Multistep substrate binding and engagement by the AAA+ ClpXP protease

Significance

AAA+ proteases play key regulatory and quality control roles in all domains of life. These destructive enzymes recognize damaged, unneeded, or regulatory proteins via specific degrons and unfold them prior to processive degradation. Here, we show that Escherichia coli ClpXP, a model AAA+ protease, recognizes ssrA-tagged substrates in a multistep binding and engagement reaction. Together with recent cryo-electron microscopy (cryo-EM) structures, our experiments reveal how ClpXP transitions from a machine that checks potential substrates for appropriate degrons to one that can unfold and translocate almost any protein. Other AAA+ proteases in organelles and bacteria are likely to use similar mechanisms to specifically identify and then destroy their target proteins.

Abstract

Escherichia coli ClpXP is one of the most thoroughly studied AAA+ proteases, but relatively little is known about the reactions that allow it to bind and then engage specific protein substrates before the adenosine triphosphate (ATP)-fueled mechanical unfolding and translocation steps that lead to processive degradation. Here, we employ a fluorescence-quenching assay to study the binding of ssrA-tagged substrates to ClpXP. Polyphasic stopped-flow association and dissociation kinetics support the existence of at least three distinct substrate-bound complexes. These kinetic data fit well to a model in which ClpXP and substrate form an initial recognition complex followed by an intermediate complex and then, an engaged complex that is competent for substrate unfolding. The initial association and dissociation steps do not require ATP hydrolysis, but subsequent forward and reverse kinetic steps are accelerated by faster ATP hydrolysis. Our results, together with recent cryo-EM structures of ClpXP bound to substrates, support a model in which the ssrA degron initially binds in the top portion of the axial channel of the ClpX hexamer and then is translocated deeper into the channel in steps that eventually pull the native portion of the substrate against the channel opening. Reversible initial substrate binding allows ClpXP to check potential substrates for degrons, potentially increasing specificity. Subsequent substrate engagement steps allow ClpXP to grip a wide variety of sequences to ensure efficient unfolding and translocation of almost any native substrate.

AAA+ proteases play important roles in quality control, protein homeostasis, and cell-cycle regulation by degrading intracellular proteins that are incomplete, damaged, unnecessary, or that repress responses to environmental or developmental signals (1, 2). These adenosine triphosphate (ATP)-fueled proteases consist of an AAA+ ring hexamer and a self-compartmentalized peptidase. Recognition and engagement of the correct protein substrates are the critical initial steps in proteolysis, as subsequent steps have little specificity. In eubacteria and eukaryotic organelles, an unstructured peptide degron at a protein terminus usually targets it for degradation. For the Escherichia coli ClpXP protease (3), for example, the AAA+ ClpX ring hexamer binds an unstructured C-terminal or N-terminal degron of a target protein within its axial channel, unfolds any native structure by repeatedly applying force, and then translocates the denatured polypeptide through the channel and into the degradation chamber of ClpP (Fig. 1A). Protein unfolding and translocation by ClpXP have been visualized by single-molecule methods (4–10), but these experiments did not capture the initial steps of substrate binding and engagement, which remain poorly characterized.

Fig. 1.

Labeling, purification, and degradation of titin substrates by ClpXP. (A) Cartoon of substrate binding, unfolding, translocation, and degradation by ClpXP. (B) Single-chain ClpX was labeled at residue 170 of one subunit with a fluorescent TAMRA dye, and a cysteine between native titin and the ssrA tag was labeled with a black-hole quencher. (C) Following labeling of the titin-ssrA construct with BHQ10 maleimide, the modified substrate was separated from unmodified protein by ion-exchange chromatography. (D) Room temperature degradation of different substrates (5 µM) by ^FSClpX (200 nM) and ClpP (600 nM) was performed in the presence of ATP (5 mM) and the extent of degradation was then assayed by sodium dodecyl sulfate polyacrylamide electrophoresis (SDS/PAGE), followed by staining with Coomassie Blue and densitometry. Values are averages of two experiments ± SEM. Rates determined from linear fits were 12 ± 0.8 nM min⁻¹ (titin-BQ-ssrA), 220 ± 14 nM min⁻¹ (^V13Ptitin-BQ-ssrA), and 640 ± 40 nM min⁻¹ (^UFtitin-BQ-ssrA).

The best-studied degron for E. coli ClpXP is the 11-residue ssrA tag, which is cotranslationally appended to the C terminus of a protein by the ubiquitous transfer-messenger RNA system when a bacterial ribosome stalls prior to completing synthesis (11, 12). Recombinant addition of this tag to almost any polypeptide or protein makes it a substrate for ClpXP degradation (13–19). ATP or ATPγS must bind to the AAA+ ring of ClpX to support binding to the ssrA tag, but hydrolysis of these nucleotides is not required for ssrA tag recognition (15, 20–24). During degradation of multidomain substrates, ClpXP often releases a partially degraded product when a very stable native domain is encountered, and single-domain substrates are also likely to be released multiple times prior to degradation (16, 19, 25–28).

Our current understanding of substrate binding and engagement by ClpX or ClpXP is relatively primitive, in part because previous binding assays have not been compatible with rapid kinetic studies. Here, we describe a robust fluorescence-quenching assay that allows dissection of these important reactions and provide evidence for a model in which an initial recognition complex (RC) is converted into an intermediate complex (IC) and then, an engaged complex (EC) prior to substrate unfolding and processive translocation/degradation. This model suggests that ClpXP initially checks potential substrates for appropriate degrons and then transitions to a machine that can unfold, translocate, and degrade a wide range of proteins. We discuss these studies in light of recent cryo-electron microscopy (cryo-EM) structures of ClpXP bound to protein substrates (28–30).

Results

Binding Assayed by Fluorescence Quenching.

To monitor binding of ClpX or ClpXP to an ssrA-tagged substrate by Förster resonance energy transfer (FRET), we labeled ClpX with a fluorescent tetramethylrhodamine (TAMRA) dye and modified an ssrA-tagged substrate containing the I27 domain of human titin by attaching a quencher (Fig. 1B). The ClpX variant was a single-chain enzyme in which each ClpX subunit had its N domain deleted, and subunits were linked by short genetically encoded tethers to form a pseudohexamer. We call this variant ^FSClpX, where F signifies fluorescent and S indicates single-chain ClpX^∆N. Single-chain ClpX pseudohexamers do not dissociate at low concentrations and support degradation of ssrA-tagged substrates by ClpP (31). One subunit of the ClpX pseudohexamer contained a D170C substitution, which represented the only solvent-exposed cysteine and allowed labeling with a fluorescent TAMRA maleimide derivative. The quencher-labeled substrate contained a native titin domain, a C-terminal ssrA tag, and an intervening cysteine to which we attached a black-hole BHQ10 quencher (titin-BQ-ssrA). After labeling with the quencher, we separated the more negatively charged quencher-modified protein from unlabeled protein by anion-exchange chromatography (Fig. 1C). Native titin is a mechanically stable protein that is difficult for ClpX to unfold and degrade (18). We also prepared and purified ^V13Ptitin-BQ-ssrA, which is less stable than the parental titin substrate, and ^UFtitin-BQ-ssrA, a variant unfolded by carboxymethylation of otherwise buried cysteines (18). As assayed by SDS/PAGE, ^FSClpXP degraded titin-BQ-ssrA very slowly, ^V13Ptitin-BQ-ssrA more rapidly, and ^UFtitin-BQ-ssrA at the fastest rate (Fig. 1D).

We titrated increasing concentrations of titin-BQ-ssrA against a fixed concentration (20 nM) of ^FSClpXP (Fig. 2A) or ^FSClpX (Fig. 2B) in the presence of saturating ATP or ATPγS, which ClpX or ClpXP hydrolyzes more slowly than ATP (21); determined the degree of quenching after fluorescence had stabilized but before significant degradation could occur; and fit the resulting binding data to a 1:1 model to determine K_1/2 values for half-maximal binding. These experiments are not at equilibrium because substrate would eventually dissociate after ATP or ATPγS was depleted by ClpXP hydrolysis. Nevertheless, the rates of ATP/ATPγS hydrolysis in our experiments are slow enough compared with the pool size that the nucleotide concentrations remain saturating, so concentrations of substrate-bound and free ClpXP change only as a function of the free substrate concentration, as they would in an equilibrium experiment.

Fig. 2.

Assays of protein substrate or nucleotide binding to ^FSClpXP or ^FSClpX. (A) Binding of titin-BQ-ssrA to ^FSClpXP (20 nM) in the presence of ATP (5 mM) or ATPγS (1 mM). Values are averages ± SD of three independent experiments, and the fits are to a quadratic tight binding model. (B) Same as A but using ^FSClpX. (C) Same as A but using titin-AQ-ssrA. (D) Activation of titin-BQ-ssrA binding to ^FSClpXP by addition of ATP or ATPγS. Values are averages ± SD of three experiments, and the fits are to the Hill equation (n is the Hill constant).

Higher maximal quenching was observed in ATPγS compared with ATP for both ^FSClpXP and ^FSClpX (Fig. 2 A and B). Quenching by FRET is sensitive to small changes in average distances near the Förster radius (32). Thus, relative to ATP, ATPγS may result in ClpX–substrate structures with a closer average distance between the fluorophore and quencher. Alternatively, ATP and ATPγS could result in differences in the fraction of ClpX enzymes capable of binding substrate, even at saturating concentrations. To distinguish between these possibilities, we modulated the quenching distance sensitivity by labeling the titin substrate with ATTO 575Q (AQ), which has a larger Förster radius when paired with TAMRA (∼62 Å) than does BHQ10 (∼46 Å). If higher quenching in ATPγS resulted from a higher fraction of binding-competent ClpX enzymes, then higher quenching should also be observed for the titin-AQ-ssrA substrate. However, in titration experiments, titin-AQ-ssrA binding to ^FSClpXP resulted in similar fluorescence quenching in the presence of ATP or ATPγS (Fig. 2C). Thus, ATP-bound ClpXP likely holds the ssrA-tagged substrate in a slightly different average conformation than ATPγS-bound ClpXP, a difference that can be resolved by the shorter-radius BHQ10–TAMRA pair but not the longer-radius AQ–TAMRA pair. The conformations of individual states in these ensembles may be identical for both nucleotides, but slower hydrolysis of ATPγS could, for example, result in a higher population of a conformation in which the fluorophore and quencher are closer together.

Binding of the titin-BQ-ssrA substrate to ^FSClpXP saturates at ATP or ATPγS concentrations that are at least 20-fold below those used for the titration experiments (Fig. 2D). ATPγS supports binding at lower concentrations than ATP and with modestly reduced positive cooperativity as measured by the Hill constant of the fit (Fig. 2D). Positive cooperativity in the ATP activation of substrate binding to a ClpX mutant defective in ATP hydrolysis was reported previously (23) and suggests that multiple subunits in the ClpX hexamer must bind ATP to stabilize the substrate binding conformation. In accord with this model, four or five subunits of the ClpX hexamer are ATP or ATPγS bound in substrate-bound cryo-EM structures (28–30). The finding that ATPγS supports substrate binding at lower concentrations than ATP and with a lower Hill constant suggests that faster hydrolysis of ATP at subsaturating concentrations results in a lower population of nucleotide-bound intermediates that can bind substrate.

Kinetics of Association and Dissociation.

Stopped-flow experiments were used to study association of ClpXP with ssrA-tagged substrate. We used a fixed concentration of ^FSClpXP and increasing concentrations of titin-BQ-ssrA to initiate binding reactions in the presence of ATP (Fig. 3A). The resulting fluorescence trajectories were polyphasic, indicating multiple reaction species. At high substrate concentrations, we observed two initial phases of decreased fluorescence, indicating elevated quenching. Fluorescence then increased in a third phase as the reaction approached steady state.

Fig. 3.

Association kinetics. (A) ^FSClpXP (final concentration 10 nM) and ATP (final concentration 2.5 mM) were mixed with the indicated concentrations of titin-BQ-ssrA to initiate binding, which was assayed by changes in fluorescence. (B) Same experiment as in A but using ATPγS (final concentration 0.5 mM). (C) Same experiment as in A but using the titin-AQ-ssrA substrate. In A and B, the black lines represent global fits to the first three steps of the model shown in Fig. 5.

In association experiments using ATPγS (Fig. 3B), the initial substrate binding phase had kinetics similar to the reaction in ATP, but the second phase of decreased fluorescence and third phase of increased fluorescence were slower. Association in ATPγS also resulted in lower final fluorescence than ATP, consistent with the differences observed in maximal quenching in steady-state binding experiments (Fig. 2A).

If the polyphasic titin-BQ-ssrA association trajectories reflect multiple substrate-bound ClpXP species, each with a different distance or orientation between the fluorophore and quencher, then fewer phases should be observed with titin-AQ-ssrA as a consequence of its larger Förster radius. An alternative possibility is that the fluorescence trajectories reflect distinct populations of ClpXP that bind and release titin-BQ-ssrA at different rates. To distinguish between these possibilities, we conducted stopped-flow experiments with titin-AQ-ssrA, as TAMRA quenching by AQ should be less sensitive to the fluorophore–quencher distance at relevant length scales, and different bound states are thus more likely to quench with similar efficiencies. Rapid mixing of ^FSClpXP and titin-AQ-ssrA caused a monotonic decrease in fluorescence (Fig. 3C), suggesting that the more complex fluorescence trajectories observed with titin-BQ-ssrA reflect differences in quenching among different substrate-bound ClpXP species that interconvert rather than distinct populations of ClpXP that bind and release substrates at different rates.

A model with multiple species of ClpXP bound to titin-BQ-ssrA, each with distinct quenching properties, also predicts that the kinetics of dissociation should be polyphasic. Thus, we allowed complexes of titin-BQ-ssrA with ^FSClpXP to form in ATP or ATPγS for at least 20 min, added a large excess of unlabeled titin-ssrA to prevent reassociation of titin-BQ-ssrA after dissociation, and then recorded changes in fluorescence (Fig. 4). A biphasic increase in fluorescence was observed in the presence of ATP or ATPγS. Fitting to a double exponential revealed that the fast phase in the ATP experiment (∼15% amplitude) had a time constant of ∼3 s and that the slow phase (∼85% amplitude) had a time constant of ∼50 s. Thus, most ClpXP–substrate complexes that form under the initial steady-state conditions of this assay dissociate relatively slowly. In ATPγS, the fast phase (17% amplitude) had a time constant of ∼2 s, and the slow phase (83% amplitude) had a time constant of ∼240 s. This result indicates that a substantial majority of ClpXP–substrate complexes formed in the presence of ATPγS are approximately fivefold more kinetically stable than those formed in ATP, whereas a minority of these complexes have comparable stabilities in the presence of ATP and ATPγS. As discussed below, these results can be explained if the initial association/dissociation of titin-BQ-ssrA is independent of the rate of nucleoside triphosphate hydrolysis, whereas a complex formed later in the engagement process is more stable when the hydrolysis rate is slower.

Fig. 4.

Dissociation kinetics. Complexes of ^FSClpXP (20 nM) and titin-BQ-ssrA (1 µM) were allowed to equilibrate in ATP (5 mM) or ATPγS (1 mM) and then mixed at time 0 with an equal volume of unlabeled titin-ssrA (50 µM) to initiate dissociation, and changes in fluorescence were recorded as a function of time. The black lines are simulations calculated using the first three steps of the Fig. 5 model and the parameters listed in Table 1.

A Multistep Binding and Engagement Model.

Fig. 5 shows a model that accounts well for the kinetic and binding results presented here and also is supported by cryo-EM structures of substrate-bound ClpXP (28–30). In this model, ClpXP associates with an ssrA-tagged substrate (k₁) to form an RC. In a subsequent step (k₂), ATP hydrolysis and translocation by ClpXP convert the RC into an IC in which the ssrA tag moves deeper into the channel, but the native portion of the substrate is not yet in contact with the ClpX ring. Additional translocation (k₃) brings the native portion of substrate in contact with the top of the ClpX ring to form an EC. From this EC state, repeated ATP-fueled power strokes can unfold the native substrate, resulting in a processive translocation complex (TC) that eventually results in substrate degradation in a multistep process subsumed under the k₄ rate constant. A final step (k₅) clears the axial channel of ClpX, regenerating free enzyme to bind another substrate. Although the overall model is relatively complex, simpler models containing fewer substrate-bound ClpX species were unable to account for our polyphasic kinetic results.

Fig. 5.

Model for substrate binding, engagement, unfolding, and processive translocation/degradation by ClpXP. As discussed in the text, the k₂, k₋₂, k₃, and k₋₃ rate constants increase as the rate of nucleoside triphosphate hydrolysis increases. The composite k₄ rate constant for ClpXP substrate unfolding and translocation also depends on the hydrolysis rate (21). ADP, adenosine diphosphate.

Fitting Association Kinetics to the Model.

ClpXP unfolding of the titin-BQ-ssrA substrate (included in the k₄ step) is expected to be very slow (18) compared with earlier steps in the model (Fig. 5). Thus, we globally fit the titin-BQ-ssrA association and dissociation data in the presence of ATP to determine rate constants for the k₁, k₋₁, k₂, k₋₂, k₃, and k₋₃ steps and quenching values for each state (Table 1). Independently, we also globally fit titin-BQ-ssrA association kinetics in the presence of ATPγS to obtain rate constants and quenching values (Table 1). Simulations based on the model and fitted rate constants and quenching values provided reasonable fits of the association data over a wide range of substrate concentrations and the dissociation data (compare black and colored lines in Figs. 3 A and B and 4).

Table 1.

Best-fit parameters obtained from global fitting of association and dissociation experiments for the titin-BQ-ssrA substrate with ATP (Figs. 3A and 4) or ATPγS (Figs. 3B and 4) using the model shown in Fig. 5

Reaction	Forward rate constant	Reverse rate constant, s−1	Molecule	Fluorescence intensity
FSClpXP, ATP, titin-BQ-ssrA
E + S ⇔ RC	k1 = 1.3 (1.2, 1.4) µM−1 s−1	k–1 = 0.48 (0.35, 1.1)	E	1.0 (defined)
RC ⇔ IC	k2 = 0.71 (0.23, 2.3) s−1	k–2 = 1.2 (0.30, 1.5)	RC	0.44 (0.41, 0.46)
IC ⇔ EC	k3 = 0.26 (0.13, 0.42) s−1	k–3 = 0.027 (0.024, 0.066)	IC	0.11 (0.00, 0.27)
			EC	0.58 (0.58, 6.3)
FSClpXP, ATPγS, titin-BQ-ssrA
E + S ⇔ RC	k1 = 1.4 (1.3, 1.5) µM−1 s−1	k–1 = 0.60 (0.51, 0.80)	E	1.0 (defined)
RC ⇔ IC	k2 = 0.15 (0.063, 0.26) s−1	k–2 = 0.27 (0.099, 0.47)	RC	0.43 (0.40, 0.45)
IC ⇔ EC	k3 = 0.051 (0.039, 0.061) s−1	k–3 = 0.0050 (0.0039, 0.0064)	IC	0.00 (0.00, 0.00)
			EC	0.35 (0.34, 0.38)

Lower and upper χ²_best/χ² = 0.95 boundaries of parameters are shown in parentheses. E, enzyme; S, substrate.

K_1/2 values calculated as (k₋₁/k₁)/(1 + (k₂/k₋₂) + (k₂k₃/k₋₂k₋₃)) from Table 1 parameters were in reasonable agreement with experimental values. For example, the calculated K_1/2 for ATP was 51 nM, similar to the experimental value of 53 nM. For ATPγS, the calculated K_1/2 was 59 nM, compared with the experimental value of 33 nM. Simulations of dissociation kinetics using parameters from the association fits also provided close agreement with the experimental biphasic dissociation trajectories (compare black and colored lines for experiments in Fig. 4), with the major slow dissociation phase arising predominantly from more prevalent ECs and the fast phase from lower concentrations of RC and IC. We do not believe that the Fig. 5 model is necessarily complete (extra steps are possible and would improve fitting) or that the fitted rate constants are highly accurate (errors of up to threefold in the values of certain rate constants were found by multidimensional search through parameter space). However, this minimal model and set of rate constants provide a plausible explanation for polyphasic association, polyphasic dissociation, and steady-state binding experiments.

The fitted rate constants for association (k₁) and dissociation (k₋₁) of titin-BQ-ssrA and ^FSClpXP were similar in the presence of ATP and ATPγS (Table 1), indicating that ATP hydrolysis does not play a substantial role in initial recognition and accounting for the fact that the fast phase of substrate dissociation in ATP and ATPγS occurs with similar kinetics (Fig. 4). For the k₂, k₋₂, k₃, and k₋₃ steps, by contrast, rate constants were substantially slower in the presence of ATPγS than ATP, suggesting that these steps involve posthydrolysis states after one or more power strokes.

Recognition, Engagement, and Degradation of Less Stable and Unfolded Substrates.

We also assayed the kinetics of ^FSClpXP association with different concentrations of ^V13Ptitin-BQ-ssrA, whose native structure is less stable than that of titin-BQ-ssrA, or ^UFtitin-BQ-ssrA, which is unfolded (Fig. 6 A and B). In both cases, the changes in fluorescence at early times (0 to 1 s) resembled those of titin-BQ-ssrA (Fig. 3A). At later times, differences between the three substrates were most pronounced at low substrate concentrations and could be attributed to faster degradation of the ^V13Ptitin-BQ-ssrA and ^UFtitin-BQ-ssrA substrates. Table 2 lists rate constants obtained by fitting the association experiments for ^V13Ptitin-BQ-ssrA and ^UFtitin-BQ-ssrA to the full model shown in Fig. 5, with initial values derived from the titin-BQ-ssrA model. The similarity of the early trajectories for the titin substrates, despite large differences in their stabilities and degradation rates, indicates that folding stability does not exert a major influence on the early steps of the binding/engagement process. For the ^V13Ptitin-BQ-ssrA substrates in the presence of ATP, we found that values for the k₁, k₋₁, k₂, k₋₂, k₃, and k₋₃ rate constants and fluorescent intensities were roughly similar to those in the titin-BQ-ssrA model (compare Tables 1 and 2) and could account for the experimental changes in fluorescence intensity. These values showed more variation for the unfolded titin substrate (Table 2). We did not attempt comprehensive error analysis for these fits because of the large number of parameters. We offer these fits as an illustration of a plausible model that can account for many of the kinetic features of substrate binding and degradation by ClpXP. Although binding and engagement are broadly similar for substrates with different stabilities, some differences may be expected. For example, the presence of native structure in the substrate may influence the kinetics and populations of microstates during the recognition and engagement steps, as cryo-EM structures show that substrate binding loops of ClpX contact native portions of substrates in complexes similar to the EC depicted in Fig. 5 (29). Because unfolding is not necessary for degradation of the ^UFtitin-BQ-ssrA substrate, the EC state and associated rate constants may also differ from those of the native substrates.

Fig. 6.

Association and degradation of less stable or unfolded titin substrates. (A) ^FSClpXP (10 nM) and ATP (2.5 mM) were mixed with the indicated concentrations of ^V13Ptitin-BQ-ssrA at time 0 to initiate binding. (B) Same as A but using the ^UFtitin-BQ-ssrA substrate. (C) Steady-state binding of ^UFtitin-BQ-ssrA to ^FSClpX (20 nM) in the presence of ATP (5 mM) or ATPγS (1 mM). (D) Changes in fluorescence during ^FSClpXP (10 nM) degradation of ^UFtitin-BQ-ssrA (0.1 or 20 µM) in the presence of ATPγS (0.5 mM). (E) Model for recognition, engagement, and processive translocation/degradation. The transition from one TC to the next (e.g., TC1 to TC2) is irreversible and involves a power stroke driven by ATP or ATPγS hydrolysis. Approximately 20 translocation steps would be required to fully translocate and degrade the titin substrate (5–9).

Table 2.

Best-fit parameters obtained from global fitting of association experiments with the ^V13Ptitin-BQ-ssrA substrate (Fig. 6A) and ^UFtitin-BQ-ssrA substrate (Fig. 6B) using the model shown in Fig. 5

Reaction	Forward rate constant	Reverse rate constant, s−1	Molecule	Fluorescence intensity
FSClpXP, ATP, V13Ptitin-BQ-ssrA
E + S ⇔ RC	k1 = 1.7 µM−1 s−1	k–1 = 0.45	E	1.0
RC ⇔ IC	k2 = 0.70 s−1	k–2 = 0.94	RC	0.44
IC ⇔ EC	k3 = 0.38 s−1	k–3 = 0.03	IC	0.14
EC ⇔ TC	k4 = 0.20 s−1	k–4 = 0.50	EC	0.59
TC → E	k5 = 0.53 s−1		TC	1.0
FSClpXP, ATP, UFtitin-BQ-ssrA
E + S ⇔ RC	k1 = 0.81 µM−1 s−1	k–1 = 0.87	E	1.0
RC ⇔ IC	k2 = 1.8 s−1	k–2 = 0.62	RC	0.39
IC ⇔ EC	k3 = 0.50 s−1	k–3 = 0.47	IC	0.0
EC ⇔ TC	k4 = 0.55 s−1	k–4 = 0.01	EC	0.9
TC → E	k5 = 1.35 s−1		TC	1.0

E, enzyme; S, substrate.

In the presence of ATP, ^FSClpX bound ^UFtitin-BQ-ssrA approximately fivefold weaker (K_1/2 = 550 nM) (Fig. 6C) than it bound titin-BQ-ssrA (K_1/2 = 110 nM) (Fig. 2B). This difference can be explained by faster substrate release of the unfolded substrate via processive translocation. If this model is correct, then slower translocation in the presence of ATPγS should reduce the observed K_1/2 differences. Indeed, ^FSClpX bound ^UFtitin-BQ-ssrA (K_1/2 = 90 nM) only ∼1.5-fold more weakly than native titin-BQ-ssrA (K_1/2 = 59 nM) in ATPγS. Comparable ^UFtitin-BQ-ssrA binding experiments using ^FSClpXP could not be performed, as rapid substrate degradation precluded steady-state binding.

To determine how impaired hydrolysis influences recognition, engagement, processive translocation, and degradation of the unfolded substrate, we conducted stopped-flow association experiments with titin^UF-BQ-ssrA and ^FSClpXP in the presence of ATPγS (Fig. 6D). Surprisingly, at high substrate concentrations, we observed oscillating fluorescence trajectories with a damped trend. At the highest substrate concentration, for example, the fluorescence initially decreased and then increased before decreasing, increasing, and decreasing again, with a periodicity of ∼200 s. This period corresponds roughly to the time required to bind, translocate, and degrade initially bound substrate and then begin this process anew but in a less synchronous manner. Oscillations are rare in biochemistry and require unusual kinetic mechanisms. One class of kinetic mechanisms that support population coherence, defined as many molecules in the same states at the same time, involves numerous irreversible steps, each of which occur on a timescale that is short compared with the time of the overall reaction. In our experiment, a plausible explanation for the oscillating fluorescence is that there are two sets of states with distinct fluorescence intensities, which can only interconvert in a step process. A likely possibility is that these states reflect initially bound and quenched states and a series of brighter translocating states, separated from each other by power strokes driven by irreversible ATPγS hydrolysis (Fig. 6E). By this model, the short steps are likely to reflect translocation of titin^UF-BQ-ssrA through the axial channel of ClpXP.

Discussion

We developed a fluorescence-quenching assay to monitor binding of ssrA-tagged substrates to ClpX or ClpXP. Initial substrate recognition requires ATP or ATPγS but is independent of the hydrolysis rate. A series of reversible and hydrolysis-dependent steps then converts the initial metastable RC into a more stable complex poised for substrate unfolding and subsequent processive translocation and degradation.

In titration experiments, K_1/2 constants for titin-BQ-ssrA binding to ^FSClpX varied from ∼30 to 100 nM. In previous studies, K_M for ClpXP degradation of titin-ssrA, by contrast, was ∼0.5 to 1.5 µM (18, 25). We suspect that the negatively charged and polyaromatic BQ quencher may stabilize binding by making additional interaction with ClpX. Intriguingly, titin-BQ-ssrA quenched ^FSClpX fluorescence to a greater extent in the presence of ATPγS than ATP. Because ClpX and ClpXP hydrolyze ATPγS more slowly than ATP (21), this result suggests that the rate of hydrolysis alters the distribution and/or fluorescent properties of substrate-bound conformations. In support of this model, maximum quenching for titin-AQ-ssrA binding to ^FSClpXP was similar in the presence of ATP and ATPγS (Fig. 3C), as expected if a larger Förster radius makes quenching for this enzyme–substrate pair more efficient over longer fluorophore–quencher distances.

We used the quenching assay to study the kinetics of substrate binding and engagement by ClpXP. Polyphasic fluorescence trajectories in association experiments using the titin-BQ-ssrA substrate (Fig. 3 A and B) could not be fit to a model with fewer than three pre-unfolding enzyme–substrate complexes, which we call RC, IC, and EC. A model including these complexes, each with its own characteristic mean fluorescence intensity, provided excellent fits of association kinetics over a 200-fold range of titin-BQ-ssrA concentrations in the presence of ATP (Fig. 3A). As in any multistep kinetic modeling scheme, more complicated models could fit the results equally well or better, and the same model with different rate constants can also provide adequate fits.

In our cartoon depictions of substrate-bound states (Figs. 5 and 6E), the native portion of the substrate is shown directly above the ClpXP channel. However, flexibility of the portion of the degron not in the ClpX channel in the RC and IC structures would allow a large ensemble of potential substrate–enzyme conformations to be explored. Thus, the fitted fluorescence intensities of the RC and IC structures probably depend on an average fluorophore–quencher distance in this distribution, which is predicted to be smaller in IC than RC. We hesitate to interpret the fluorescence of the EC state only in terms of average quencher–fluorophore distance, as the microenvironment of the quencher in this state is likely to affect energy transfer via orientation effects. For example, the negatively charged quencher is expected to be in or near the top of the ClpX axial channel, where it could interact with positively charged ClpX arginine-lysine-histidine loops, and the fluorophore on the top face of ClpX may contact the native titin domain.

Both titin-BQ-ssrA association and dissociation were slower in the presence of ATPγS than ATP, and the fitted values of the k₂, k₋₂, k₃, and k₋₃ rate constants from the ATPγS experiments were substantially smaller than those determined in the ATP experiments (Table 1). These results suggest that the rate of ATP/ATPγS hydrolysis is an important factor in determining the rates of the associated conformational transitions. Under the conditions of our experiments, the maximum steady-state rate of ATP hydrolysis by ClpXP is ∼3.6 s⁻¹ (28). The fitted values of the k₂, k₋₂, k₃, and k₋₃ rate constants in experiments performed using ATP were all smaller than this hydrolysis rate (Table 1), and thus, these steps might include multiple hydrolysis reactions and/or slower conformational changes that occur subsequent to hydrolysis.

The number of states uniquely resolvable in our stopped-flow experiments limits the complexity of our model. Moreover, kinetic states in the model may not be unique conformations but rather, ensembles of conformations explored in a semisynchronous manner upon rapid mixing. For example, in our fits, the fluorescence value of the EC is lower in ATPγS than in ATP. Thus, it is likely that this complex represents an ensemble average of low-fluorescence pre-ATP hydrolysis states (more highly populated in ATPγS) and higher-fluorescence post-ATP hydrolysis states (relatively more populated in ATP).

In experiments with natively destabilized or unfolded titin substrates, we explored the kinetic relationships between binding, engagement, and degradation. Binding and engagement of the destabilized ^V13Ptitin-BQ-ssrA substrate by ^FSClpXP resembled that of titin-BQ-ssrA. At high ^V13Ptitin-BQ-ssrA concentrations, transitions relating to binding and engagement were largely complete with 5 s of mixing, whereas unfolding, processive translocation, and degradation were substantially slower. Association experiments with ^UFtitin-BQ-ssrA and ^FSClpXP also revealed rapid association and engagement, relative to degradation. The intermediate steady-state fluorescence observed at late time points with high ^UFtitin-BQ-ssrA concentrations presumably reflects a mixture of different ClpXP–substrate complexes, distinct from the ensemble of states observed with titin-BQ-ssrA (prior to unfolding), and thus, is refractory to direct comparison.

Specific degron recognition by ClpXP and related AAA+ proteases is necessary to avoid uncontrolled protein destruction, as later steps including substrate unfolding and processive translocation through the axial channel have little specificity (18, 19, 33). The specificity of recognition and promiscuity of processive translocation seem at first to be at odds. A highly specific AAA+ protease might be expected to bind tightly to a limited set of sequences, potentially limiting processivity, whereas a highly processive protease that can grip most sequences for translocation might be expected to have limited specificity. As outlined below, the kinetic studies presented here together with recent cryo-EM structures of substrate-bound ClpXP (28–30) suggest a mechanism by which ClpXP achieves specificity and processivity.

Our studies provide evidence for three sequentially populated RC, IC, and EC complexes prior to substrate unfolding (Fig. 5). For the titin-BQ-ssrA substrate in the presence of ATP, the relative steady-state populations of RC, IC, and EC are ∼14, 8, and 78%, respectively. Thus, IC is a higher-energy state than RC, and EC is the lowest-energy state. RC is likely to be similar to a cryo-EM structure in which an ssrA degron binds in the top portion of an otherwise closed ClpX channel with specific packing and hydrogen-bond interactions between ClpX side chains and the Ala-Ala-coo^– of the ssrA tag (29). IC appears to correspond to a cryo-EM structure in which the ssrA degron moves six residues deeper into a now open ClpX channel, but the native portion of the substrate is not in contact with the top of the ClpX ring (29). EC appears to be similar to a cryo-EM structure in which the degron has moved deeply enough into the channel to draw the native portion of the substrate against the entrance to the ClpX channel (28). Substrate contacts in the channels of the IC-like and EC-like structures appear to be generally similar to each other and nonspecific (28, 29). Hence, the lower energy of EC relative to IC is likely to result from contacts between the native portion of the substrate and ClpX.

The RC-like conformation of ClpXP is ideally suited for “checking” short unstructured peptides in proteins for degrons, as contacts with the Ala-Ala-coo^– of the ssrA degron in this complex are sequence dependent (29). Although contacts in the downstream IC and EC complexes are nonspecific, our results indicate that both of these structures can revert to RC. For example, after titin-BQ-ssrA reaches IC, it is approximately fourfold more likely to revert back to RC than to move forward to EC. Even though the ssrA degron has evolved to ensure rapid ClpXP degradation of attached proteins, only ∼20% of titin-BQ-ssrA substrates that bind initially to form RC advance directly to EC, largely as a consequence of this back step. Because this reversible IC → RC step is ATP dependent, this cycle could potentially act as a type of kinetic proofreading (34) by providing multiple opportunities for substrate discrimination. Indeed, the multiple ATP hydrolysis-dependent steps that we observe prior to substrate engagement should ensure that only substrates with proper degrons have a substantial chance of being unfolded and degraded. Thus, degron interactions in the RC-like structure appear to mediate ClpXP specificity, whereas ATP-dependent translocation of the degron deeper into the axial channel allows ClpXP to reversibly transition into a more promiscuous and ultimately, processive unfoldase/translocase. We note that the detailed mechanism of substrate engagement by ClpXP is likely to depend upon degron length (29), and it will be important to determine how changes in tag length and sequence affect mechanism using the methods developed here. Other AAA+ proteases may use similar multistep engagement mechanisms to ensure specific and robust destruction of their targets.

Methods

Proteins.

All proteins contained His₆ tags, were expressed from plasmids in E. coli, and were purified using established procedures (35, 36). A single-chain E. coli ^C169SClpX^∆N pseudohexamer with a single solvent-exposed Cys¹⁷⁰ residue in subunit 1 was constructed by sortase-mediated ligation of appropriate single-chain trimers as described (36). This pseudohexamer was labeled with TAMRA-maleimide (ThermoFisher Scientific) to produce ^FSClpX. BHQ10 maleimide was prepared from BHQ10 succinimidyl ester (Biosearch Technologies) as described for the related BHQ3 molecule (36). Our ssrA-tagged titin I27 construct contained the N-terminal sequence Met-His-Glu-Gly, residues 1 to 89 of the I27 domain of human titin (National Center for Biotechnology Information accession no. 1530732605 or 6I0Y_z), followed by the sequence Gly-Cys-Gly-(His)₆-Ala-Ala-Asn-Asp-Glu-Asn-Tyr-Ala-Leu-Ala-Ala (ssrA tag underlined). To produce titin-BQ-ssrA, this protein was labeled for 30 min at room temperature with three molar equivalents of BHQ10 maleimide in buffer A (20 mM N-(2-hydroxyethyl)piperazine-N-2-ethanesulfonic acid [Hepes], pH 7.7, 30 mM KCl, 0.1 mM ethylenediaminetetraacetic acid [EDTA], 10% glycerol) supplemented with 5 mM Tris, pH 8.0. Following labeling, the protein was desalted into buffer A using a PD10 column, and titin-BQ-ssrA was separated from unlabeled protein by SOURCE-15Q chromatography using a gradient from 0 to 40% buffer B (20 mM Hepes, pH 7.7, 1 M KCl, 0.1 mM EDTA, 10% glycerol). Fractions containing pure titin-BQ-ssrA were pooled, concentrated, and flash frozen. To produce titin-AQ-ssrA, our ssrA-tagged titin protein was reacted at room temperature with three equivalents of AQ maleimide (ATTO-TEC; GMBH) for 30 min in buffer A supplemented with 5 mM Tris, pH 8.0. The reaction was quenched by addition of 15 mM dithiothreitol, desalted using a PD-10 column, desalted further by Sephadex G-25 chromatography, then concentrated, and flash frozen. The concentrations of titin-BQ-ssrA and titin-AQ-ssrA were measured by the BCA assay (ThermoFisher Scientific).

Binding and Kinetic Assays.

Equilibrium and kinetic experiments as well as degradation assays were performed at room temperature in PD buffer (25 mM Hepes, pH 7.5, 100 mM KCl, 10 mM MgCl₂, 0.1 mM EDTA, 10% glycerol) supplemented with ATP (5 mM) or ATPγS (1 mM) as indicated. Experiments performed with ATP also contained a creatine/creatine-kinase regeneration system (16 mM phosphocreatine, 0.032 mg/mL creatine kinase; Sigma). For equilibrium binding, fixed concentrations of ^FSClpX (typically 20 nM) without or with ClpP (60 nM) were initially mixed in the absence of nucleotide for 2 min with different concentrations of titin-BQ-ssrA or titin-AQ-ssrA in a microtiter plate. Under these conditions, ^FSClpX does not bind substrate. Fluorescence emission spectra from 560 to 590 nm (excitation 530 nm) were recorded using a SpectraMax M5 plate reader (Molecular Devices). Next, ATP or ATPγS was added. After incubation for 5 min (ATP experiments) or 15 min (ATPγS experiments), a second fluorescence emission spectrum was recorded. Areas under emission curves were calculated without nucleotide (F_buf) and with nucleotide (F_nuc), and the equilibrium fluorescence quenching value was defined as 1 − F_nuc/F_buf − Q_buf, where Q_buf was the slight increase in fluorescence that occurred with dilution of the nucleotide-free control.

Kinetic experiments were performed using an SF-300X stopped-flow instrument (KinTek). For association kinetics, one syringe contained a fixed concentration of ^FSClpX, ClpP, ATP, or ATPγS, and a second syringe contained varying concentrations of titin-BQ-ssrA or titin-AQ-ssrA. For dissociation experiments, titin-BQ-ssrA was incubated for 15 min with ^FSClpX, ClpP, and ATP or ATPγS and loaded into one syringe. The second syringe contained 50 µM unlabeled I27 titin-ssrA. Following mixing of the syringe contents (dead time ∼ 1 ms), the fluorescence trajectory was recorded (excitation 550 nm; emission passed through a Newport 580-/10-nm band-pass filter).

Stopped-flow kinetic data were globally fit using KinTek Explorer software (KinTek). The general model was developed from the ClpXP and titin-BQ-ssrA association and dissociation experiments conducted in ATP (Fig. 3A). We adjusted the dissociation quenching data by a linear factor to account for day-to-day changes in fluorimeter values. The fluorescence traces were fit to models with one, two, three, and four bound states, each with distinct fluorescence intensities. For simplicity, strictly linear progression or retrogression between states was enforced. Forward and reverse rate constants and state-specific fluorescent intensities were allowed to vary freely during the fitting process, and the parameter space was comprehensively explored. We were not able to obtain satisfactory fits with models that included fewer than three bound states.

We fit the data from the ClpXP/titin-BQ-ssrA/ATPγS association experiments (Fig. 3B) to the same model with three substrate bound states. To obtain satisfactory fits, we needed to vary not just the rate constants controlling the interchange between states but also, the state-specific fluorescent intensities, suggesting that the ensembles of conformations reflected by each bound state differ between ATP and ATPγS.

Satisfactory fits of data from ^V13Ptitin-BQ-ssrA and ^UFtitin-BQ-ssrA association/degradation experiments (Fig. 6 A and B, respectively) required an additional bound state and an irreversible proteolytic step. We maintained the rates and fluorescent intensities for the bimolecular association and the interconversion between the first three bound states as close to those of the titin-BQ-ssrA model as possible while maintaining a satisfactory fit. Models with many parameters have many possible solutions, and thus, our goal was to find example parameter combinations that were broadly consistent with the experimental data.

Error analysis was conducted with the Kintek FitSpace Explorer 2D feature, which assesses the extent to which pairwise combinations of parameters can covary and maintain a good fit. For each parameter, we identified the minimum and maximum values that, with all other values floating independently, enabled a fit 95% as good as our best fit, as assessed by χ²_best/χ². We did not conduct error analysis with the ^V13Ptitin-BQ-ssrA and ^UFtitin-BQ-ssrA experiments because the parameters are substantially underconstrained.

Data Availability.

All study data are included in the article.