C mutation.  These data suggest that pilus biogenesis requires specific PilN and PilO configurations that are perturbed by these mutations.
Because T4P function relies on the dynamic assembly and disassembly of pili, it was important to assess motility as well as piliation levels (Fig. 4B). Strains forming covalently crosslinked dimers (PilN R142C /PilO D175C , PilO A158C/I178C , and PilO Y154C/I178C ) were incapable of twitching. PilO S166C/D175C and PilO V161C/D175C strains had approximately half of wildtype motility, consistent with their decrease in piliation levels, although no in vivo cross-linked dimers were observed for these mutants. The remaining Cys mutants twitched similarly to the wild type.
Lack of Piliation Is Caused by Inefficient Assembly-Any pili assembled in strains lacking the retraction ATPase, PilT, become trapped on the cell surface (46). Loss of pilus retraction in alignment subcomplex mutants was shown previously (14) to allow for capture of a few pili on the surface of the cells. Because three of the double Cys mutants had reductions in the levels of surface pili (PilN R142C /PilO D175C , PilO A158C/I178C , and PilO Y154C/I178C ), we hypothesized that they would assemble more pili in a retraction-deficient background, similar to pilN or pilO mutants. We disrupted pilT in these backgrounds and tested the resulting levels of surface piliation. As predicted, all mutants assembled more pili when retraction was disabled (Fig.  4C). This result suggests that all mutants are capable of extending pili but that assembly in those backgrounds is inefficient, especially in the PilO Y154C/I178C mutant, resulting in net pilus retraction when PilT is active.
Specific Single Cys Mutations Affect Function of the T4P System-Because moderate reductions in twitching and piliation were observed in two PilO double Cys mutant strains that share a common substitution (PilO S166C/D175C and PilO V161C/D175C ), we next investigated whether individual Cys substitutions disrupted T4P production and/or function. The stability of alignment subcomplex components in strains with single Cys substitutions was tested using specific antibodies, and the levels of PilMNOP were found to be comparable with wild type (Fig. 5A). Each single Cys mutant produced pili at levels similar to wild type, with the exception of PilN R142C that had a modest reduction in piliation (Fig. 5B). Interestingly, neither PilN R142C nor PilO D175C could twitch (Fig. 5C), even though both produced pili; this phenotype is a hallmark of retraction defects. When the retraction ATPase PilT was deleted from PilN R142C and PilO D175C , their piliation levels were similar to the pilT control, showing they are capable of wild-type pilus assembly (Fig. 5D). Motility, and therefore pilus retraction, of all other single Cys mutants was unaffected.
Single Cys Substitutions Allow for in Vivo Homodimer Formation, but Only a Subset Has Functional Consequences-To better understand why PilN R142C and PilO D175C single mutants could extend pili but not twitch, we hypothesized that these substitutions might trap PilN or PilO in specific homodimer configurations, impairing function. Whole cell lysates of all single Cys mutants were analyzed in reducing and non-reducing conditions followed by immunoblotting with PilN and PilO antibodies. We were surprised to see that all single Cys mutants formed covalent homodimers in vivo to varying degrees, with PilN R142C and PilO Y154C forming the most, and fewer dimers captured for other Cys mutants such as PilO D175C and PilO V161C , where dimer formation is harder to detect (Fig. 5, E and F). These data suggest that with specific exceptions (PilN R142C and PilO D175C ), covalent in vivo homodimerization of PilN and PilO can occur without functional consequences.
Heterodimer Formation of PilN R142C and PilO D175C Is Abolished in a pilM Background-As proposed for the T2S system (34), partner switching, or reorientation of PilN and PilO interfaces, might explain how PilN and PilO can form both homoand heterodimers in vivo. We hypothesized that PilN might respond to conformational changes in its cytoplasmic partner PilM, repositioning it relative to PilO. To test whether formation of covalent dimers requires PilM, the pilM gene was deleted in each of the PilN R142C /PilO D175C , PilO A158C/I178C , and PilO Y154C/I178C strains. The absence of PilM was confirmed by Western blot (Fig. 6A), and whole cell lysates of Cys mutant strains lacking PilM were analyzed in reducing PilN ⌬44 and PilO ⌬51 co-eluted at a volume similar to conalbumin as indicated on graph with the red x The elution volumes (V e ) of the standards are divided by the void volume of the column (V 0 ) and plotted against the log of the standards' molecular weights to give the slope of the line that was used to solve for the molecular weight of the eluted PilN ⌬44 /PilO ⌬51_His complex (70.4 kDa). PilN residues indicated in blue text. and non-reducing conditions, followed by immunoblot analysis with PilN or PilO antibodies. Although both PilO A158C/I178C and PilO Y154C/I178C formed covalent homodimers in the absence of PilM, PilN R142C /PilO D175C no longer formed covalent heterodimers (Fig. 6, B and C).
Heterodimer Formation by PilN R142C and PilO D175C Is Also Abolished in a pilP Background-To determine whether the remaining member of the alignment subcomplex, PilP, had effects on PilNO dimer formation, we created a pilP strain expressing the Cys pairs of interest (PilN R142C /PilO D175C , PilO A158C/I178C , and PilO Y154C/I178C ), confirmed loss of PilP by Western blot (Fig. 6A), and analyzed whole cell lysates as described above. The PilN R142C /PilO D175C strain failed to form covalent PilNO heterodimers in the absence of PilP; however, PilN and PilO formed homodimers (Fig. 6, D and E), as in the pilM mutant. To investigate whether the absence of PilP alone affected the PilNO interface, or whether the change in dimerization state was due to the lack of PilNO tethering via PilP to the N0 domain of PilQ (19), we created a pilQ strain expressing the PilN R142C /PilO D175C Cys pair, confirmed loss of PilQ by Western blot (Fig. 6A), and tested lysates as described above. Absence of PilQ had no effect on homo-or heterodimer formation (Fig. 6, F and G), suggesting that PilP specifically contributes to the formation of an interface conducive to the formation of a covalent bond between PilN R142C /PilO D175C .
Other T4P Components Do Not Modulate the PilNO Interface-Recently, interactions between PilM and motor ATPases of the T4P system were identified for both P. aeruginosa and M. xanthus (40,41). Because PilM was important for formation of covalent PilN R142C /PilO D175C heterodimers, we hypothesized that conformational changes in PilM, driven by its interaction with other cytoplasmic components of the T4P motor while bound to the N terminus of PilN, might be propagated to and affect the orientation of periplasmic PilNO interfaces. Three AAA ϩ -ATPases modulate extension (PilB) and retraction (PilTU) of the pilus fiber (5,47,48). We created pilB, pilT, and pilU mutants containing the Cys pairs of interest (PilN R142C /PilO D175C , PilO A158C/I178C , and PilO Y154C/I178C ) and confirmed the absence of each ATPase by Western blotting (Fig. 7). Unexpectedly, all strains formed covalent homo-and heterodimers (Fig. 8, B, D, and E), suggesting that the ATPases do not influence the PilNO interface.
We further investigated whether other T4P components associated with PilM, PilN, or PilO, including PilA, the major component of the pilus fiber and PilC, the IM platform protein, influenced the dimerization state of PilN and PilO. Strains with disruptions in pilA or pilC, plus the above Cys substitutions (Fig. 7), formed covalent homo-and heterodimers similar to the wild type (Fig. 8, A and C). These data suggest that only interactions among members of the alignment subcomplex modulate the interface between PilN and PilO.

Discussion
The PilMNOP alignment subcomplex is highly conserved among T4aP-expressing bacteria (5). Previous studies demonstrated that PilN and PilO form stable heterodimers in a 1:1 stoichiometry (18); however, the precise interface(s) between PilN and PilO has not yet been elucidated. We used disulfidebonding analysis to study the interactions between these components and to assess the functional consequences of covalently cross-linking PilN and/or PilO. Our results revealed that PilN and PilO form both homo-and heterodimers in vivo under . Sheared surface preparations revealed that the PilN R142C /PilO D175C and PilO A158C/I178C mutants exhibited severely reduced levels of surface pili, although the PilO Y154C/I178C mutant exhibited no surface-exposed pili, and all were devoid of twitching motility similar to negative controls. Large disruptions in motility and piliation in the remaining PilNO Cys pairs were not observed. The ability of PilNO or PilOO Cys pairs to form Cys cross-links is indicated below (Y). C, select Cys mutants were introduced into a retraction-deficient background, which revealed that all three double Cys mutants assemble more pili without the retraction ATPase, PilT. PilT presence (ϩ) or absence (Ϫ) in each strain is indicated below. Images are representative of at least three independent experiments. physiological conditions and that conformational flexibility of these states is necessary for T4P function.
PilN and PilO May Form Functional Dimers of Dimers in Vivo-Studies in the T2S system showed that PilN-and PilOlike proteins (GspL and GspM, respectively) form homo-and heterodimers, possibly via a process called "partner switching," implying large rotations in the TMS and core interfaces (34). Here, we captured PilO homodimers, previously observed only upon overexpression in trans (17,18), when expressed at native levels and stoichiometry from the Pseudomonas chromosome, confirming that the PilO homodimer interface revealed by structural studies is physiological (Fig. 2C). Unexpectedly, while trying to understand whether PilNO heterodimers share the same interface as PilO homodimers, we identified multiple species on an anti-PilN Western blot (Fig. 2B), with masses corresponding to PilNO heterodimers and PilN homodimers. Although a PilN homodimer (bound to PilM) from the T4P system of T. thermophilus was observed in vitro (29), to our knowledge this is the first report of PilN and PilO homodimers in their native context.
When Cys-substituted soluble periplasmic fragments of PilN or PilO (PilN R142C and PilO D175C , Fig. 3, A and B) were expressed in E. coli, they formed covalent homodimers. Coexpression of PilO D175C with PilN R142C led to capture of three subpopulations: PilN R142C homodimers, PilO D175C homodimers, and PilN R142C /PilO D175C heterodimers (Fig. 3, A  and B), suggesting that there are (at least) two possible interfaces for each protein. Interestingly, a large amount of PilO monomers remained in the non-reducing lanes, indicating that a substantial proportion of PilO was either present as a monomer or participating in non-covalent dimers, possibly as a result of non-permissive positioning of a PilO interaction interface for disulfide bond formation.
P. aeruginosa PilN expressed in vitro without PilO is mostly insoluble (20), which has precluded studies of PilN alone. We hypothesized that formation of covalent PilN homodimers might stabilize it in the absence of PilO, and we attempted to purify PilN ⌬44_His R142C under non-reducing conditions, but the protein remained mostly insoluble. However, untagged PilN R142C homodimers co-purified with tagged PilO, and the molecular mass of the complex was ϳ75 kDa (Fig. 3C), likely corresponding to a 2:2 PilN ⌬44 /PilO ⌬51_His complex. This stoichiometry is consistent with our previous data, where PilNOP were shown by gel filtration to elute as either a trimer or a dimer of trimers (20). A dimeric complex of PilMNO (2:2:2) was reported for T. thermophilus (29), fitting with the idea that P. aeruginosa PilN and PilO form a dimer of dimers (17,19,20,29).
Only two of four predicted PilO Cys mutant pairs, and one of three PilNO pairs, formed disulfide bonds, suggesting that the FIGURE 5. PilN R142C and PilO D175C single Cys substitutions cause T4P dysfunction in wild-type and retraction-deficient backgrounds. Lysates for the wild type (wt), negative controls (Ϫ), and the Cys mutants were subject to SDS-PAGE and then blotted with specific antibodies to each alignment subcomplex component (␣-PilM, ␣-PilN, ␣-PilO, and ␣-PilP) to confirm stability of each protein in the Cys mutants (A). Each mutant along with wild type, non-piliated (AϪ), hyperpiliated (TϪ), and negative controls (N-and O-) were tested for sheared surface pili (B), and twitching motility (C). PilN R142C had less pili and motility than WT, whereas PilO D175C had similar level of pili but no motility. D, PilN R142C and PilO D175C were introduced into a retraction-deficient background, which revealed that both Cys mutants can assemble more pili without the retraction ATPase, PilT. PilT presence (ϩ) or absence (Ϫ) in each strain is indicated below. Whole cell lysates were then subjected to SDS-PAGE under non-reducing (nr) and reducing (r) conditions and then blotted with specific antibodies to PilN (E), and PilO (F). All single Cys mutants formed homodimers to varying extents. Images are representative of at least three independent experiments. SEPTEMBER 16, 2016 • VOLUME 291 • NUMBER 38 truncated PilO ⌬68 structure (Fig. 1A) (18) and the PilN ⌬57 / PilO ⌬68 heterodimer homology model derived from it (Fig. 1B) do not exactly mirror the in vivo conformation of the proteins. However, the ability to cross-link both homo-and heterodimers of PilN and PilO in vivo confirms that select residues are close enough and in the appropriate orientation to allow for disulfide bond formation. The recent cryoelectron tomography study of the M. xanthus T4P system (49), with resolutions averaging 4 nm, now offers the exciting possibility that we could use similar approaches to image the T4P systems of our P. aeruginosa Cys mutants in situ, to visualize how cross-linking of PilNO affects the configuration of the machinery.

In Vivo PilN and PilO Dimerization
Interestingly, the relative position of the Cys residue pairs able to cross-link PilO homodimers in vivo were at the "top" of the predicted interface in most cases, although they were at the "bottom" for the PilNO heterodimer (Fig. 2D), suggesting that the homo-and heterodimer interfaces might be angled differently. Models of the core interaction interface between PilN and PilO in T. thermophilus (29) and M. xanthus (49) show them as tilted, possibly because of longer predicted coiled coils in PilO (18). The persistence of PilN and PilO monomers (Figs.  2, B and C, and 3, A and B) not involved in covalent interactions suggests the co-existence of multiple PilN and PilO interaction species. This scenario was reported for type II secretion system PilP-and PilQ-like proteins (GspC and GspD, respectively), where contacts were established at three distinct interfaces in vivo (50). These observations, combined with our results, suggest that homo-and heterodimers may occur simultaneously. Based on these new data, we propose that the oppositely charged coiled coils of PilNO (18) could serve as the primary interaction interface between heterodimers (Fig. 9B), whereas the core regions could mediate homodimer interactions (Fig.  9C). This revised model better explains our previous findings, where up to six combined point mutations along the PilN core (now proposed to be the homodimer interface) failed to disrupt PilNO heterodimers in a bacterial two-hybrid assay (42). Furthermore, this new model could explain why PilN R142C and PilO D75C residues at the bottom of the original model (Fig. 1) adjacent to the coiled coils participate in both homo-and heterodimers (Figs. 2, B and C, and 3, A and B). A tilted PilNO core interface could position these residues in close enough proximity to form a disulfide bond with either partner. Importantly, our data do not support the proposal of large rotations in the core-core interfaces, as suggested previously for the T2S system (34).

PilN and PilO Homodimers May Represent a Functional Rather Than Assembly Intermediate State of the T4P System-
Our data show that covalent PilN or PilO homodimers are not necessarily detrimental to function. With the exception of PilN R142C and PilO D175C , both of which assembled pili but had retraction defects that abrogated motility, none of the single Cys mutants had obvious piliation or motility defects (Fig. 5, B and C), despite their capacity to form covalent homodimers in vivo (Fig. 5, E and F). The relative abundance of covalent homodimers formed by each mutant did not correlate with T4P dysfunction, as PilO Y154C formed the most dimers, but without functional consequences. This result suggests that cross-linking of only specific residues affects function and supports the idea that PilN and PilO homodimers represent a normal functional state of the T4P system.
Three covalent pairs (PilN R142C /PilO D175C , PilO A158C/I178C , and PilO Y154C/I178C ) caused significant loss of surface piliation and complete loss of motility (Fig. 4, A and B). This result suggests that some configurations of PilNO can affect pilus function. In a retraction-deficient background, where assembled pili are trapped on the surface (46), all three double Cys mutants had more pili (Fig. 4C), confirming that covalently linked dimers permit pilus assembly. Thus, lack of motility in these strains more likely arises from extension-retraction imbalances. We showed previously (42) that substitution of specific residues at the coiled-coil interface of PilN and PilO also confers an assembly-competent yet motility-deficient phenotype. Together, these data emphasize that PilNO contribute to both pilus extension and retraction and that altering their interfaces could affect the ability of the T4P machinery to switch between those states.
Previous work in T. thermophilus suggested that a dimeric complex of PilMN (two PilMs and two PilNs) forms an intermediate transmembrane platform onto which other alignment subcomplex components dock (29). The addition of PilO to the PilMN subassembly was proposed to dissociate PilN homodimers, sequestering each monomer into a PilNO heterodimer. Our data support an alternative model, in which PilN and PilO might homodimerize via their core domains (Fig.  9C), and simultaneously heterodimerize via their coiled coils and potentially their TMS (Fig. 9B) (42). This scenario could explain how untagged covalent PilN homodimers were co-purified with tagged PilO homodimers, as both constructs retain their coiled coils (Fig. 3). Alternatively, tetrameric arrangement of PilN and PilO homodimers could suggest an alternative PilNO interface on the "side" of the core homodimer interface (Fig. 9B); however, we have no evidence to support that idea. The PilN R142C and PilO D175C residues are located near the base of the modeled core interfaces and could be oriented more laterally (Fig. 9B) than depicted by our static models (Fig. 1), explaining why these particular residues were the only ones that successfully formed cross-linked homo-and heterodimers.
Modulation of the PilNO Core Interface by the Other Alignment Subcomplex Components-Cross-linking of Cys residues at select locations within the PilNO equivalents (GspLM) in the T2S system suggested rotation of the proteins through their TMS and periplasmic cores (34). In the T2S system, PilM orthologs interact with the single cytoplasmic motor ATPase (51)(52)(53)(54), and more recently, interactions between PilM and the T4P motor ATPases in P. aeruginosa and M. xanthus were confirmed (40,41). Thus, we tested whether interactions with the cytoplasmic component of the alignment subcomplex, PilM, , and Cys mutants were tested for expression and stability of each protein using protein-specific antisera. Minus signs indicate the gene was disrupted via an FRT insertion; ⌬ indicates the gene was produced by a deletion of the gene; and ⍀ indicates the gene was disrupted by an ⍀ cassette insertion.
which is bound to the N terminus of PilN, modulated PilNO interfaces. PilN R142C /PilO D175C heterodimers, but not homodimers, were lost in the absence of PilM (Fig. 6, B and C), suggesting that the PilNO interface no longer adopts an orientation permissive for disulfide bond formation. New crystal structures of a P. aeruginosa PilM homodimer and of monomeric PilM fused to part of its binding partner, PilN (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12), revealed that in the absence of PilN, the seven N-terminal residues of PilM (PilM (1-7)) bind in the PilN binding cleft of a second PilM monomer (41). Thus, PilM dimerizes in the absence of PilN. In this scenario, the lack of PilM-PilN interaction may favor PilN and PilO homodimer formation, whereas when PilM becomes monomeric upon binding PilN's N terminus, a corresponding shift in the position of the PilNO interfaces could induce heterodimer formation. PilM's interactions with its cytoplasmic partners, PilB, PilT, and/or PilC (41), differ depending on whether it is bound to PilN, potentially revealing a mechanism for switching between active and resting systems or between pilus assembly and disassembly. These differential interactions might similarly modulate PilNO homoversus heterodimer states in the periplasm. Disruption of pilB, pilT, or pilU had no effect on PilN and PilO dimerization (Fig. 8, B, D, and E). We considered that in the absence of one ATPase (e.g. PilB), one of the other two (PilT or PilU) may interact with PilM, stabilizing a configuration permissive for the formation of PilNO homoand heterodimers. However, removal of PilC, the proposed interaction partner for all three ATPases, had no effect on dimerization states (Fig. 8C).
The major pilin interacts with PilNOPQ in P. aeruginosa (19), but its specific interaction partner(s) are unknown. Transmission electron microscopy studies of T. thermophilus revealed a stable complex composed of soluble fragments of PilMNO-PilA (29) with PilN interacting directly with PilA. One could envision that PilN and PilO might help extract PilA from the IM during assembly, priming the base of the pilus fiber for incorporation of the next PilA subunit. However, loss of pilA had no effect on PilNO dimerization states (Fig. 8A).
In M. xanthus, PilQ recruits PilP, which binds PilN and PilO, and PilN binds PilM, creating a transenvelope complex (55). In P. aeruginosa, a similar pathway has been proposed, where PilP binds PilQ as well as PilNO but does not bind PilO homodimers (20). Here, loss of PilP prevented formation of PilN R142C / PilO D175C heterodimers (Fig. 6, D and E), although loss of PilQ had no effect (Fig. 6, F and G). The core regions of PilN or PilO may homodimerize (Fig. 9C), and the large electrostatic difference predicted for the coiled coils of PilN and PilO (18) could promote heterodimerization (Fig. 9B). The unstructured N terminus of PilP could interact with the core or coiled-coil regions of PilNO heterodimers, stabilizing the interface in an orientation that allowed for disulfide bond formation.
PilN and PilO Rearrangement during "Active" and "Resting" States of the T4P System-Elegant cryoelectron tomography studies of the M. xanthus T4P system allowed for derivation of a "pseudo-atomic" model that implied differences between active and resting states of the machinery (49). The active state occurs during assembly or disassembly of the fiber, when a PilB or PilT ATPase is engaged at the base of the system, whereas the resting state was ATPase-free. PilM was proposed to form a ring surrounding a dimer of PilC, and expansion of the ring upon binding of PilB or PilT to PilC during the active state, accompanied by rearrangement of the density associated with PilN and PilO's coiled coils, was observed (49). Without PilM, density corresponding to PilC and the ATPases was missing (49). In our work, only PilM or PilP could modulate the PilNO core interface. We speculate that subtle reorientation of PilN and PilO may occur during the transition between resting and active or "extension" versus "retraction" states of the T4P system, consistent with the hypothesis that PilM may disengage from PilN (41). Although a ratcheting-type mechanism, in which PilNO conformational dynamics might contribute to the stepwise addition or removal of pilin subunits cannot yet be ruled out, our data do not support this idea.

Experimental Procedures
Strains, Media, and Growth Conditions-Bacterial strains used in this study are listed in Table 1. E. coli and P. aeruginosa were grown at 37°C in Luria-Bertani (LB) media supplemented with antibiotics at the following final concentrations when necessary (g⅐ml Ϫ1 ): ampicillin, 100; carbapenem (Cb), 200; kanamycin, 50; gentamicin (Gm), 15 for E. coli and 30 for P. aeruginosa, unless otherwise specified. Bacterial strains were stored at Ϫ80°C in LB with 20% glycerol (v/v). Plasmids were transformed by heat shock into chemically competent cells. All constructs were verified by DNA sequencing (MOBIX, McMaster University).
Generation of Cys Mutants-Sets of double Cys mutations in PilO, as well as single Cys mutations in PilN and PilO, were chosen with reference to the PilO ⌬68 homodimer crystal structure Protein Data Bank code 2RJZ (18) and a PilN ⌬57 /PilO ⌬68 heterodimer homology model, created by substituting the Phyre 2 generated PilN ⌬57 monomer in place of a PilO subunit in the PilO homodimer structure (18,45). Distances were measured in PyMOL, and residues were selected to ensure that the distance between C␤ atoms was less than 8 Å (Fig. 1). Codons for selected Cys substitutions were introduced into either a pEX18Gm::pilMNOP construct (for PilNO Cys pairs) or a pEX18Gm::pilNOP construct (for double PilO Cys mutants) using the QuikChange site-directed mutagenesis kit (Stratagene) following the manufacturer's protocol. Mutations were verified by DNA sequencing (MOBIX).
Generation of pilN and pilO Cys Mutations onto the Chromosome of P. aeruginosa PAK-Cys-encoding chromosomal mutants of P. aeruginosa were generated by biparental mating with E. coli SM10 using a Flp-FRT (FLP recombination target) system as described previously (18). Constructs containing the desired Cys substitutions, pEX18Gm::pilMNOP for PilNO Cys mutants and pEX18Gm::pilNOP for PilO Cys mutants, were introduced into E. coli SM10 cells for biparental mating with PAK pilN::FRT and PAK pilO::FRT, respectively (12). The E. coli SM10 donor was counter-selected by plating on Pseu-  SEPTEMBER 16, 2016 • VOLUME 291 • NUMBER 38 domonas isolation agar (PIA; Difco) containing Gm (100 g⅐ml Ϫ1 ). Gm-resistant P. aeruginosa isolates were streaked on LB-no salt sucrose plates (1% w/v bacto-tryptone, 0.5% w/v bacto-yeast extract, 5% w/v sucrose) and then incubated for 16 h at 30°C. Select colonies were restreaked in parallel on LB and LB plates supplemented with Gm. Gm-sensitive colonies were screened by PCR using pilN or pilO primer pairs (Table 2) to confirm replacement of the original FRT-disrupted gene, and PCR products of the expected size were DNA-sequenced (MOBIX) to confirm incorporation of the desired mutations.

In Vivo PilN and PilO Dimerization
Preparation of Whole Cell Lysates-P. aeruginosa strains were grown on LB agar plates overnight at 37°C. Cells were scraped from the surface and resuspended in 1ϫ PBS (pH 7.4) to an A 600 of 0.6. A 1-ml aliquot of cells was collected by centrifugation at 2292 ϫ g for 3 min in a microcentrifuge. The cell pellet was resuspended in 100 l of either 1ϫ SDS-PAGE loading buffer (125 mM Tris, pH 6.8, 2% (v/v) ␤-ME, 20% (v/v) glycerol, 4% (w/v) SDS, and 0.001% (w/v) bromphenol blue) or 1ϫ non-reducing SDS-PAGE loading buffer (no ␤-ME) and subjected to Western blotting analysis. Biological and technical replicates of all strains were performed n Ն 3.
Western Blotting Analysis-Samples were separated on 15% SDS-polyacrylamide gels for 1.5 h at 160 V and transferred to nitrocellulose membranes for 1 h at 225 mA. Membranes were blocked using a 5% (w/v) low fat skim milk powder in 1ϫ PBS for 1 h at room temperature on a shaking platform, followed by incubation with protein-specific antibodies at the following dilutions: ␣-PilM, ␣-PilN, ␣-PilO, and ␣-PilP at dilutions of 1:1000 (17); ␣-PilB, ␣-PilT, and ␣-PilU at 1:2000, 1:500, and 1:5000, respectively (56); ␣-PilA at 1:5000 (57), and ␣-PilC at  (14). Blots were incubated in primary antibody ϳ16 h at 4°C on a rocking platform. The membranes were washed twice in 1ϫ PBS for 5 min and then incubated in goat anti-rabbit IgG-alkaline phosphatase-conjugated secondary antibody (Bio-Rad) at a dilution of 1:3000 for 1 h at room temperature. The membranes were washed twice in 1ϫ PBS for 5 min and visualized with alkaline phosphatase developing reagent (Bio-Rad) following the manufacturer's protocol. Western blottings were performed n Ն 3. Expression and Purification of PilN ⌬44 and PilO ⌬51 -N-terminally truncated versions of PilN and PilO were previously cloned into the EcoRI/HindIII and NdeI/XhoI cloning sites, respectively, of a pET28a vector, creating a vector that overexpresses untagged PilN ⌬44 and C-terminal His 6 -tagged PilO ⌬51_His (18,20). Cys mutations were introduced into this construct using the QuikChange site-directed mutagenesis kit (Stratagene) following the manufacturer's instructions to create pET28a::pilN ⌬44 /pilO ⌬51_His containing pilN ⌬44 R142C , pilO ⌬51 D175C , or a double pilN ⌬44 R142C /pilO ⌬51 D175C construct. Constructs were transformed into E. coli Origami2 (DE3) pLysS cells and plated on LB agar plates supplemented with 50 g⅐ml Ϫ1 kanamycin. A single colony was used to inoculate 20 ml of LB containing 50 mg⅐ml Ϫ1 kanamycin and incubated overnight at 37°C with shaking. The overnight culture was used to inoculate 1 liter of fresh LB (1:100 dilution) containing antibiotic, and the cells were grown at 37°C, with shaking, to an A 600 of 0.6. Protein expression was induced by adding isopropyl ␤-D-1-thiogalactopyranoside (Sigma) to the culture at a final concentration of 1 mM. The cells were allowed to grow for 4 h at 37°C, prior to being harvested by centrifugation (3993 ϫ g, 15 min, 4°C) in an Avanti J-26 XPI centrifuge (Beckman Coulter). Bacterial pellets were frozen at Ϫ80°C until further use.
Bacterial pellets were thawed and resuspended in 10 ml nickel A buffer (20 mM Tris-HCl, pH 7.2, 500 mM KCl, 10 mM imidazole, 10% (v/v) glycerol) and then combined in a 50-ml screw cap tube with one complete EDTA-free protease inhibitor tablet (Roche Applied Science). Cells were lysed via sonication on ice, on setting 4, for 2 min with cycles of 10 s on and 10 s off (Sonicator 3000; Misonix). The lysates were centrifuged (11,000 ϫ g, 30 min, 4°C) in an Avanti J-26 XPI centrifuge (Beckman Coulter) to remove intact cells and other cellular debris. Pelleted material was retained for analysis by SDS-PAGE, and supernatants were filtered through 0.22-m Acrodisc syringe filter (Pall Corp.). The lysate was purified using nickel-nitrilotriacetic acid affinity chromatography on an AKTA start FPLC (VWR Scientific). Protein lysate was flowed through a 1-ml His-Trap TM FF column (GE Healthcare) preequilibrated with buffer and washed with nickel A buffer and then increasing amounts of nickel B buffer (20 mM Tris-HCl, pH 7.2, 500 mM KCl, 300 mM imidazole, 10% (v/v) glycerol) in a linear gradient. The bound protein was eluted from the column in pure nickel B buffer and collected in 1-ml fractions. An aliquot of each fraction was mixed 1:1 with 2ϫ reducing or nonreducing SDS-PAGE loading and electrophoresed on a 15% SDS-polyacrylamide gel. The gel was stained using Coomassie Blue Staining solution (0.1% Coomassie Brilliant Blue R-250,