Structure

Article

Structural Determinantsof Polymerization Reactivityof the P pilus Adaptor Subunit PapFDenis Verger,1,2 Rebecca J. Rose,3 Emanuele Paci,3 Greg Costakes,1 Tina Daviter,1,2 Scott Hultgren,4 Han Remaut,1,2

Alison E. Ashcroft,3,* Sheena E. Radford,3,* and Gabriel Waksman1,2,5,*1Institute of Structural and Molecular Biology, University College London, Birkbeck College, Malet Street, London WC1E 7HX, UK2School of Crystallography, Malet Street, London WC1E 7HX, UK3Astbury Centre for Structural Molecular Biology, Astbury Building, University of Leeds, Leeds LS2 9JT, UK4Department of Microbiology and Center for Women’s Infectious Disease Research, Washington University Medical School,

St. Louis, MO 63011, USA5Research Department of Structural and Molecular Biology, University College London, Gower Street, London WC1E 6BT, UK*Correspondence: g.waksman@bbk.ac.uk or g.waksman@ucl.ac.uk (G.W.), S.E.Radford@leeds.ac.uk (S.E.R.),

A.E.Ashcroft@leeds.ac.uk (A.E.A.)

DOI 10.1016/j.str.2008.08.012

SUMMARY

P pili are important adhesive fibers involved in kidneyinfection by uropathogenic Escherichia coli. Pilussubunits are characterized by a large groove resultingfrom lack of a b strand. Polymerization of pilus sub-units occurs via the donor-strand exchange (DSE)mechanism initiated when the N terminus of an in-coming subunit interacts with the P5 region/pocketof the previously assembled subunit groove. Here,we solve the structure of the PapD:PapF complex inorder to understand why PapF undergoes slowDSE. The structure reveals that the PapF P5 pocketis partially obstructed. MD simulations show this re-gion of PapF is flexible compared with its equivalentin PapH, a subunit that also has an obstructed P5pocket and is unable to undergo DSE. Using electro-spray-ionization mass spectrometry, we show thatmutations in the P5 region result in increased DSErates. Thus, partial obstruction of the P5 pocketserves as a modulating mechanism of DSE.

INTRODUCTION

Uropathogenic E. coli (UPEC) are responsible for urinary tract

infections, notably cystitis (infection of the bladder) or pyelone-

phritis (infection of the kidney) (Hooton and Stamm, 1997;

Roberts et al., 1994). Survival and persistence of these bacteria

in the urinary tract require the expression, among other virulence

factors, of type 1 and P pili. P pili are required for the ability of

UPEC to bind Gal-a(1-4)-Gal moieties on the surface of human

kidney cells and thus are important virulence determinants of

pyelonephritis (Dodson et al., 2001; Kuehn et al., 1992; Roberts

et al., 1994; Soto and Hultgren, 1999). P pili are encoded by the

pap gene cluster and are assembled via the highly conserved

chaperone-usher pathway, involving the periplasmic immuno-

globulin-like chaperone PapD and an outer membrane usher

PapC (Barnhart et al., 2000; Dodson et al., 1993; Holmgren and

1724 Structure 16, 1724–1731, November 12, 2008 ª2008 Elsevier L

Branden, 1989; Remaut et al., 2008; Sauer et al., 2004; Saulino

et al., 1998; Slonim et al., 1992; Soto et al., 1998; Thanassi

et al., 1998). P pili consist of six types of subunits making up

a composite fiber with the Gal-a(1-4)-Gal -specific adhesin,

PapG, at the tip, a short tip fibrillum formed of 5–10 PapE subunits

joined to a more rigid helical rod formed of about a thousand

PapA subunits and finally the terminator subunit PapH, last to

be assembled (Baga et al., 1987; Bullitt and Makowski, 1995;

Gong and Makowski, 1992; Jacob-Dubuisson et al., 1993; Kuehn

et al., 1992) (Figure 1A). In addition, there are two subunits termed

‘‘adaptor subunits,’’ PapF and PapK. The adaptor subunit PapF is

located between the adhesin and the tip fibrillum, and its incorpo-

ration within the pilus is thought to induce the polymerization of

PapE and also to orientate PapG relative to the PapE fibrillum

for optimal presentation to the kidney receptor (Jacob-Dubuis-

son et al., 1993; Lee et al., 2007). The other adaptor subunit,

PapK, is found between the PapE tip fibrillum and the PapA rod

(Jacob-Dubuisson et al., 1993) (Figure 1A).

All pilin subunits adopt a truncated immunoglobulin (Ig)-like

fold where the C-terminal G b strand is lacking, thus producing

a large hydrophobic groove on the side of the protein (Choudhury

et al., 1999; Sauer et al., 1999). In a process termed donor-strand

complementation, the chaperone PapD provides the structural

information lacking in the pilus subunit by inserting one of its

b strands, the G1 b strand, into the subunit’s groove, thereby

completing the Ig fold of the subunit (Barnhart et al., 2000;

Choudhury et al., 1999; Sauer et al., 1999). Four alternating

residues in the chaperone’s G1 b strand, termed P1 to P4 resi-

dues, contact the groove in four contiguous regions of the

groove, named P1 to P4 binding regions/pockets. Pilus subunit

assembly proceeds via a noncovalent polymerization process

called donor-strand exchange (Sauer et al., 2002, 2004; Soto

et al., 1998; Vetsch et al., 2006; Zavialov et al., 2003) . All subunits,

except the adhesin PapG, possess an N-terminal extension (Nte)

peptide (see sequence alignment in Figure 1C) that is disordered

and not part of their Ig-like structure. The Nte comprises a highly

conserved array of alternating hydrophobic residues, termed

P2 to P5 residues. DSE proceeds via the formation of an interme-

diate consisting of the chaperone-subunit complex bound to the

Nte of the subunit coming next in assembly. The formation of this

td All rights reserved

Structure

Crystal Structure of the P Pilus Subunit PapF

complex is crucially dependent on the identity of the Nte P5

residue. Insertion of this P5 residue into a distinct pocket in the

subunit groove, termed the P5 pocket, initiates a zip-in-zip-out

process whereby the chaperone’s G1 strand is progressively dis-

placed by the Nte peptide, resulting in subunit-subunit interaction

and release of the chaperone (Remaut et al., 2006). Recently, the

significance of this process was confirmed by Verger et al. (2006),

who showed that PapH has an obstructed P5 pocket and is thus

unable to undergo DSE.

The defined order by which Pap subunits are assembled to

form the P pilus has recently been investigated and shown to

be strongly correlated with the interaction kinetics between

subunits’ grooves and Nte peptides, which, in turn, are dictated

by the C-terminal sequence of the Nte peptides, where the P5

residue is located (Rose et al., 2008a). In the study by Rose

and coworkers, all six chaperone-subunit complexes that the

Pap subunits can form with their cognate chaperone PapD

(PapD:PapG, F, E, K, A, and H) were challenged in vitro by

each of the five possible Nte peptides (as indicated above,

PapG does not have an Nte) and DSE was monitored using elec-

trospray-ionization mass spectrometry (ESI-MS) (Rose et al.,

2008a). Remarkably, each of the complexes reacted most

Figure 1. P Pilus Organization and Function

(A) Schematic diagram of the P pilus. Each

subunit, the chaperone, and the usher are labeled

using a one-letter code (e.g., G for PapG). For

clarity, only one protomer in the usher is shown

(ushers are dimeric). A growing and a mature P pi-

lus are shown at left and right, respectively.

(B) t50 values for each DSE reaction involving

cognate Ntes. In each case, a 10-fold molar

excess of the peptide Nte was added over the

concentration of PapD:PapX (where X denotes

the subunit under consideration) and the concen-

tration of PapD:PapX remaining was monitored as

a function of time using ESI-MS (see Experimental

Procedures). The rates of DSE, denoted here as

the time for the reaction to reach 50% completion

(t50), are shown for PapD:PapG with FNte, PapD:

PapF with ENte, PapD:PapE with KNte, PapD:PapK

with ANte, and PapD:PapA with ANte. The data

were taken from Rose et al. (2008a).

(C) Sequence alignment of the Pap subunits.

Identical residues are boxed in red, and similar

residues in purple. PapA and PapF secondary

structures are shown in orange and blue, respec-

tively, with b strands, a helices, and loops dis-

played as arrows, boxes, and lines, respectively.

The only helix in PapA is shown in red. Nte resi-

dues corresponding to the P2, P3, P4 (purple),

and P5 (red) positions are indicated. Pro32 is

boxed in magenta whereas Arg126 and Glu33

are boxed in green. Arrows in the corresponding

colors are positioned under the corresponding

boxes to help locate these residues.

rapidly when challenged by the Nte pep-

tide derived from the subunit known to

come next in the assembly of the P pilus

(termed ‘‘cognate’’ Nte), indicating that

the kinetics of groove-Nte interaction re-

capitulate the order of assembly (Rose et al., 2008a). PapG

and PapF, the two most distal subunits in the P pilus, undergo

slow DSE compared with all other subunits (Rose et al., 2008a)

(Figure 1B). The DSE reaction where PapD:PapF is challenged

by the Nte peptide of PapE (ENte) takes about 76 hr to reach

50% completion (t50 = 76 hr). DSE reaction with PapD:PapG

challenged with the Nte peptide of PapF (FNte) is even slower

with a t50 of 120 hr. In contrast, the DSE reactions for PapD:PapK

when challenged with the Nte peptide of PapA (ANte) or for

PapD:PapE when challenged with the Nte peptide of PapK

(KNte) have t50 values of 12.5 and 3.2 hr, respectively (Rose

et al., 2008a) (Figure 1B). Thus PapG and PapF are slow-ex-

changing subunits whereas the others (except PapH, which was

confirmed to be unable to undergo DSE) are fast exchanging.

From the structures of PapK, PapE, PapH, and recently PapA,

it appears that fast-exchanging subunits (PapA, PapK, and

PapE) all have an open (PapA and PapK) or disordered (PapE)

P5 pocket. Only in PapH is the P5 pocket closed or obstructed

(Sauer et al., 1999, 2002; Verger et al., 2006, 2007).

In this study, we sought to understand the slow-exchanging

behavior of the PapF subunit. By combining structural, computa-

tional, and biophysical methods, we show that the P5 pocket of

Structure 16, 1724–1731, November 12, 2008 ª2008 Elsevier Ltd All rights reserved 1725

Structure

Crystal Structure of the P Pilus Subunit PapF

PapF is partially obstructed by residues in two loops. Molecular

dynamics (MD) simulations show that these loops are flexible,

whereas the equivalent regions in the PapH structure are rigid.

Moreover, mutating contact residues between the two loops in

PapF to decrease their interaction results in faster DSE reac-

tions. These results show that intrinsic loop flexibility at and

around the P5 pocket modulates P5 pocket accessibility, which

in turn determines the rates at which subunits undergo DSE.

RESULTS

Structure of PapF and Comparisonwith Other Pap SubunitsWe first asked whether the structure of the PapD:PapF complex

could provide insights that could help understand why this com-

plex exhibits slow rates of DSE in vitro when challenged by its

cognate Nte peptide, ENte, in vitro. The full-length PapD:PapF

was thus purified but did not yield crystals because of a tendency

to aggregate at high concentration. We hypothesized that aggre-

gation of PapF may occur through DSE, and thus we sought to

prevent aggregation by mutating the conserved P4 glycine in

the Nte to Asn (Figure 1C). The P4 glycine residue in the sub-

unit-Nte complex lies on top of a bulky residue in the subunit’s

groove, and thus Asn at P4 should provide the steric clashes

that prevent the Nte from fitting into the subunit’s groove. The

resulting PapD:PapFG8N did not aggregate and yielded crystals

diffracting to 2.2 A resolution. Its structure was solved by molec-

ular replacement using the PapD:PapK structure as the search

model (see Experimental Procedures).

The PapD:PapFG8N structure is very similar to the already

known PapD:PapK, PapD:PapENtd, PapD:PapHNtd1, or PapD:

PapANtd1_G15N structures. Like in PapK, PapE, PapH, or PapA,

PapF exhibits an Ig-like fold where the seventh strand is missing

(Figure 2). Like in all other Pap complexes, the chaperone PapD

complements the PapF structure by donating its G1 strand to

complement the PapF’s fold (not shown).

When comparing the structures of PapA and PapF (Figure 2A),

two major differences are apparent in the region between

strands A2 and B (the A2-B loop) and between strands B and

C1 (the B-C1 loop). The B-C1 loop in PapF is short and disor-

dered (electron density for residues 53 and 54 is weak and these

residues could not be built), but is long and also disordered in the

PapD-bound PapA subunit of the PapD:PapA2 complex (Verger

et al., 2007) (Figures 1 and 2A). In PapA, the B-C1 loop becomes

ordered upon DSE, extending the subunit’s groove to accommo-

date the long PapA Nte. Among all Pap subunits, the only other

subunit with a long B-C1 loop is PapH, and in the structure of this

subunit, it is ordered (Verger et al., 2006) (Figure 2B).

Differences in the A2-B loop of the structure are likely to be

functionally significant as this loop forms part of the P5 pocket,

a major initiation point for DSE and a specificity-determining

DSE site (Remaut et al., 2006; Rose et al., 2008a). In PapF,

Pro32 obstructs the P5 pocket by contacting residues in the

bD1 strand (Phe81) and the E-F loop (Leu132, Gly135) (Figure 3B).

The A2-B loop is stabilized by a salt bridge between Glu33 in the

loop and Arg126 at the end of the bE strand (Figures 3A and 3B).

In Figure 3, the structure of the region around the P5 pocket

is shown in all Pap subunits for which the structure is known

(except PapE, where the regions forming the P5 pocket are dis-

1726 Structure 16, 1724–1731, November 12, 2008 ª2008 Elsevier

ordered). PapA and PapK both exhibit well-defined P5 pockets

(Figure 3C). These subunits are known to undergo DSE at very

similar rates (with a reaction midpoint [t50] of 15.2 and 12.5 hr, re-

spectively; Figure 1B). In PapE, the A2-B loop is disordered and

thus the region around the P5 pocket is presumably flexible. Re-

markably, PapE is known to be a fast-exchanging pilus subunit;

that is, it undergoes DSE at the highest rate (t50 of 3.2 hr) among

Pap subunits (Rose et al., 2008a) (Figure 1B). In contrast, in PapH

and PapF, the P5 pocket is obstructed by residues in the A2-B

loop (Thr52 in PapH and Pro32 in PapF) (Figure 2B). PapH is

unable to undergo DSE while PapF undergoes DSE at a very

slow rate (Figure 1B). Thus, there appears to be a correlation

Figure 2. Comparison of the PapF Structure with the PapA and PapH

Structures

(A) Superposition of the structures of PapFG8N (cyan) and PapANtd1_G15N

(orange). These are shown in ribbon representations with b strands indicated

as arrows and a-helices indicated as cylinders. Secondary structures referred

to in the main text are labeled. The a-helix and the A2-B and B-C1 loops are

indicated, as well as Pro32 (in stick representation).

(B) Superposition of the structures of PapFG8N (cyan) and PapHNtd1 (magenta).

Representation is the same as in panel A. Residues Pro32 of PapF, and Thr52

and Pro53 of PapH, are shown in stick representation. The a-helix and the

A2-B and E-F loops are indicated.

Ltd All rights reserved

Structure

Crystal Structure of the P Pilus Subunit PapF

between the level of accessibility and flexibility of the P5 pocket

and the rate of DSE: the more accessible and flexible the region

around the P5 pocket is, the faster the exchange reaction.

The A2-B Loop of PapF Is Flexible,Whereas that of PapH Is RigidFor PapF to undergo DSE, accessibility to the P5 pocket is

required. Yet, the P5 pocket of PapF is obstructed in the crystal

structure. Thus, in view of the crystal structure alone, PapF

should not be able to undergo DSE. The fact that PapF is able

to participate in the exchange reaction, albeit at a slow rate,

suggests that its P5 pocket must transition to an open state.

To test this hypothesis, MD simulations were carried out on the

PapD:PapF and PapD:PapH complex structures (Figure 4).

Each structure was stable during simulations, as indicated by

an rmsd from the crystal structure not exceeding 3 A.

The walls of the P5 pocket are formed by residues in the A2-B

loop on one side of the P5 pocket and residues in the E-F loop

and the N terminus of strand D1 on the other side of the pocket

(Figure 2B). Any opening and closing transitions in the P5 pocket

should result in fluctuations in the distances between the A2-B

loop residues and those in the E-F loop or in the N terminus of

strand D1. Thus, we monitored the following distances. In PapH

(Figure 4, top panel), we monitored the distance between Thr52

(A2-B loop) and Ile114 (strand D1), and between Thr52 and

Gly160 (E-F loop). In PapF (Figure 4, lower panel), we monitored

the distance between Pro32 (A2-B loop) and Phe81 (strand D1)

and between Pro32 and Leu132 (E-F loop). These simulations

reveal that in PapH, the distances between Thr52 and Gly160

or Ile114 do not change substantially, whereas in PapF the

distances between Pro32 and Phe81 or Leu132 fluctuate signif-

Figure 3. Comparison of the P5 Regions of

the Pap Subunits

(A and C) The structures are in surface representa-

tion. Each structure is identified with the name of

the molecule above it. In each, the location of the

conserved aromatic P4 pocket residue is indi-

cated and serves to position each panel relative

to one another. In each panel, the P5 pocket is

identified by a green circle, either in a continuous

line when the P5 pocket is accessible (PapA,

PapK) or in dashed line when it is obstructed

(PapF, PapH). Finally, for PapF and PapH, the

location of the residues mentioned in the text is

indicated.

(B) A stereo diagram of the P5 region of PapF is

shown with residues Pro32, Glu33, Arg126,

Phe81, Gly135, and Leu132 in the P5 region

(shown in sticks). The location of the P4 pocket

residue is represented in sticks and color coded

in red.

icantly. Interestingly, fluctuations in the

positioning of Pro32 relative to Phe81

and Leu132 in PapF appears to be re-

stricted to a limited number of conforma-

tional states, closed as seen in the crystal

structure and open where the P5 pocket

appears to be no longer obstructed. The

comparison of rmsd per residue between PapF and PapH is

also instructive. It clearly shows that fluctuations are much larger

in the P5 region of PapF than in the equivalent region of PapH:

Pro32, the residue in the A2-B loop obstructing the P5 pocket

in PapF, has an rmsd of 4.2 A, whereas the corresponding

Thr52 in PapH has an rmsd of 1.2 A. Overall, whereas the P5

pocket appears blocked in both the PapH and PapF crystal

structures, MD simulations suggest that the P5 pocket of PapF

becomes accessible intermittently whereas that of PapH does

not. These results thus provide an explanation as to why PapF

undergoes slow DSE and PapH does not undergo DSE at all.

Mutants Destabilizing the A2-B Loop DisplayHigher DSE RatesTo test the hypothesis that the A2-B loop flexibility is important in

controlling the rate of DSE in PapF, two mutations anticipated to

lower the stability of this loop were introduced and their effect on

DSE was monitored. Pro32 was mutated to Gly, a mutation that

would lower the contact area between the A2-B and E-F loops in

this region, and Arg126 was mutated to Ala, to remove the salt

bridge stabilizing the A2-B loop (see above). Both mutations

are expected to increase structural flexibility in the P5 pocket

region. DSE was monitored using the ESI-MS assay described

in Remaut et al. (2006) for SafA and Rose et al. (2008a) for all

Pap subunits. Because the structure determined here was of

PapFG8N, this amino acid substitution was also introduced in

the proteins. Complexes (wild-type PapD:PapF, PapD:PapFG8N,

PapD:PapFG8N_P32G, and PapD:PapFG8N_R126A) were monitored

in the absence of the attacking Nte and were shown to be stable

over the incubation period (Figure 5A). DSE reactions were initi-

ated by the addition of the Nte of PapE (ENte) and monitored over

Structure 16, 1724–1731, November 12, 2008 ª2008 Elsevier Ltd All rights reserved 1727

Structure

Crystal Structure of the P Pilus Subunit PapF

a period of 200 hr at room temperature. Control experiments

comparing wild-type PapD:PapF with the mutant PapD:PapFG8N

confirmed that the G8N mutation does not impact DSE rates

(Figure 5B). In the presence of ENte, PapD:PapFG8N undergoes

DSE with the t50 of 81.5 hr, comparable to that obtained by

Rose et al. (2008a). The double mutants PapFG8N_P32G and

PapFG8N_R126A undergo DSE at significantly higher rates than

PapFG8N with a t50 of 37.0 hr and 42.2 hr, respectively (Figure 5B).

It would thus appear that mutations expected to increase the

flexibility of the A2-B loop result in increased rates of DSE.

DISCUSSION

PapF is an important adaptor protein that is not only essential

in P pilus biogenesis, but also appears, together with PapG, to

act as a checkpoint in this process (Dodson et al., 2001;

Jacob-Dubuisson et al., 1993; Lee et al., 2007). Indeed, it was

recently shown that the in vitro DSE rates for PapG and PapF

are by far the slowest among all Pap subunits (Rose et al.,

2008a) (Figure 1B). Thus, commitment to fully-fledged pilus bio-

genesis only occurs after the two slowest subunits have reacted,

PapG with the Nte of PapF, and PapF with the Nte of the first

incoming PapE subunit. Once this is achieved, DSE proceeds

rapidly until termination by the PapH subunit.

The molecular basis for the slow DSE kinetics of the PapF

subunit was unclear. The structure of this subunit determined

here provides a molecular interpretation of these observations,

showing that PapF has a partially obstructed P5 pocket, which

appears to prevent access by the P5 residue of the attacking

PapE Nte peptide. In prior studies, we have demonstrated that

the P5 region of subunits’ grooves is an essential region serving

as the first point of contact for the Nte peptide of the subunit next

in assembly (Remaut et al., 2006). Not only does the P5 pocket of

receiving subunits act as an initiation site for DSE, but also we

have shown that this early interaction event determines, to a large

Figure 4. Molecular Dynamics Simulations

Fluctuations of distances between pairs of residues in PapH (top panel) and in

PapF (bottom panel). The residues monitored during simulations are indicated

in the inset in each panel.

1728 Structure 16, 1724–1731, November 12, 2008 ª2008 Elsevier L

extent, the specificity of the DSE reaction (Remaut et al., 2006;

Rose et al., 2008a). Among all Pap subunits, PapH is the only

subunit unable to undergo DSE. This is because it has an ob-

structed P5 pocket and, as a result, lacks a DSE-initiating site.

Thus, obstruction of the P5 pocket is a brake on DSE. In PapF,

partial obstruction of the P5 pocket would also act as a brake,

but DSE still proceeds in this case, albeit at a slow rate. This

implies that the PapF P5 pocket, in contrast to that of PapH, is

intermittently accessible for DSE initiation. This conclusion is

borne out by the MD simulations on SafA (Rose et al., 2008b)

and on the PapF and PapH subunits (this work), showing clear

differences between PapF and PapH in the region where the

P5 pocket lies: it is flexible in PapF and rigid in PapH.

The differences in flexibility between the two proteins in the P5

region are likely due to the combined effects of two structural

features: the first is the presence of the only a-helix in the Ig

fold of PapH, which is absent in PapF, and the second is the

Figure 5. DSE ESI-MS Assay of the Wild-Type and Mutant PapD:

PapF Complexes

The DSE reactions are monitored in the absence of any Nte (A) or after adding

ENte (B). Wild-type PapD:PapF is shown in black, PapD:PapFG8N in red, PapD-

PapFG8N_P32G in green, and PapD-PapFG8N_R126A in blue. Average error is

± 2.5%.

td All rights reserved

Structure

Crystal Structure of the P Pilus Subunit PapF

location of residue Pro53 in the PapH structure, a residue that

immediately follows the helix (Figure 2B). Pro53 introduces

a kink in the loop between the helix and the subsequent strand

and this kink projects Thr 52 into the PapH P5 pocket, thereby

obstructing it completely. Thr52 is thus rigidly maintained within

the P5 pocket by the combined action of the helix and Pro53. In

contrast, although PapF also contains a proline in that region

(Pro32), this proline aligns with Thr52 of PapH and not Pro53 in

the superposition of the two structures (Figure 2B). It is also

located in a long flexible region that does not contain an a-helix

like in PapH. Thus, obstruction of the P5 pocket of PapF by

Pro32 is solely due to the interaction of Pro32 with the residues

in the bD1 strand and the E-F loop, an interaction that is relatively

unstable, as the MD simulations presented here suggest. As a

result, Pro32 moves in and out of the P5 pocket. Note that the

presence of an a-helix between the bA2 and bB strands is not

unique to PapH. It is also present in PapA (Figure 2A) and in

PapK. However, in PapA and PapK, it provides a rigid foundation

for a well-defined P5 pocket (Figure 3C).

The mutational study presented here confirms our structural in-

terpretation of the observed DSE kinetics of PapF. Reducing the

contacts and electrostatic interactions between the A2B and the

EF loops residues results in faster DSE reactions. Thus, changing

the residue composition of the P5 pocket changes the DSE rates.

In a recent mutational study, Remaut et al. (2006) showed that in

the Saf system, the chaperone occupies the P5 site but appears

to do so in an equilibrium with an unbound, accessible state. Mu-

tating the chaperone’s G1 residue occupying the P5 site toa larger

hydrophobic side chain shifts the equilibrium toward the closed

state, suppresses formation of the chaperone-subunit-Nte inter-

mediate, and drastically reduces the DSE rate. Here, for the first

time, we show that mutating the P5 pocket itself results in predict-

able changes in DSE rates. Thus, this study unravels an exquisite

control system centered on the accessibility of the P5 pocket to

the attacking Nte, which modulates DSE reactions for all Pap

subunits during pilus biogenesis.

These studies altogether converge to confirm the P5 initiation

model first proposed by Remaut et al. (2006). Whether a similar

model applies to other systems where strand-zippering has

been observed (see review by Remaut and Waksman, 2006)

remains to be seen. Given the general importance of strand

addition and insertion in macromolecular interactions, thorough

mechanistic studies of these systems are overdue. However, it is

expected that they will likely share common mechanistic details

with DSE, including the need for an initiation point not only

driving the process but also contributing to specificity.

EXPERIMENTAL PROCEDURES

Plasmids

The PapD:PapF construct (pTrc-DF) was designed as described previously

(Rose et al., 2008a). The PapD:PapFG8N mutant (pTrc-DFG8N) was made

from the pTrc-DF plasmid, using oligonucleotides overlapping the sequence

encoding the G8 residue, with the glycine 8-encoding codon replaced by an

asparagine one. Similarly, the double mutants PapD:PapFG8N_P32G and PapD:

PapFG8N_R126A were prepared from the pTrc-DFG8N plasmid, using oligonucle-

otides overlapping the codon encoding P32 or R126, with the P32 or R126

codons replaced by a glycine (P32G) or an alanine (R126A), respectively.

(Stratagene QuickChange Site-Directed Mutagenesis Kit; see Table 1 for

oligonucleotide design).

Structure 16, 1724–17

Expression and Purification of PapD:PapF, PapD:PapFG8N,

PapD:PapFG8N_P32G, and PapD:PapFG8N_R126A

PapD:PapF, PapD:PapFG8N, PapD:PapFG8N_P32G and PapD:PapFG8N_R126A

were expressed and purified as previously described by Rose et al. (2008a).

Each complex was concentrated to around 10 mg/ml and some of the PapD:

PapFG8N complex was used for crystallization trials. For mass spectrometry

analysis, all the complexes were dialyzed against 5mM ammonium acetate

pH 5.5 and used at a concentration of 40 mM.

Mass Spectrometry Analysis

A 40 micromolar solution of PapD:PapF complex (wild-type or mutant) was in-

cubated with 400 mM of a peptide derived from the Nte of PapE (sequence

VDNLTFRGKLII, to which 2 lysines at the C terminus were added in order to

render the peptide soluble), in 5mM ammonium acetate buffer (pH 5.5) at

room temperature. As a control, the complex was also incubated in the

same buffer and at the same temperature in the absence of any peptide.

One-microliter aliquots were removed from the reaction mixture at given

time points and analyzed by mass spectrometry. Nano-electrospray ionization

was used, where samples (1–2 mL) are introduced into gold-plated borosilicate

vials and analyzed using a Q-Tof mass spectrometer (Waters UK Ltd.,

Manchester, UK). Data were acquired over the m/z range of 1500–5000 with

a scan speed of 2.4 s, and CsI clusters were used as an external calibrant to

ensure mass accuracy of within 0.01%. Peaks relating to the various PapD:

PapF complexes were integrated and summed. For each time point, the ratio

of these peaks to the total area of peaks in the spectrum was calculated. These

data were normalized between the maximum value (before addition of peptide)

and the baseline, and fitted using a single exponential.

PapD:PapFG8N Crystallization, Structure Determination,

and Refinement

The PapD:PapFG8N complex was crystallized at room temperature in a hanging

drop equilibrated against a reservoir solution containing 20% glycerol and

2.0–2.2 M ammonium sulfate with no buffer (original condition obtained from

solution C6, Jena Biosciences screen 6). The space group of the crystals is

P41212, with one complex per asymmetric unit (44% solvent). The cell dimen-

sions are as follows: a = 65.4 A, b = 65.4 A, and c = 166.1 A. The data for a single

crystal were processed to a resolution of 2.2 A (see Table 2). The structure was

solved by molecular replacement using PapD:PapK structure (PDB entry

1PDK) as a search model using the program AMoRe (Navaza, 2001). Succes-

sive cycles of manual rebuilding with O and conjugate gradient minimization

using crystallography and NMR system were performed (Brunger et al.,

1998; Jones et al., 1991). B factors were refined individually. Water molecules

were added in agreement with expected hydrogen bond distances and Fo �Fc density superior or equal to 3.5s. The refinement converged to the final

values of R = 23.55% and Rfree = 27.73% (20–2.2 A range; F/sF R 0.0) with

good stereochemistry (see Table 2 for additional statistics). The final model

includes all PapD main chain atoms except the last 3 residues (216–218);

the following residues in PapFG8N were missing from the electron density:

1 to 8, 53 to 54, and 98 to 114. The final model includes 2 sulfate ions and

85 water molecules.

Table 1. Oligonucleotides Used in this Study

Construct Forward primer Reverse primer

PapFG8N GTG CAG ATT AAC

ATC AGG AAC AAT

GTT TAT ATC CCC

CC

ATA TAA ACA TTG

TTC CTG ATG TTA

ATC TGC ACA TCA

GCC

PapFG8N_P32G GAT TTT GGG AAT

ATT AAT GGT GAG

CAT GTG GAC AAC

TCA CG

GAG TTG TCC ACA

TGC TCA CCA TTA

ATA TTC CCA AAA

TCA AC

PapFG8N_R126A CT TCA GTG CCC

TTT GCG AAT GGC

AGC GGG ATA CTG

AAT GG

CAG TAT CCC GCT

GCC ATT CGC AAA

GGG CAC TGA AGT

AAA GG

31, November 12, 2008 ª2008 Elsevier Ltd All rights reserved 1729

Structure

Crystal Structure of the P Pilus Subunit PapF

Molecular Dynamics Simulations

Simulations were performed using an atomistic model with implicit solvent

(Lazaridis and Karplus, 1999). The crystallographic structures of the PapD:

PapHNtd1 (PDB code 2J2Z) and PapD:PapFG8N (this work) chaperone subunit

complexes were used as initial conformations for the simulations. In PapD:

PapFG8N, the loop residues K53 and S54 missing from the model were built

as alanines. Missing residues 98–114, belonging to a long and structurally con-

served loop region among all subunits, were reconstructed by structural align-

ment based on the equivalent region of the PapK model (PDB code 1PDK).

These additional modeled regions were energy minimized, and the resulting

model showed good stereochemistry (data not shown). After minimization

and equilibration, 100 ns of Langevin dynamics at 300K with a friction coeffi-

cient of 1 ps�1 were produced for each complex.

ACCESSION NUMBERS

The coordinates of the PapD:PapFG8N complex were submitted to the Protein

Data Bank (entry code 2w07).

ACKNOWLEDGMENTS

R.J.R. was funded by a BBSRC/CASE PhD studentship and Waters UK Ltd.

We thank R. H. Bateman (Waters UK Ltd.) for continued support. The

Table 2. Data Collection and Refinement Statistics

Data collection

Data set PapD/PapFG8N

(ESRF ID14)

Resolution 20.0�2.2 A

Total/unique 138,425/19,108

reflections

Completeness (%)a 99.8 (100)

Rsym(%)b 5.3 (29.3)

I/s(I)c 9.4 (2.6)

Refinement

Resolution (A) 20.0�2.2

Number of reflectionsd 19,065 (980)

Total number of atomse 2626

R/Rfree (%)f 23.6/27.7

rmsdg Bonds (A)

0.006

Angles (�)

1.26

B values (A2)

Main chain Side chain

1.5 2.2a Completeness for I/s(I) > 0, high-resolution shell (2.32�2.2 A) in paren-

theses.b Rsym =

PjI� hIij

PI, where I = observed intensity and hIi = average

intensity from multiple observations of symmetry related reflections;

high-resolution shell in parentheses.c I/s(I), high-resolution shell in parentheses.d Total number of reflections (F/s(F) > 0.0), number in the ‘‘test set’’ in

parentheses.e Includes water molecules and sulfate ions.f Rfree was calculated on the basis of 5% of the total number of reflections

(‘‘test set’’) randomly omitted from the refinement.g Deviations from ideal bond lengths and angles and in B factors of

bonded atoms.

1730 Structure 16, 1724–1731, November 12, 2008 ª2008 Elsevier L

Q-Tof1 mass spectrometer was funded by the BBSRC. The work was sup-

ported with funds from NIH, the BBSRC, and the Wellcome Trust.

Received: July 14, 2008

Revised: August 13, 2008

Accepted: August 14, 2008

Published: November 11, 2008

REFERENCES

Baga, M., Norgren, M., and Normark, S. (1987). Biogenesis of E. coli Pap pili:

PapH, a minor pilin subunit involved in cell anchoring and length modulation.

Cell 49, 241–251.

Barnhart, M.M., Pinkner, J.S., Soto, G.E., Sauer, F.G., Langermann, S., Waks-

man, G., Frieden, C., and Hultgren, S.J. (2000). PapD-like chaperones provide

the missing information for folding of pilin proteins. Proc. Natl. Acad. Sci. USA

97, 7709–7714.

Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-

Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al.

(1998). Crystallography & NMR system: A new software suite for macromolec-

ular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921.

Bullitt, E., and Makowski, L. (1995). Structural polymorphism of bacterial adhe-

sion pili. Nature 373, 164–167.

Choudhury, D., Thompson, A., Stojanoff, V., Langermann, S., Pinkner, J.,

Hultgren, S.J., and Knight, S.D. (1999). X-ray structure of the FimC-FimH

chaperone-adhesin complex from uropathogenic Escherichia coli. Science

285, 1061–1066.

Dodson, K.W., Jacob-Dubuisson, F., Striker, R.T., and Hultgren, S.J. (1993).

Outer-membrane PapC molecular usher discriminately recognizes periplas-

mic chaperone-pilus subunit complexes. Proc. Natl. Acad. Sci. USA 90,

3670–3674.

Dodson, K.W., Pinkner, J.S., Rose, T., Magnusson, G., Hultgren, S.J., and

Waksman, G. (2001). Structural basis of the interaction of the pyelonephritic

E. coli adhesin to its human kidney receptor. Cell 105, 733–743.

Gong, M., and Makowski, L. (1992). Helical structure of P pili from Escherichia

coli. Evidence from X-ray fiber diffraction and scanning transmission electron

microscopy. J. Mol. Biol. 228, 735–742.

Holmgren, A., and Branden, C.I. (1989). Crystal structure of chaperone protein

PapD reveals an immunoglobulin fold. Nature 342, 248–251.

Hooton, T.M., and Stamm, W.E. (1997). Diagnosis and treatment of uncompli-

cated urinary tract infection. Infect. Dis. Clin. North Am. 11, 551–581.

Jacob-Dubuisson, F., Heuser, J., Dodson, K., Normark, S., and Hultgren, S.

(1993). Initiation of assembly and association of the structural elements of

a bacterial pilus depend on two specialized tip proteins. EMBO J. 12, 837–847.

Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991). Improved

methods for building protein models in electron density maps and the location

of errors in these models. Acta Crystallogr. A 47, 110–119.

Kuehn, M.J., Heuser, J., Normark, S., and Hultgren, S.J. (1992). P pili in uropa-

thogenic E. coli are composite fibres with distinct fibrillar adhesive tips. Nature

356, 252–255.

Lazaridis, T., and Karplus, M. (1999). Effective energy function for proteins in

solution. Proteins 35, 133–152.

Lee, Y.M., Dodson, K.W., and Hultgren, S.J. (2007). Adaptor function of PapF

depends on donor strand exchange in P-pilus biogenesis of Escherichia coli.

J. Bacteriol. 189, 5276–5283.

Navaza, J. (2001). Implementation of molecular replacement in AMoRe. Acta

Crystallogr. D Biol. Crystallogr. 57, 1367–1372.

Remaut, H., Rose, R.J., Hannan, T.J., Hultgren, S.J., Radford, S.E., Ashcroft,

A.E., and Waksman, G. (2006). Donor-strand exchange in chaperone-assisted

pilus assembly proceeds through a concerted beta strand displacement

mechanism. Mol. Cell 22, 831–842.

Remaut, H., and Waksman, G. (2006). Protein-protein interaction through

beta-strand addition. Trends Biochem. Sci. 31, 436–444.

td All rights reserved

Structure

Crystal Structure of the P Pilus Subunit PapF

Remaut, H., Tang, C., Henderson, N.S., Pinkner, J.S., Wang, T., Hultgren, S.J.,

Thanassi, D.G., Waksman, G., and Li, H. (2008). Fiber formation across the

bacterial outer membrane by the chaperone/usher pathway. Cell 133,

640–652.

Roberts, J.A., Marklund, B.I., Ilver, D., Haslam, D., Kaack, M.B., Baskin, G.,

Louis, M., Mollby, R., Winberg, J., and Normark, S. (1994). The Gal(alpha

1-4)Gal-specific tip adhesin of Escherichia coli P-fimbriae is needed for pyelo-

nephritis to occur in the normal urinary tract. Proc. Natl. Acad. Sci. USA 91,

11889–11893.

Rose, R.J., Verger, D., Daviter, T., Remaut, H., Paci, E., Waksman, G.,

Ashcroft, A.E., and Radford, S.E. (2008a). Unravelling the molecular basis of

subunit specificity in P pilus assembly by mass spectrometry. Proc. Natl.

Acad. Sci. USA 105, 12873–12875.

Rose, R.J., Welsh, T.S., Waksman, G., Ashcroft, A.E., Radford, S.E., and Paci,

E. (2008b). Donor-strand exchange in chaperone-assisted pilus assembly

revealed in atomic detail by molecular dynamics. J. Mol. Biol. 375, 908–919.

Sauer, F.G., Futterer, K., Pinkner, J.S., Dodson, K.W., Hultgren, S.J., and

Waksman, G. (1999). Structural basis of chaperone function and pilus biogen-

esis. Science 285, 1058–1061.

Sauer, F.G., Pinkner, J.S., Waksman, G., and Hultgren, S.J. (2002). Chaperone

priming of pilus subunits facilitates a topological transition that drives fiber

formation. Cell 111, 543–551.

Sauer, F.G., Remaut, H., Hultgren, S.J., and Waksman, G. (2004). Fiber

assembly by the chaperone-usher pathway. Biochim. Biophys. Acta 1694,

259–267.

Saulino, E.T., Thanassi, D.G., Pinkner, J.S., and Hultgren, S.J. (1998). Ramifi-

cations of kinetic partitioning on usher-mediated pilus biogenesis. EMBO J.

17, 2177–2185.

Structure 16, 1724–17

Slonim, L.N., Pinkner, J.S., Branden, C.I., and Hultgren, S.J. (1992). Interactive

surface in the PapD chaperone cleft is conserved in pilus chaperone super-

family and essential in subunit recognition and assembly. EMBO J. 11,

4747–4756.

Soto, G.E., and Hultgren, S.J. (1999). Bacterial adhesins: common themes and

variations in architecture and assembly. J. Bacteriol. 181, 1059–1071.

Soto, G.E., Dodson, K.W., Ogg, D., Liu, C., Heuser, J., Knight, S., Kihlberg, J.,

Jones, C.H., and Hultgren, S.J. (1998). Periplasmic chaperone recognition

motif of subunits mediates quaternary interactions in the pilus. EMBO J. 17,

6155–6167.

Thanassi, D.G., Saulino, E.T., Lombardo, M.J., Roth, R., Heuser, J., and Hultg-

ren, S.J. (1998). The PapC usher forms an oligomeric channel: implications for

pilus biogenesis across the outer membrane. Proc. Natl. Acad. Sci. USA 95,

3146–3151.

Verger, D., Miller, E., Remaut, H., Waksman, G., and Hultgren, S. (2006).

Molecular mechanism of P pilus termination in uropathogenic Escherichia

coli. EMBO Rep. 7, 1228–1232.

Verger, D., Bullitt, E., Hultgren, S.J., and Waksman, G. (2007). Crystal structure

of the P pilus rod subunit PapA. PLoS Pathog. 3, e73.

Vetsch, M., Erilov, D., Moliere, N., Nishiyama, M., Ignatov, O., and Glock-

shuber, R. (2006). Mechanism of fibre assembly through the chaperone-usher

pathway. EMBO Rep. 7, 734–738.

Zavialov, A.V., Berglund, J., Pudney, A.F., Fooks, L.J., Ibrahim, T.M., MacIn-

tyre, S., and Knight, S.D. (2003). Structure and biogenesis of the capsular F1

antigen from Yersinia pestis: preserved folding energy drives fiber formation.

Cell 113, 587–596.

31, November 12, 2008 ª2008 Elsevier Ltd All rights reserved 1731