The activated ClpP peptidase forcefully grips a protein substrate

Abstract

ATPases associated with diverse cellular activities (AAA+) proteases power the maintenance of protein homeostasis by coupling ATP hydrolysis to mechanical protein unfolding, translocation, and ultimately degradation. Although ATPase activity drives a large portion of the mechanical work these molecular machines perform, how the peptidase contributes to the forceful denaturation and processive threading of substrates remains unknown. Here, using single-molecule optical trapping, we examine the mechanical activity of the caseinolytic peptidase P (ClpP) from Escherichia coli in the absence of a partner ATPase and in the presence of an activating small-molecule acyldepsipeptide. We demonstrate that ClpP grips protein substrate under mechanical loads exceeding 40 pN, which are greater than those observed for the AAA+ unfoldase ClpX and the AAA+ protease complexes ClpXP and ClpAP. We further characterize substrate-ClpP bond lifetimes and rupture forces under varying loads. We find that the resulting slip bond behavior does not depend on ClpP peptidase activity. In addition, we find that unloaded bond lifetimes between ClpP and protein substrate are on a timescale relevant to unfolding times (up to ∼160 s) for difficult to unfold model substrate proteins. These direct measurements of the substrate-peptidase bond under load define key properties required by AAA+ proteases to mechanically unfold and degrade protein substrates.

Significance

Energy-dependent proteases drive essential protein degradation to maintain cellular homeostasis and to rapidly regulate protein levels in response to changes in cellular environment. Using single-molecule optical tweezers, several studies demonstrate that the molecular process of degradation involves the mechanical unfolding and translocation of protein substrates by the ATP hydrolyzing enzyme component of these protease complexes. This study provides evidence that the chambered peptidase component of these molecular machines also contributes to the mechanical process of degradation by gripping the substrate under load in a manner independent of peptide hydrolysis. Our results suggest that the peptidase actively contributes to the biophysical mechanisms underlying processive protein degradation by energy-dependent proteolytic machines.

Introduction

AAA+ (ATPases associated with diverse cellular activities) proteases power protein degradation in the cell to eliminate damaged or misfolded proteins and control cellular processes by modulating protein levels. These proteolytic molecular machines comprise ATP-dependent, ring-shaped motor proteins (i.e., AAA+ protein unfoldases) that recognize, unfold, and translocate substrate protein into a self-compartmentalized peptidase (1,2). In Escherichia coli, the ClpP peptidase pairs with the AAA+ unfoldases ClpA and ClpX to form functional AAA+ proteases (3,4). Importantly, protein degradation is processive (i.e., the enzyme translocates along the protein substrate without dissociating until degradation is complete). Robust processivity requires that the probability of substrate dissociation is low and that subunits within the translocating machinery maintain enzymatic cycles out of phase such that polypeptide does not dissociate during substrate unfolding and translocation.

Single-molecule studies in the past decade have illuminated how the AAA+ proteases ClpXP and ClpAP function during processive protein unfolding and translocation. More specifically, optical trapping experiments combined with solution biochemistry revealed how AAA+ proteases generate force, coordinate ATPase cycles, grip the protein substrate, and translocate along the polypeptide track (5, 6, 7, 8, 9, 10, 11, 12, 13). Because the AAA+ unfoldase behaves as a motor protein (i.e., couples chemical energy in the form of ATP hydrolysis to physical translocation along a macromolecular track) and unfolds and translocates substrate protein in the absence of its proteolytic partner, the peptidase has been overlooked as contributing to the chemomechanical cycle of protein degradation by AAA+ proteases. In fact, there is little direct evidence that ClpP generates force during the process of protein degradation. However, the AAA+ motors show mechanical defects when ClpP is not present. For example, ClpA unfolds a dimeric substrate more slowly (14) and takes slower kinetic steps in the absence of ClpP (15,16). Moreover, ClpX, which grips the substrate via its pore-1 loops (10,11,17), slips on the substrate more often than ClpXP (10). Although differences in motor function can be partially explained by changes in ATPase activity in the presence of ClpP (18,19), we sought to test if ClpP participates mechanically during protein degradation since direct observation of ClpP mechanics has not been demonstrated. In addition, another group used atomic force microscopy to characterize the mechanics of the architecturally similar proteasomal 20S core particle engaging a model protein substrate in the absence of a AAA+ partner. The study concluded that the 20S active site threonine significantly contributes to observed high-force rupture events measured between peptidase and substrate (20).

Obtaining direct evidence of ClpP mechanical degradation has been complicated as ClpP poorly degrades large unfolded protein substrates in the absence of AAA+ motors due to gating by its N-terminal loops (21,22). AAA+ unfoldase binding activates ClpP by opening the axial pore and allowing the polypeptide to be threaded into the ClpP chamber for degradation. However, a class of natural products called acyldepsipeptides (ADEPs) activate ClpP in the absence of motor proteins (23,24), allowing the direct observation of protein degradation by ClpP (Fig. 1 A). ADEPs bind to the same hydrophobic pocket on ClpP to which motors bind (25) and are thought to activate ClpP in a similar manner, i.e., by causing rearrangement of the ClpP N-terminal loops (26). In cells, this activation leads to indiscriminate proteolysis and ultimately cell death, making ClpP a promising target for developing antibiotics that hamper pathogenic biofilm formation (27) and anticancer therapies due to ClpP’s conserved role in mitochondrial protein homeostasis (28,29). In addition, ADEPs disrupt bacterial cell division through the specific degradation of the protein FtsZ by ClpP (30). Furthermore, using purified proteins, ADEP-activated ClpP appears to unfold and degrade FtsZ in vitro without the need for motor proteins (31). Taken together, these data suggest that ClpP is capable of mechanically engaging and degrading the substrate in the absence of a AAA+ unfoldase.

Figure 1.

Measuring single-molecule ClpP mechanics by optical trapping. (A) Cartoon of ADEP1 activation of ClpP. ADEP1 (yellow) binds to the ClpP tetradecamer (blue), opening its central pore and allowing metastable protein substrates (orange) to enter and be degraded. (B) Schematic of optical trapping assays. ClpP is immobilized to one laser-trapped bead and engages a CM-titin substrate bound to a different bead using a 3500 bp DNA linker. (C) Example schematic of the flow cell used for optical trapping showing a top-down view. Solutions were prepared separately and flown into the channels as shown. Single substrate- and ClpP-coated beads were captured in channels 1 and 4, respectively. The stage was then moved to channel 3 where experiments were performed in the presence of 10 μM ADEP1.

Here, we aim to provide evidence of ClpP’s contribution to mechanical protein degradation in the absence of motor proteins. We hypothesize that ClpP grips and degrades substrate against external force, and that the active site serine contributes to substrate grip. Using single-molecule optical trapping, we do not observe denatured substrate translocation but demonstrate that ClpP grips the substrate against applied loads in excess of 40 pN when activated by ADEP1. We find that substrate-ClpP bond lifetimes and rupture forces decrease as external load increases consistent with slip bond behavior. We further show that active site inactivation does not significantly affect substrate grip and discuss what other portions of ClpP likely account for its mechanical behavior. To our knowledge, this study provides the first direct evidence that ClpP maintains a force-dependent grip on protein substrates without a motor protein and suggests additional activities that ClpP may contribute to ATP-dependent processive translocation and protein degradation outside of its peptidase activity.

Materials and methods

Biochemical purification of ClpP and substrate proteins

Full-length E. coli ClpP with a C-terminal hexahistidine tag and a terminal cysteine residue, and a substrate protein comprising an N-terminal HaloTag domain, four variant titin^I27 domains (V13P), a C-terminal hexahistidine tag, and an 11-amino acid ssrA degron were cloned and purified as described previously (32). In brief, ClpP was cloned into the pQE70 plasmid and expressed in JK-10 cells, which lack endogenous ClpP (33). Cells were initially grown to OD₆₀₀ ∼0.6 in LB broth at 30°C, cooled to 18°C, and induced with 0.5 mM IPTG for expression overnight. Cells were harvested and resuspended in lysis buffer (50 mM sodium phosphate [pH 8.0], 1 M NaCl, 5 mM imidazole, 10% glycerol), frozen in liquid nitrogen, and stored at −80°C. For purification, all the following steps were performed at 4°C unless noted otherwise. Cells were thawed and lysed with two passes through an Emulsiflex high-pressure homogenizer (Avestin, Canada). The lysate was clarified by centrifugation at 30,000 × g for 30 min. Clarified lysate was passed through an INDIGO-Ni (Cube Biotech, Germany) affinity column, washed, and eluted with lysis buffer containing 500 mM imidazole. Fractions were analyzed by SDS-PAGE, pooled, and concentrated to ∼1 mL. Concentrated protein was further purified by size-exclusion chromatography using a HiPrep 16/60 Sephacryl S300 HR column (Cytiva, Marlborough, MA, USA) equilibrated with storage buffer (50 mM Tris-HCl [pH 8.0], 150 mM KCl, 0.5 mM EDTA, 10% glycerol). Fractions were analyzed by 12% SDS-PAGE and appropriate fractions pooled, concentrated using an Amicon Ultra-15 10-kDa MWCO centrifugal filter (Millipore Sigma, Burlington, MA, USA), flash frozen, and stored at −80°C. ClpP was biotinylated at the terminal C-terminal cysteine using EZ-Link Maleimide-PEG2-Biotin (Thermo Scientific, Waltham, MA, USA). First a 20-mM stock of the biotin-maleimide was made in storage buffer and added to a final concentration of 20× molar excess to ClpP. The sample was left rotating at 4°C overnight and buffer exchanged into storage buffer containing 1 mM DTT before freezing with liquid nitrogen. Protein concentration was determined in storage buffer using ε₂₈₀ = 125,160 M⁻¹ cm⁻¹ for the ClpP tetradecamer.

The substrate protein was cloned into a pFN18A plasmid (Promega, Madison, WI, USA) and expressed in BL21(DE3) cells. Cells were grown to OD₆₀₀ ∼0.6 in LB broth at 37°C, cooled to 25°C, and induced with 1 mM IPTG for 3 h. Cells were harvested by centrifugation at 4000 × g for 15 min, resuspended in lysis buffer (50 mM sodium phosphate [pH 8.0], 500 mM NaCl, 10% glycerol, 10 mM β-mercaptoethanol, 20 mM imidazole) and flash frozen in liquid nitrogen for storage at −80°C. Lysis, clarification, and INDIGO-Ni affinity was performed as described above, eluting with 250 mM imidazole. Fractions were analyzed by 12% SDS-PAGE. Pure fractions were pooled, concentrated, flash frozen in small aliquots. Protein concentration was determined using ε₂₈₀ = 89,380 M⁻¹ cm⁻¹. For titin domain carboxymethylation, aliquots of substrate were first unfolded using 2 M guanidine-HCl at room temperature for 1.5 h. Then, a fresh stock of 0.5 M iodoacetic acid was added to a final concentration of 2.5 mM. After another 1.5-h incubation, the reaction was quenched by adding excess 1 M DTT to a final concentration of 10 mM. Samples were buffer exchanged into PD buffer (25 mM Hepes [pH 7.6], 100 mM KCl, 10% glycerol, 10 mM MgCl₂, 0.1% Tween-20) and flash frozen for storage at −80°C.

Single-molecule optical trapping of ClpP-substrate complexes

For trapping experiments, biotinylated ClpP and substrate were immobilized onto 1.25-μm streptavidin beads (Spherotech) in PD buffer supplemented with 1 mg/mL BSA (PD-BSA). For substrate, we constructed a 3500-bp linker with a 3′ 20 bp overhang from the M13mp18 plasmid (Bayou Biolabs) by PCR using these primers (Integrated DNA Technologies): TTTCCCGTGTCCCTCTCGA-T/idSp/TTGAAATACCGACCGTGTGA, and AATCCGCTTTGCTTCTGAC with 5′ biotin. The complement to the 20-bp overhang with sequence ATCGAGAGGGACACGGGAAA contained a 5′ phosphate and 3′ amine to which a HaloTag substrate was conjugated using a HaloTag succinimidyl ester O4 ligand (Promega). The 3500-bp DNA linker was ligated in the presence of CM-titin at room temperature for >1 h before conjugating to streptavidin beads.

All optical trapping data were collected using a dual-laser m-Trap Optical Tweezers system (LUMICKS, Amsterdam, the Netherlands) equipped with a five-channel laminar flow microfluidics device. Before experiments, the microfluidic chip was washed extensively with ddH₂O and equilibrated with PD-BSA for >30 min. ClpP and substrate beads were washed and resuspended in PD-BSA containing an oxygen scavenging system (0.25 mg/mL glucose oxidase, 0.03 mg/mL catalase, 3 mg/mL glucose; PD-BSA-OX). ClpP-bound beads and substrate beads were flown into the second and fifth channels, with the third and fourth containing PD-BSA-OX and PD-BSA-OX supplemented with 10 μM ADEP1 (Cayman Chemical, Ann Arbor, MI, USA), respectively. Custom Python scripts were written and used to automate bead capture, force ramp/clamp control, and to perform data analysis. For rupture force experiments, after capturing beads with each trap and forming a tether, the script moved the beads at a constant velocity until a rupture occurs. Similarly, for lifetime traces the script steers the trap until a defined force is reached, after which it pauses until the tether ruptures. Traces with incorrect contour lengths or multiple ruptures were discarded to avoid nonspecific interactions and multiple tethers.

Optical trapping data analysis

Data analysis for both lifetime and rupture force experiments were carried out using custom Python scripts with the LUMICKS Bluelake software. For rupture force measurements, force data were first downsampled to 15 Hz using a moving mean to match distance measurements. Then, rupture forces were found by using the first derivative of the force data. For lifetimes, the end of the force ramp and the terminal rupture were both reported using the first derivative of the force data. For lifetimes, data were downsampled further to 5 Hz, which was necessary to automate detection of the last point of the force ramp. Then, lifetimes were calculated by taking the difference between the two timepoints and the force was averaged over the duration of each lifetime.

After finding rupture forces, plotted data were fit to the Evans-Ritchie (34) and Dudko-Hummer-Szabo (35) models for molecular adhesions to extract the intrinsic time constant and distance to transition state. Where τ is the intrinsic lifetime, r is the loading rate, x^‡ is the distance to transition state, and v is a variable that represents the shape of the free energy barrier (1/2 for a cusp and 2/3 for a linear-cubic). From these fits, we also calculated the most probable rupture forces (F^∗). We fit the distance to transition state globally for both models as it varied little with the loading rate.

Evans-Ritchie: $P\left(F\right)={\tau }^{-1}{r}^{-1}{e}^{\left\{\frac{F\text{x}‡}{k_{B}T}-\frac{k_{B}T}{\tau r\text{x}‡}\left(e^{\frac{F\text{x}‡}{k_{B}T}}-1\right)\right\}}$

$$F^{\ast }=\frac{K_{B}T}{{\text{x}}^{‡}}\text{ln}\left(\frac{r{\text{x}}^{‡}\text{τ}}{k_{B}T}\right)$$

Dudko-Hummer-Szabo: $P\left(F\right)=r^{-1}\ast k\left(F\right)e^{\left(\frac{1}{\tau {\text{x}}^{‡}r}\right)}{e}^{-\left[\frac{k\left(F\right)}{{\text{x}}^{‡}r}\right]{\left[1-\left(\frac{vF{\text{x}}^{‡}}{\text{Δ}{\text{G}}^{‡}}\right)\right]}^{1-1/v}}.$

With $k\left(F\right)={\tau }^{-1}{\left(1-\frac{vF{\text{x}}^{‡}}{\text{Δ}{\text{G}}^{‡}}\right)}^{\frac{1}{v}-1}{e}^{\text{Δ}{\text{G}}^{‡}{\left[1-\left(1-\frac{vF{\text{x}}^{‡}}{\text{Δ}{\text{G}}^{‡}}\right)\right]}^{\frac{1}{v}}}.$

Biochemical degradation assays

FITC-Casein (Pierce Fluorescent Protease Assay Kit, Thermo Scientific) was prepared as directed to 5 mg/mL in ddH₂O and stored at −20°C. For experiments, reactions were made in PD buffer without FITC-Casein and incubated at 30°C for >10 min. After incubation, FITC-Casein was added to final concentration of 0.1 mg/mL and 50 μL reactions were pipetted into a 384-well plate (Greiner Bio-One, Kremsmünster, Austria). The fluorescence was tracked using a Biotek Syngergy HTX multi-mode plate reader (Agilent Technologies, Tempe, AZ, USA) every 30 s with excitation/emission wavelengths of 502/528 nm. Similarly, for CM-titin degradation, reactions were made in PD buffer without substrate and incubated at 30°C for >10 min, after which CM-titin was added to a final concentration of 2 μM and time started. Each time point taken was quenched with a final concentration of 2× SDS-PAGE assay buffer and flash frozen in liquid nitrogen. Samples were boiled at 95°C for 5 min and 12% SDS-PAGE followed by staining with Coomassie brilliant blue.

Results

Single-molecule mechanics of the ClpP peptidase

To probe the single-molecule mechanics of ClpP engaging an unfolded protein substrate, we used a dual-laser optical trap in passive mode (Fig. 1, B and C) without force feedback to maintain constant force (36). We immobilized biotinylated ClpP to one streptavidin-coated bead and a model multidomain substrate to separate streptavidin-coated beads. The substrate comprised a HaloTag domain at its N-terminus, which was conjugated to a biotinylated 3500-bp DNA linker, tandem repeats of a variant of the I27 domain of human titin (titin^I27) that were chemically denatured by carboxymethylating buried cysteine residues (CM-titin), and a C-terminal ssrA degron tag. For optical trapping experiments, we used a microfluidic device to introduce ClpP and substrate beads into a flow cell for staged assembly in laminar flow of the ClpP-substrate complex in the absence and presence of saturating concentrations of ADEP1 (Fig. 1 C). Since tethers rarely formed in the absence of ADEP1, ADEP1 was present in all optical trapping experiments. We use ADEP1 here under the reasonable assumption that ADEP binding faithfully mimics how AAA+ motors influence ClpP behavior as recent cryo-electron microscopy studies of ClpXP and ClpAP reveal nearly superimposable ClpP structures when bound to ADEPs or motors (37, 38, 39).

First, we measured the lifetimes of the interaction between ClpP and substrate at constant force at various applied loads (Fig. 2 A). The interactions between ClpP and substrate were extremely stable and did not show any translocation during the experimental timecourse up to 200–300 s. The distribution of lifetimes as a function of applied load followed a slip bond behavior and fit to the Bell model of force-dependent bond rupture between two molecules separated by a potential barrier (40) (Fig. 2 B). Specifically, we were interested in measuring the unloaded lifetime (τ₀ = 1/k) and the distance to the transition state (x^‡). From this fit, we obtained an unloaded lifetime (τ₀) of 158 ± 39 s and a distance to the transition state (x^‡) of 0.3 ± 0.1 nm (mean ± SE, N = 43). We note that this unloaded lifetime is much longer than the unfolding time constants of several substrate domains from single-molecule studies of ClpAP and ClpXP, which vary between 0.3 and 55 s for several variants of titin^I27, 0.03–3.4 s for filamin domains, and 9.1–19 s for GFP (5, 6, 7, 8,12,17,41). The ability of ADEP-ClpP to grip the substrate on timescales relevant for degradation suggests that ClpP could contribute to degradation by gripping substrates within the ClpAP/XP complexes thus preventing the substrate backslipping observed with the AAA+ motor in the absence of ClpP (5,6) or premature dissociation.

Figure 2.

ADEP1-ClpP forms long-lived interactions with protein substrate under load. (A) Example time course of ClpP interaction with CM-titin as a function of applied load. Data were downsampled to 700 Hz (gray) and 50 Hz (black). A constant speed force ramp was applied until the target force was reached, after which the trap position remained constant until tether rupture back to 0 pN. (B) Tether lifetimes as a function of applied load showing the mean ± SEM in both x and y (N = 8, 4, 8, 2, 9, and 3 for each force from low to high). The solid line is the fit to the Bell-Evans model for a slip bond (see materials and methods) yielding τ = 158 ± 39 s and x^‡ = 0.3 ± 0.1 nm (fit mean ± SE).

Because of the observed long timescales of the experiments above and lack of translocation, we also measured rupture force as a function of loading rate to probe the mechanical strength of the interaction between ClpP and substrate. A linear force ramp was applied to the ClpP-substrate tether until a terminal rupture occurred (Fig. 3 A). We observed a bimodal behavior in ClpP-substrate interactions, as demonstrated by the distributions of rupture forces under different loading rates (Fig. 3 B). To exclude the possibility of artifacts arising from tethering our enzyme-substrate complex via a DNA linker or from the experimental dual bead geometry, we examined the force-induced rupture of an oligo annealed to DNA to ensure the validity of our dual-bead assay. Our results using the same 3500-bp DNA linker with a 20-bp overhang with ligation were infrequent and occurred at forces different than ClpP-substrate tethers. Furthermore, the ruptures without ligation agreed with previously published literature on DNA shearing (42) and yielded different rupture forces than any of the rupture peaks observed for ClpP-substrate tethers, giving us confidence in our experimental design (Fig. S1).

Figure 3.

ADEP1-ClpP grips the substrate against external load. (A) Example rupture force traces of ADEP1-ClpP engaging CM-titin. Data were downsampled to 700 Hz (gray) and 50 Hz (black). Example traces are offset on the y-axis for clarity. A constant speed force ramp is applied until the interaction ruptures to 0 pN. (B) Violin plots of ClpP-substrate rupture forces in the presence of ADEP1 are shown at the indicated loading rates. Data points represent unique tethers with a terminal rupture to 0 pN. Vertical lines mark the median and quartiles of each distribution and N of each loading rate is shown. (C) Histograms of the rupture forces shown in (B). Fits to the Evans-Ritchie model (34) for each loading rate are shown as solid black lines along with fits to the Dudko-Hummer-Szabo model (35), where v = 1/2 (dashed lines) and v = 2/3 (dotted lines). (D) Most probable rupture force is plotted as a function of loading rate for wild-type ClpP-substrate interactions in the presence of ADEP1. Dashed lines are 95% CI of the fits. The most probable rupture forces shown are derived from fits of data in (C) to the Evans-Ritchie model. The data were fit to a semi log line with parameters b = 7.6 ± 7.3 pN and m = 13.4 ± 4.9 (fit mean ± SE).

For ClpP-substrate interactions, the rupture force distributions at different loading rates were fit to several models of bond rupture based on Kramer’s theory, the Evans-Ritchie model (34), and the Dudko-Hummer-Szabo model (35) (Fig. 3 C). From these fits, we obtained the thermal off rate (k), distance to the transition state (x^‡), and the free energy of activation (ΔG^‡) for the ClpP-substrate complex in the presence of ADEP1 (Table 1). For the Evans-Ritchie model, the fits for the unloaded lifetimes (τ₀ = 1/k) varied between 1 and 10 s depending on the loading rate, although this variability may represent the crossing of an energy barrier as seen for other molecular interactions (43,44). The Dudko-Hummer-Szabo model takes the shape of the transition-state surface into account, which can be either a cusp (v = 1/2) or linear-cubic (v = 2/3). The (τ₀) of these fits were similar to the Evans-Ritchie parameters, whereas the (x^‡) was similar for the linear-cubic and slightly larger for the cusp profile, 0.3 ± 0.1 and 0.4 ± 0.1 nm, respectively (mean ± SE). The free energy of activation varied with the loading rate yielding values between 3 and 5 k_BT (Table 1). Finally, we fit the most probable rupture forces calculated from the Evans-Ritchie model as a function of loading rate to a semi-log line with intercept (b) = 8.9 ± 5.2 pN and slope (m) = 12.6 ± 3.5 (Fig. 3 D). This slope represents the force sensitivity of the interaction as it is the force necessary to increase the dissociation rate k_off by e-fold.

Table 1.

Fits for rupture force distributions shown in Figs. 3 and 4

	Evans-Ritchie		Dudko-Hummer-Szabo
			v = 1/2			v = 2/3
Loading rate (pN/s)	t (s)	x‡ (nm)	t (s)	x‡ (nm)	ΔG (kT)	t (s)	x‡ (nm)	ΔG (kT)
ClpPwt
5	8.59 ± 1.44	0.22 ± 0.03	9.16 ± 1.90	0.37 ± 0.07	2.74 ± 0.57	8.97 ± 2.26	0.25 ± 0.07	2.93 ± 3.39
17	5.10 ± 1.16		6.88 ± 2.06		4.73 ± 2.13	5.89 ± 2.17		4.95 ± 7.33
40	2.11 ± 0.47		2.43 ± 0.77		3.19 ± 0.71	3.22 ± 1.56		2.93 ± 0.82
105	1.08 ± 0.27		1.42 ± 0.48		4.43 ± 1.59	2.24 ± 1.43		3.18 ± 0.86
ClpPS97A
5	9.68 ± 2.11	0.25 ± 0.04	10.49 ± 3.04	0.41 ± 0.10	3.19 ± 1.13	9.34 ± 2.91	0.26 ± 0.07	4.06 ± 8.33
17	6.60 ± 1.99		9.67 ± 4.27		5.47 ± 3.06	7.48 ± 4.33		unstable
40	2.90 ± 0.88		3.41 ± 1.66		3.19 ± 0.95	3.92 ± 2.34		2.96 ± 0.74
105	1.22 ± 0.38		1.56 ± 0.72		4.29 ± 1.67	2.78 ± 2.14		3.10 ± 0.81

Table of fit parameters to the Evans-Ritchie and Dudko-Hummer-Szabo models (see Materials and methods). Values shown are the best fit mean ± SE. ClpP-substrate bond rupture model parameters.

Contribution of the ClpP active site serine to substrate grip

Having observed the ability of ClpP to grip protein substrate under load, we asked if we could determine what domain of ClpP contributes to the peptidase’s mechanical behavior. We hypothesized that the ClpP active site serine might affect substrate grip through formation of a covalent intermediate, as observed for all serine proteases, by coordinating substrate binding, or through a combination of both mechanisms. Therefore, we first mutated the active site serine to alanine (ClpPS97A) to assess how the active site affects substrate grip. ClpPS97A inactivation was verified by monitoring the degradation of a fluorescently labeled unfolded substrate (FITC-Casein) and of our model CM-titin substrate (Fig. S2). In rupture force experiments, the shape of ClpPS97A distributions remained similar to wild-type ClpP (Fig. 4 A). We fit the distribution of rupture forces to the Evans-Ritchie and Dudko-Hummer-Szabo models, which yielded similar off rates, distance to the transition state, and free energy of activation as wild-type ClpP (Fig. 4 B and Table 1). Finally, we fit the most probable rupture forces as a function of loading rate for ClpPS97A and found that the slope and intercept also remained similar to wild-type ClpP. These data suggest that the ClpP active site serine does not contribute to substrate grip. In addition to the active site mutation, we used diisopropylfluorophosphate (DFP) to verify if the ClpP active site serine contributes to grip, as DFP chemically inactivates ClpP. We found that the shape of the distribution still remained similar to both wild-type ClpP and ClpPS97A at the tested loading rate (Fig. S2). This further supports the conclusion that the ClpP active site serine is not coupled to mechanical gripping of protein substrate. Using the parameters from the Dudko-Hummer-Szabo model, we recreated the transition-state energy landscapes of the observed ClpP-substrate interactions with energy wells represented as harmonic potentials (Fig. 5). We compared these with other measured protein-protein interactions (45) and found that the free energy of activation and distance to the transition state are comparable with the interaction between fluorescein and an antifluorescein antibody. Interestingly, the transition-state distance is also similar (∼0.7–1 nm) to those observed for force-dependent ClpXP translocation along the substrate polypeptide (5,11).

Figure 4.

ClpP active site inactivation does not affect substrate grip. (A) Violin plots of ClpPS97A-substrate rupture forces in the presence of ADEP1 are shown at indicated loading rates. Data points represent unique tethers with a terminal rupture to 0 pN. Vertical lines mark the median and quartiles of each distribution. (B) Histograms of the rupture forces shown in (A). Fits to the Evans-Ritchie model for each loading rate are shown as solid black lines, whereas the Dudko-Hummer-Szabo model fits with v = 1/2 (dashed lines) and v = 2/3 (dotted lines). (C) Most probable rupture force is plotted as a function of loading rate for wild-type ClpP- and ClpPS97A-substrate interactions in the presence of ADEP1. Dashed lines are 95% CI of the fits. The most probable rupture forces are derived from fits of data in (B) to the Evans-Ritchie model. For ClpPS97A, the data were fit to a semi log line with parameters b = 13.9 ± 6.4 pN and m = 10.5 ± 4.3 (fit mean ± SE).

Figure 5.

Free energy diagram of the ClpP-substrate interactions. Free energy diagrams were constructed according to the Dudko-Hummer-Szabo model (35) assuming a cusp shape (v = 1/2). Each curve represents a different interaction graphed in x until the distance to the transition state, at which point the curve stops. Both wild-type ClpP and ClpPS97A show a similar energy barrier to antifluorescein-fluorescein. Values for several other interactions were taken from (45,46) for comparison.

Discussion

Protein degradation by AAA+ proteases is an essential cellular process that requires the coordination of ATP-dependent motor proteins with their peptidase components. Based on single-molecule studies (1,2,5, 6, 7, 8,10,11), the AAA+ proteases ClpXP and ClpAP produce pN forces to unfold and translocate protein substrates. The motors grip the substrate using distinct pore loops (10,11,17) and drive polypeptide translocation through conformational changes between motor subunits (7,8,47). However, whether the peptidase, ClpP, aids mechanically to this reaction remained a question of interest. Here, using single-molecule optical trapping, we show the first evidence that ClpP grips an unfolded protein substrate against significant external load, in the absence of a AAA+ motor. The force-dependent lifetimes follow characteristic slip bond behavior (Fig. 2), with an average unloaded lifetime of 157 s. Notably, this is similar to ClpAP- and ClpXP-mediated unfolding lifetimes of difficult-to-unfold substrates, such as the wild-type titin^I27 domain (5, 6, 7, 8,10,12,17,41), and helps explain the trapping of substrates by proteolytically inactive ClpP variants in proteomic studies (48, 49, 50). We hypothesize that the observed force-dependent interaction between ClpP and substrate contributes to overall substrate grip during AAA+ protease degradation and likely aids in preventing slipping of protein substrates. Our hypothesis is consistent with experimental data demonstrating that addition of an unfolded region before the folded domain of GFP-ssrA increases the unfolding and degradation speed by ClpXP (51), that slipping events are readily observed during unfolding and translocation of substrate by ClpX in the absence of ClpP (5,6,10), and that a longer unstructured substrate tail capable of reaching into ClpP compensates for reduced substrate grip by pore 1-loop variants of ClpXP (10).

Furthermore, we characterized ClpP-substrate rupture forces and fit them to models of force-dependent protein-ligand interaction based on Kramers theory (34,35). The fits yield small distances to the transition state (x^‡ = 0.2–0.3 nm), whereas unloaded bond lifetimes (τ = 1/k₀) vary between 1 and 10 s depending on the loading rate (Fig. 3 and Table 1). Based on these parameters, we obtain a free energy of ClpP-substrate interaction, ΔG ≈ 4 k_BT, which is similar to estimates of work (∼5 k_BT) produced by ClpXP and ClpAP during a power stroke (5,12). Therefore, the energetics contributing to substrate grip by ClpP could provide a partial failsafe for the AAA+ protease to remain bound to a difficult-to-unfold substrate or under conditions of limiting ATP concentration, such as during stationary phase in bacteria (52). During these periods of slow growth due to unfavorable conditions, such as nutrient limitation, proteins could evade degradation by refolding and releasing before the motor has a chance to unfold and translocate (53).

Interestingly, we find that rupture forces distribute in a bimodal fashion. Several hypotheses account for bimodality. First, at least two populations arise from interactions of the substrate with distinct ClpP conformers, such as those observed in structural studies (54, 55, 56, 57, 58). Second, ClpP possesses two substrate binding regions or sites, each responding to external force differently. We conjectured that one site would be the active site since the protease reaction proceeds through a covalent intermediate (Fig. 4). Here, we show that mutating the active site serine to alanine did not significantly affect substrate grip by ClpP. Third, the N-terminal loops of ClpP present another candidate site mediating substrate grip, as they already play a role in substrate gating (21) and modulate the activity of the active site serine (22). Likewise, different populations could arise due to a combination of multiple ClpP subunits engaging the substrate. Finally, the bimodality could be caused by the substrate’s conformation as it enters the chamber of ClpP. For example, unfolded CM-titin possibly enters as or forms partially folded intermediates within ClpP that require greater force to rupture. Such structures are capable of being degraded by ClpXP as previous studies show that two polypeptide chains linked by a disulfide bond (4) and knotted protein substrates are degraded by ClpXP (59, 60, 61). Ultimately, ClpP grips the protein substrate in the absence of a AAA+ motor protein, although the molecular details defining ClpP grip require further study.

Substrate grip exhibited by ClpP might have important implications for ClpAP and ClpXP. For example, ClpA’s unfolding speed increases when in complex with ClpP in a manner not fully accounted for by ClpA’s ATPase activation when ClpP is present, i.e., sevenfold unfolding speed increase yet twofold ATPase increase with ClpP (14). In addition, a ClpX variant with unfolding defects is rescued when in complex with wild-type ClpP, ClpPS97A, and DFP-labeled ClpP (62). Furthermore, ClpX slips back farther and more frequently in the absence of ClpP at the single-molecule level (5,6,10,12). While the exact mechanisms are unclear, we hypothesize that substrate grip provided by ClpP helps ClpA and ClpX unfold substrate and suppress backslips by ClpX. Our data suggest that ClpP plays a more active role in degradation by aiding in a substrate grip that may result in increased degradation efficiency or unfolding speed by preventing reversible folding and premature release. Likewise, the ability of ClpP to grip protein substrates would be predicted to enhance processivity of these proteolytic enzymes by decreasing the probability of substrate release.

Processive degradation by AAA+ proteases prevents the release of partially degraded products that would be detrimental to cellular function, since these products could bind and inhibit protein partners or lead to aggregation and cell death. However, many AAA+ enzymes are weakly or not processive and do not need to fulfill their cellular functions. For example, spastin and katanin only partially unfold tubulin dimers to sever microtubules (63), NSF disassembles SNARE complexes without entirely unfolding and translocating individual SNARE proteins (64), and mitochondrial ClpX remodels the heme biosynthetic enzyme ALAS through partial unfolding (65). Despite similarities in structure among the various AAA+ unfoldases characterized to date (66), AAA+ proteases must possess some unique property to ensure processive translocation and degradation of substrates. While many studies of ClpXP and ClpAP focus on how the motor contributes to processivity, we propose that the combination of ATPase and peptidase makes a complete processive machine. The ability of the peptidase to perhaps act as a processivity factor is likely a general feature of AAA+ proteases, as results from Classen et al. show that the proteasomal 20S core particle also maintains protein substrate grip under load (20). However, the 20S active site threonine contributes to the observed mechanical behavior of the enzyme, which we do not observe here for the homologous ClpP active site serine. Therefore, there are likely key differences to how the peptidase components of AAA+ proteases contribute to overall mechanical degradation.

Author contributions

S.D.W. and A.O.O. designed research. S.D.W. performed research. S.D.W. contributed analytical tools. S.W. and A.O.O. analyzed data. S.D.W. and A.O. wrote the manuscript.

Editor: Doug Barrick.