IDL Unfolding Occurs at Unprecedented Forces.

There are several limitations inherent to the mechanical study of purified pili. Pili do not exclusively contain the SpaA shaft proteins, and measurements may be confounded by incorporated adhesins and other ancillary proteins (13). Moreover, the attachment of pili to the surface and cantilever relies on nonspecific interactions, such that the vector of the force is not explicitly defined. To overcome these limitations, we produced an engineered heteropolyprotein of the SpaA pilin with a C-terminal Cys for covalent attachment to gold-coated cantilevers, along with an N-terminal HaloTag protein that allows covalent tethering to functionalized glass surfaces (22, 23) (Fig. 3A). Such covalent anchorage ensures single-molecule tethers that can withstand forces close to 1 nN (23). Importantly, SpaA is pulled with the same geometry in the heteropolyprotein as in native pili.

Open in a separate window

Fig. 3.

Gram-positive pilins are extensible at unprecedented unfolding forces. Schematic representations of the SpaA polyprotein (A) and the FimA polyprotein (D), which include CnaA (A), CnaB (B), and N-terminal (N, SpaA only) domains. Red lines represent the IDL between the CnaA and CnaB domains (Fig. 1C). Force-extension recordings of the SpaA (B) and FimA (E) polyproteins with characteristic sawtooth unfolding patterns. Worm-like chain fits have 14-nm (red, IDL) or 29-nm (blue, I27) contour length increments. Histograms of the SpaA (C, n = 25) and FimA (F, n = 96) unfolding forces for the peaks that preceded the 14-nm contour length increments are shown. Black lines show Gaussian fits.

Initial attempts to make a synthetic pilus-like HaloTag-(SpaA)₄-Cys heteropolyprotein were complicated by poor protein expression. In turn, we flanked SpaA pilins with three I27 modules, which serve to enhance expression and also to provide an additional mechanical fingerprint (11) (Fig. 3 A and B). In force extension, we observed a distinct population of sawtooth peaks with 13.6 ± 0.4-nm contour length increments at a force of 525 ± 65 pN (Fig. 3C and Fig. S4A). These results with the engineered SpaA heteropolyprotein are consistent with the measurements on native SpaA-type pili and with our structure-based predictions, and confirm that unfolding in SpaA is limited to the 53 residues not protected behind isopeptide bonds.

Open in a separate window

Fig. S4.

Force-extension contour length increments identify unfolding of the IDL. (A) Schematic representation of the SpaA heteropolyprotein (above) and histogram of the WLC contour length increments in force extension (n = 154; additionally, two events at >80 nm are not shown in the histogram). The shapes in the schematic are matched to their respective populations, with SpaA divided between an N-terminal domain unfolding (N, gray triangle) and an IDL unfolding [red line between CnaA (A) and CnaB (B) domains]. Gaussians curves are fit to the data between contour length increments of 0–20 nm, 20–40 nm, 40–60 nm, or 60–80 nm (black lines). (B) Schematic representation of the FimA heteropolyprotein (above) and histogram of the WLC contour length increments in force extension (n = 128; additionally, 16 events at >30 nm are not shown in the histogram). The black line is a Gaussian fit to the data.

Of note, an additional population with unfolding increments of 51.4 ± 1.2 nm at a force of 71 ± 17 pN was observed (Figs. S4A and ​andS5),S5), which is in agreement with the predicted extension upon unfolding of the SpaA N-terminal domain (138 residues × 0.4 nm⋅residue⁻¹ − 3.1 nm of initial extension = 52.1 nm). This N-terminal domain is not subject to axial force in the native SpaA-type pilus, where force propagates from one SpaA subunit to the next via an intermolecular isopeptide bond with Lys190, which is located immediately distal to this domain (Fig. 1C). As such, the role of the N-terminal domain is likely tangential to the mechanical function of the pilus shaft.

Open in a separate window

Fig. S5.

Mechanical stability of the SpaA N-terminal domain. A histogram of the force of unfolding the 48- to 56-nm population of extensions in force extension of the SpaA polyprotein, which corresponds to the N-terminal domain, which falls outside of the axial force conduction pathway in the native pilus (n = 12), is shown.

To explore whether such mechanical features are a general feature of CnaA-type pilin domains, we expanded our study to the shaft FimA pilin of the type 2 pilus of A. oris (24). We synthesized a pilus-like FimA polyprotein construct of four FimA pilins in series, along with an N-terminal HaloTag and a C-terminal Cys for covalent tethering (23) (Fig. 3D). When extended at a constant velocity of 400 nm⋅s⁻¹, the FimA construct yields a sawtooth-like pattern in the force vs. extension trace, with four peaks of equal spacing (Fig. 3E). The FimA pilin unfolds at an extraordinarily high force of 690 ± 70 pN, with a contour length increment of 14.2 ± 0.6 nm (Fig. 3F and Fig. S4B). In the FimA crystal structure, 51 residues are unprotected by the isopeptide bonds, and the N and C termini are separated by 6.9 nm, giving a predicted unfolding extension of 15.5 nm (51 residues × 0.4 nm⋅residue⁻¹ + 2 isopeptides × 1.0 nm⋅isopeptide⁻¹ − 6.9 nm).

Thus, in both the SpaA and FimA shaft pilins, unfolding extensions are limited to the fraction of residues not protected behind isopeptide bonds. In the SpaA structure, 42 of the 53 unprotected residues fall in the IDL, with the remainder located before or after the isopeptides at the termini. In the FimA structure, 44 of 51 unprotected residues fall in the IDL. Moreover, the unprotected segments at the termini are already in extended or close to extended conformations in the SpaA and FimA crystal structures. Thus, it is most likely that the unfolding extensions and high mechanical stabilities owe largely to the unfolding of the IDLs of SpaA and FimA.

SpaA and FimA Refold to a Mechanically Weak Intermediate.

The addition of internal covalent bonds has been demonstrated to accelerate the rate of folding (20, 25). We therefore asked whether the IDL, in addition to its high mechanical stability, had rapid refolding kinetics.

To measure SpaA and FimA refolding, we applied denature–quench–probe protocols using AFM-based force spectroscopy in “force-clamp” mode (26, 27). On the denature pulse, a high force (≥350 pN) is applied to unfold the polyprotein, with individual events observed as ∼12.5-nm steps in the extension vs. time trace. Next, a quench pulse with the force set at 0 pN allows for domain refolding. Domains that refold during the quench pulse are detected by their subsequent unfolding during the probe pulse. In a representative recording of FimA refolding, four steps are observed in the initial unfolding pulse and three steps are observed in the subsequent probe pulse, indicating that three of four domains refolded during the interspaced 50-ms quench pulse (Fig. 4A). Similarly, in a representative recording of the SpaA polyprotein, both SpaA domains unfold on the denature pulse and again on the probe pulse, having refolded during the 200-ms quench pulse (Fig. 4C). Unfolding of the I27 modules can be observed as 25-nm steps, although the kinetics of I27 refolding are slower and take several seconds (20).

Open in a separate window

Fig. 4.

FimA and SpaA refold to mechanically weaker intermediates. (A) Representative force-clamp recording of the FimA polyprotein showing a high-force unfolding pulse, followed by a 50-ms low-force refolding pulse (ΔtQ) and a second high-force unfolding pulse. FimA IDL unfolding is identifiable as ∼12.5-nm stepwise length increments (red slashes). (Scale bar, 1 s.) (B) Faded traces show individual FimA unfolding traces on a 500-pN denature pulse (red, 27 recordings) and on a 500-pN probe pulse (blue, 20 recordings), with the normalized sums depicted in solid colors. All recordings on the probe pulse follow a 1-s force quench. (C) Representative force-clamp recording of the SpaA polyprotein. Two forces are applied on both the denature and probe pulses, an initial low-force pulse (165 pN) to unfold the I27 domains and a later high-force pulse (430 pN) to unfold the SpaA pilins. I27 unfolding is identifiable as ∼25-nm stepwise length increments (blue circle), and SpaA unfolding is identifiable as ∼12.5-nm stepwise length increments (red slashes). (Scale bar, 4 s.) Steps identified with asterisks correspond to entropic increases in the length of the polyprotein due to the change in the applied force. (D) Faded traces show individual SpaA unfolding traces on a 430-pN denature pulse (red, 11 recordings) and on a 430-pN probe pulse (blue, 20 recordings), with the normalized sums depicted in solid colors. All recordings on the probe pulse follow a force quench of ≤1 s.

We summed 47 recordings of FimA unfolding at 500 pN on denature pulses or on probe pulses (Fig. 4B). Unfolding events occur more rapidly on the probe pulse than on the denature pulse, suggesting a decreased mechanical stability of the refolded domains. To measure the respective unfolding rates, we summed all of the extension vs. time recordings and then fit this sum with a single exponential. FimA unfolds at a rate of 0.58 ± 0.12 s⁻¹ on the denature pulse at 500 pN, and at a rate of 17.43 ± 2.37 s⁻¹ on the probe pulse at 500 pN (Fig. 4B). A similar phenomenon was observed for SpaA, which unfolds at a rate of 1.63 ± 0.41 s⁻¹ on the denature pulse at 430 pN, and at a rate of 30.38 ± 7.98 s⁻¹ on the probe pulse at 430 pN (Fig. 4D). This discrepancy in unfolding rates between the denature pulse and the probe pulse indicates that both FimA and SpaA pilins exist in a weaker state immediately upon refolding.

We resolved the refolding kinetics of both pilins by varying the time of the quench pulse, ΔtQ, and counting the number of steps recovered on the probe pulse. Refolding of SpaA and FimA proceeds rapidly, approaching completion after 100 ms. However, refolding is only observed in proteins that collapse from their mechanically extended state (Fig. S6A). Only after a polypeptide collapses beyond ∼60% of its unfolded length does protein folding become appreciable (Fig. S6 B and D). We have proposed before that mechanical folding of proteins is a two-step process, where actual folding is preceded by the hydrophobic collapse of the extended polypeptide (28, 29). We therefore considered the time course of refolding in only those traces that collapsed more than 60% of their unfolded lengths. Beyond that threshold, the FimA pilin refolds at a rate of at least 17 s⁻¹, whereas the SpaA pilin refolds at a rate of at least 14 s⁻¹ (Fig. S6 C and E). These rates are lower bound estimates of the actual folding rate, because our measurements are limited by the time it takes for the polypeptide to collapse. Even so, these refolding rates are one order of magnitude faster than other mechanical proteins studied using force spectroscopy, such as the titin I27 domain and ubiquitin (20, 29). Internal isopeptide bonds are likely critical to this rapid refolding; indeed, refolding accelerates when internal covalent bonds are engineered into proteins (20, 25).

Open in a separate window

Fig. S6.

Polypeptide collapse is rate-limiting for pilin refolding. (A) Two force-clamp recordings of the FimA polyprotein, with 0-pN force quench pulses of variable durations (ΔtQ). With longer quench pulses, the polypeptide has more time to collapse and domains have more time to refold, seen as steps on the subsequent high-force pulse. At the same quench pulse of 50 ms, the polyprotein that collapses by 71% of its total extended length exhibits refolding (Right), whereas the polyprotein that collapses by 49% does not (Left). The percentage of collapse is calculated as the maximum distance of collapse (indicated by the black arrows) divided by the total extended length. (B) FimA refolding as a function of the percentage of tether collapse at low force. Error bars represent the SEM (n ≥ 20 for each data point). (C) Time course of FimA refolding, with a single exponential fit (black line) to the aggregated refolding data for traces with at least 60% tether collapse. Asterisks indicate time points with fewer than five events in the measurement. These time points were excluded from the fitting of the data. All other time points represent >40 events. Error bars represent the SEM. (D) SpaA refolding as a function of the percentage of tether collapse at low force. Error bars represent the SEM (n ≥ 17 for each data point). (E) Time course of SpaA refolding, with a single exponential fit (black line) to the aggregated refolding data for traces with at least 60% tether collapse. All time points represent ≥16 events. Error bars represent the SEM.

Maturation of CnaA Domains to the Native Fold Is Slow.

The native unfolding on the denature pulse and the weaker unfolding on the probe pulse are each observed as individual ∼12.5-nm steps in force clamp. To differentiate between the two populations, we implemented refolding experiments using a linear increasing ramp in the force (100 pN⋅s⁻¹), or “force ramp.” This technique allows us to discern the relative mechanical stability of domains in the denature and probe pulses by measuring the force at which they unfold on the ramp (27). In a representative recording of a force-ramp refolding experiment, a single FimA polyprotein is subject to three unfolding/refolding cycles alternating 1-s and 20-s quench pulses (Fig. 5A). Four ∼12.5-nm steps are observed on the denature pulse and on each subsequent probe pulse, indicating 100% refolding on the three quench pulses. In this trace, mechanical stability of FimA appears to depend on the duration of the quench pulse. All four events on the denature pulse are seen at >500 pN, whereas all four events occur at <500 pN after a 1-s quench pulse. With a 20-s quench pulse, two of four events unfold at forces above 500 pN, yet all return to the weak state after a subsequent 1-s quench pulse.

Open in a separate window

Fig. 5.

Pilin recovery of native-like mechanical stability with time or ligand binding. (A) Representative trace showing unfolding force-ramp pulses separated by refolding pulses. (Bottom) Spikes in the force vs. time trace indicate the forces of IDL unfolding. Events above 500 pN are colored in red, and events below 500 pN are colored in blue. (B) Histograms of the force-ramp FimA unfolding forces on the first unfolding force pulse and after refolding pulses of variable durations. Events above 500 pN are colored in red, and events below 500 pN are colored in blue. The black lines represent fits to a double Gaussian function (n = 16 at the 120-s time point, n ≥ 70 at other time points). (C) Histograms of the force-ramp unfolding forces on the initial denature ramp (gray) and on the probe ramp (black) following a 1-s quench pulse for both the SpaA and FimA pilins. Experiments were carried out in standard AFM buffer with 1 mM EDTA (Top) or were supplemented with 10 mM CaCl₂ (Bottom). Black lines indicate Gaussian fits to the data (n ≥ 59 in each histogram).

We aggregated the data from 135 recordings to explore the kinetics of maturation of the mechanical stability of FimA. FimA unfolding steps occur at 563 ± 46 pN on the denature force ramp, which drops to 438 ± 38 pN on the probe force ramp following a 1-s quench pulse (Fig. 5B). The fraction of events that occur at high force increases with the duration of the quench pulse (Fig. 5B). Considering these two populations of stabilities as two states of FimA, a native strong state that is present on the initial denature pulse and an intermediate weak state only present on the probe pulse, we modeled a first-order kinetic process for the maturation of the FimA intermediate state at a rate of 0.011 s⁻¹ (Fig. S7B). Similarly, SpaA unfolding events occur at a higher force on a denature force ramp, 453 ± 46 pN, than on a probe force ramp after a ΔtQ of 1 s, 369 ± 41 pN (Fig. 5C). We also observed a time-dependent maturation of the mechanical stability of SpaA, although at a slower rate than FimA (Fig. S7 A and B).

Open in a separate window

Fig. S7.

Kinetics of SpaA and FimA maturation. (A) Histograms of the force-ramp SpaA unfolding forces on the first unfolding force pulse and after refolding pulses of variable durations. Events above 400 pN are colored in red, and events below 400 pN are colored in blue. The black lines represent fits to a double Gaussian with centroids and widths that are fixed based on the distribution of unfolding forces on the initial denature force ramp and on the 1-s refolding force ramp (n ≥ 10 events at all time points). (B) Relative fraction of domains unfolding with “native-like” mechanical stability on a force-ramp probe pulse of 100 pN⋅s⁻¹ following force quench pulses of variable duration (“Time” axis, logarithmic scale). The maturation of FimA domains (blue triangles) is fit to a single exponential kinetic with a rate constant of 0.011 s⁻¹. The maturation of SpaA domains (red diamonds) is fit to a single exponential kinetic with a rate constant of 0.005 s⁻¹. All FimA data points represent at least 16 events, and all SpaA data points represent at least 10 events. Error bars depict the SD.

Construction of SpaA-K199A.

SpaA carrying a K199A mutation was constructed by using inverse PCR. Using pHTT93 coding for SpaA WT as a template, a Phusion-based PCR was performed by using a phosphorylated forward primer containing the K199A mutation (5′-GCAACTGCTGTGGATCCGGATGCC-3′) and a phosphorylated reverse primer (5′-CACCGGCTCAGACAAAGCCTGGTG-3′). The 8.4-Kb PCR product was purified and ligated before being transformed into Escherichia coli DH5α. The DNA was sequenced to confirm the K199A mutation before being transformed into C. diphtheriae (ΔspaA/spaB/spaC/spaG).

Protein Expression and Purification.

The genes for monomeric SpaA and FimA were synthesized with codons optimized for expression in E. coli. Genes for the SpaA and FimA polyproteins were constructed via a multistep cloning process involving BamHI, BglII, and KpnI restriction sites (29). The genes were then cloned into a pFN18a vector with an N-terminal HaloTag gene and a C-terminal His₆-Cys sequence (23). The pFN18a vector was cloned into E. coli ERL or BLR(DE3) cells, and protein expression was induced with 1 mM isopropyl β-d-1-thiogalactopyranoside, overnight at 25 °C. Cells were pelleted and lysed by French press in 50 mM sodium phosphate (pH 7.0), 300 mM NaCl, 10% (vol/vol) glycerol, and 0.5 mM DTT buffer. Protein purification was achieved using a Ni²⁺-NTA His GraviTrap affinity column (GE Healthcare), and a gel filtration step in a Superdex 200 FPLC column (GE Healthcare) in 10 mM Hepes (pH 7.2), 150 mM NaCl, 1 mM EDTA, and 10% (vol/vol) glycerol buffer.

Single-Molecule AFM.

All single-molecule AFM experiments were conducted using a custom-built AFM setup, as described previously (26), or on a commercially available atomic force microscope (Luigs and Neumann AG). Cantilevers were purchased from Brucker (model MLCT), with a typical spring constant of 13–16 pN⋅nm⁻¹. Experiments were carried out at room temperature in the AFM buffer of 10 mM Hepes (pH 7.2), 150 mM NaCl, and 1 mM EDTA. For experiments with purified pili, aliquots of ∼5–10 μL were deposited directly onto freshly evaporated gold coverslips and allowed to stand for ≥15 min before washing with AFM buffer and the start of experimentation. For experiments with purified polyproteins, coverslips were instead functionalized with HaloLigand (Promega), as described previously (26), and gold was evaporated onto the cantilever tips. Aliquots of 20–40 μL were deposited onto the functionalized coverslips and allowed to stand for ≥30 min before washing with AFM buffer. In force-extension experiments, the cantilever was first pushed against the surface to a force of 2 nN, before being retracted at a rate of 400 nm⋅s⁻¹. In force-clamp and force-ramp experiments, the cantilever was first pushed against the surface to a force of 2 nN and held for 1–2 s. During retraction, the force on the cantilever was held to a set value via proportional-integral-derivative feedback on the piezoelectric AFM stage with a feedback response of <5 ms. In all force-ramp experiments, the force was linearly increased at 100 pN⋅s⁻¹ to a maximal value of 650 pN. For experiments on the calcium-dependent maturation of the pilins, the AFM sample buffer was supplemented with 10 mM CaCl₂.

This work was funded by NIH Grants HL66030, HL61228 (to J.M.F.), and {"type":"entrez-nucleotide","attrs":{"text":"AI106072","term_id":"3461175","term_text":"AI106072"}}AI106072 (to J.A.-C.); by National Institute of Dental and Craniofacial Research Grants DE017382 and {"type":"entrez-nucleotide","attrs":{"text":"DE025015","term_id":"62268485","term_text":"DE025015"}}DE025015 (to H.T.-T.); and by Marie Curie International Incoming Fellowship FP7-PEOPLE-2010-COFUND-267149 (to J.A.-C.).